Fixes build for no-default-features

* Move for_each functions from Scorer to Weight.

* Specialized foreach / foreach_pruning for union of termscorer.

.github/FUNDING.yml

@@ -0,0 +1,12 @@
+# These are supported funding model platforms
+
+github: fulmicoton
+patreon: # Replace with a single Patreon username
+open_collective: # Replace with a single Open Collective username
+ko_fi: # Replace with a single Ko-fi username
+tidelift: # Replace with a single Tidelift platform-name/package-name e.g., npm/babel
+community_bridge: # Replace with a single Community Bridge project-name e.g., cloud-foundry
+liberapay: # Replace with a single Liberapay username
+issuehunt: # Replace with a single IssueHunt username
+otechie: # Replace with a single Otechie username
+custom: # Replace with up to 4 custom sponsorship URLs e.g., ['link1', 'link2']

.travis.yml

@@ -10,7 +10,7 @@ env:
   global:
     - CRATE_NAME=tantivy
     - TRAVIS_CARGO_NIGHTLY_FEATURE=""
-    - secure: eC8HjTi1wgRVCsMAeXEXt8Ckr0YBSGOEnQkkW4/Nde/OZ9jJjz2nmP1ELQlDE7+czHub2QvYtDMG0parcHZDx/Kus0yvyn08y3g2rhGIiE7y8OCvQm1Mybu2D/p7enm6shXquQ6Z5KRfRq+18mHy80wy9ABMA/ukEZdvnfQ76/Een8/Lb0eHaDoXDXn3PqLVtByvSfQQ7OhS60dEScu8PWZ6/l1057P5NpdWbMExBE7Ro4zYXNhkJeGZx0nP/Bd4Jjdt1XfPzMEybV6NZ5xsTILUBFTmOOt603IsqKGov089NExqxYu5bD3K+S4MzF1Nd6VhomNPJqLDCfhlymJCUj5n5Ku4yidlhQbM4Ej9nGrBalJnhcjBjPua5tmMF2WCxP9muKn/2tIOu1/+wc0vMf9Yd3wKIkf5+FtUxCgs2O+NslWvmOMAMI/yD25m7hb4t1IwE/4Bk+GVcWJRWXbo0/m6ZUHzRzdjUY2a1qvw7C9udzdhg7gcnXwsKrSWi2NjMiIVw86l+Zim0nLpKIN41sxZHLaFRG63Ki8zQ/481LGn32awJ6i3sizKS0WD+N1DfR2qYMrwYHaMN0uR0OFXYTJkFvTFttAeUY3EKmRKAuMhmO2YRdSr4/j/G5E9HMc1gSGJj6PxgpQU7EpvxRsmoVAEJr0mszmOj9icGHep/FM=
+    # - secure: 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
 
 addons:
   apt:
@@ -38,20 +38,21 @@ matrix:
     # Linux
     #- env: TARGET=aarch64-unknown-linux-gnu
     #- env: TARGET=i686-unknown-linux-gnu
-    - env: TARGET=x86_64-unknown-linux-gnu CODECOV=1
+    - env: TARGET=x86_64-unknown-linux-gnu CODECOV=1 #UPLOAD_DOCS=1
     # - env: TARGET=x86_64-unknown-linux-musl CODECOV=1
-
     # OSX
-    - env: TARGET=x86_64-apple-darwin
-      os: osx
+    #- env: TARGET=x86_64-apple-darwin
+    #  os: osx
 
 before_install:
   - set -e
   - rustup self update
+  - rustup component add rustfmt
 
 install:
   - sh ci/install.sh
   - source ~/.cargo/env || true
+  - env | grep "TRAVIS"
 
 before_script:
   - export PATH=$HOME/.cargo/bin:$PATH
@@ -60,14 +61,25 @@ before_script:
 
 script:
   - bash ci/script.sh
+  - cargo fmt --all -- --check
 
 before_deploy:
   - sh ci/before_deploy.sh
 
-cache: cargo
-before_cache:
-  # Travis can't cache files that are not readable by "others"
-  - chmod -R a+r $HOME/.cargo
+after_success:
+  # Needs GH_TOKEN env var to be set in travis settings
+  - if [[ -v GH_TOKEN ]]; then echo "GH TOKEN IS SET"; else echo "GH TOKEN NOT SET"; fi
+  - if [[ -v UPLOAD_DOCS ]]; then cargo doc; cargo doc-upload; else echo "doc upload disabled."; fi
+
+#cache: cargo
+#before_cache:
+#  # Travis can't cache files that are not readable by "others"
+#  - chmod -R a+r $HOME/.cargo
+#  - find ./target/debug -type f -maxdepth 1 -delete
+#  - rm -f  ./target/.rustc_info.json
+#  - rm -fr ./target/debug/{deps,.fingerprint}/tantivy*
+#  - rm -r target/debug/examples/
+#  - ls -1 examples/ | sed -e 's/\.rs$//' | xargs -I "{}" find target/* -name "*{}*" -type f -delete
 
 #branches:
 #  only:
@@ -77,4 +89,4 @@ before_cache:
 
 notifications:
   email:
-    on_success: never
+    on_success: never

CHANGELOG.md

@@ -1,5 +1,200 @@
+Tantivy 0.13.0
+======================
+- Bugfix in `FuzzyTermQuery` not matching terms by prefix when it should (@Peachball)
+- Relaxed constraints on the custom/tweak score functions. At the segment level, they can be mut, and they are not required to be Sync + Send.
+- `MMapDirectory::open` does not return a `Result` anymore.
+- Change in the DocSet and Scorer API. (@fulmicoton). 
+A freshly created DocSet point directly to their first doc. A sentinel value called TERMINATED marks the end of a DocSet.
+`.advance()` returns the new DocId. `Scorer::skip(target)` has been replaced by `Scorer::seek(target)` and returns the resulting DocId.
+As a result, iterating through DocSet now looks as follows
+```rust
+let mut doc = docset.doc();
+while doc != TERMINATED {
+   // ...
+   doc = docset.advance();
+}
+```
+The change made it possible to greatly simplify a lot of the docset's code.
+- Misc internal optimization and introduction of the `Scorer::for_each_pruning` function. (@fulmicoton)
+- Added an offset option to the Top(.*)Collectors. (@robyoung)
+
+Tantivy 0.12.0
+======================
+- Removing static dispatch in tokenizers for simplicity. (#762)
+- Added backward iteration for `TermDictionary` stream. (@halvorboe)
+- Fixed a performance issue when searching for the posting lists of a missing term (@audunhalland)
+- Added a configurable maximum number of docs (10M by default) for a segment to be considered for merge (@hntd187, landed by @halvorboe #713) 
+- Important Bugfix #777, causing tantivy to retain memory mapping. (diagnosed by @poljar)
+- Added support for field boosting. (#547, @fulmicoton)
+
+## How to update?
+
+Crates relying on custom tokenizer, or registering tokenizer in the manager will require some 
+minor changes. Check https://github.com/tantivy-search/tantivy/blob/master/examples/custom_tokenizer.rs
+to check for some code sample.
+
+Tantivy 0.11.3
+=======================
+- Fixed DateTime as a fast field (#735)
+
+Tantivy 0.11.2
+=======================
+- The future returned by `IndexWriter::merge` does not borrow `self` mutably anymore (#732)
+- Exposing a constructor for `WatchHandle` (#731)
+
+Tantivy 0.11.1
+=====================
+- Bug fix #729
+
+
+Tantivy 0.11.0
+=====================
+
+- Added f64 field. Internally reuse u64 code the same way i64 does (@fdb-hiroshima)
+- Various bugfixes in the query parser.
+    - Better handling of hyphens in query parser. (#609)
+    - Better handling of whitespaces.
+- Closes #498 - add support for Elastic-style unbounded range queries for alphanumeric types eg. "title:>hello", "weight:>=70.5", "height:<200" (@petr-tik)
+- API change around `Box<BoxableTokenizer>`. See detail in #629
+- Avoid rebuilding Regex automaton whenever a regex query is reused. #639 (@brainlock)
+- Add footer with some metadata to index files. #605 (@fdb-hiroshima)
+- Add a method to check the compatibility of the footer in the index with the running version of tantivy (@petr-tik)
+- TopDocs collector: ensure stable sorting on equal score. #671 (@brainlock)
+- Added handling of pre-tokenized text fields (#642), which will enable users to
+  load tokens created outside tantivy. See usage in examples/pre_tokenized_text. (@kkoziara)
+- Fix crash when committing multiple times with deleted documents. #681 (@brainlock)
+
+## How to update?
+
+- The index format is changed. You are required to reindex your data to use tantivy 0.11. 
+- `Box<dyn BoxableTokenizer>` has been replaced by a `BoxedTokenizer` struct.
+- Regex are now compiled when the `RegexQuery` instance is built. As a result, it can now return
+an error and handling the `Result` is required.
+- `tantivy::version()` now returns a `Version` object. This object implements `ToString()`
+
+Tantivy 0.10.2
+=====================
+
+- Closes #656. Solving memory leak.
+
+Tantivy 0.10.1
+=====================
+
+- Closes #544.  A few users experienced problems with the directory watching system.
+Avoid watching the mmap directory until someone effectively creates a reader that uses
+this functionality.
+
+
+Tantivy 0.10.0
+=====================
+
+*Tantivy 0.10.0 index format is compatible with the index format in 0.9.0.*
+
+- Added an API to easily tweak or entirely replace the 
+ default score. See `TopDocs::tweak_score`and `TopScore::custom_score` (@pmasurel)
+- Added an ASCII folding filter (@drusellers)
+- Bugfix in `query.count` in presence of deletes (@pmasurel)
+- Added `.explain(...)` in `Query` and `Weight` to (@pmasurel)
+- Added an efficient way to `delete_all_documents` in `IndexWriter` (@petr-tik). 
+  All segments are simply removed.
+
+Minor
+---------
+- Switched to Rust 2018 (@uvd)
+- Small simplification of the code. 
+Calling .freq() or .doc() when .advance() has never been called
+on segment postings should panic from now on.
+- Tokens exceeding `u16::max_value() - 4` chars are discarded silently instead of panicking.
+- Fast fields are now preloaded when the `SegmentReader` is created.
+- `IndexMeta` is now public.  (@hntd187)
+- `IndexWriter` `add_document`, `delete_term`. `IndexWriter` is `Sync`, making it possible to use it with a `
+Arc<RwLock<IndexWriter>>`. `add_document` and `delete_term` can 
+only require a read lock. (@pmasurel)
+- Introducing `Opstamp` as an expressive type alias for `u64`. (@petr-tik)
+- Stamper now relies on `AtomicU64` on all platforms (@petr-tik)
+- Bugfix - Files get deleted slightly earlier
+- Compilation resources improved (@fdb-hiroshima)
+
+## How to update?
+
+Your program should be usable as is.
+
+### Fast fields
+
+Fast fields used to be accessed directly from the `SegmentReader`.
+The API changed, you are now required to acquire your fast field reader via the
+`segment_reader.fast_fields()`, and use one of the typed method: 
+- `.u64()`, `.i64()` if your field is single-valued ;
+- `.u64s()`, `.i64s()` if your field is multi-valued ;
+- `.bytes()` if your field is bytes fast field.
+
+
+
+Tantivy 0.9.0
+=====================
+*0.9.0 index format is not compatible with the 
+previous index format.*
+- MAJOR BUGFIX : 
+  Some `Mmap` objects were being leaked, and would never get released. (@fulmicoton)
+- Removed most unsafe (@fulmicoton)
+- Indexer memory footprint improved. (VInt comp, inlining the first block. (@fulmicoton)
+- Stemming in other language possible (@pentlander)
+- Segments with no docs are deleted earlier (@barrotsteindev)
+- Added grouped add and delete operations. 
+  They are guaranteed to happen together (i.e. they cannot be split by a commit). 
+  In addition, adds are guaranteed to happen on the same segment. (@elbow-jason)
+- Removed `INT_STORED` and `INT_INDEXED`. It is now possible to use `STORED` and `INDEXED`
+  for int fields. (@fulmicoton)
+- Added DateTime field (@barrotsteindev)
+- Added IndexReader. By default, index is reloaded automatically upon new commits (@fulmicoton)
+- SIMD linear search within blocks (@fulmicoton)
+
+## How to update ?
+
+tantivy 0.9 brought some API breaking change.
+To update from tantivy 0.8, you will need to go through the following steps.
+
+- `schema::INT_INDEXED` and `schema::INT_STORED`  should be replaced by `schema::INDEXED` and `schema::INT_STORED`.
+- The index now does not hold the pool of searcher anymore. You are required to create an intermediary object called 
+`IndexReader` for this. 
+    
+    ```rust
+    // create the reader. You typically need to create 1 reader for the entire
+    // lifetime of you program.
+    let reader = index.reader()?;
+    
+    // Acquire a searcher (previously `index.searcher()`) is now written:
+    let searcher = reader.searcher();
+    
+    // With the default setting of the reader, you are not required to 
+    // call `index.load_searchers()` anymore.
+    //
+    // The IndexReader will pick up that change automatically, regardless
+    // of whether the update was done in a different process or not.
+    // If this behavior is not wanted, you can create your reader with 
+    // the `ReloadPolicy::Manual`, and manually decide when to reload the index
+    // by calling `reader.reload()?`.
+  
+    ```
+
+
+Tantivy 0.8.2
+=====================
+Fixing build for x86_64 platforms. (#496)
+No need to update from 0.8.1 if tantivy
+is building on your platform.
+
+
 Tantivy 0.8.1
 =====================
+Hotfix of #476.
+
+Merge was reflecting deletes before commit was passed. 
+Thanks @barrotsteindev  for reporting the bug.
+
+
+Tantivy 0.8.0
+=====================
 *No change in the index format*
 - API Breaking change in the collector API. (@jwolfe, @fulmicoton)
 - Multithreaded search (@jwolfe, @fulmicoton) 

Cargo.toml

@@ -1,61 +1,67 @@
 [package]
 name = "tantivy"
-version = "0.8.0-dev"
+version = "0.12.0"
 authors = ["Paul Masurel <paul.masurel@gmail.com>"]
 license = "MIT"
 categories = ["database-implementations", "data-structures"]
 description = """Search engine library"""
-documentation = "https://tantivy-search.github.io/tantivy/tantivy/index.html"
+documentation = "https://docs.rs/tantivy/"
 homepage = "https://github.com/tantivy-search/tantivy"
 repository = "https://github.com/tantivy-search/tantivy"
 readme = "README.md"
 keywords = ["search", "information", "retrieval"]
+edition = "2018"
 
 [dependencies]
-base64 = "0.10.0"
+base64 = "0.12.0"
 byteorder = "1.0"
-lazy_static = "1"
-regex = "1.0"
-fst = {version="0.3", default-features=false}
-fst-regex = { version="0.2" }
+crc32fast = "1.2.0"
+once_cell = "1.0"
+regex ={version = "1.3.0", default-features = false, features = ["std"]}
+tantivy-fst = "0.3"
+memmap = {version = "0.7", optional=true}
 lz4 = {version="1.20", optional=true}
-snap = {version="0.2"}
+snap = "1"
 atomicwrites = {version="0.2.2", optional=true}
 tempfile = "3.0"
 log = "0.4"
-combine = "3"
-tempdir = "0.3"
-serde = "1.0"
-serde_derive = "1.0"
+serde = {version="1.0", features=["derive"]}
 serde_json = "1.0"
 num_cpus = "1.2"
-itertools = "0.7"
-levenshtein_automata = {version="0.1", features=["fst_automaton"]}
-bit-set = "0.5"
-uuid = { version = "0.7", features = ["v4", "serde"] }
-crossbeam = "0.5"
-futures = "0.1"
-futures-cpupool = "0.1"
+fs2={version="0.4", optional=true}
+levenshtein_automata = "0.2"
+notify = {version="4", optional=true}
+uuid = { version = "0.8", features = ["v4", "serde"] }
+crossbeam = "0.7"
+futures = {version = "0.3",  features=["thread-pool"] }
 owning_ref = "0.4"
 stable_deref_trait = "1.0.0"
-rust-stemmers = "1"
-downcast = { version="0.9" }
-matches = "0.1"
-bitpacking = "0.5"
-census = "0.1"
+rust-stemmers = "1.2"
+downcast-rs = { version="1.0" }
+tantivy-query-grammar = { version="0.13", path="./query-grammar" }
+bitpacking = {version="0.8", default-features = false, features=["bitpacker4x"]}
+census = "0.4"
 fnv = "1.0.6"
 owned-read = "0.4"
 failure = "0.1"
 htmlescape = "0.3.1"
-fail = "0.2"
-scoped-pool = "1.0"
+fail = "0.4"
+murmurhash32 = "0.2"
+chrono = "0.4"
+smallvec = "1.0"
+rayon = "1"
 
 [target.'cfg(windows)'.dependencies]
-winapi = "0.2"
+winapi = "0.3"
 
 [dev-dependencies]
-rand = "0.6"
+rand = "0.7"
 maplit = "1"
+matches = "0.1.8"
+
+[dev-dependencies.fail]
+version = "0.4"
+features = ["failpoints"]
 
 [profile.release]
 opt-level = 3
@@ -67,14 +73,27 @@ debug-assertions = true
 overflow-checks = true
 
 [features]
-# by default no-fail is disabled. We manually enable it when running test.
-default = ["mmap", "no_fail"]
-mmap = ["fst/mmap", "atomicwrites"]
+default = ["mmap"]
+mmap = ["atomicwrites", "fs2", "memmap", "notify"]
 lz4-compression = ["lz4"]
-no_fail = ["fail/no_fail"]
+failpoints = ["fail/failpoints"]
 unstable = [] # useful for benches.
+wasm-bindgen = ["uuid/wasm-bindgen"]
+
+[workspace]
+members = ["query-grammar"]
 
 [badges]
 travis-ci = { repository = "tantivy-search/tantivy" }
 
-
+# Following the "fail" crate best practises, we isolate
+# tests that define specific behavior in fail check points
+# in a different binary.
+#
+# We do that because, fail rely on a global definition of
+# failpoints behavior and hence, it is incompatible with
+# multithreading.
+[[test]]
+name = "failpoints"
+path = "tests/failpoints/mod.rs"
+required-features = ["fail/failpoints"]

Makefile

@@ -0,0 +1,3 @@
+test:
+	echo "Run test only... No examples."
+	cargo test --tests --lib

README.md

@@ -4,6 +4,7 @@
 [![Join the chat at https://gitter.im/tantivy-search/tantivy](https://badges.gitter.im/tantivy-search/tantivy.svg)](https://gitter.im/tantivy-search/tantivy?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=badge)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
 [![Build status](https://ci.appveyor.com/api/projects/status/r7nb13kj23u8m9pj/branch/master?svg=true)](https://ci.appveyor.com/project/fulmicoton/tantivy/branch/master)
+[![Crates.io](https://img.shields.io/crates/v/tantivy.svg)](https://crates.io/crates/tantivy)
 [![Say Thanks!](https://img.shields.io/badge/Say%20Thanks-!-1EAEDB.svg)](https://saythanks.io/to/fulmicoton)
 
 ![Tantivy](https://tantivy-search.github.io/logo/tantivy-logo.png)
@@ -17,74 +18,123 @@
 [![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/6)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/6)
 [![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/7)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/7)
 
+[![Become a patron](https://c5.patreon.com/external/logo/become_a_patron_button.png)](https://www.patreon.com/fulmicoton)
 
 
-**Tantivy** is a **full text search engine library** written in rust.
+**Tantivy** is a **full text search engine library** written in Rust.
 
-It is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elastic Search](https://www.elastic.co/products/elasticsearch) and [Apache Solr](https://lucene.apache.org/solr/) in the sense it is not
+It is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elasticsearch](https://www.elastic.co/products/elasticsearch) or [Apache Solr](https://lucene.apache.org/solr/) in the sense it is not
 an off-the-shelf search engine server, but rather a crate that can be used
 to build such a search engine.
 
 Tantivy is, in fact, strongly inspired by Lucene's design.
 
+# Benchmark
+
+The following [benchmark](https://tantivy-search.github.io/bench/) break downs 
+performance for different type of queries / collection.
+
+
+In general, Tantivy tends to be 
+- slower than Lucene on union with a Top-K due to Block-WAND optimization.
+- faster than Lucene on intersection and phrase queries. 
+
+Your mileage WILL vary depending on the nature of queries and their load.
+
 # Features
 
 - Full-text search
+- Configurable tokenizer (stemming available for 17 Latin languages with third party support for Chinese ([tantivy-jieba](https://crates.io/crates/tantivy-jieba) and [cang-jie](https://crates.io/crates/cang-jie)), Japanese ([lindera](https://github.com/lindera-morphology/lindera-tantivy) and [tantivy-tokenizer-tiny-segmente](https://crates.io/crates/tantivy-tokenizer-tiny-segmenter)) and Korean ([lindera](https://github.com/lindera-morphology/lindera-tantivy) + [lindera-ko-dic-builder](https://github.com/lindera-morphology/lindera-ko-dic-builder))
 - Fast (check out the :racehorse: :sparkles: [benchmark](https://tantivy-search.github.io/bench/) :sparkles: :racehorse:)
 - Tiny startup time (<10ms), perfect for command line tools
-- BM25 scoring (the same as lucene)
-- Natural query language `(michael AND jackson) OR "king of pop"`
-- Phrase queries search (`"michael jackson"`)
+- BM25 scoring (the same as Lucene)
+- Natural query language (e.g. `(michael AND jackson) OR "king of pop"`)
+- Phrase queries search (e.g. `"michael jackson"`)
 - Incremental indexing
 - Multithreaded indexing (indexing English Wikipedia takes < 3 minutes on my desktop)
 - Mmap directory
-- SIMD integer compression when the platform/CPU includes the SSE2 instruction set.
-- Single valued and multivalued u64 and i64 fast fields (equivalent of doc values in Lucene)
+- SIMD integer compression when the platform/CPU includes the SSE2 instruction set
+- Single valued and multivalued u64, i64, and f64 fast fields (equivalent of doc values in Lucene)
 - `&[u8]` fast fields
+- Text, i64, u64, f64, dates, and hierarchical facet fields
 - LZ4 compressed document store
 - Range queries
 - Faceted search
 - Configurable indexing (optional term frequency and position indexing)
 - Cheesy logo with a horse
 
-# Non-features
+## Non-features
 
-- Distributed search is out of the scope of tantivy. That being said, tantivy is meant as a
+- Distributed search is out of the scope of Tantivy. That being said, Tantivy is a
 library upon which one could build a distributed search. Serializable/mergeable collector state for instance, 
-are within the scope of tantivy.
+are within the scope of Tantivy.
 
 
-# Supported OS and compiler
-
-Tantivy works on stable rust (>= 1.27) and supports Linux, MacOS and Windows.
-
 # Getting started
 
-- [tantivy's simple search example](http://fulmicoton.com/tantivy-examples/simple_search.html)
-- [tantivy-cli and its tutorial](https://github.com/tantivy-search/tantivy-cli).
-`tantivy-cli` is an actual command line interface that makes it easy for you to create a search engine,
-index documents and search via the CLI or a small server with a REST API.
-It will walk you through getting a wikipedia search engine up and running in a few minutes.
-- [reference doc]
-    - [For the last released version](https://docs.rs/tantivy/)
-    - [For the last master branch](https://tantivy-search.github.io/tantivy/tantivy/index.html)
+Tantivy works on stable Rust (>= 1.27) and supports Linux, MacOS, and Windows.
 
-# Compiling
+- [Tantivy's simple search example](https://tantivy-search.github.io/examples/basic_search.html)
+- [tantivy-cli and its tutorial](https://github.com/tantivy-search/tantivy-cli) - `tantivy-cli` is an actual command line interface that makes it easy for you to create a search engine,
+index documents, and search via the CLI or a small server with a REST API.
+It walks you through getting a wikipedia search engine up and running in a few minutes.
+- [Reference doc for the last released version](https://docs.rs/tantivy/)
 
-## Development
+# How can I support this project?
 
-Tantivy compiles on stable rust but requires `Rust >= 1.27`.
-To check out and run tests, you can simply run :
+There are many ways to support this project. 
 
-    git clone git@github.com:tantivy-search/tantivy.git
+- Use Tantivy and tell us about your experience on [Gitter](https://gitter.im/tantivy-search/tantivy) or by email (paul.masurel@gmail.com)
+- Report bugs
+- Write a blog post
+- Help with documentation by asking questions or submitting PRs
+- Contribute code (you can join [our Gitter](https://gitter.im/tantivy-search/tantivy))
+- Talk about Tantivy around you
+- Drop a word on on [![Say Thanks!](https://img.shields.io/badge/Say%20Thanks-!-1EAEDB.svg)](https://saythanks.io/to/fulmicoton) or even [![Become a patron](https://c5.patreon.com/external/logo/become_a_patron_button.png)](https://www.patreon.com/fulmicoton)
+
+# Contributing code
+
+We use the GitHub Pull Request workflow: reference a GitHub ticket and/or include a comprehensive commit message when opening a PR.
+
+## Clone and build locally
+
+Tantivy compiles on stable Rust but requires `Rust >= 1.27`.
+To check out and run tests, you can simply run:
+
+```bash
+    git clone https://github.com/tantivy-search/tantivy.git
     cd tantivy
     cargo build
+```
 
-## Running tests
+## Run tests
 
 Some tests will not run with just `cargo test` because of `fail-rs`.
-To run the tests exhaustively, run `./run-tests.sh`. 
+To run the tests exhaustively, run `./run-tests.sh`.
 
-# Contribute
+## Debug
 
-Send me an email (paul.masurel at gmail.com) if you want to contribute to tantivy.
+You might find it useful to step through the programme with a debugger.
+
+### A failing test
+
+Make sure you haven't run `cargo clean` after the most recent `cargo test` or `cargo build` to guarantee that the `target/` directory exists. Use this bash script to find the name of the most recent debug build of Tantivy and run it under `rust-gdb`:
+
+```bash
+find target/debug/ -maxdepth 1 -executable -type f -name "tantivy*" -printf '%TY-%Tm-%Td %TT %p\n' | sort -r | cut -d " " -f 3 | xargs -I RECENT_DBG_TANTIVY rust-gdb RECENT_DBG_TANTIVY
+```
+
+Now that you are in `rust-gdb`, you can set breakpoints on lines and methods that match your source code and run the debug executable with flags that you normally pass to `cargo test` like this:
+
+```bash
+$gdb run --test-threads 1 --test $NAME_OF_TEST
+```
+
+### An example
+
+By default, `rustc` compiles everything in the `examples/` directory in debug mode. This makes it easy for you to make examples to reproduce bugs:
+
+```bash
+rust-gdb target/debug/examples/$EXAMPLE_NAME
+$ gdb run
+```

appveyor.yml

@@ -18,5 +18,5 @@ install:
 build: false
 
 test_script:
-  - REM SET RUST_LOG=tantivy,test & cargo test --verbose --no-default-features --features mmap -- --test-threads 1
+  - REM SET RUST_LOG=tantivy,test & cargo test --all --verbose --no-default-features --features mmap
   - REM SET RUST_BACKTRACE=1 & cargo build --examples

ci/script.sh

@@ -7,7 +7,7 @@ set -ex
 main() {
     if [ ! -z $CODECOV ]; then
         echo "Codecov"
-        cargo build --verbose && cargo coverage --verbose && bash <(curl -s https://codecov.io/bash) -s target/kcov
+        cargo build --verbose && cargo coverage --verbose --all && bash <(curl -s https://codecov.io/bash) -s target/kcov
     else
         echo "Build"
         cross build --target $TARGET
@@ -15,7 +15,8 @@ main() {
             return
         fi
         echo "Test"
-        cross test --target $TARGET --no-default-features --features mmap -- --test-threads 1
+        cross test --target $TARGET --no-default-features --features mmap
+        cross test --target $TARGET --no-default-features --features mmap query-grammar
     fi
     for example in $(ls examples/*.rs)
     do

examples/basic_search.rs

@@ -5,27 +5,23 @@
 //
 // We will :
 // - define our schema
-// = create an index in a directory
-// - index few documents in our index
-// - search for the best document matchings "sea whale"
-// - retrieve the best document original content.
-
-extern crate tempdir;
+// - create an index in a directory
+// - index a few documents into our index
+// - search for the best document matching a basic query
+// - retrieve the best document's original content.
 
 // ---
 // Importing tantivy...
-#[macro_use]
-extern crate tantivy;
 use tantivy::collector::TopDocs;
 use tantivy::query::QueryParser;
 use tantivy::schema::*;
-use tantivy::Index;
-use tempdir::TempDir;
+use tantivy::{doc, Index, ReloadPolicy};
+use tempfile::TempDir;
 
 fn main() -> tantivy::Result<()> {
     // Let's create a temporary directory for the
     // sake of this example
-    let index_path = TempDir::new("tantivy_example_dir")?;
+    let index_path = TempDir::new()?;
 
     // # Defining the schema
     //
@@ -34,7 +30,7 @@ fn main() -> tantivy::Result<()> {
     // and for each field, its type and "the way it should
     // be indexed".
 
-    // first we need to define a schema ...
+    // First we need to define a schema ...
     let mut schema_builder = Schema::builder();
 
     // Our first field is title.
@@ -49,7 +45,7 @@ fn main() -> tantivy::Result<()> {
     //
     // `STORED` means that the field will also be saved
     // in a compressed, row-oriented key-value store.
-    // This store is useful to reconstruct the
+    // This store is useful for reconstructing the
     // documents that were selected during the search phase.
     schema_builder.add_text_field("title", TEXT | STORED);
 
@@ -58,8 +54,7 @@ fn main() -> tantivy::Result<()> {
     // need to be able to be able to retrieve it
     // for our application.
     //
-    // We can make our index lighter and
-    // by omitting `STORED` flag.
+    // We can make our index lighter by omitting the `STORED` flag.
     schema_builder.add_text_field("body", TEXT);
 
     let schema = schema_builder.build();
@@ -72,7 +67,7 @@ fn main() -> tantivy::Result<()> {
     // with our schema in the directory.
     let index = Index::create_in_dir(&index_path, schema.clone())?;
 
-    // To insert document we need an index writer.
+    // To insert a document we will need an index writer.
     // There must be only one writer at a time.
     // This single `IndexWriter` is already
     // multithreaded.
@@ -106,37 +101,37 @@ fn main() -> tantivy::Result<()> {
     // For convenience, tantivy also comes with a macro to
     // reduce the boilerplate above.
     index_writer.add_document(doc!(
-        title => "Of Mice and Men",
-        body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
-                bank and runs deep and green. The water is warm too, for it has slipped twinkling \
-                over the yellow sands in the sunlight before reaching the narrow pool. On one \
-                side of the river the golden foothill slopes curve up to the strong and rocky \
-                Gabilan Mountains, but on the valley side the water is lined with trees—willows \
-                fresh and green with every spring, carrying in their lower leaf junctures the \
-                debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
-                limbs and branches that arch over the pool"
+    title => "Of Mice and Men",
+    body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
+            bank and runs deep and green. The water is warm too, for it has slipped twinkling \
+            over the yellow sands in the sunlight before reaching the narrow pool. On one \
+            side of the river the golden foothill slopes curve up to the strong and rocky \
+            Gabilan Mountains, but on the valley side the water is lined with trees—willows \
+            fresh and green with every spring, carrying in their lower leaf junctures the \
+            debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
+            limbs and branches that arch over the pool"
     ));
 
     index_writer.add_document(doc!(
-        title => "Of Mice and Men",
-        body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
-                bank and runs deep and green. The water is warm too, for it has slipped twinkling \
-                over the yellow sands in the sunlight before reaching the narrow pool. On one \
-                side of the river the golden foothill slopes curve up to the strong and rocky \
-                Gabilan Mountains, but on the valley side the water is lined with trees—willows \
-                fresh and green with every spring, carrying in their lower leaf junctures the \
-                debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
-                limbs and branches that arch over the pool"
+    title => "Of Mice and Men",
+    body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
+            bank and runs deep and green. The water is warm too, for it has slipped twinkling \
+            over the yellow sands in the sunlight before reaching the narrow pool. On one \
+            side of the river the golden foothill slopes curve up to the strong and rocky \
+            Gabilan Mountains, but on the valley side the water is lined with trees—willows \
+            fresh and green with every spring, carrying in their lower leaf junctures the \
+            debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
+            limbs and branches that arch over the pool"
     ));
 
     // Multivalued field just need to be repeated.
     index_writer.add_document(doc!(
-       title => "Frankenstein",
-       title => "The Modern Prometheus",
-       body => "You will rejoice to hear that no disaster has accompanied the commencement of an \
-                enterprise which you have regarded with such evil forebodings.  I arrived here \
-                yesterday, and my first task is to assure my dear sister of my welfare and \
-                increasing confidence in the success of my undertaking."
+    title => "Frankenstein",
+    title => "The Modern Prometheus",
+    body => "You will rejoice to hear that no disaster has accompanied the commencement of an \
+             enterprise which you have regarded with such evil forebodings.  I arrived here \
+             yesterday, and my first task is to assure my dear sister of my welfare and \
+             increasing confidence in the success of my undertaking."
     ));
 
     // This is an example, so we will only index 3 documents
@@ -150,8 +145,8 @@ fn main() -> tantivy::Result<()> {
     // At this point our documents are not searchable.
     //
     //
-    // We need to call .commit() explicitly to force the
-    // index_writer to finish processing the documents in the queue,
+    // We need to call `.commit()` explicitly to force the
+    // `index_writer` to finish processing the documents in the queue,
     // flush the current index to the disk, and advertise
     // the existence of new documents.
     //
@@ -163,31 +158,40 @@ fn main() -> tantivy::Result<()> {
     // persistently indexed.
     //
     // In the scenario of a crash or a power failure,
-    // tantivy behaves as if has rolled back to its last
+    // tantivy behaves as if it has rolled back to its last
     // commit.
 
     // # Searching
     //
     // ### Searcher
     //
-    // Let's search our index. Start by reloading
-    // searchers in the index. This should be done
-    // after every `commit()`.
-    index.load_searchers()?;
+    // A reader is required first in order to search an index.
+    // It acts as a `Searcher` pool that reloads itself,
+    // depending on a `ReloadPolicy`.
+    //
+    // For a search server you will typically create one reader for the entire lifetime of your
+    // program, and acquire a new searcher for every single request.
+    //
+    // In the code below, we rely on the 'ON_COMMIT' policy: the reader
+    // will reload the index automatically after each commit.
+    let reader = index
+        .reader_builder()
+        .reload_policy(ReloadPolicy::OnCommit)
+        .try_into()?;
 
     // We now need to acquire a searcher.
-    // Some search experience might require more than
-    // one query.
     //
-    // The searcher ensure that we get to work
-    // with a consistent version of the index.
+    // A searcher points to a snapshotted, immutable version of the index.
+    //
+    // Some search experience might require more than
+    // one query. Using the same searcher ensures that all of these queries will run on the
+    // same version of the index.
     //
     // Acquiring a `searcher` is very cheap.
     //
-    // You should acquire a searcher every time you
-    // start processing a request and
+    // You should acquire a searcher every time you start processing a request and
     // and release it right after your query is finished.
-    let searcher = index.searcher();
+    let searcher = reader.searcher();
 
     // ### Query
 
@@ -197,7 +201,7 @@ fn main() -> tantivy::Result<()> {
     // in both title and body.
     let query_parser = QueryParser::for_index(&index, vec![title, body]);
 
-    // QueryParser may fail if the query is not in the right
+    // `QueryParser` may fail if the query is not in the right
     // format. For user facing applications, this can be a problem.
     // A ticket has been opened regarding this problem.
     let query = query_parser.parse_query("sea whale")?;
@@ -213,7 +217,7 @@ fn main() -> tantivy::Result<()> {
     //
     // We are not interested in all of the documents but
     // only in the top 10. Keeping track of our top 10 best documents
-    // is the role of the TopDocs.
+    // is the role of the `TopDocs` collector.
 
     // We can now perform our query.
     let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
@@ -224,7 +228,6 @@ fn main() -> tantivy::Result<()> {
     // Since the body field was not configured as stored,
     // the document returned will only contain
     // a title.
-
     for (_score, doc_address) in top_docs {
         let retrieved_doc = searcher.doc(doc_address)?;
         println!("{}", schema.to_json(&retrieved_doc));

examples/custom_collector.rs

@@ -7,19 +7,14 @@
 // Of course, you can have a look at the tantivy's built-in collectors
 // such as the `CountCollector` for more examples.
 
-extern crate tempdir;
-
 // ---
 // Importing tantivy...
-#[macro_use]
-extern crate tantivy;
 use tantivy::collector::{Collector, SegmentCollector};
 use tantivy::fastfield::FastFieldReader;
 use tantivy::query::QueryParser;
 use tantivy::schema::Field;
-use tantivy::schema::{Schema, FAST, INT_INDEXED, TEXT};
-use tantivy::Index;
-use tantivy::SegmentReader;
+use tantivy::schema::{Schema, FAST, INDEXED, TEXT};
+use tantivy::{doc, Index, SegmentReader, TantivyError};
 
 #[derive(Default)]
 struct Stats {
@@ -75,9 +70,18 @@ impl Collector for StatsCollector {
     fn for_segment(
         &self,
         _segment_local_id: u32,
-        segment: &SegmentReader,
+        segment_reader: &SegmentReader,
     ) -> tantivy::Result<StatsSegmentCollector> {
-        let fast_field_reader = segment.fast_field_reader(self.field)?;
+        let fast_field_reader = segment_reader
+            .fast_fields()
+            .u64(self.field)
+            .ok_or_else(|| {
+                let field_name = segment_reader.schema().get_field_name(self.field);
+                TantivyError::SchemaError(format!(
+                    "Field {:?} is not a u64 fast field.",
+                    field_name
+                ))
+            })?;
         Ok(StatsSegmentCollector {
             fast_field_reader,
             stats: Stats::default(),
@@ -137,7 +141,7 @@ fn main() -> tantivy::Result<()> {
     // products, and with a name, a description, and a price.
     let product_name = schema_builder.add_text_field("name", TEXT);
     let product_description = schema_builder.add_text_field("description", TEXT);
-    let price = schema_builder.add_u64_field("price", INT_INDEXED | FAST);
+    let price = schema_builder.add_u64_field("price", INDEXED | FAST);
     let schema = schema_builder.build();
 
     // # Indexing documents
@@ -170,9 +174,9 @@ fn main() -> tantivy::Result<()> {
         price => 5_200u64
     ));
     index_writer.commit()?;
-    index.load_searchers()?;
 
-    let searcher = index.searcher();
+    let reader = index.reader()?;
+    let searcher = reader.searcher();
     let query_parser = QueryParser::for_index(&index, vec![product_name, product_description]);
 
     // here we want to get a hit on the 'ken' in Frankenstein

examples/custom_tokenizer.rs

@@ -2,14 +2,11 @@
 //
 // In this example, we'll see how to define a tokenizer pipeline
 // by aligning a bunch of `TokenFilter`.
-
-#[macro_use]
-extern crate tantivy;
 use tantivy::collector::TopDocs;
 use tantivy::query::QueryParser;
 use tantivy::schema::*;
 use tantivy::tokenizer::NgramTokenizer;
-use tantivy::Index;
+use tantivy::{doc, Index};
 
 fn main() -> tantivy::Result<()> {
     // # Defining the schema
@@ -68,12 +65,12 @@ fn main() -> tantivy::Result<()> {
     // heap for the indexer can increase its throughput.
     let mut index_writer = index.writer(50_000_000)?;
     index_writer.add_document(doc!(
-        title => "The Old Man and the Sea",
-        body => "He was an old man who fished alone in a skiff in the Gulf Stream and \
-         he had gone eighty-four days now without taking a fish."
+    title => "The Old Man and the Sea",
+    body => "He was an old man who fished alone in a skiff in the Gulf Stream and \
+     he had gone eighty-four days now without taking a fish."
     ));
     index_writer.add_document(doc!(
-       title => "Of Mice and Men",
+    title => "Of Mice and Men",
        body => r#"A few miles south of Soledad, the Salinas River drops in close to the hillside
                 bank and runs deep and green. The water is warm too, for it has slipped twinkling
                 over the yellow sands in the sunlight before reaching the narrow pool. On one
@@ -84,16 +81,16 @@ fn main() -> tantivy::Result<()> {
                 limbs and branches that arch over the pool"#
     ));
     index_writer.add_document(doc!(
-        title => "Frankenstein",
+    title => "Frankenstein",
         body => r#"You will rejoice to hear that no disaster has accompanied the commencement of an
                 enterprise which you have regarded with such evil forebodings.  I arrived here
                 yesterday, and my first task is to assure my dear sister of my welfare and
                 increasing confidence in the success of my undertaking."#
     ));
     index_writer.commit()?;
-    index.load_searchers()?;
 
-    let searcher = index.searcher();
+    let reader = index.reader()?;
+    let searcher = reader.searcher();
 
     // The query parser can interpret human queries.
     // Here, if the user does not specify which

examples/deleting_updating_documents.rs

@@ -8,18 +8,19 @@
 //
 // ---
 // Importing tantivy...
-#[macro_use]
-extern crate tantivy;
 use tantivy::collector::TopDocs;
 use tantivy::query::TermQuery;
 use tantivy::schema::*;
-use tantivy::Index;
+use tantivy::{doc, Index, IndexReader};
 
 // A simple helper function to fetch a single document
 // given its id from our index.
 // It will be helpful to check our work.
-fn extract_doc_given_isbn(index: &Index, isbn_term: &Term) -> tantivy::Result<Option<Document>> {
-    let searcher = index.searcher();
+fn extract_doc_given_isbn(
+    reader: &IndexReader,
+    isbn_term: &Term,
+) -> tantivy::Result<Option<Document>> {
+    let searcher = reader.searcher();
 
     // This is the simplest query you can think of.
     // It matches all of the documents containing a specific term.
@@ -85,12 +86,12 @@ fn main() -> tantivy::Result<()> {
        isbn => "978-9176370711",
     ));
     index_writer.commit()?;
-    index.load_searchers()?;
+    let reader = index.reader()?;
 
     let frankenstein_isbn = Term::from_field_text(isbn, "978-9176370711");
 
     // Oops our frankenstein doc seems mispelled
-    let frankenstein_doc_misspelled = extract_doc_given_isbn(&index, &frankenstein_isbn)?.unwrap();
+    let frankenstein_doc_misspelled = extract_doc_given_isbn(&reader, &frankenstein_isbn)?.unwrap();
     assert_eq!(
         schema.to_json(&frankenstein_doc_misspelled),
         r#"{"isbn":["978-9176370711"],"title":["Frankentein"]}"#,
@@ -129,10 +130,10 @@ fn main() -> tantivy::Result<()> {
     // Everything happened as if the document was updated.
     index_writer.commit()?;
     // We reload our searcher to make our change available to clients.
-    index.load_searchers()?;
+    reader.reload()?;
 
     // No more typo!
-    let frankenstein_new_doc = extract_doc_given_isbn(&index, &frankenstein_isbn)?.unwrap();
+    let frankenstein_new_doc = extract_doc_given_isbn(&reader, &frankenstein_isbn)?.unwrap();
     assert_eq!(
         schema.to_json(&frankenstein_new_doc),
         r#"{"isbn":["978-9176370711"],"title":["Frankenstein"]}"#,

examples/faceted_search.rs

@@ -10,71 +10,103 @@
 // - search for the best document matchings "sea whale"
 // - retrieve the best document original content.
 
-extern crate tempdir;
-
 // ---
 // Importing tantivy...
-#[macro_use]
-extern crate tantivy;
 use tantivy::collector::FacetCollector;
-use tantivy::query::AllQuery;
+use tantivy::query::{AllQuery, TermQuery};
 use tantivy::schema::*;
-use tantivy::Index;
+use tantivy::{doc, Index};
 
 fn main() -> tantivy::Result<()> {
-    // Let's create a temporary directory for the
-    // sake of this example
-    let index_path = TempDir::new("tantivy_facet_example_dir")?;
+    // Let's create a temporary directory for the sake of this example
     let mut schema_builder = Schema::builder();
 
-    schema_builder.add_text_field("name", TEXT | STORED);
-
-    // this is our faceted field
-    schema_builder.add_facet_field("tags");
+    let name = schema_builder.add_text_field("felin_name", TEXT | STORED);
+    // this is our faceted field: its scientific classification
+    let classification = schema_builder.add_facet_field("classification");
 
     let schema = schema_builder.build();
+    let index = Index::create_in_ram(schema);
 
-    let index = Index::create_in_dir(&index_path, schema.clone())?;
-
-    let mut index_writer = index.writer(50_000_000)?;
-
-    let name = schema.get_field("name").unwrap();
-    let tags = schema.get_field("tags").unwrap();
+    let mut index_writer = index.writer(30_000_000)?;
 
     // For convenience, tantivy also comes with a macro to
     // reduce the boilerplate above.
     index_writer.add_document(doc!(
-        name => "the ditch",
-        tags => Facet::from("/pools/north")
+        name => "Cat",
+        classification => Facet::from("/Felidae/Felinae/Felis")
     ));
-
     index_writer.add_document(doc!(
-        name => "little stacey",
-        tags => Facet::from("/pools/south")
+        name => "Canada lynx",
+        classification => Facet::from("/Felidae/Felinae/Lynx")
+    ));
+    index_writer.add_document(doc!(
+        name => "Cheetah",
+        classification => Facet::from("/Felidae/Felinae/Acinonyx")
+    ));
+    index_writer.add_document(doc!(
+        name => "Tiger",
+        classification => Facet::from("/Felidae/Pantherinae/Panthera")
+    ));
+    index_writer.add_document(doc!(
+        name => "Lion",
+        classification => Facet::from("/Felidae/Pantherinae/Panthera")
+    ));
+    index_writer.add_document(doc!(
+        name => "Jaguar",
+        classification => Facet::from("/Felidae/Pantherinae/Panthera")
+    ));
+    index_writer.add_document(doc!(
+        name => "Sunda clouded leopard",
+        classification => Facet::from("/Felidae/Pantherinae/Neofelis")
+    ));
+    index_writer.add_document(doc!(
+        name => "Fossa",
+        classification => Facet::from("/Eupleridae/Cryptoprocta")
     ));
-
     index_writer.commit()?;
 
-    index.load_searchers()?;
+    let reader = index.reader()?;
+    let searcher = reader.searcher();
+    {
+        let mut facet_collector = FacetCollector::for_field(classification);
+        facet_collector.add_facet("/Felidae");
+        let facet_counts = searcher.search(&AllQuery, &facet_collector)?;
+        // This lists all of the facet counts, right below "/Felidae".
+        let facets: Vec<(&Facet, u64)> = facet_counts.get("/Felidae").collect();
+        assert_eq!(
+            facets,
+            vec![
+                (&Facet::from("/Felidae/Felinae"), 3),
+                (&Facet::from("/Felidae/Pantherinae"), 4),
+            ]
+        );
+    }
 
-    let searcher = index.searcher();
+    // Facets are also searchable.
+    //
+    // For instance a common UI pattern is to allow the user someone to click on a facet link
+    // (e.g: `Pantherinae`) to drill down and filter the current result set with this subfacet.
+    //
+    // The search would then look as follows.
 
-    let mut facet_collector = FacetCollector::for_field(tags);
-    facet_collector.add_facet("/pools");
-
-    let facet_counts = searcher.search(&AllQuery, &facet_collector).unwrap();
-
-    // This lists all of the facet counts
-    let facets: Vec<(&Facet, u64)> = facet_counts.get("/pools").collect();
-    assert_eq!(
-        facets,
-        vec![
-            (&Facet::from("/pools/north"), 1),
-            (&Facet::from("/pools/south"), 1),
-        ]
-    );
+    // Check the reference doc for different ways to create a `Facet` object.
+    {
+        let facet = Facet::from_text("/Felidae/Pantherinae");
+        let facet_term = Term::from_facet(classification, &facet);
+        let facet_term_query = TermQuery::new(facet_term, IndexRecordOption::Basic);
+        let mut facet_collector = FacetCollector::for_field(classification);
+        facet_collector.add_facet("/Felidae/Pantherinae");
+        let facet_counts = searcher.search(&facet_term_query, &facet_collector)?;
+        let facets: Vec<(&Facet, u64)> = facet_counts.get("/Felidae/Pantherinae").collect();
+        assert_eq!(
+            facets,
+            vec![
+                (&Facet::from("/Felidae/Pantherinae/Neofelis"), 1),
+                (&Facet::from("/Felidae/Pantherinae/Panthera"), 3),
+            ]
+        );
+    }
 
     Ok(())
 }
-
-use tempdir::TempDir;

examples/faceted_search_with_tweaked_score.rs

@@ -0,0 +1,98 @@
+use std::collections::HashSet;
+use tantivy::collector::TopDocs;
+use tantivy::doc;
+use tantivy::query::BooleanQuery;
+use tantivy::schema::*;
+use tantivy::{DocId, Index, Score, SegmentReader};
+
+fn main() -> tantivy::Result<()> {
+    let mut schema_builder = Schema::builder();
+
+    let title = schema_builder.add_text_field("title", STORED);
+    let ingredient = schema_builder.add_facet_field("ingredient");
+
+    let schema = schema_builder.build();
+    let index = Index::create_in_ram(schema.clone());
+
+    let mut index_writer = index.writer(30_000_000)?;
+
+    index_writer.add_document(doc!(
+        title => "Fried egg",
+        ingredient => Facet::from("/ingredient/egg"),
+        ingredient => Facet::from("/ingredient/oil"),
+    ));
+    index_writer.add_document(doc!(
+        title => "Scrambled egg",
+        ingredient => Facet::from("/ingredient/egg"),
+        ingredient => Facet::from("/ingredient/butter"),
+        ingredient => Facet::from("/ingredient/milk"),
+        ingredient => Facet::from("/ingredient/salt"),
+    ));
+    index_writer.add_document(doc!(
+        title => "Egg rolls",
+        ingredient => Facet::from("/ingredient/egg"),
+        ingredient => Facet::from("/ingredient/garlic"),
+        ingredient => Facet::from("/ingredient/salt"),
+        ingredient => Facet::from("/ingredient/oil"),
+        ingredient => Facet::from("/ingredient/tortilla-wrap"),
+        ingredient => Facet::from("/ingredient/mushroom"),
+    ));
+    index_writer.commit()?;
+
+    let reader = index.reader()?;
+    let searcher = reader.searcher();
+    {
+        let facets = vec![
+            Facet::from("/ingredient/egg"),
+            Facet::from("/ingredient/oil"),
+            Facet::from("/ingredient/garlic"),
+            Facet::from("/ingredient/mushroom"),
+        ];
+        let query = BooleanQuery::new_multiterms_query(
+            facets
+                .iter()
+                .map(|key| Term::from_facet(ingredient, &key))
+                .collect(),
+        );
+        let top_docs_by_custom_score =
+            TopDocs::with_limit(2).tweak_score(move |segment_reader: &SegmentReader| {
+                let mut ingredient_reader = segment_reader.facet_reader(ingredient).unwrap();
+                let facet_dict = ingredient_reader.facet_dict();
+
+                let query_ords: HashSet<u64> = facets
+                    .iter()
+                    .filter_map(|key| facet_dict.term_ord(key.encoded_str()))
+                    .collect();
+
+                let mut facet_ords_buffer: Vec<u64> = Vec::with_capacity(20);
+
+                move |doc: DocId, original_score: Score| {
+                    ingredient_reader.facet_ords(doc, &mut facet_ords_buffer);
+                    let missing_ingredients = facet_ords_buffer
+                        .iter()
+                        .filter(|ord| !query_ords.contains(ord))
+                        .count();
+                    let tweak = 1.0 / 4_f32.powi(missing_ingredients as i32);
+
+                    original_score * tweak
+                }
+            });
+        let top_docs = searcher.search(&query, &top_docs_by_custom_score)?;
+
+        let titles: Vec<String> = top_docs
+            .iter()
+            .map(|(_, doc_id)| {
+                searcher
+                    .doc(*doc_id)
+                    .unwrap()
+                    .get_first(title)
+                    .unwrap()
+                    .text()
+                    .unwrap()
+                    .to_owned()
+            })
+            .collect();
+        assert_eq!(titles, vec!["Fried egg", "Egg rolls"]);
+    }
+    Ok(())
+}

examples/integer_range_search.rs

@@ -0,0 +1,39 @@
+// # Searching a range on an indexed int field.
+//
+// Below is an example of creating an indexed integer field in your schema
+// You can use RangeQuery to get a Count of all occurrences in a given range.
+use tantivy::collector::Count;
+use tantivy::query::RangeQuery;
+use tantivy::schema::{Schema, INDEXED};
+use tantivy::{doc, Index, Result};
+
+fn run() -> Result<()> {
+    // For the sake of simplicity, this schema will only have 1 field
+    let mut schema_builder = Schema::builder();
+
+    // `INDEXED` is a short-hand to indicate that our field should be "searchable".
+    let year_field = schema_builder.add_u64_field("year", INDEXED);
+    let schema = schema_builder.build();
+    let index = Index::create_in_ram(schema);
+    let reader = index.reader()?;
+    {
+        let mut index_writer = index.writer_with_num_threads(1, 6_000_000)?;
+        for year in 1950u64..2019u64 {
+            index_writer.add_document(doc!(year_field => year));
+        }
+        index_writer.commit()?;
+        // The index will be a range of years
+    }
+    reader.reload()?;
+    let searcher = reader.searcher();
+    // The end is excluded i.e. here we are searching up to 1969
+    let docs_in_the_sixties = RangeQuery::new_u64(year_field, 1960..1970);
+    // Uses a Count collector to sum the total number of docs in the range
+    let num_60s_books = searcher.search(&docs_in_the_sixties, &Count)?;
+    assert_eq!(num_60s_books, 10);
+    Ok(())
+}
+
+fn main() {
+    run().unwrap()
+}

examples/iterating_docs_and_positions.rs

@@ -9,11 +9,8 @@
 
 // ---
 // Importing tantivy...
-#[macro_use]
-extern crate tantivy;
 use tantivy::schema::*;
-use tantivy::Index;
-use tantivy::{DocId, DocSet, Postings};
+use tantivy::{doc, DocSet, Index, Postings, TERMINATED};
 
 fn main() -> tantivy::Result<()> {
     // We first create a schema for the sake of the
@@ -33,9 +30,9 @@ fn main() -> tantivy::Result<()> {
     index_writer.add_document(doc!(title => "The modern Promotheus"));
     index_writer.commit()?;
 
-    index.load_searchers()?;
+    let reader = index.reader()?;
 
-    let searcher = index.searcher();
+    let searcher = reader.searcher();
 
     // A tantivy index is actually a collection of segments.
     // Similarly, a searcher just wraps a list `segment_reader`.
@@ -65,12 +62,11 @@ fn main() -> tantivy::Result<()> {
         {
             // this buffer will be used to request for positions
             let mut positions: Vec<u32> = Vec::with_capacity(100);
-            while segment_postings.advance() {
-                // the number of time the term appears in the document.
-                let doc_id: DocId = segment_postings.doc(); //< do not try to access this before calling advance once.
-
+            let mut doc_id = segment_postings.doc();
+            while doc_id != TERMINATED {
                 // This MAY contains deleted documents as well.
                 if segment_reader.is_deleted(doc_id) {
+                    doc_id = segment_postings.advance();
                     continue;
                 }
 
@@ -89,6 +85,7 @@ fn main() -> tantivy::Result<()> {
                 // Doc 2: TermFreq 1: [0]
                 // ```
                 println!("Doc {}: TermFreq {}: {:?}", doc_id, term_freq, positions);
+                doc_id = segment_postings.advance();
             }
         }
     }

examples/multiple_producer.rs

@@ -0,0 +1,100 @@
+// # Indexing from different threads.
+//
+// It is fairly common to have to index from different threads.
+// Tantivy forbids to create more than one `IndexWriter` at a time.
+//
+// This `IndexWriter` itself has its own multithreaded layer, so managing your own
+// indexing threads will not help. However, it can still be useful for some applications.
+//
+// For instance, if preparing documents to send to tantivy before indexing is the bottleneck of
+// your application, it is reasonable to have multiple threads.
+//
+// Another very common reason to want to index from multiple threads, is implementing a webserver
+// with CRUD capabilities. The server framework will most likely handle request from
+// different threads.
+//
+// The recommended way to address both of these use case is to wrap your `IndexWriter` into a
+// `Arc<RwLock<IndexWriter>>`.
+//
+// While this is counterintuitive, adding and deleting documents do not require mutability
+// over the `IndexWriter`, so several threads will be able to do this operation concurrently.
+//
+// The example below does not represent an actual real-life use case (who would spawn thread to
+// index a single document?), but aims at demonstrating the mechanism that makes indexing
+// from several threads possible.
+
+// ---
+// Importing tantivy...
+use std::sync::{Arc, RwLock};
+use std::thread;
+use std::time::Duration;
+use tantivy::schema::{Schema, STORED, TEXT};
+use tantivy::{doc, Index, IndexWriter, Opstamp};
+
+fn main() -> tantivy::Result<()> {
+    // # Defining the schema
+    let mut schema_builder = Schema::builder();
+    let title = schema_builder.add_text_field("title", TEXT | STORED);
+    let body = schema_builder.add_text_field("body", TEXT);
+    let schema = schema_builder.build();
+
+    let index = Index::create_in_ram(schema);
+    let index_writer: Arc<RwLock<IndexWriter>> = Arc::new(RwLock::new(index.writer(50_000_000)?));
+
+    // # First indexing thread.
+    let index_writer_clone_1 = index_writer.clone();
+    thread::spawn(move || {
+        // we index 100 times the document... for the sake of the example.
+        for i in 0..100 {
+            let opstamp = index_writer_clone_1
+                .read().unwrap() //< A read lock is sufficient here.
+                .add_document(
+                    doc!(
+                        title => "Of Mice and Men",
+                        body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
+                        bank and runs deep and green. The water is warm too, for it has slipped twinkling \
+                        over the yellow sands in the sunlight before reaching the narrow pool. On one \
+                        side of the river the golden foothill slopes curve up to the strong and rocky \
+                        Gabilan Mountains, but on the valley side the water is lined with trees—willows \
+                        fresh and green with every spring, carrying in their lower leaf junctures the \
+                        debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
+                        limbs and branches that arch over the pool"
+                    ));
+            println!("add doc {} from thread 1 - opstamp {}", i, opstamp);
+            thread::sleep(Duration::from_millis(20));
+        }
+    });
+
+    // # Second indexing thread.
+    let index_writer_clone_2 = index_writer.clone();
+    // For convenience, tantivy also comes with a macro to
+    // reduce the boilerplate above.
+    thread::spawn(move || {
+        // we index 100 times the document... for the sake of the example.
+        for i in 0..100 {
+            // A read lock is sufficient here.
+            let opstamp = {
+                let index_writer_rlock = index_writer_clone_2.read().unwrap();
+                index_writer_rlock.add_document(doc!(
+                    title => "Manufacturing consent",
+                    body => "Some great book description..."
+                ))
+            };
+            println!("add doc {} from thread 2 - opstamp {}", i, opstamp);
+            thread::sleep(Duration::from_millis(10));
+        }
+    });
+
+    // # In the main thread, we commit 10 times, once every 500ms.
+    for _ in 0..10 {
+        let opstamp: Opstamp = {
+            // Committing or rollbacking on the other hand requires write lock. This will block other threads.
+            let mut index_writer_wlock = index_writer.write().unwrap();
+            index_writer_wlock.commit().unwrap()
+        };
+        println!("committed with opstamp {}", opstamp);
+        thread::sleep(Duration::from_millis(500));
+    }
+
+    Ok(())
+}

examples/pre_tokenized_text.rs

@@ -0,0 +1,139 @@
+// # Pre-tokenized text example
+//
+// This example shows how to use pre-tokenized text. Sometimes yout might
+// want to index and search through text which is already split into
+// tokens by some external tool.
+//
+// In this example we will:
+// - use tantivy tokenizer to create tokens and load them directly into tantivy,
+// - import tokenized text straight from json,
+// - perform a search on documents with pre-tokenized text
+
+use tantivy::collector::{Count, TopDocs};
+use tantivy::query::TermQuery;
+use tantivy::schema::*;
+use tantivy::tokenizer::{PreTokenizedString, SimpleTokenizer, Token, Tokenizer};
+use tantivy::{doc, Index, ReloadPolicy};
+use tempfile::TempDir;
+
+fn pre_tokenize_text(text: &str) -> Vec<Token> {
+    let mut token_stream = SimpleTokenizer.token_stream(text);
+    let mut tokens = vec![];
+    while token_stream.advance() {
+        tokens.push(token_stream.token().clone());
+    }
+    tokens
+}
+
+fn main() -> tantivy::Result<()> {
+    let index_path = TempDir::new()?;
+
+    let mut schema_builder = Schema::builder();
+
+    schema_builder.add_text_field("title", TEXT | STORED);
+    schema_builder.add_text_field("body", TEXT);
+
+    let schema = schema_builder.build();
+
+    let index = Index::create_in_dir(&index_path, schema.clone())?;
+
+    let mut index_writer = index.writer(50_000_000)?;
+
+    // We can create a document manually, by setting the fields
+    // one by one in a Document object.
+    let title = schema.get_field("title").unwrap();
+    let body = schema.get_field("body").unwrap();
+
+    let title_text = "The Old Man and the Sea";
+    let body_text = "He was an old man who fished alone in a skiff in the Gulf Stream";
+
+    // Content of our first document
+    // We create `PreTokenizedString` which contains original text and vector of tokens
+    let title_tok = PreTokenizedString {
+        text: String::from(title_text),
+        tokens: pre_tokenize_text(title_text),
+    };
+
+    println!(
+        "Original text: \"{}\" and tokens: {:?}",
+        title_tok.text, title_tok.tokens
+    );
+
+    let body_tok = PreTokenizedString {
+        text: String::from(body_text),
+        tokens: pre_tokenize_text(body_text),
+    };
+
+    // Now lets create a document and add our `PreTokenizedString`
+    let old_man_doc = doc!(title => title_tok, body => body_tok);
+
+    // ... now let's just add it to the IndexWriter
+    index_writer.add_document(old_man_doc);
+
+    // Pretokenized text can also be fed as JSON
+    let short_man_json = r#"{
+        "title":[{
+            "text":"The Old Man",
+            "tokens":[
+                {"offset_from":0,"offset_to":3,"position":0,"text":"The","position_length":1},
+                {"offset_from":4,"offset_to":7,"position":1,"text":"Old","position_length":1},
+                {"offset_from":8,"offset_to":11,"position":2,"text":"Man","position_length":1}
+            ]
+        }]
+    }"#;
+
+    let short_man_doc = schema.parse_document(&short_man_json)?;
+
+    index_writer.add_document(short_man_doc);
+
+    // Let's commit changes
+    index_writer.commit()?;
+
+    // ... and now is the time to query our index
+
+    let reader = index
+        .reader_builder()
+        .reload_policy(ReloadPolicy::OnCommit)
+        .try_into()?;
+
+    let searcher = reader.searcher();
+
+    // We want to get documents with token "Man", we will use TermQuery to do it
+    // Using PreTokenizedString means the tokens are stored as is avoiding stemming
+    // and lowercasing, which preserves full words in their original form
+    let query = TermQuery::new(
+        Term::from_field_text(title, "Man"),
+        IndexRecordOption::Basic,
+    );
+
+    let (top_docs, count) = searcher
+        .search(&query, &(TopDocs::with_limit(2), Count))
+        .unwrap();
+
+    assert_eq!(count, 2);
+
+    // Now let's print out the results.
+    // Note that the tokens are not stored along with the original text
+    // in the document store
+    for (_score, doc_address) in top_docs {
+        let retrieved_doc = searcher.doc(doc_address)?;
+        println!("Document: {}", schema.to_json(&retrieved_doc));
+    }
+
+    // In contrary to the previous query, when we search for the "man" term we
+    // should get no results, as it's not one of the indexed tokens. SimpleTokenizer
+    // only splits text on whitespace / punctuation.
+
+    let query = TermQuery::new(
+        Term::from_field_text(title, "man"),
+        IndexRecordOption::Basic,
+    );
+
+    let (_top_docs, count) = searcher
+        .search(&query, &(TopDocs::with_limit(2), Count))
+        .unwrap();
+
+    assert_eq!(count, 0);
+
+    Ok(())
+}

examples/snippet.rs

@@ -4,23 +4,19 @@
 // your hit result.
 // Snippet are an extracted of a target document, and returned in HTML format.
 // The keyword searched by the user are highlighted with a `<b>` tag.
-extern crate tempdir;
 
 // ---
 // Importing tantivy...
-#[macro_use]
-extern crate tantivy;
 use tantivy::collector::TopDocs;
 use tantivy::query::QueryParser;
 use tantivy::schema::*;
-use tantivy::Index;
-use tantivy::{Snippet, SnippetGenerator};
-use tempdir::TempDir;
+use tantivy::{doc, Index, Snippet, SnippetGenerator};
+use tempfile::TempDir;
 
 fn main() -> tantivy::Result<()> {
     // Let's create a temporary directory for the
     // sake of this example
-    let index_path = TempDir::new("tantivy_example_dir")?;
+    let index_path = TempDir::new()?;
 
     // # Defining the schema
     let mut schema_builder = Schema::builder();
@@ -35,28 +31,27 @@ fn main() -> tantivy::Result<()> {
 
     // we'll only need one doc for this example.
     index_writer.add_document(doc!(
-        title => "Of Mice and Men",
-        body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
-                bank and runs deep and green. The water is warm too, for it has slipped twinkling \
-                over the yellow sands in the sunlight before reaching the narrow pool. On one \
-                side of the river the golden foothill slopes curve up to the strong and rocky \
-                Gabilan Mountains, but on the valley side the water is lined with trees—willows \
-                fresh and green with every spring, carrying in their lower leaf junctures the \
-                debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
-                limbs and branches that arch over the pool"
+    title => "Of Mice and Men",
+    body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
+            bank and runs deep and green. The water is warm too, for it has slipped twinkling \
+            over the yellow sands in the sunlight before reaching the narrow pool. On one \
+            side of the river the golden foothill slopes curve up to the strong and rocky \
+            Gabilan Mountains, but on the valley side the water is lined with trees—willows \
+            fresh and green with every spring, carrying in their lower leaf junctures the \
+            debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
+            limbs and branches that arch over the pool"
     ));
     // ...
     index_writer.commit()?;
 
-    index.load_searchers()?;
-
-    let searcher = index.searcher();
+    let reader = index.reader()?;
+    let searcher = reader.searcher();
     let query_parser = QueryParser::for_index(&index, vec![title, body]);
     let query = query_parser.parse_query("sycamore spring")?;
 
     let top_docs = searcher.search(&query, &TopDocs::with_limit(10))?;
 
-    let snippet_generator = SnippetGenerator::new(&searcher, &*query, body)?;
+    let snippet_generator = SnippetGenerator::create(&searcher, &*query, body)?;
 
     for (score, doc_address) in top_docs {
         let doc = searcher.doc(doc_address)?;

examples/stop_words.rs

@@ -9,17 +9,13 @@
 // - add a few stop words
 // - index few documents in our index
 
-extern crate tempdir;
-
 // ---
 // Importing tantivy...
-#[macro_use]
-extern crate tantivy;
 use tantivy::collector::TopDocs;
 use tantivy::query::QueryParser;
 use tantivy::schema::*;
 use tantivy::tokenizer::*;
-use tantivy::Index;
+use tantivy::{doc, Index};
 
 fn main() -> tantivy::Result<()> {
     // this example assumes you understand the content in `basic_search`
@@ -54,7 +50,7 @@ fn main() -> tantivy::Result<()> {
 
     // This tokenizer lowers all of the text (to help with stop word matching)
     // then removes all instances of `the` and `and` from the corpus
-    let tokenizer = SimpleTokenizer
+    let tokenizer = TextAnalyzer::from(SimpleTokenizer)
         .filter(LowerCaser)
         .filter(StopWordFilter::remove(vec![
             "the".to_string(),
@@ -72,33 +68,33 @@ fn main() -> tantivy::Result<()> {
     title => "The Old Man and the Sea",
     body => "He was an old man who fished alone in a skiff in the Gulf Stream and \
      he had gone eighty-four days now without taking a fish."
-  ));
+    ));
 
     index_writer.add_document(doc!(
-      title => "Of Mice and Men",
-      body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
-              bank and runs deep and green. The water is warm too, for it has slipped twinkling \
-              over the yellow sands in the sunlight before reaching the narrow pool. On one \
-              side of the river the golden foothill slopes curve up to the strong and rocky \
-              Gabilan Mountains, but on the valley side the water is lined with trees—willows \
-              fresh and green with every spring, carrying in their lower leaf junctures the \
-              debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
-              limbs and branches that arch over the pool"
-  ));
+    title => "Of Mice and Men",
+    body => "A few miles south of Soledad, the Salinas River drops in close to the hillside \
+            bank and runs deep and green. The water is warm too, for it has slipped twinkling \
+            over the yellow sands in the sunlight before reaching the narrow pool. On one \
+            side of the river the golden foothill slopes curve up to the strong and rocky \
+            Gabilan Mountains, but on the valley side the water is lined with trees—willows \
+            fresh and green with every spring, carrying in their lower leaf junctures the \
+            debris of the winter’s flooding; and sycamores with mottled, white, recumbent \
+            limbs and branches that arch over the pool"
+    ));
 
     index_writer.add_document(doc!(
-       title => "Frankenstein",
-       body => "You will rejoice to hear that no disaster has accompanied the commencement of an \
-                enterprise which you have regarded with such evil forebodings.  I arrived here \
-                yesterday, and my first task is to assure my dear sister of my welfare and \
-                increasing confidence in the success of my undertaking."
+    title => "Frankenstein",
+    body => "You will rejoice to hear that no disaster has accompanied the commencement of an \
+             enterprise which you have regarded with such evil forebodings.  I arrived here \
+             yesterday, and my first task is to assure my dear sister of my welfare and \
+             increasing confidence in the success of my undertaking."
     ));
 
     index_writer.commit()?;
 
-    index.load_searchers()?;
+    let reader = index.reader()?;
 
-    let searcher = index.searcher();
+    let searcher = reader.searcher();
 
     let query_parser = QueryParser::for_index(&index, vec![title, body]);
 

examples/working_with_json.rs

@@ -1,4 +1,4 @@
-extern crate tantivy;
+use tantivy;
 use tantivy::schema::*;
 
 // # Document from json
@@ -12,7 +12,7 @@ fn main() -> tantivy::Result<()> {
     let mut schema_builder = Schema::builder();
     schema_builder.add_text_field("title", TEXT | STORED);
     schema_builder.add_text_field("body", TEXT);
-    schema_builder.add_u64_field("year", INT_INDEXED);
+    schema_builder.add_u64_field("year", INDEXED);
     let schema = schema_builder.build();
 
     // Let's assume we have a json-serialized document.

query-grammar/Cargo.toml

@@ -0,0 +1,16 @@
+[package]
+name = "tantivy-query-grammar"
+version = "0.13.0"
+authors = ["Paul Masurel <paul.masurel@gmail.com>"]
+license = "MIT"
+categories = ["database-implementations", "data-structures"]
+description = """Search engine library"""
+documentation = "https://tantivy-search.github.io/tantivy/tantivy/index.html"
+homepage = "https://github.com/tantivy-search/tantivy"
+repository = "https://github.com/tantivy-search/tantivy"
+readme = "README.md"
+keywords = ["search", "information", "retrieval"]
+edition = "2018"
+
+[dependencies]
+combine = {version="4", default-features=false, features=[] }

query-grammar/README.md

@@ -0,0 +1,3 @@
+# Tantivy Query Grammar
+
+This crate is used by tantivy to parse queries.

query-grammar/src/lib.rs

@@ -0,0 +1,15 @@
+mod occur;
+mod query_grammar;
+mod user_input_ast;
+use combine::parser::Parser;
+
+pub use crate::occur::Occur;
+use crate::query_grammar::parse_to_ast;
+pub use crate::user_input_ast::{UserInputAST, UserInputBound, UserInputLeaf, UserInputLiteral};
+
+pub struct Error;
+
+pub fn parse_query(query: &str) -> Result<UserInputAST, Error> {
+    let (user_input_ast, _remaining) = parse_to_ast().parse(query).map_err(|_| Error)?;
+    Ok(user_input_ast)
+}

query-grammar/src/occur.rs

@@ -1,5 +1,8 @@
+use std::fmt;
+use std::fmt::Write;
+
 /// Defines whether a term in a query must be present,
-/// should be present or must not be present.
+/// should be present or must be not present.
 #[derive(Debug, Clone, Hash, Copy, Eq, PartialEq)]
 pub enum Occur {
     /// For a given document to be considered for scoring,
@@ -18,32 +21,38 @@ impl Occur {
     /// - `Should` => '?',
     /// - `Must` => '+'
     /// - `Not` => '-'
-    pub fn to_char(self) -> char {
+    fn to_char(self) -> char {
         match self {
             Occur::Should => '?',
             Occur::Must => '+',
             Occur::MustNot => '-',
         }
     }
-}
 
-/// Compose two occur values.
-pub fn compose_occur(left: Occur, right: Occur) -> Occur {
-    match left {
-        Occur::Should => right,
-        Occur::Must => {
-            if right == Occur::MustNot {
-                Occur::MustNot
-            } else {
-                Occur::Must
+    /// Compose two occur values.
+    pub fn compose(left: Occur, right: Occur) -> Occur {
+        match left {
+            Occur::Should => right,
+            Occur::Must => {
+                if right == Occur::MustNot {
+                    Occur::MustNot
+                } else {
+                    Occur::Must
+                }
             }
-        }
-        Occur::MustNot => {
-            if right == Occur::MustNot {
-                Occur::Must
-            } else {
-                Occur::MustNot
+            Occur::MustNot => {
+                if right == Occur::MustNot {
+                    Occur::Must
+                } else {
+                    Occur::MustNot
+                }
             }
         }
     }
 }
+
+impl fmt::Display for Occur {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        f.write_char(self.to_char())
+    }
+}

query-grammar/src/query_grammar.rs

@@ -0,0 +1,510 @@
+use super::user_input_ast::{UserInputAST, UserInputBound, UserInputLeaf, UserInputLiteral};
+use crate::Occur;
+use combine::error::StringStreamError;
+use combine::parser::char::{char, digit, letter, space, spaces, string};
+use combine::parser::Parser;
+use combine::{
+    attempt, choice, eof, many, many1, one_of, optional, parser, satisfy, skip_many1, value,
+};
+
+fn field<'a>() -> impl Parser<&'a str, Output = String> {
+    (
+        letter(),
+        many(satisfy(|c: char| c.is_alphanumeric() || c == '_')),
+    )
+        .skip(char(':'))
+        .map(|(s1, s2): (char, String)| format!("{}{}", s1, s2))
+}
+
+fn word<'a>() -> impl Parser<&'a str, Output = String> {
+    (
+        satisfy(|c: char| {
+            !c.is_whitespace()
+                && !['-', '^', '`', ':', '{', '}', '"', '[', ']', '(', ')'].contains(&c)
+        }),
+        many(satisfy(|c: char| {
+            !c.is_whitespace() && ![':', '^', '{', '}', '"', '[', ']', '(', ')'].contains(&c)
+        })),
+    )
+        .map(|(s1, s2): (char, String)| format!("{}{}", s1, s2))
+        .and_then(|s: String| match s.as_str() {
+            "OR" | "AND " | "NOT" => Err(StringStreamError::UnexpectedParse),
+            _ => Ok(s),
+        })
+}
+
+fn term_val<'a>() -> impl Parser<&'a str, Output = String> {
+    let phrase = char('"').with(many1(satisfy(|c| c != '"'))).skip(char('"'));
+    phrase.or(word())
+}
+
+fn term_query<'a>() -> impl Parser<&'a str, Output = UserInputLiteral> {
+    let term_val_with_field = negative_number().or(term_val());
+    (field(), term_val_with_field).map(|(field_name, phrase)| UserInputLiteral {
+        field_name: Some(field_name),
+        phrase,
+    })
+}
+
+fn literal<'a>() -> impl Parser<&'a str, Output = UserInputLeaf> {
+    let term_default_field = term_val().map(|phrase| UserInputLiteral {
+        field_name: None,
+        phrase,
+    });
+    attempt(term_query())
+        .or(term_default_field)
+        .map(UserInputLeaf::from)
+}
+
+fn negative_number<'a>() -> impl Parser<&'a str, Output = String> {
+    (
+        char('-'),
+        many1(digit()),
+        optional((char('.'), many1(digit()))),
+    )
+        .map(|(s1, s2, s3): (char, String, Option<(char, String)>)| {
+            if let Some(('.', s3)) = s3 {
+                format!("{}{}.{}", s1, s2, s3)
+            } else {
+                format!("{}{}", s1, s2)
+            }
+        })
+}
+
+fn spaces1<'a>() -> impl Parser<&'a str, Output = ()> {
+    skip_many1(space())
+}
+
+/// Function that parses a range out of a Stream
+/// Supports ranges like:
+/// [5 TO 10], {5 TO 10}, [* TO 10], [10 TO *], {10 TO *], >5, <=10
+/// [a TO *], [a TO c], [abc TO bcd}
+fn range<'a>() -> impl Parser<&'a str, Output = UserInputLeaf> {
+    let range_term_val = || {
+        word()
+            .or(negative_number())
+            .or(char('*').with(value("*".to_string())))
+    };
+
+    // check for unbounded range in the form of <5, <=10, >5, >=5
+    let elastic_unbounded_range = (
+        choice([
+            attempt(string(">=")),
+            attempt(string("<=")),
+            attempt(string("<")),
+            attempt(string(">")),
+        ])
+        .skip(spaces()),
+        range_term_val(),
+    )
+        .map(
+            |(comparison_sign, bound): (&str, String)| match comparison_sign {
+                ">=" => (UserInputBound::Inclusive(bound), UserInputBound::Unbounded),
+                "<=" => (UserInputBound::Unbounded, UserInputBound::Inclusive(bound)),
+                "<" => (UserInputBound::Unbounded, UserInputBound::Exclusive(bound)),
+                ">" => (UserInputBound::Exclusive(bound), UserInputBound::Unbounded),
+                // default case
+                _ => (UserInputBound::Unbounded, UserInputBound::Unbounded),
+            },
+        );
+    let lower_bound = (one_of("{[".chars()), range_term_val()).map(
+        |(boundary_char, lower_bound): (char, String)| {
+            if lower_bound == "*" {
+                UserInputBound::Unbounded
+            } else if boundary_char == '{' {
+                UserInputBound::Exclusive(lower_bound)
+            } else {
+                UserInputBound::Inclusive(lower_bound)
+            }
+        },
+    );
+    let upper_bound = (range_term_val(), one_of("}]".chars())).map(
+        |(higher_bound, boundary_char): (String, char)| {
+            if higher_bound == "*" {
+                UserInputBound::Unbounded
+            } else if boundary_char == '}' {
+                UserInputBound::Exclusive(higher_bound)
+            } else {
+                UserInputBound::Inclusive(higher_bound)
+            }
+        },
+    );
+    // return only lower and upper
+    let lower_to_upper = (
+        lower_bound.skip((spaces(), string("TO"), spaces())),
+        upper_bound,
+    );
+
+    (
+        optional(field()).skip(spaces()),
+        // try elastic first, if it matches, the range is unbounded
+        attempt(elastic_unbounded_range).or(lower_to_upper),
+    )
+        .map(|(field, (lower, upper))|
+             // Construct the leaf from extracted field (optional)
+             // and bounds
+             UserInputLeaf::Range {
+                 field,
+                 lower,
+                 upper
+    })
+}
+
+fn negate(expr: UserInputAST) -> UserInputAST {
+    expr.unary(Occur::MustNot)
+}
+
+fn leaf<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
+    parser(|input| {
+        char('(')
+            .with(ast())
+            .skip(char(')'))
+            .or(char('*').map(|_| UserInputAST::from(UserInputLeaf::All)))
+            .or(attempt(
+                string("NOT").skip(spaces1()).with(leaf()).map(negate),
+            ))
+            .or(attempt(range().map(UserInputAST::from)))
+            .or(literal().map(UserInputAST::from))
+            .parse_stream(input)
+            .into_result()
+    })
+}
+
+fn occur_symbol<'a>() -> impl Parser<&'a str, Output = Occur> {
+    char('-')
+        .map(|_| Occur::MustNot)
+        .or(char('+').map(|_| Occur::Must))
+}
+
+fn occur_leaf<'a>() -> impl Parser<&'a str, Output = (Option<Occur>, UserInputAST)> {
+    (optional(occur_symbol()), boosted_leaf())
+}
+
+fn positive_float_number<'a>() -> impl Parser<&'a str, Output = f32> {
+    (many1(digit()), optional((char('.'), many1(digit())))).map(
+        |(int_part, decimal_part_opt): (String, Option<(char, String)>)| {
+            let mut float_str = int_part;
+            if let Some((chr, decimal_str)) = decimal_part_opt {
+                float_str.push(chr);
+                float_str.push_str(&decimal_str);
+            }
+            float_str.parse::<f32>().unwrap()
+        },
+    )
+}
+
+fn boost<'a>() -> impl Parser<&'a str, Output = f32> {
+    (char('^'), positive_float_number()).map(|(_, boost)| boost)
+}
+
+fn boosted_leaf<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
+    (leaf(), optional(boost())).map(|(leaf, boost_opt)| match boost_opt {
+        Some(boost) if (boost - 1.0).abs() > std::f32::EPSILON => {
+            UserInputAST::Boost(Box::new(leaf), boost)
+        }
+        _ => leaf,
+    })
+}
+
+#[derive(Clone, Copy)]
+enum BinaryOperand {
+    Or,
+    And,
+}
+
+fn binary_operand<'a>() -> impl Parser<&'a str, Output = BinaryOperand> {
+    string("AND")
+        .with(value(BinaryOperand::And))
+        .or(string("OR").with(value(BinaryOperand::Or)))
+}
+
+fn aggregate_binary_expressions(
+    left: UserInputAST,
+    others: Vec<(BinaryOperand, UserInputAST)>,
+) -> UserInputAST {
+    let mut dnf: Vec<Vec<UserInputAST>> = vec![vec![left]];
+    for (operator, operand_ast) in others {
+        match operator {
+            BinaryOperand::And => {
+                if let Some(last) = dnf.last_mut() {
+                    last.push(operand_ast);
+                }
+            }
+            BinaryOperand::Or => {
+                dnf.push(vec![operand_ast]);
+            }
+        }
+    }
+    if dnf.len() == 1 {
+        UserInputAST::and(dnf.into_iter().next().unwrap()) //< safe
+    } else {
+        let conjunctions = dnf.into_iter().map(UserInputAST::and).collect();
+        UserInputAST::or(conjunctions)
+    }
+}
+
+fn operand_leaf<'a>() -> impl Parser<&'a str, Output = (BinaryOperand, UserInputAST)> {
+    (
+        binary_operand().skip(spaces()),
+        boosted_leaf().skip(spaces()),
+    )
+}
+
+pub fn ast<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
+    let boolean_expr = (boosted_leaf().skip(spaces()), many1(operand_leaf()))
+        .map(|(left, right)| aggregate_binary_expressions(left, right));
+    let whitespace_separated_leaves = many1(occur_leaf().skip(spaces().silent())).map(
+        |subqueries: Vec<(Option<Occur>, UserInputAST)>| {
+            if subqueries.len() == 1 {
+                let (occur_opt, ast) = subqueries.into_iter().next().unwrap();
+                match occur_opt.unwrap_or(Occur::Should) {
+                    Occur::Must | Occur::Should => ast,
+                    Occur::MustNot => UserInputAST::Clause(vec![(Some(Occur::MustNot), ast)]),
+                }
+            } else {
+                UserInputAST::Clause(subqueries.into_iter().collect())
+            }
+        },
+    );
+    let expr = attempt(boolean_expr).or(whitespace_separated_leaves);
+    spaces().with(expr).skip(spaces())
+}
+
+pub fn parse_to_ast<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
+    spaces()
+        .with(optional(ast()).skip(eof()))
+        .map(|opt_ast| opt_ast.unwrap_or_else(UserInputAST::empty_query))
+}
+
+#[cfg(test)]
+mod test {
+
+    use super::*;
+    use combine::parser::Parser;
+
+    pub fn nearly_equals(a: f32, b: f32) -> bool {
+        (a - b).abs() < 0.0005 * (a + b).abs()
+    }
+
+    fn assert_nearly_equals(expected: f32, val: f32) {
+        assert!(
+            nearly_equals(val, expected),
+            "Got {}, expected {}.",
+            val,
+            expected
+        );
+    }
+
+    #[test]
+    fn test_occur_symbol() {
+        assert_eq!(super::occur_symbol().parse("-"), Ok((Occur::MustNot, "")));
+        assert_eq!(super::occur_symbol().parse("+"), Ok((Occur::Must, "")));
+    }
+
+    #[test]
+    fn test_positive_float_number() {
+        fn valid_parse(float_str: &str, expected_val: f32, expected_remaining: &str) {
+            let (val, remaining) = positive_float_number().parse(float_str).unwrap();
+            assert_eq!(remaining, expected_remaining);
+            assert_nearly_equals(val, expected_val);
+        }
+        fn error_parse(float_str: &str) {
+            assert!(positive_float_number().parse(float_str).is_err());
+        }
+        valid_parse("1.0", 1.0f32, "");
+        valid_parse("1", 1.0f32, "");
+        valid_parse("0.234234 aaa", 0.234234f32, " aaa");
+        error_parse(".3332");
+        error_parse("1.");
+        error_parse("-1.");
+    }
+
+    fn test_parse_query_to_ast_helper(query: &str, expected: &str) {
+        let query = parse_to_ast().parse(query).unwrap().0;
+        let query_str = format!("{:?}", query);
+        assert_eq!(query_str, expected);
+    }
+
+    fn test_is_parse_err(query: &str) {
+        assert!(parse_to_ast().parse(query).is_err());
+    }
+
+    #[test]
+    fn test_parse_empty_to_ast() {
+        test_parse_query_to_ast_helper("", "<emptyclause>");
+    }
+
+    #[test]
+    fn test_parse_query_to_ast_hyphen() {
+        test_parse_query_to_ast_helper("\"www-form-encoded\"", "\"www-form-encoded\"");
+        test_parse_query_to_ast_helper("www-form-encoded", "\"www-form-encoded\"");
+        test_parse_query_to_ast_helper("www-form-encoded", "\"www-form-encoded\"");
+    }
+
+    #[test]
+    fn test_parse_query_to_ast_not_op() {
+        assert_eq!(
+            format!("{:?}", parse_to_ast().parse("NOT")),
+            "Err(UnexpectedParse)"
+        );
+        test_parse_query_to_ast_helper("NOTa", "\"NOTa\"");
+        test_parse_query_to_ast_helper("NOT a", "(-\"a\")");
+    }
+
+    #[test]
+    fn test_boosting() {
+        assert!(parse_to_ast().parse("a^2^3").is_err());
+        assert!(parse_to_ast().parse("a^2^").is_err());
+        test_parse_query_to_ast_helper("a^3", "(\"a\")^3");
+        test_parse_query_to_ast_helper("a^3 b^2", "(*(\"a\")^3 *(\"b\")^2)");
+        test_parse_query_to_ast_helper("a^1", "\"a\"");
+    }
+
+    #[test]
+    fn test_parse_query_to_ast_binary_op() {
+        test_parse_query_to_ast_helper("a AND b", "(+\"a\" +\"b\")");
+        test_parse_query_to_ast_helper("a OR b", "(?\"a\" ?\"b\")");
+        test_parse_query_to_ast_helper("a OR b AND c", "(?\"a\" ?(+\"b\" +\"c\"))");
+        test_parse_query_to_ast_helper("a AND b         AND c", "(+\"a\" +\"b\" +\"c\")");
+        assert_eq!(
+            format!("{:?}", parse_to_ast().parse("a OR b aaa")),
+            "Err(UnexpectedParse)"
+        );
+        assert_eq!(
+            format!("{:?}", parse_to_ast().parse("a AND b aaa")),
+            "Err(UnexpectedParse)"
+        );
+        assert_eq!(
+            format!("{:?}", parse_to_ast().parse("aaa a OR b ")),
+            "Err(UnexpectedParse)"
+        );
+        assert_eq!(
+            format!("{:?}", parse_to_ast().parse("aaa ccc a OR b ")),
+            "Err(UnexpectedParse)"
+        );
+    }
+
+    #[test]
+    fn test_parse_elastic_query_ranges() {
+        test_parse_query_to_ast_helper("title: >a", "title:{\"a\" TO \"*\"}");
+        test_parse_query_to_ast_helper("title:>=a", "title:[\"a\" TO \"*\"}");
+        test_parse_query_to_ast_helper("title: <a", "title:{\"*\" TO \"a\"}");
+        test_parse_query_to_ast_helper("title:<=a", "title:{\"*\" TO \"a\"]");
+        test_parse_query_to_ast_helper("title:<=bsd", "title:{\"*\" TO \"bsd\"]");
+
+        test_parse_query_to_ast_helper("weight: >70", "weight:{\"70\" TO \"*\"}");
+        test_parse_query_to_ast_helper("weight:>=70", "weight:[\"70\" TO \"*\"}");
+        test_parse_query_to_ast_helper("weight: <70", "weight:{\"*\" TO \"70\"}");
+        test_parse_query_to_ast_helper("weight:<=70", "weight:{\"*\" TO \"70\"]");
+        test_parse_query_to_ast_helper("weight: >60.7", "weight:{\"60.7\" TO \"*\"}");
+
+        test_parse_query_to_ast_helper("weight: <= 70", "weight:{\"*\" TO \"70\"]");
+
+        test_parse_query_to_ast_helper("weight: <= 70.5", "weight:{\"*\" TO \"70.5\"]");
+    }
+
+    #[test]
+    fn test_occur_leaf() {
+        let ((occur, ast), _) = super::occur_leaf().parse("+abc").unwrap();
+        assert_eq!(occur, Some(Occur::Must));
+        assert_eq!(format!("{:?}", ast), "\"abc\"");
+    }
+
+    #[test]
+    fn test_range_parser() {
+        // testing the range() parser separately
+        let res = range().parse("title: <hello").unwrap().0;
+        let expected = UserInputLeaf::Range {
+            field: Some("title".to_string()),
+            lower: UserInputBound::Unbounded,
+            upper: UserInputBound::Exclusive("hello".to_string()),
+        };
+        let res2 = range().parse("title:{* TO hello}").unwrap().0;
+        assert_eq!(res, expected);
+        assert_eq!(res2, expected);
+        let expected_weight = UserInputLeaf::Range {
+            field: Some("weight".to_string()),
+            lower: UserInputBound::Inclusive("71.2".to_string()),
+            upper: UserInputBound::Unbounded,
+        };
+
+        let res3 = range().parse("weight: >=71.2").unwrap().0;
+        let res4 = range().parse("weight:[71.2 TO *}").unwrap().0;
+        assert_eq!(res3, expected_weight);
+        assert_eq!(res4, expected_weight);
+    }
+
+    #[test]
+    fn test_parse_query_to_triming_spaces() {
+        test_parse_query_to_ast_helper("   abc", "\"abc\"");
+        test_parse_query_to_ast_helper("abc ", "\"abc\"");
+        test_parse_query_to_ast_helper("(  a OR abc)", "(?\"a\" ?\"abc\")");
+        test_parse_query_to_ast_helper("(a  OR abc)", "(?\"a\" ?\"abc\")");
+        test_parse_query_to_ast_helper("(a OR  abc)", "(?\"a\" ?\"abc\")");
+        test_parse_query_to_ast_helper("a OR abc ", "(?\"a\" ?\"abc\")");
+        test_parse_query_to_ast_helper("(a OR abc )", "(?\"a\" ?\"abc\")");
+        test_parse_query_to_ast_helper("(a OR  abc) ", "(?\"a\" ?\"abc\")");
+    }
+
+    #[test]
+    fn test_parse_query_single_term() {
+        test_parse_query_to_ast_helper("abc", "\"abc\"");
+    }
+
+    #[test]
+    fn test_parse_query_default_clause() {
+        test_parse_query_to_ast_helper("a b", "(*\"a\" *\"b\")");
+    }
+
+    #[test]
+    fn test_parse_query_must_default_clause() {
+        test_parse_query_to_ast_helper("+(a b)", "(*\"a\" *\"b\")");
+    }
+
+    #[test]
+    fn test_parse_query_must_single_term() {
+        test_parse_query_to_ast_helper("+d", "\"d\"");
+    }
+
+    #[test]
+    fn test_single_term_with_field() {
+        test_parse_query_to_ast_helper("abc:toto", "abc:\"toto\"");
+    }
+
+    #[test]
+    fn test_single_term_with_float() {
+        test_parse_query_to_ast_helper("abc:1.1", "abc:\"1.1\"");
+    }
+
+    #[test]
+    fn test_must_clause() {
+        test_parse_query_to_ast_helper("(+a +b)", "(+\"a\" +\"b\")");
+    }
+
+    #[test]
+    fn test_parse_test_query_plus_a_b_plus_d() {
+        test_parse_query_to_ast_helper("+(a b) +d", "(+(*\"a\" *\"b\") +\"d\")");
+    }
+
+    #[test]
+    fn test_parse_test_query_other() {
+        test_parse_query_to_ast_helper("(+a +b) d", "(*(+\"a\" +\"b\") *\"d\")");
+        test_parse_query_to_ast_helper("+abc:toto", "abc:\"toto\"");
+        test_parse_query_to_ast_helper("(+abc:toto -titi)", "(+abc:\"toto\" -\"titi\")");
+        test_parse_query_to_ast_helper("-abc:toto", "(-abc:\"toto\")");
+        test_parse_query_to_ast_helper("abc:a b", "(*abc:\"a\" *\"b\")");
+        test_parse_query_to_ast_helper("abc:\"a b\"", "abc:\"a b\"");
+        test_parse_query_to_ast_helper("foo:[1 TO 5]", "foo:[\"1\" TO \"5\"]");
+    }
+
+    #[test]
+    fn test_parse_query_with_range() {
+        test_parse_query_to_ast_helper("[1 TO 5]", "[\"1\" TO \"5\"]");
+        test_parse_query_to_ast_helper("foo:{a TO z}", "foo:{\"a\" TO \"z\"}");
+        test_parse_query_to_ast_helper("foo:[1 TO toto}", "foo:[\"1\" TO \"toto\"}");
+        test_parse_query_to_ast_helper("foo:[* TO toto}", "foo:{\"*\" TO \"toto\"}");
+        test_parse_query_to_ast_helper("foo:[1 TO *}", "foo:[\"1\" TO \"*\"}");
+        test_parse_query_to_ast_helper("foo:[1.1 TO *}", "foo:[\"1.1\" TO \"*\"}");
+        test_is_parse_err("abc +    ");
+    }
+}

query-grammar/src/user_input_ast.rs

@@ -1,8 +1,9 @@
 use std::fmt;
 use std::fmt::{Debug, Formatter};
 
-use query::Occur;
+use crate::Occur;
 
+#[derive(PartialEq)]
 pub enum UserInputLeaf {
     Literal(UserInputLiteral),
     All,
@@ -14,7 +15,7 @@ pub enum UserInputLeaf {
 }
 
 impl Debug for UserInputLeaf {
-    fn fmt(&self, formatter: &mut Formatter) -> Result<(), fmt::Error> {
+    fn fmt(&self, formatter: &mut Formatter<'_>) -> Result<(), fmt::Error> {
         match self {
             UserInputLeaf::Literal(literal) => literal.fmt(formatter),
             UserInputLeaf::Range {
@@ -35,13 +36,14 @@ impl Debug for UserInputLeaf {
     }
 }
 
+#[derive(PartialEq)]
 pub struct UserInputLiteral {
     pub field_name: Option<String>,
     pub phrase: String,
 }
 
 impl fmt::Debug for UserInputLiteral {
-    fn fmt(&self, formatter: &mut fmt::Formatter) -> Result<(), fmt::Error> {
+    fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
         match self.field_name {
             Some(ref field_name) => write!(formatter, "{}:\"{}\"", field_name, self.phrase),
             None => write!(formatter, "\"{}\"", self.phrase),
@@ -49,23 +51,27 @@ impl fmt::Debug for UserInputLiteral {
     }
 }
 
+#[derive(PartialEq)]
 pub enum UserInputBound {
     Inclusive(String),
     Exclusive(String),
+    Unbounded,
 }
 
 impl UserInputBound {
-    fn display_lower(&self, formatter: &mut fmt::Formatter) -> Result<(), fmt::Error> {
+    fn display_lower(&self, formatter: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
         match *self {
             UserInputBound::Inclusive(ref word) => write!(formatter, "[\"{}\"", word),
             UserInputBound::Exclusive(ref word) => write!(formatter, "{{\"{}\"", word),
+            UserInputBound::Unbounded => write!(formatter, "{{\"*\""),
         }
     }
 
-    fn display_upper(&self, formatter: &mut fmt::Formatter) -> Result<(), fmt::Error> {
+    fn display_upper(&self, formatter: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
         match *self {
             UserInputBound::Inclusive(ref word) => write!(formatter, "\"{}\"]", word),
             UserInputBound::Exclusive(ref word) => write!(formatter, "\"{}\"}}", word),
+            UserInputBound::Unbounded => write!(formatter, "\"*\"}}"),
         }
     }
 
@@ -73,38 +79,40 @@ impl UserInputBound {
         match *self {
             UserInputBound::Inclusive(ref contents) => contents,
             UserInputBound::Exclusive(ref contents) => contents,
+            UserInputBound::Unbounded => &"*",
         }
     }
 }
 
 pub enum UserInputAST {
-    Clause(Vec<UserInputAST>),
-    Unary(Occur, Box<UserInputAST>),
-    //    Not(Box<UserInputAST>),
-    //    Should(Box<UserInputAST>),
-    //    Must(Box<UserInputAST>),
+    Clause(Vec<(Option<Occur>, UserInputAST)>),
     Leaf(Box<UserInputLeaf>),
+    Boost(Box<UserInputAST>, f32),
 }
 
 impl UserInputAST {
     pub fn unary(self, occur: Occur) -> UserInputAST {
-        UserInputAST::Unary(occur, Box::new(self))
+        UserInputAST::Clause(vec![(Some(occur), self)])
     }
 
     fn compose(occur: Occur, asts: Vec<UserInputAST>) -> UserInputAST {
-        assert!(occur != Occur::MustNot);
+        assert_ne!(occur, Occur::MustNot);
         assert!(!asts.is_empty());
         if asts.len() == 1 {
             asts.into_iter().next().unwrap() //< safe
         } else {
             UserInputAST::Clause(
                 asts.into_iter()
-                    .map(|ast: UserInputAST| ast.unary(occur))
+                    .map(|ast: UserInputAST| (Some(occur), ast))
                     .collect::<Vec<_>>(),
             )
         }
     }
 
+    pub fn empty_query() -> UserInputAST {
+        UserInputAST::Clause(Vec::default())
+    }
+
     pub fn and(asts: Vec<UserInputAST>) -> UserInputAST {
         UserInputAST::compose(Occur::Must, asts)
     }
@@ -114,42 +122,6 @@ impl UserInputAST {
     }
 }
 
-/*
-impl UserInputAST {
-
-    fn compose_occur(self, occur: Occur) -> UserInputAST {
-        match self {
-            UserInputAST::Not(other) => {
-                let new_occur = compose_occur(Occur::MustNot, occur);
-                other.simplify()
-            }
-            _ => {
-                self
-            }
-        }
-    }
-
-    pub fn simplify(self) -> UserInputAST {
-        match self {
-            UserInputAST::Clause(els) => {
-                if els.len() == 1 {
-                    return els.into_iter().next().unwrap();
-                } else {
-                    return self;
-                }
-            }
-            UserInputAST::Not(els) => {
-                if els.len() == 1 {
-                    return els.into_iter().next().unwrap();
-                } else {
-                    return self;
-                }
-            }
-        }
-    }
-}
-*/
-
 impl From<UserInputLiteral> for UserInputLeaf {
     fn from(literal: UserInputLiteral) -> UserInputLeaf {
         UserInputLeaf::Literal(literal)
@@ -162,26 +134,38 @@ impl From<UserInputLeaf> for UserInputAST {
     }
 }
 
+fn print_occur_ast(
+    occur_opt: Option<Occur>,
+    ast: &UserInputAST,
+    formatter: &mut fmt::Formatter,
+) -> fmt::Result {
+    if let Some(occur) = occur_opt {
+        write!(formatter, "{}{:?}", occur, ast)?;
+    } else {
+        write!(formatter, "*{:?}", ast)?;
+    }
+    Ok(())
+}
+
 impl fmt::Debug for UserInputAST {
-    fn fmt(&self, formatter: &mut fmt::Formatter) -> Result<(), fmt::Error> {
+    fn fmt(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
         match *self {
             UserInputAST::Clause(ref subqueries) => {
                 if subqueries.is_empty() {
                     write!(formatter, "<emptyclause>")?;
                 } else {
                     write!(formatter, "(")?;
-                    write!(formatter, "{:?}", &subqueries[0])?;
+                    print_occur_ast(subqueries[0].0, &subqueries[0].1, formatter)?;
                     for subquery in &subqueries[1..] {
-                        write!(formatter, " {:?}", subquery)?;
+                        write!(formatter, " ")?;
+                        print_occur_ast(subquery.0, &subquery.1, formatter)?;
                     }
                     write!(formatter, ")")?;
                 }
                 Ok(())
             }
-            UserInputAST::Unary(ref occur, ref subquery) => {
-                write!(formatter, "{}({:?})", occur.to_char(), subquery)
-            }
             UserInputAST::Leaf(ref subquery) => write!(formatter, "{:?}", subquery),
+            UserInputAST::Boost(ref leaf, boost) => write!(formatter, "({:?})^{}", leaf, boost),
         }
     }
 }

run-tests.sh

@@ -1,2 +1,2 @@
 #!/bin/bash
-cargo test --no-default-features --features mmap -- --test-threads 1
+cargo test

src/collector/count_collector.rs

@@ -1,58 +1,40 @@
 use super::Collector;
-use collector::SegmentCollector;
-use DocId;
-use Result;
-use Score;
-use SegmentLocalId;
-use SegmentReader;
+use crate::collector::SegmentCollector;
+use crate::DocId;
+use crate::Score;
+use crate::SegmentLocalId;
+use crate::SegmentReader;
 
 /// `CountCollector` collector only counts how many
 /// documents match the query.
 ///
 /// ```rust
-/// #[macro_use]
-/// extern crate tantivy;
-/// use tantivy::schema::{Schema, TEXT};
-/// use tantivy::{Index, Result};
 /// use tantivy::collector::Count;
 /// use tantivy::query::QueryParser;
+/// use tantivy::schema::{Schema, TEXT};
+/// use tantivy::{doc, Index};
 ///
-/// # fn main() { example().unwrap(); }
-/// fn example() -> Result<()> {
-///     let mut schema_builder = Schema::builder();
-///     let title = schema_builder.add_text_field("title", TEXT);
-///     let schema = schema_builder.build();
-///     let index = Index::create_in_ram(schema);
-///     {
-///         let mut index_writer = index.writer(3_000_000)?;
-///         index_writer.add_document(doc!(
-///             title => "The Name of the Wind",
-///         ));
-///         index_writer.add_document(doc!(
-///             title => "The Diary of Muadib",
-///         ));
-///         index_writer.add_document(doc!(
-///             title => "A Dairy Cow",
-///         ));
-///         index_writer.add_document(doc!(
-///             title => "The Diary of a Young Girl",
-///         ));
-///         index_writer.commit().unwrap();
-///     }
+/// let mut schema_builder = Schema::builder();
+/// let title = schema_builder.add_text_field("title", TEXT);
+/// let schema = schema_builder.build();
+/// let index = Index::create_in_ram(schema);
 ///
-///     index.load_searchers()?;
-///     let searcher = index.searcher();
+/// let mut index_writer = index.writer(3_000_000).unwrap();
+/// index_writer.add_document(doc!(title => "The Name of the Wind"));
+/// index_writer.add_document(doc!(title => "The Diary of Muadib"));
+/// index_writer.add_document(doc!(title => "A Dairy Cow"));
+/// index_writer.add_document(doc!(title => "The Diary of a Young Girl"));
+/// assert!(index_writer.commit().is_ok());
 ///
-///     {
-///         let query_parser = QueryParser::for_index(&index, vec![title]);
-///         let query = query_parser.parse_query("diary")?;
-///         let count = searcher.search(&query, &Count).unwrap();
+/// let reader = index.reader().unwrap();
+/// let searcher = reader.searcher();
 ///
-///         assert_eq!(count, 2);
-///     }
+/// // Here comes the important part
+/// let query_parser = QueryParser::for_index(&index, vec![title]);
+/// let query = query_parser.parse_query("diary").unwrap();
+/// let count = searcher.search(&query, &Count).unwrap();
 ///
-///     Ok(())
-/// }
+/// assert_eq!(count, 2);
 /// ```
 pub struct Count;
 
@@ -61,7 +43,11 @@ impl Collector for Count {
 
     type Child = SegmentCountCollector;
 
-    fn for_segment(&self, _: SegmentLocalId, _: &SegmentReader) -> Result<SegmentCountCollector> {
+    fn for_segment(
+        &self,
+        _: SegmentLocalId,
+        _: &SegmentReader,
+    ) -> crate::Result<SegmentCountCollector> {
         Ok(SegmentCountCollector::default())
     }
 
@@ -69,7 +55,7 @@ impl Collector for Count {
         false
     }
 
-    fn merge_fruits(&self, segment_counts: Vec<usize>) -> Result<usize> {
+    fn merge_fruits(&self, segment_counts: Vec<usize>) -> crate::Result<usize> {
         Ok(segment_counts.into_iter().sum())
     }
 }
@@ -94,8 +80,8 @@ impl SegmentCollector for SegmentCountCollector {
 #[cfg(test)]
 mod tests {
     use super::{Count, SegmentCountCollector};
-    use collector::Collector;
-    use collector::SegmentCollector;
+    use crate::collector::Collector;
+    use crate::collector::SegmentCollector;
 
     #[test]
     fn test_count_collect_does_not_requires_scoring() {
@@ -125,5 +111,4 @@ mod tests {
             assert_eq!(count_collector.harvest(), 2);
         }
     }
-
 }

src/collector/custom_score_top_collector.rs

@@ -0,0 +1,125 @@
+use crate::collector::top_collector::{TopCollector, TopSegmentCollector};
+use crate::collector::{Collector, SegmentCollector};
+use crate::{DocAddress, DocId, Score, SegmentReader};
+
+pub(crate) struct CustomScoreTopCollector<TCustomScorer, TScore = Score> {
+    custom_scorer: TCustomScorer,
+    collector: TopCollector<TScore>,
+}
+
+impl<TCustomScorer, TScore> CustomScoreTopCollector<TCustomScorer, TScore>
+where
+    TScore: Clone + PartialOrd,
+{
+    pub(crate) fn new(
+        custom_scorer: TCustomScorer,
+        collector: TopCollector<TScore>,
+    ) -> CustomScoreTopCollector<TCustomScorer, TScore> {
+        CustomScoreTopCollector {
+            custom_scorer,
+            collector,
+        }
+    }
+}
+
+/// A custom segment scorer makes it possible to define any kind of score
+/// for a given document belonging to a specific segment.
+///
+/// It is the segment local version of the [`CustomScorer`](./trait.CustomScorer.html).
+pub trait CustomSegmentScorer<TScore>: 'static {
+    /// Computes the score of a specific `doc`.
+    fn score(&mut self, doc: DocId) -> TScore;
+}
+
+/// `CustomScorer` makes it possible to define any kind of score.
+///
+/// The `CustomerScorer` itself does not make much of the computation itself.
+/// Instead, it helps constructing `Self::Child` instances that will compute
+/// the score at a segment scale.
+pub trait CustomScorer<TScore>: Sync {
+    /// Type of the associated [`CustomSegmentScorer`](./trait.CustomSegmentScorer.html).
+    type Child: CustomSegmentScorer<TScore>;
+    /// Builds a child scorer for a specific segment. The child scorer is associated to
+    /// a specific segment.
+    fn segment_scorer(&self, segment_reader: &SegmentReader) -> crate::Result<Self::Child>;
+}
+
+impl<TCustomScorer, TScore> Collector for CustomScoreTopCollector<TCustomScorer, TScore>
+where
+    TCustomScorer: CustomScorer<TScore>,
+    TScore: 'static + PartialOrd + Clone + Send + Sync,
+{
+    type Fruit = Vec<(TScore, DocAddress)>;
+
+    type Child = CustomScoreTopSegmentCollector<TCustomScorer::Child, TScore>;
+
+    fn for_segment(
+        &self,
+        segment_local_id: u32,
+        segment_reader: &SegmentReader,
+    ) -> crate::Result<Self::Child> {
+        let segment_scorer = self.custom_scorer.segment_scorer(segment_reader)?;
+        let segment_collector = self
+            .collector
+            .for_segment(segment_local_id, segment_reader)?;
+        Ok(CustomScoreTopSegmentCollector {
+            segment_collector,
+            segment_scorer,
+        })
+    }
+
+    fn requires_scoring(&self) -> bool {
+        false
+    }
+
+    fn merge_fruits(&self, segment_fruits: Vec<Self::Fruit>) -> crate::Result<Self::Fruit> {
+        self.collector.merge_fruits(segment_fruits)
+    }
+}
+
+pub struct CustomScoreTopSegmentCollector<T, TScore>
+where
+    TScore: 'static + PartialOrd + Clone + Send + Sync + Sized,
+    T: CustomSegmentScorer<TScore>,
+{
+    segment_collector: TopSegmentCollector<TScore>,
+    segment_scorer: T,
+}
+
+impl<T, TScore> SegmentCollector for CustomScoreTopSegmentCollector<T, TScore>
+where
+    TScore: 'static + PartialOrd + Clone + Send + Sync,
+    T: 'static + CustomSegmentScorer<TScore>,
+{
+    type Fruit = Vec<(TScore, DocAddress)>;
+
+    fn collect(&mut self, doc: DocId, _score: Score) {
+        let score = self.segment_scorer.score(doc);
+        self.segment_collector.collect(doc, score);
+    }
+
+    fn harvest(self) -> Vec<(TScore, DocAddress)> {
+        self.segment_collector.harvest()
+    }
+}
+
+impl<F, TScore, T> CustomScorer<TScore> for F
+where
+    F: 'static + Send + Sync + Fn(&SegmentReader) -> T,
+    T: CustomSegmentScorer<TScore>,
+{
+    type Child = T;
+
+    fn segment_scorer(&self, segment_reader: &SegmentReader) -> crate::Result<Self::Child> {
+        Ok((self)(segment_reader))
+    }
+}
+
+impl<F, TScore> CustomSegmentScorer<TScore> for F
+where
+    F: 'static + FnMut(DocId) -> TScore,
+{
+    fn score(&mut self, doc: DocId) -> TScore {
+        (self)(doc)
+    }
+}

src/collector/facet_collector.rs

@@ -1,9 +1,13 @@
-use collector::Collector;
-use collector::SegmentCollector;
-use docset::SkipResult;
-use fastfield::FacetReader;
-use schema::Facet;
-use schema::Field;
+use crate::collector::Collector;
+use crate::collector::SegmentCollector;
+use crate::fastfield::FacetReader;
+use crate::schema::Facet;
+use crate::schema::Field;
+use crate::DocId;
+use crate::Score;
+use crate::SegmentLocalId;
+use crate::SegmentReader;
+use crate::TantivyError;
 use std::cmp::Ordering;
 use std::collections::btree_map;
 use std::collections::BTreeMap;
@@ -12,11 +16,6 @@ use std::collections::BinaryHeap;
 use std::collections::Bound;
 use std::iter::Peekable;
 use std::{u64, usize};
-use DocId;
-use Result;
-use Score;
-use SegmentLocalId;
-use SegmentReader;
 
 struct Hit<'a> {
     count: u64,
@@ -26,13 +25,13 @@ struct Hit<'a> {
 impl<'a> Eq for Hit<'a> {}
 
 impl<'a> PartialEq<Hit<'a>> for Hit<'a> {
-    fn eq(&self, other: &Hit) -> bool {
+    fn eq(&self, other: &Hit<'_>) -> bool {
         self.count == other.count
     }
 }
 
 impl<'a> PartialOrd<Hit<'a>> for Hit<'a> {
-    fn partial_cmp(&self, other: &Hit) -> Option<Ordering> {
+    fn partial_cmp(&self, other: &Hit<'_>) -> Option<Ordering> {
         Some(self.cmp(other))
     }
 }
@@ -80,15 +79,12 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
 ///
 ///
 /// ```rust
-/// #[macro_use]
-/// extern crate tantivy;
-/// use tantivy::schema::{Facet, Schema, TEXT};
-/// use tantivy::{Index, Result};
 /// use tantivy::collector::FacetCollector;
 /// use tantivy::query::AllQuery;
+/// use tantivy::schema::{Facet, Schema, TEXT};
+/// use tantivy::{doc, Index};
 ///
-/// # fn main() { example().unwrap(); }
-/// fn example() -> Result<()> {
+/// fn example() -> tantivy::Result<()> {
 ///     let mut schema_builder = Schema::builder();
 ///
 ///     // Facet have their own specific type.
@@ -122,17 +118,16 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
 ///             facet => Facet::from("/lang/en"),
 ///             facet => Facet::from("/category/biography")
 ///         ));
-///         index_writer.commit().unwrap();
+///         index_writer.commit()?;
 ///     }
-///
-///     index.load_searchers()?;
-///     let searcher = index.searcher();
+///     let reader = index.reader()?;
+///     let searcher = reader.searcher();
 ///
 ///     {
-///			let mut facet_collector = FacetCollector::for_field(facet);
+///         let mut facet_collector = FacetCollector::for_field(facet);
 ///         facet_collector.add_facet("/lang");
 ///         facet_collector.add_facet("/category");
-///         let facet_counts = searcher.search(&AllQuery, &facet_collector).unwrap();
+///         let facet_counts = searcher.search(&AllQuery, &facet_collector)?;
 ///
 ///         // This lists all of the facet counts
 ///         let facets: Vec<(&Facet, u64)> = facet_counts
@@ -145,9 +140,9 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
 ///     }
 ///
 ///     {
-///			let mut facet_collector = FacetCollector::for_field(facet);
+///         let mut facet_collector = FacetCollector::for_field(facet);
 ///         facet_collector.add_facet("/category/fiction");
-///         let facet_counts = searcher.search(&AllQuery, &facet_collector).unwrap();
+///         let facet_counts = searcher.search(&AllQuery, &facet_collector)?;
 ///
 ///         // This lists all of the facet counts
 ///         let facets: Vec<(&Facet, u64)> = facet_counts
@@ -160,10 +155,10 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
 ///         ]);
 ///     }
 ///
-///    {
-///			let mut facet_collector = FacetCollector::for_field(facet);
+///     {
+///         let mut facet_collector = FacetCollector::for_field(facet);
 ///         facet_collector.add_facet("/category/fiction");
-///         let facet_counts = searcher.search(&AllQuery, &facet_collector).unwrap();
+///         let facet_counts = searcher.search(&AllQuery, &facet_collector)?;
 ///
 ///         // This lists all of the facet counts
 ///         let facets: Vec<(&Facet, u64)> = facet_counts.top_k("/category/fiction", 1);
@@ -174,6 +169,7 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
 ///
 ///     Ok(())
 /// }
+/// # assert!(example().is_ok());
 /// ```
 pub struct FacetCollector {
     field: Field,
@@ -191,23 +187,28 @@ pub struct FacetSegmentCollector {
     collapse_facet_ords: Vec<u64>,
 }
 
+enum SkipResult {
+    Found,
+    NotFound,
+}
+
 fn skip<'a, I: Iterator<Item = &'a Facet>>(
     target: &[u8],
     collapse_it: &mut Peekable<I>,
 ) -> SkipResult {
     loop {
         match collapse_it.peek() {
-            Some(facet_bytes) => match facet_bytes.encoded_bytes().cmp(target) {
+            Some(facet_bytes) => match facet_bytes.encoded_str().as_bytes().cmp(target) {
                 Ordering::Less => {}
                 Ordering::Greater => {
-                    return SkipResult::OverStep;
+                    return SkipResult::NotFound;
                 }
                 Ordering::Equal => {
-                    return SkipResult::Reached;
+                    return SkipResult::Found;
                 }
             },
             None => {
-                return SkipResult::End;
+                return SkipResult::NotFound;
             }
         }
         collapse_it.next();
@@ -264,8 +265,11 @@ impl Collector for FacetCollector {
         &self,
         _: SegmentLocalId,
         reader: &SegmentReader,
-    ) -> Result<FacetSegmentCollector> {
-        let facet_reader = reader.facet_reader(self.field)?;
+    ) -> crate::Result<FacetSegmentCollector> {
+        let field_name = reader.schema().get_field_name(self.field);
+        let facet_reader = reader.facet_reader(self.field).ok_or_else(|| {
+            TantivyError::SchemaError(format!("Field {:?} is not a facet field.", field_name))
+        })?;
 
         let mut collapse_mapping = Vec::new();
         let mut counts = Vec::new();
@@ -281,7 +285,7 @@ impl Collector for FacetCollector {
                     // is positionned on a term that has not been processed yet.
                     let skip_result = skip(facet_streamer.key(), &mut collapse_facet_it);
                     match skip_result {
-                        SkipResult::Reached => {
+                        SkipResult::Found => {
                             // we reach a facet we decided to collapse.
                             let collapse_depth = facet_depth(facet_streamer.key());
                             let mut collapsed_id = 0;
@@ -301,7 +305,7 @@ impl Collector for FacetCollector {
                             }
                             break;
                         }
-                        SkipResult::End | SkipResult::OverStep => {
+                        SkipResult::NotFound => {
                             collapse_mapping.push(0);
                             if !facet_streamer.advance() {
                                 break;
@@ -327,7 +331,7 @@ impl Collector for FacetCollector {
         false
     }
 
-    fn merge_fruits(&self, segments_facet_counts: Vec<FacetCounts>) -> Result<FacetCounts> {
+    fn merge_fruits(&self, segments_facet_counts: Vec<FacetCounts>) -> crate::Result<FacetCounts> {
         let mut facet_counts: BTreeMap<Facet, u64> = BTreeMap::new();
         for segment_facet_counts in segments_facet_counts {
             for (facet, count) in segment_facet_counts.facet_counts {
@@ -369,7 +373,8 @@ impl SegmentCollector for FacetSegmentCollector {
             let mut facet = vec![];
             let facet_ord = self.collapse_facet_ords[collapsed_facet_ord];
             facet_dict.ord_to_term(facet_ord as u64, &mut facet);
-            facet_counts.insert(unsafe { Facet::from_encoded(facet) }, count);
+            // TODO
+            facet_counts.insert(Facet::from_encoded(facet).unwrap(), count);
         }
         FacetCounts { facet_counts }
     }
@@ -394,7 +399,7 @@ impl<'a> Iterator for FacetChildIterator<'a> {
 }
 
 impl FacetCounts {
-    pub fn get<T>(&self, facet_from: T) -> FacetChildIterator
+    pub fn get<T>(&self, facet_from: T) -> FacetChildIterator<'_>
     where
         Facet: From<T>,
     {
@@ -403,12 +408,13 @@ impl FacetCounts {
         let right_bound = if facet.is_root() {
             Bound::Unbounded
         } else {
-            let mut facet_after_bytes: Vec<u8> = facet.encoded_bytes().to_owned();
-            facet_after_bytes.push(1u8);
-            let facet_after = unsafe { Facet::from_encoded(facet_after_bytes) }; // ok logic
+            let mut facet_after_bytes: String = facet.encoded_str().to_owned();
+            facet_after_bytes.push('\u{1}');
+            let facet_after = Facet::from_encoded_string(facet_after_bytes);
             Bound::Excluded(facet_after)
         };
-        let underlying: btree_map::Range<_, _> = self.facet_counts.range((left_bound, right_bound));
+        let underlying: btree_map::Range<'_, _, _> =
+            self.facet_counts.range((left_bound, right_bound));
         FacetChildIterator { underlying }
     }
 
@@ -449,12 +455,14 @@ impl FacetCounts {
 #[cfg(test)]
 mod tests {
     use super::{FacetCollector, FacetCounts};
-    use core::Index;
-    use query::AllQuery;
+    use crate::collector::Count;
+    use crate::core::Index;
+    use crate::query::{AllQuery, QueryParser, TermQuery};
+    use crate::schema::{Document, Facet, Field, IndexRecordOption, Schema};
+    use crate::Term;
     use rand::distributions::Uniform;
     use rand::prelude::SliceRandom;
     use rand::{thread_rng, Rng};
-    use schema::{Document, Facet, Field, Schema};
     use std::iter;
 
     #[test]
@@ -474,15 +482,16 @@ mod tests {
                 n /= 4;
                 let leaf = n % 5;
                 Facet::from(&format!("/top{}/mid{}/leaf{}", top, mid, leaf))
-            }).collect();
+            })
+            .collect();
         for i in 0..num_facets * 10 {
             let mut doc = Document::new();
             doc.add_facet(facet_field, facets[i % num_facets].clone());
             index_writer.add_document(doc);
         }
         index_writer.commit().unwrap();
-        index.load_searchers().unwrap();
-        let searcher = index.searcher();
+        let reader = index.reader().unwrap();
+        let searcher = reader.searcher();
         let mut facet_collector = FacetCollector::for_field(facet_field);
         facet_collector.add_facet(Facet::from("/top1"));
         let counts = searcher.search(&AllQuery, &facet_collector).unwrap();
@@ -500,20 +509,18 @@ mod tests {
                     ("/top1/mid2", 50),
                     ("/top1/mid3", 50),
                 ]
-                    .iter()
-                    .map(|&(facet_str, count)| (String::from(facet_str), count))
-                    .collect::<Vec<_>>()
+                .iter()
+                .map(|&(facet_str, count)| (String::from(facet_str), count))
+                .collect::<Vec<_>>()
             );
         }
     }
 
     #[test]
-    #[should_panic(
-        expected = "Tried to add a facet which is a descendant of \
-                    an already added facet."
-    )]
+    #[should_panic(expected = "Tried to add a facet which is a descendant of \
+                               an already added facet.")]
     fn test_misused_facet_collector() {
-        let mut facet_collector = FacetCollector::for_field(Field(0));
+        let mut facet_collector = FacetCollector::for_field(Field::from_field_id(0));
         facet_collector.add_facet(Facet::from("/country"));
         facet_collector.add_facet(Facet::from("/country/europe"));
     }
@@ -532,8 +539,8 @@ mod tests {
             facet_field => Facet::from_text(&"/subjects/B/b"),
         ));
         index_writer.commit().unwrap();
-        index.load_searchers().unwrap();
-        let searcher = index.searcher();
+        let reader = index.reader().unwrap();
+        let searcher = reader.searcher();
         assert_eq!(searcher.num_docs(), 1);
         let mut facet_collector = FacetCollector::for_field(facet_field);
         facet_collector.add_facet("/subjects");
@@ -542,9 +549,59 @@ mod tests {
         assert_eq!(facets[0].1, 1);
     }
 
+    #[test]
+    fn test_doc_search_by_facet() {
+        let mut schema_builder = Schema::builder();
+        let facet_field = schema_builder.add_facet_field("facet");
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+        index_writer.add_document(doc!(
+            facet_field => Facet::from_text(&"/A/A"),
+        ));
+        index_writer.add_document(doc!(
+            facet_field => Facet::from_text(&"/A/B"),
+        ));
+        index_writer.add_document(doc!(
+            facet_field => Facet::from_text(&"/A/C/A"),
+        ));
+        index_writer.add_document(doc!(
+            facet_field => Facet::from_text(&"/D/C/A"),
+        ));
+        index_writer.commit().unwrap();
+        let reader = index.reader().unwrap();
+        let searcher = reader.searcher();
+        assert_eq!(searcher.num_docs(), 4);
+
+        let count_facet = |facet_str: &str| {
+            let term = Term::from_facet(facet_field, &Facet::from_text(facet_str));
+            searcher
+                .search(&TermQuery::new(term, IndexRecordOption::Basic), &Count)
+                .unwrap()
+        };
+
+        assert_eq!(count_facet("/"), 4);
+        assert_eq!(count_facet("/A"), 3);
+        assert_eq!(count_facet("/A/B"), 1);
+        assert_eq!(count_facet("/A/C"), 1);
+        assert_eq!(count_facet("/A/C/A"), 1);
+        assert_eq!(count_facet("/C/A"), 0);
+        {
+            let query_parser = QueryParser::for_index(&index, vec![]);
+            {
+                let query = query_parser.parse_query("facet:/A/B").unwrap();
+                assert_eq!(1, searcher.search(&query, &Count).unwrap());
+            }
+            {
+                let query = query_parser.parse_query("facet:/A").unwrap();
+                assert_eq!(3, searcher.search(&query, &Count).unwrap());
+            }
+        }
+    }
+
     #[test]
     fn test_non_used_facet_collector() {
-        let mut facet_collector = FacetCollector::for_field(Field(0));
+        let mut facet_collector = FacetCollector::for_field(Field::from_field_id(0));
         facet_collector.add_facet(Facet::from("/country"));
         facet_collector.add_facet(Facet::from("/countryeurope"));
     }
@@ -563,13 +620,15 @@ mod tests {
                 let facet = Facet::from(&format!("/facet/{}", c));
                 let doc = doc!(facet_field => facet);
                 iter::repeat(doc).take(count)
-            }).map(|mut doc| {
+            })
+            .map(|mut doc| {
                 doc.add_facet(
                     facet_field,
                     &format!("/facet/{}", thread_rng().sample(&uniform)),
                 );
                 doc
-            }).collect();
+            })
+            .collect();
         docs[..].shuffle(&mut thread_rng());
 
         let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
@@ -577,9 +636,7 @@ mod tests {
             index_writer.add_document(doc);
         }
         index_writer.commit().unwrap();
-        index.load_searchers().unwrap();
-
-        let searcher = index.searcher();
+        let searcher = index.reader().unwrap().searcher();
 
         let mut facet_collector = FacetCollector::for_field(facet_field);
         facet_collector.add_facet("/facet");
@@ -597,19 +654,18 @@ mod tests {
             );
         }
     }
-
 }
 
 #[cfg(all(test, feature = "unstable"))]
 mod bench {
 
-    use collector::FacetCollector;
-    use query::AllQuery;
-    use rand::{thread_rng, Rng};
-    use schema::Facet;
-    use schema::Schema;
+    use crate::collector::FacetCollector;
+    use crate::query::AllQuery;
+    use crate::schema::{Facet, Schema};
+    use crate::Index;
+    use rand::seq::SliceRandom;
+    use rand::thread_rng;
     use test::Bencher;
-    use Index;
 
     #[bench]
     fn bench_facet_collector(b: &mut Bencher) {
@@ -626,17 +682,16 @@ mod bench {
             }
         }
         // 40425 docs
-        thread_rng().shuffle(&mut docs[..]);
+        docs[..].shuffle(&mut thread_rng());
 
         let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
         for doc in docs {
             index_writer.add_document(doc);
         }
         index_writer.commit().unwrap();
-        index.load_searchers().unwrap();
-
+        let reader = index.reader().unwrap();
         b.iter(|| {
-            let searcher = index.searcher();
+            let searcher = reader.searcher();
             let facet_collector = FacetCollector::for_field(facet_field);
             searcher.search(&AllQuery, &facet_collector).unwrap();
         });

src/collector/int_facet_collector.rs

@@ -82,18 +82,20 @@ mod tests {
         let mut schema_builder = schema::Schema::builder();
         let num_field_i64 = schema_builder.add_i64_field("num_i64", FAST);
         let num_field_u64 = schema_builder.add_u64_field("num_u64", FAST);
+        let num_field_f64 = schema_builder.add_f64_field("num_f64", FAST);
         let text_field = schema_builder.add_text_field("text", STRING);
         let schema = schema_builder.build();
 
         let index = Index::create_in_ram(schema.clone());
 
         {
-            let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
+            let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
             {
                 for i in 0u64..10u64 {
                     index_writer.add_document(doc!(
                         num_field_i64 => ((i as i64) % 3i64) as i64,
                         num_field_u64 => (i % 2u64) as u64,
+                        num_field_f64 => (i % 4u64) as f64,
                         text_field => "text"
                     ));
                 }
@@ -101,14 +103,14 @@ mod tests {
             assert_eq!(index_writer.commit().unwrap(), 10u64);
         }
 
-        index.load_searchers().unwrap();
-        let searcher = index.searcher();
+        let searcher = index.reader().searcher();
         let mut ffvf_i64: IntFacetCollector<I64FastFieldReader> = IntFacetCollector::new(num_field_i64);
         let mut ffvf_u64: IntFacetCollector<U64FastFieldReader> = IntFacetCollector::new(num_field_u64);
+        let mut ffvf_f64: IntFacetCollector<F64FastFieldReader> = IntFacetCollector::new(num_field_f64);
 
         {
             // perform the query
-            let mut facet_collectors = chain().push(&mut ffvf_i64).push(&mut ffvf_u64);
+            let mut facet_collectors = chain().push(&mut ffvf_i64).push(&mut ffvf_u64).push(&mut ffvf_f64);
             let mut query_parser = QueryParser::for_index(index, vec![text_field]);
             let query = query_parser.parse_query("text:text").unwrap();
             query.search(&searcher, &mut facet_collectors).unwrap();
@@ -118,6 +120,8 @@ mod tests {
         assert_eq!(ffvf_u64.counters[&1], 5);
         assert_eq!(ffvf_i64.counters[&0], 4);
         assert_eq!(ffvf_i64.counters[&1], 3);
+        assert_eq!(ffvf_f64.counters[&0.0], 3);
+        assert_eq!(ffvf_f64.counters[&2.0], 2);
 
     }
 }

src/collector/mod.rs

@@ -35,7 +35,6 @@ The resulting `Fruit` will then be a typed tuple with each collector's original
 in their respective position.
 
 ```rust
-# extern crate tantivy;
 # use tantivy::schema::*;
 # use tantivy::*;
 # use tantivy::query::*;
@@ -53,9 +52,9 @@ use tantivy::collector::{Count, TopDocs};
 #     index_writer.add_document(doc!(
 #        title => "The Diary of Muadib",
 #     ));
-#     index_writer.commit().unwrap();
-#     index.load_searchers()?;
-#     let searcher = index.searcher();
+#     index_writer.commit()?;
+#     let reader = index.reader()?;
+#     let searcher = reader.searcher();
 #     let query_parser = QueryParser::for_index(&index, vec![title]);
 #     let query = query_parser.parse_query("diary")?;
 let (doc_count, top_docs): (usize, Vec<(Score, DocAddress)>) =
@@ -66,7 +65,7 @@ let (doc_count, top_docs): (usize, Vec<(Score, DocAddress)>) =
 
 The `Collector` trait is implemented for up to 4 collectors.
 If you have more than 4 collectors, you can either group them into
-tuples of tuples `(a,(b,(c,d)))`, or rely on `MultiCollector`'s.
+tuples of tuples `(a,(b,(c,d)))`, or rely on [`MultiCollector`](./struct.MultiCollector.html).
 
 # Combining several collectors dynamically
 
@@ -85,12 +84,11 @@ See the `custom_collector` example.
 
 */
 
-use downcast;
-use DocId;
-use Result;
-use Score;
-use SegmentLocalId;
-use SegmentReader;
+use crate::DocId;
+use crate::Score;
+use crate::SegmentLocalId;
+use crate::SegmentReader;
+use downcast_rs::impl_downcast;
 
 mod count_collector;
 pub use self::count_collector::Count;
@@ -103,17 +101,21 @@ mod top_collector;
 mod top_score_collector;
 pub use self::top_score_collector::TopDocs;
 
-mod top_field_collector;
-pub use self::top_field_collector::TopDocsByField;
+mod custom_score_top_collector;
+pub use self::custom_score_top_collector::{CustomScorer, CustomSegmentScorer};
+
+mod tweak_score_top_collector;
+pub use self::tweak_score_top_collector::{ScoreSegmentTweaker, ScoreTweaker};
 
 mod facet_collector;
 pub use self::facet_collector::FacetCollector;
+use crate::query::Weight;
 
 /// `Fruit` is the type for the result of our collection.
 /// e.g. `usize` for the `Count` collector.
-pub trait Fruit: Send + downcast::Any {}
+pub trait Fruit: Send + downcast_rs::Downcast {}
 
-impl<T> Fruit for T where T: Send + downcast::Any {}
+impl<T> Fruit for T where T: Send + downcast_rs::Downcast {}
 
 /// Collectors are in charge of collecting and retaining relevant
 /// information from the document found and scored by the query.
@@ -145,14 +147,37 @@ pub trait Collector: Sync {
         &self,
         segment_local_id: SegmentLocalId,
         segment: &SegmentReader,
-    ) -> Result<Self::Child>;
+    ) -> crate::Result<Self::Child>;
 
     /// Returns true iff the collector requires to compute scores for documents.
     fn requires_scoring(&self) -> bool;
 
     /// Combines the fruit associated to the collection of each segments
     /// into one fruit.
-    fn merge_fruits(&self, segment_fruits: Vec<Self::Fruit>) -> Result<Self::Fruit>;
+    fn merge_fruits(&self, segment_fruits: Vec<Self::Fruit>) -> crate::Result<Self::Fruit>;
+
+    /// Created a segment collector and
+    fn collect_segment(
+        &self,
+        weight: &dyn Weight,
+        segment_ord: u32,
+        reader: &SegmentReader,
+    ) -> crate::Result<<Self::Child as SegmentCollector>::Fruit> {
+        let mut segment_collector = self.for_segment(segment_ord as u32, reader)?;
+
+        if let Some(delete_bitset) = reader.delete_bitset() {
+            weight.for_each(reader, &mut |doc, score| {
+                if delete_bitset.is_alive(doc) {
+                    segment_collector.collect(doc, score);
+                }
+            })?;
+        } else {
+            weight.for_each(reader, &mut |doc, score| {
+                segment_collector.collect(doc, score);
+            })?;
+        }
+        Ok(segment_collector.harvest())
+    }
 }
 
 /// The `SegmentCollector` is the trait in charge of defining the
@@ -183,7 +208,11 @@ where
     type Fruit = (Left::Fruit, Right::Fruit);
     type Child = (Left::Child, Right::Child);
 
-    fn for_segment(&self, segment_local_id: u32, segment: &SegmentReader) -> Result<Self::Child> {
+    fn for_segment(
+        &self,
+        segment_local_id: u32,
+        segment: &SegmentReader,
+    ) -> crate::Result<Self::Child> {
         let left = self.0.for_segment(segment_local_id, segment)?;
         let right = self.1.for_segment(segment_local_id, segment)?;
         Ok((left, right))
@@ -196,7 +225,7 @@ where
     fn merge_fruits(
         &self,
         children: Vec<(Left::Fruit, Right::Fruit)>,
-    ) -> Result<(Left::Fruit, Right::Fruit)> {
+    ) -> crate::Result<(Left::Fruit, Right::Fruit)> {
         let mut left_fruits = vec![];
         let mut right_fruits = vec![];
         for (left_fruit, right_fruit) in children {
@@ -238,7 +267,11 @@ where
     type Fruit = (One::Fruit, Two::Fruit, Three::Fruit);
     type Child = (One::Child, Two::Child, Three::Child);
 
-    fn for_segment(&self, segment_local_id: u32, segment: &SegmentReader) -> Result<Self::Child> {
+    fn for_segment(
+        &self,
+        segment_local_id: u32,
+        segment: &SegmentReader,
+    ) -> crate::Result<Self::Child> {
         let one = self.0.for_segment(segment_local_id, segment)?;
         let two = self.1.for_segment(segment_local_id, segment)?;
         let three = self.2.for_segment(segment_local_id, segment)?;
@@ -249,7 +282,7 @@ where
         self.0.requires_scoring() || self.1.requires_scoring() || self.2.requires_scoring()
     }
 
-    fn merge_fruits(&self, children: Vec<Self::Fruit>) -> Result<Self::Fruit> {
+    fn merge_fruits(&self, children: Vec<Self::Fruit>) -> crate::Result<Self::Fruit> {
         let mut one_fruits = vec![];
         let mut two_fruits = vec![];
         let mut three_fruits = vec![];
@@ -297,7 +330,11 @@ where
     type Fruit = (One::Fruit, Two::Fruit, Three::Fruit, Four::Fruit);
     type Child = (One::Child, Two::Child, Three::Child, Four::Child);
 
-    fn for_segment(&self, segment_local_id: u32, segment: &SegmentReader) -> Result<Self::Child> {
+    fn for_segment(
+        &self,
+        segment_local_id: u32,
+        segment: &SegmentReader,
+    ) -> crate::Result<Self::Child> {
         let one = self.0.for_segment(segment_local_id, segment)?;
         let two = self.1.for_segment(segment_local_id, segment)?;
         let three = self.2.for_segment(segment_local_id, segment)?;
@@ -312,7 +349,7 @@ where
             || self.3.requires_scoring()
     }
 
-    fn merge_fruits(&self, children: Vec<Self::Fruit>) -> Result<Self::Fruit> {
+    fn merge_fruits(&self, children: Vec<Self::Fruit>) -> crate::Result<Self::Fruit> {
         let mut one_fruits = vec![];
         let mut two_fruits = vec![];
         let mut three_fruits = vec![];
@@ -358,10 +395,7 @@ where
     }
 }
 
-#[allow(missing_docs)]
-mod downcast_impl {
-    downcast!(super::Fruit);
-}
+impl_downcast!(Fruit);
 
 #[cfg(test)]
 pub mod tests;

src/collector/multi_collector.rs

@@ -1,30 +1,29 @@
 use super::Collector;
 use super::SegmentCollector;
-use collector::Fruit;
-use downcast::Downcast;
+use crate::collector::Fruit;
+use crate::DocId;
+use crate::Score;
+use crate::SegmentLocalId;
+use crate::SegmentReader;
+use crate::TantivyError;
 use std::marker::PhantomData;
-use DocId;
-use Result;
-use Score;
-use SegmentLocalId;
-use SegmentReader;
-use TantivyError;
+use std::ops::Deref;
 
 pub struct MultiFruit {
-    sub_fruits: Vec<Option<Box<Fruit>>>,
+    sub_fruits: Vec<Option<Box<dyn Fruit>>>,
 }
 
 pub struct CollectorWrapper<TCollector: Collector>(TCollector);
 
 impl<TCollector: Collector> Collector for CollectorWrapper<TCollector> {
-    type Fruit = Box<Fruit>;
-    type Child = Box<BoxableSegmentCollector>;
+    type Fruit = Box<dyn Fruit>;
+    type Child = Box<dyn BoxableSegmentCollector>;
 
     fn for_segment(
         &self,
         segment_local_id: u32,
         reader: &SegmentReader,
-    ) -> Result<Box<BoxableSegmentCollector>> {
+    ) -> crate::Result<Box<dyn BoxableSegmentCollector>> {
         let child = self.0.for_segment(segment_local_id, reader)?;
         Ok(Box::new(SegmentCollectorWrapper(child)))
     }
@@ -33,37 +32,41 @@ impl<TCollector: Collector> Collector for CollectorWrapper<TCollector> {
         self.0.requires_scoring()
     }
 
-    fn merge_fruits(&self, children: Vec<<Self as Collector>::Fruit>) -> Result<Box<Fruit>> {
+    fn merge_fruits(
+        &self,
+        children: Vec<<Self as Collector>::Fruit>,
+    ) -> crate::Result<Box<dyn Fruit>> {
         let typed_fruit: Vec<TCollector::Fruit> = children
             .into_iter()
             .map(|untyped_fruit| {
-                Downcast::<TCollector::Fruit>::downcast(untyped_fruit)
+                untyped_fruit
+                    .downcast::<TCollector::Fruit>()
                     .map(|boxed_but_typed| *boxed_but_typed)
-                    .map_err(|e| {
-                        let err_msg = format!("Failed to cast child collector fruit. {:?}", e);
-                        TantivyError::InvalidArgument(err_msg)
+                    .map_err(|_| {
+                        TantivyError::InvalidArgument("Failed to cast child fruit.".to_string())
                     })
-            }).collect::<Result<_>>()?;
+            })
+            .collect::<crate::Result<_>>()?;
         let merged_fruit = self.0.merge_fruits(typed_fruit)?;
         Ok(Box::new(merged_fruit))
     }
 }
 
-impl SegmentCollector for Box<BoxableSegmentCollector> {
-    type Fruit = Box<Fruit>;
+impl SegmentCollector for Box<dyn BoxableSegmentCollector> {
+    type Fruit = Box<dyn Fruit>;
 
     fn collect(&mut self, doc: u32, score: f32) {
         self.as_mut().collect(doc, score);
     }
 
-    fn harvest(self) -> Box<Fruit> {
+    fn harvest(self) -> Box<dyn Fruit> {
         BoxableSegmentCollector::harvest_from_box(self)
     }
 }
 
 pub trait BoxableSegmentCollector {
     fn collect(&mut self, doc: u32, score: f32);
-    fn harvest_from_box(self: Box<Self>) -> Box<Fruit>;
+    fn harvest_from_box(self: Box<Self>) -> Box<dyn Fruit>;
 }
 
 pub struct SegmentCollectorWrapper<TSegmentCollector: SegmentCollector>(TSegmentCollector);
@@ -75,7 +78,7 @@ impl<TSegmentCollector: SegmentCollector> BoxableSegmentCollector
         self.0.collect(doc, score);
     }
 
-    fn harvest_from_box(self: Box<Self>) -> Box<Fruit> {
+    fn harvest_from_box(self: Box<Self>) -> Box<dyn Fruit> {
         Box::new(self.0.harvest())
     }
 }
@@ -88,76 +91,67 @@ pub struct FruitHandle<TFruit: Fruit> {
 impl<TFruit: Fruit> FruitHandle<TFruit> {
     pub fn extract(self, fruits: &mut MultiFruit) -> TFruit {
         let boxed_fruit = fruits.sub_fruits[self.pos].take().expect("");
-        *Downcast::<TFruit>::downcast(boxed_fruit).expect("Failed")
+        *boxed_fruit
+            .downcast::<TFruit>()
+            .map_err(|_| ())
+            .expect("Failed to downcast collector fruit.")
     }
 }
 
 /// Multicollector makes it possible to collect on more than one collector.
 /// It should only be used for use cases where the Collector types is unknown
 /// at compile time.
-/// If the type of the collectors is known, you should prefer to use `ChainedCollector`.
+///
+/// If the type of the collectors is known, you can just group yours collectors
+/// in a tuple. See the
+/// [Combining several collectors section of the collector documentation](./index.html#combining-several-collectors).
 ///
 /// ```rust
-/// #[macro_use]
-/// extern crate tantivy;
-/// use tantivy::schema::{Schema, TEXT};
-/// use tantivy::{Index, Result};
 /// use tantivy::collector::{Count, TopDocs, MultiCollector};
 /// use tantivy::query::QueryParser;
+/// use tantivy::schema::{Schema, TEXT};
+/// use tantivy::{doc, Index};
 ///
-/// # fn main() { example().unwrap(); }
-/// fn example() -> Result<()> {
-///     let mut schema_builder = Schema::builder();
-///     let title = schema_builder.add_text_field("title", TEXT);
-///     let schema = schema_builder.build();
-///     let index = Index::create_in_ram(schema);
-///     {
-///         let mut index_writer = index.writer(3_000_000)?;
-///         index_writer.add_document(doc!(
-///             title => "The Name of the Wind",
-///         ));
-///         index_writer.add_document(doc!(
-///             title => "The Diary of Muadib",
-///         ));
-///         index_writer.add_document(doc!(
-///             title => "A Dairy Cow",
-///         ));
-///         index_writer.add_document(doc!(
-///             title => "The Diary of a Young Girl",
-///         ));
-///         index_writer.commit().unwrap();
-///     }
+/// let mut schema_builder = Schema::builder();
+/// let title = schema_builder.add_text_field("title", TEXT);
+/// let schema = schema_builder.build();
+/// let index = Index::create_in_ram(schema);
 ///
-///     index.load_searchers()?;
-///     let searcher = index.searcher();
+/// let mut index_writer = index.writer(3_000_000).unwrap();
+/// index_writer.add_document(doc!(title => "The Name of the Wind"));
+/// index_writer.add_document(doc!(title => "The Diary of Muadib"));
+/// index_writer.add_document(doc!(title => "A Dairy Cow"));
+/// index_writer.add_document(doc!(title => "The Diary of a Young Girl"));
+/// assert!(index_writer.commit().is_ok());
 ///
-///     let mut collectors = MultiCollector::new();
-///     let top_docs_handle = collectors.add_collector(TopDocs::with_limit(2));
-///     let count_handle = collectors.add_collector(Count);
-///     let query_parser = QueryParser::for_index(&index, vec![title]);
-///     let query = query_parser.parse_query("diary")?;
-///     let mut multi_fruit = searcher.search(&query, &collectors)?;
+/// let reader = index.reader().unwrap();
+/// let searcher = reader.searcher();
 ///
-///     let count = count_handle.extract(&mut multi_fruit);
-///     let top_docs = top_docs_handle.extract(&mut multi_fruit);
+/// let mut collectors = MultiCollector::new();
+/// let top_docs_handle = collectors.add_collector(TopDocs::with_limit(2));
+/// let count_handle = collectors.add_collector(Count);
+/// let query_parser = QueryParser::for_index(&index, vec![title]);
+/// let query = query_parser.parse_query("diary").unwrap();
+/// let mut multi_fruit = searcher.search(&query, &collectors).unwrap();
 ///
-///     # assert_eq!(count, 2);
-///     # assert_eq!(top_docs.len(), 2);
+/// let count = count_handle.extract(&mut multi_fruit);
+/// let top_docs = top_docs_handle.extract(&mut multi_fruit);
 ///
-///     Ok(())
-/// }
+/// assert_eq!(count, 2);
+/// assert_eq!(top_docs.len(), 2);
 /// ```
+#[allow(clippy::type_complexity)]
+#[derive(Default)]
 pub struct MultiCollector<'a> {
-    collector_wrappers:
-        Vec<Box<Collector<Child = Box<BoxableSegmentCollector>, Fruit = Box<Fruit>> + 'a>>,
+    collector_wrappers: Vec<
+        Box<dyn Collector<Child = Box<dyn BoxableSegmentCollector>, Fruit = Box<dyn Fruit>> + 'a>,
+    >,
 }
 
 impl<'a> MultiCollector<'a> {
     /// Create a new `MultiCollector`
-    pub fn new() -> MultiCollector<'a> {
-        MultiCollector {
-            collector_wrappers: Vec::new(),
-        }
+    pub fn new() -> Self {
+        Default::default()
     }
 
     /// Add a new collector to our `MultiCollector`.
@@ -183,21 +177,24 @@ impl<'a> Collector for MultiCollector<'a> {
         &self,
         segment_local_id: SegmentLocalId,
         segment: &SegmentReader,
-    ) -> Result<MultiCollectorChild> {
+    ) -> crate::Result<MultiCollectorChild> {
         let children = self
             .collector_wrappers
             .iter()
             .map(|collector_wrapper| collector_wrapper.for_segment(segment_local_id, segment))
-            .collect::<Result<Vec<_>>>()?;
+            .collect::<crate::Result<Vec<_>>>()?;
         Ok(MultiCollectorChild { children })
     }
 
     fn requires_scoring(&self) -> bool {
-        self.collector_wrappers.iter().any(|c| c.requires_scoring())
+        self.collector_wrappers
+            .iter()
+            .map(Deref::deref)
+            .any(Collector::requires_scoring)
     }
 
-    fn merge_fruits(&self, segments_multifruits: Vec<MultiFruit>) -> Result<MultiFruit> {
-        let mut segment_fruits_list: Vec<Vec<Box<Fruit>>> = (0..self.collector_wrappers.len())
+    fn merge_fruits(&self, segments_multifruits: Vec<MultiFruit>) -> crate::Result<MultiFruit> {
+        let mut segment_fruits_list: Vec<Vec<Box<dyn Fruit>>> = (0..self.collector_wrappers.len())
             .map(|_| Vec::with_capacity(segments_multifruits.len()))
             .collect::<Vec<_>>();
         for segment_multifruit in segments_multifruits {
@@ -213,13 +210,14 @@ impl<'a> Collector for MultiCollector<'a> {
             .zip(segment_fruits_list)
             .map(|(child_collector, segment_fruits)| {
                 Ok(Some(child_collector.merge_fruits(segment_fruits)?))
-            }).collect::<Result<_>>()?;
+            })
+            .collect::<crate::Result<_>>()?;
         Ok(MultiFruit { sub_fruits })
     }
 }
 
 pub struct MultiCollectorChild {
-    children: Vec<Box<BoxableSegmentCollector>>,
+    children: Vec<Box<dyn BoxableSegmentCollector>>,
 }
 
 impl SegmentCollector for MultiCollectorChild {
@@ -246,12 +244,12 @@ impl SegmentCollector for MultiCollectorChild {
 mod tests {
 
     use super::*;
-    use collector::{Count, TopDocs};
-    use query::TermQuery;
-    use schema::IndexRecordOption;
-    use schema::{Schema, TEXT};
-    use Index;
-    use Term;
+    use crate::collector::{Count, TopDocs};
+    use crate::query::TermQuery;
+    use crate::schema::IndexRecordOption;
+    use crate::schema::{Schema, TEXT};
+    use crate::Index;
+    use crate::Term;
 
     #[test]
     fn test_multi_collector() {
@@ -271,8 +269,7 @@ mod tests {
             index_writer.add_document(doc!(text=>"abc"));
             index_writer.commit().unwrap();
         }
-        index.load_searchers().unwrap();
-        let searcher = index.searcher();
+        let searcher = index.reader().unwrap().searcher();
         let term = Term::from_field_text(text, "abc");
         let query = TermQuery::new(term, IndexRecordOption::Basic);
 

src/collector/tests.rs

@@ -1,12 +1,20 @@
 use super::*;
-use core::SegmentReader;
-use fastfield::BytesFastFieldReader;
-use fastfield::FastFieldReader;
-use schema::Field;
-use DocAddress;
-use DocId;
-use Score;
-use SegmentLocalId;
+use crate::core::SegmentReader;
+use crate::fastfield::BytesFastFieldReader;
+use crate::fastfield::FastFieldReader;
+use crate::schema::Field;
+use crate::DocAddress;
+use crate::DocId;
+use crate::Score;
+use crate::SegmentLocalId;
+
+pub const TEST_COLLECTOR_WITH_SCORE: TestCollector = TestCollector {
+    compute_score: true,
+};
+
+pub const TEST_COLLECTOR_WITHOUT_SCORE: TestCollector = TestCollector {
+    compute_score: true,
+};
 
 /// Stores all of the doc ids.
 /// This collector is only used for tests.
@@ -14,7 +22,9 @@ use SegmentLocalId;
 ///
 /// actise, as it does not store
 /// the segment ordinals
-pub struct TestCollector;
+pub struct TestCollector {
+    pub compute_score: bool,
+}
 
 pub struct TestSegmentCollector {
     segment_id: SegmentLocalId,
@@ -32,7 +42,6 @@ impl TestFruit {
     pub fn docs(&self) -> &[DocAddress] {
         &self.docs[..]
     }
-
     pub fn scores(&self) -> &[Score] {
         &self.scores[..]
     }
@@ -46,7 +55,7 @@ impl Collector for TestCollector {
         &self,
         segment_id: SegmentLocalId,
         _reader: &SegmentReader,
-    ) -> Result<TestSegmentCollector> {
+    ) -> crate::Result<TestSegmentCollector> {
         Ok(TestSegmentCollector {
             segment_id,
             fruit: TestFruit::default(),
@@ -54,10 +63,10 @@ impl Collector for TestCollector {
     }
 
     fn requires_scoring(&self) -> bool {
-        true
+        self.compute_score
     }
 
-    fn merge_fruits(&self, mut children: Vec<TestFruit>) -> Result<TestFruit> {
+    fn merge_fruits(&self, mut children: Vec<TestFruit>) -> crate::Result<TestFruit> {
         children.sort_by_key(|fruit| {
             if fruit.docs().is_empty() {
                 0
@@ -114,11 +123,15 @@ impl Collector for FastFieldTestCollector {
     fn for_segment(
         &self,
         _: SegmentLocalId,
-        reader: &SegmentReader,
-    ) -> Result<FastFieldSegmentCollector> {
+        segment_reader: &SegmentReader,
+    ) -> crate::Result<FastFieldSegmentCollector> {
+        let reader = segment_reader
+            .fast_fields()
+            .u64(self.field)
+            .expect("Requested field is not a fast field.");
         Ok(FastFieldSegmentCollector {
             vals: Vec::new(),
-            reader: reader.fast_field_reader(self.field)?,
+            reader,
         })
     }
 
@@ -126,7 +139,7 @@ impl Collector for FastFieldTestCollector {
         false
     }
 
-    fn merge_fruits(&self, children: Vec<Vec<u64>>) -> Result<Vec<u64>> {
+    fn merge_fruits(&self, children: Vec<Vec<u64>>) -> crate::Result<Vec<u64>> {
         Ok(children.into_iter().flat_map(|v| v.into_iter()).collect())
     }
 }
@@ -170,11 +183,14 @@ impl Collector for BytesFastFieldTestCollector {
     fn for_segment(
         &self,
         _segment_local_id: u32,
-        segment: &SegmentReader,
-    ) -> Result<BytesFastFieldSegmentCollector> {
+        segment_reader: &SegmentReader,
+    ) -> crate::Result<BytesFastFieldSegmentCollector> {
         Ok(BytesFastFieldSegmentCollector {
             vals: Vec::new(),
-            reader: segment.bytes_fast_field_reader(self.field)?,
+            reader: segment_reader
+                .fast_fields()
+                .bytes(self.field)
+                .expect("Field is not a bytes fast field."),
         })
     }
 
@@ -182,7 +198,7 @@ impl Collector for BytesFastFieldTestCollector {
         false
     }
 
-    fn merge_fruits(&self, children: Vec<Vec<u8>>) -> Result<Vec<u8>> {
+    fn merge_fruits(&self, children: Vec<Vec<u8>>) -> crate::Result<Vec<u8>> {
         Ok(children.into_iter().flat_map(|c| c.into_iter()).collect())
     }
 }
@@ -191,7 +207,7 @@ impl SegmentCollector for BytesFastFieldSegmentCollector {
     type Fruit = Vec<u8>;
 
     fn collect(&mut self, doc: u32, _score: f32) {
-        let data = self.reader.get_val(doc);
+        let data = self.reader.get_bytes(doc);
         self.vals.extend(data);
     }
 

src/collector/top_collector.rs

@@ -1,52 +1,63 @@
+use crate::DocAddress;
+use crate::DocId;
+use crate::SegmentLocalId;
+use crate::SegmentReader;
 use serde::export::PhantomData;
 use std::cmp::Ordering;
 use std::collections::BinaryHeap;
-use DocAddress;
-use DocId;
-use Result;
-use SegmentLocalId;
-use SegmentReader;
 
 /// Contains a feature (field, score, etc.) of a document along with the document address.
 ///
 /// It has a custom implementation of `PartialOrd` that reverses the order. This is because the
 /// default Rust heap is a max heap, whereas a min heap is needed.
 ///
+/// Additionally, it guarantees stable sorting: in case of a tie on the feature, the document
+/// address is used.
+///
 /// WARNING: equality is not what you would expect here.
 /// Two elements are equal if their feature is equal, and regardless of whether `doc`
 /// is equal. This should be perfectly fine for this usage, but let's make sure this
 /// struct is never public.
-struct ComparableDoc<T, D> {
-    feature: T,
-    doc: D,
+pub(crate) struct ComparableDoc<T, D> {
+    pub feature: T,
+    pub doc: D,
 }
 
-impl<T: PartialOrd, D> PartialOrd for ComparableDoc<T, D> {
+impl<T: PartialOrd, D: PartialOrd> PartialOrd for ComparableDoc<T, D> {
     fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
         Some(self.cmp(other))
     }
 }
 
-impl<T: PartialOrd, D> Ord for ComparableDoc<T, D> {
+impl<T: PartialOrd, D: PartialOrd> Ord for ComparableDoc<T, D> {
     #[inline]
     fn cmp(&self, other: &Self) -> Ordering {
-        other
+        // Reversed to make BinaryHeap work as a min-heap
+        let by_feature = other
             .feature
             .partial_cmp(&self.feature)
-            .unwrap_or_else(|| Ordering::Equal)
+            .unwrap_or(Ordering::Equal);
+
+        let lazy_by_doc_address = || self.doc.partial_cmp(&other.doc).unwrap_or(Ordering::Equal);
+
+        // In case of a tie on the feature, we sort by ascending
+        // `DocAddress` in order to ensure a stable sorting of the
+        // documents.
+        by_feature.then_with(lazy_by_doc_address)
     }
 }
 
-impl<T: PartialOrd, D> PartialEq for ComparableDoc<T, D> {
+impl<T: PartialOrd, D: PartialOrd> PartialEq for ComparableDoc<T, D> {
     fn eq(&self, other: &Self) -> bool {
         self.cmp(other) == Ordering::Equal
     }
 }
 
-impl<T: PartialOrd, D> Eq for ComparableDoc<T, D> {}
+impl<T: PartialOrd, D: PartialOrd> Eq for ComparableDoc<T, D> {}
 
 pub(crate) struct TopCollector<T> {
-    limit: usize,
+    pub limit: usize,
+    pub offset: usize,
     _marker: PhantomData<T>,
 }
 
@@ -62,33 +73,37 @@ where
         if limit < 1 {
             panic!("Limit must be strictly greater than 0.");
         }
-        TopCollector {
+        Self {
             limit,
+            offset: 0,
             _marker: PhantomData,
         }
     }
 
-    pub fn limit(&self) -> usize {
-        self.limit
+    /// Skip the first "offset" documents when collecting.
+    ///
+    /// This is equivalent to `OFFSET` in MySQL or PostgreSQL and `start` in
+    /// Lucene's TopDocsCollector.
+    pub fn and_offset(mut self, offset: usize) -> TopCollector<T> {
+        self.offset = offset;
+        self
     }
 
     pub fn merge_fruits(
         &self,
         children: Vec<Vec<(T, DocAddress)>>,
-    ) -> Result<Vec<(T, DocAddress)>> {
+    ) -> crate::Result<Vec<(T, DocAddress)>> {
         if self.limit == 0 {
             return Ok(Vec::new());
         }
         let mut top_collector = BinaryHeap::new();
         for child_fruit in children {
             for (feature, doc) in child_fruit {
-                if top_collector.len() < self.limit {
+                if top_collector.len() < (self.limit + self.offset) {
                     top_collector.push(ComparableDoc { feature, doc });
-                } else {
-                    if let Some(mut head) = top_collector.peek_mut() {
-                        if head.feature < feature {
-                            *head = ComparableDoc { feature, doc };
-                        }
+                } else if let Some(mut head) = top_collector.peek_mut() {
+                    if head.feature < feature {
+                        *head = ComparableDoc { feature, doc };
                     }
                 }
             }
@@ -96,16 +111,33 @@ where
         Ok(top_collector
             .into_sorted_vec()
             .into_iter()
+            .skip(self.offset)
             .map(|cdoc| (cdoc.feature, cdoc.doc))
             .collect())
     }
 
-    pub(crate) fn for_segment(
+    pub(crate) fn for_segment<F: PartialOrd>(
         &self,
         segment_id: SegmentLocalId,
         _: &SegmentReader,
-    ) -> Result<TopSegmentCollector<T>> {
-        Ok(TopSegmentCollector::new(segment_id, self.limit))
+    ) -> crate::Result<TopSegmentCollector<F>> {
+        Ok(TopSegmentCollector::new(
+            segment_id,
+            self.limit + self.offset,
+        ))
+    }
+
+    /// Create a new TopCollector with the same limit and offset.
+    ///
+    /// Ideally we would use Into but the blanket implementation seems to cause the Scorer traits
+    /// to fail.
+    #[doc(hidden)]
+    pub(crate) fn into_tscore<TScore: PartialOrd + Clone>(self) -> TopCollector<TScore> {
+        TopCollector {
+            limit: self.limit,
+            offset: self.offset,
+            _marker: PhantomData,
+        }
     }
 }
 
@@ -142,7 +174,8 @@ impl<T: PartialOrd + Clone> TopSegmentCollector<T> {
                     comparable_doc.feature,
                     DocAddress(segment_id, comparable_doc.doc),
                 )
-            }).collect()
+            })
+            .collect()
     }
 
     /// Return true iff at least K documents have gone through
@@ -179,8 +212,7 @@ impl<T: PartialOrd + Clone> TopSegmentCollector<T> {
 #[cfg(test)]
 mod tests {
     use super::{TopCollector, TopSegmentCollector};
-    use DocAddress;
-    use Score;
+    use crate::DocAddress;
 
     #[test]
     fn test_top_collector_not_at_capacity() {
@@ -218,8 +250,134 @@ mod tests {
     }
 
     #[test]
-    #[should_panic]
-    fn test_top_0() {
-        let _collector: TopCollector<Score> = TopCollector::with_limit(0);
+    fn test_top_segment_collector_stable_ordering_for_equal_feature() {
+        // given that the documents are collected in ascending doc id order,
+        // when harvesting we have to guarantee stable sorting in case of a tie
+        // on the score
+        let doc_ids_collection = [4, 5, 6];
+        let score = 3.14;
+
+        let mut top_collector_limit_2 = TopSegmentCollector::new(0, 2);
+        for id in &doc_ids_collection {
+            top_collector_limit_2.collect(*id, score);
+        }
+
+        let mut top_collector_limit_3 = TopSegmentCollector::new(0, 3);
+        for id in &doc_ids_collection {
+            top_collector_limit_3.collect(*id, score);
+        }
+
+        assert_eq!(
+            top_collector_limit_2.harvest(),
+            top_collector_limit_3.harvest()[..2].to_vec(),
+        );
+    }
+
+    #[test]
+    fn test_top_collector_with_limit_and_offset() {
+        let collector = TopCollector::with_limit(2).and_offset(1);
+
+        let results = collector
+            .merge_fruits(vec![vec![
+                (0.9, DocAddress(0, 1)),
+                (0.8, DocAddress(0, 2)),
+                (0.7, DocAddress(0, 3)),
+                (0.6, DocAddress(0, 4)),
+                (0.5, DocAddress(0, 5)),
+            ]])
+            .unwrap();
+
+        assert_eq!(
+            results,
+            vec![(0.8, DocAddress(0, 2)), (0.7, DocAddress(0, 3)),]
+        );
+    }
+
+    #[test]
+    fn test_top_collector_with_limit_larger_than_set_and_offset() {
+        let collector = TopCollector::with_limit(2).and_offset(1);
+
+        let results = collector
+            .merge_fruits(vec![vec![(0.9, DocAddress(0, 1)), (0.8, DocAddress(0, 2))]])
+            .unwrap();
+
+        assert_eq!(results, vec![(0.8, DocAddress(0, 2)),]);
+    }
+
+    #[test]
+    fn test_top_collector_with_limit_and_offset_larger_than_set() {
+        let collector = TopCollector::with_limit(2).and_offset(20);
+
+        let results = collector
+            .merge_fruits(vec![vec![(0.9, DocAddress(0, 1)), (0.8, DocAddress(0, 2))]])
+            .unwrap();
+
+        assert_eq!(results, vec![]);
+    }
+}
+
+#[cfg(all(test, feature = "unstable"))]
+mod bench {
+    use super::TopSegmentCollector;
+    use test::Bencher;
+
+    #[bench]
+    fn bench_top_segment_collector_collect_not_at_capacity(b: &mut Bencher) {
+        let mut top_collector = TopSegmentCollector::new(0, 400);
+
+        b.iter(|| {
+            for i in 0..100 {
+                top_collector.collect(i, 0.8);
+            }
+        });
+    }
+
+    #[bench]
+    fn bench_top_segment_collector_collect_at_capacity(b: &mut Bencher) {
+        let mut top_collector = TopSegmentCollector::new(0, 100);
+
+        for i in 0..100 {
+            top_collector.collect(i, 0.8);
+        }
+
+        b.iter(|| {
+            for i in 0..100 {
+                top_collector.collect(i, 0.8);
+            }
+        });
+    }
+
+    #[bench]
+    fn bench_top_segment_collector_collect_and_harvest_many_ties(b: &mut Bencher) {
+        b.iter(|| {
+            let mut top_collector = TopSegmentCollector::new(0, 100);
+
+            for i in 0..100 {
+                top_collector.collect(i, 0.8);
+            }
+
+            // it would be nice to be able to do the setup N times but still
+            // measure only harvest(). We can't since harvest() consumes
+            // the top_collector.
+            top_collector.harvest()
+        });
+    }
+
+    #[bench]
+    fn bench_top_segment_collector_collect_and_harvest_no_tie(b: &mut Bencher) {
+        b.iter(|| {
+            let mut top_collector = TopSegmentCollector::new(0, 100);
+            let mut score = 1.0;
+
+            for i in 0..100 {
+                score += 1.0;
+                top_collector.collect(i, score);
+            }
+
+            // it would be nice to be able to do the setup N times but still
+            // measure only harvest(). We can't since harvest() consumes
+            // the top_collector.
+            top_collector.harvest()
+        });
     }
 }

src/collector/top_field_collector.rs

@@ -1,250 +0,0 @@
-use super::Collector;
-use collector::top_collector::TopCollector;
-use collector::top_collector::TopSegmentCollector;
-use collector::SegmentCollector;
-use fastfield::FastFieldReader;
-use fastfield::FastValue;
-use schema::Field;
-use DocAddress;
-use Result;
-use SegmentLocalId;
-use SegmentReader;
-
-/// The Top Field Collector keeps track of the K documents
-/// sorted by a fast field in the index
-///
-/// The implementation is based on a `BinaryHeap`.
-/// The theorical complexity for collecting the top `K` out of `n` documents
-/// is `O(n log K)`.
-///
-/// ```rust
-/// #[macro_use]
-/// extern crate tantivy;
-/// # use tantivy::schema::{Schema, Field, FAST, TEXT};
-/// # use tantivy::{Index, Result, DocAddress};
-/// # use tantivy::query::{Query, QueryParser};
-/// use tantivy::collector::TopDocs;
-///
-/// # fn main() {
-/// #   let mut schema_builder = Schema::builder();
-/// #   let title = schema_builder.add_text_field("title", TEXT);
-/// #   let rating = schema_builder.add_u64_field("rating", FAST);
-/// #   let schema = schema_builder.build();
-/// #   let index = Index::create_in_ram(schema);
-/// #   let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
-/// #   index_writer.add_document(doc!(
-/// #       title => "The Name of the Wind",
-/// #       rating => 92u64,
-/// #   ));
-/// #   index_writer.add_document(doc!(title => "The Diary of Muadib", rating => 97u64));
-/// #   index_writer.add_document(doc!(title => "A Dairy Cow", rating => 63u64));
-/// #   index_writer.add_document(doc!(title => "The Diary of a Young Girl", rating => 80u64));
-/// #   index_writer.commit().unwrap();
-/// #   index.load_searchers().unwrap();
-///	#   let query = QueryParser::for_index(&index, vec![title]).parse_query("diary").unwrap();
-/// #   let top_docs = docs_sorted_by_rating(&index, &query, rating).unwrap();
-/// #   assert_eq!(top_docs,
-/// #            vec![(97u64, DocAddress(0u32, 1)),
-/// #                 (80u64, DocAddress(0u32, 3))]);
-/// # }
-/// #
-/// /// Searches the document matching the given query, and
-/// /// collects the top 10 documents, order by the `field`
-/// /// given in argument.
-/// ///
-/// /// `field` is required to be a FAST field.
-/// fn docs_sorted_by_rating(index: &Index, query: &Query, sort_by_field: Field)
-///     -> Result<Vec<(u64, DocAddress)>> {
-///
-///     // This is where we build our collector!
-///     let top_docs_by_rating = TopDocs::with_limit(2).order_by_field(sort_by_field);
-///
-///     // ... and here is our documents. Not this is a simple vec.
-///     // The `u64` in the pair is the value of our fast field for each documents.
-///     index.searcher()
-///          .search(query, &top_docs_by_rating)
-/// }
-/// ```
-pub struct TopDocsByField<T> {
-    collector: TopCollector<T>,
-    field: Field,
-}
-
-impl<T: FastValue + PartialOrd + Clone> TopDocsByField<T> {
-    /// Creates a top field collector, with a number of documents equal to "limit".
-    ///
-    /// The given field name must be a fast field, otherwise the collector have an error while
-    /// collecting results.
-    ///
-    /// # Panics
-    /// The method panics if limit is 0
-    pub(crate) fn new(field: Field, limit: usize) -> TopDocsByField<T> {
-        TopDocsByField {
-            collector: TopCollector::with_limit(limit),
-            field,
-        }
-    }
-}
-
-impl<T: FastValue + PartialOrd + Send + Sync + 'static> Collector for TopDocsByField<T> {
-    type Fruit = Vec<(T, DocAddress)>;
-
-    type Child = TopFieldSegmentCollector<T>;
-
-    fn for_segment(
-        &self,
-        segment_local_id: SegmentLocalId,
-        reader: &SegmentReader,
-    ) -> Result<TopFieldSegmentCollector<T>> {
-        let collector = self.collector.for_segment(segment_local_id, reader)?;
-        let reader = reader.fast_field_reader(self.field)?;
-        Ok(TopFieldSegmentCollector { collector, reader })
-    }
-
-    fn requires_scoring(&self) -> bool {
-        false
-    }
-
-    fn merge_fruits(
-        &self,
-        segment_fruits: Vec<Vec<(T, DocAddress)>>,
-    ) -> Result<Vec<(T, DocAddress)>> {
-        self.collector.merge_fruits(segment_fruits)
-    }
-}
-
-pub struct TopFieldSegmentCollector<T: FastValue + PartialOrd> {
-    collector: TopSegmentCollector<T>,
-    reader: FastFieldReader<T>,
-}
-
-impl<T: FastValue + PartialOrd + Send + Sync + 'static> SegmentCollector
-    for TopFieldSegmentCollector<T>
-{
-    type Fruit = Vec<(T, DocAddress)>;
-
-    fn collect(&mut self, doc: u32, _score: f32) {
-        let field_value = self.reader.get(doc);
-        self.collector.collect(doc, field_value);
-    }
-
-    fn harvest(self) -> Vec<(T, DocAddress)> {
-        self.collector.harvest()
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::TopDocsByField;
-    use collector::Collector;
-    use collector::TopDocs;
-    use query::Query;
-    use query::QueryParser;
-    use schema::Field;
-    use schema::IntOptions;
-    use schema::{Schema, FAST, TEXT};
-    use DocAddress;
-    use Index;
-    use IndexWriter;
-    use TantivyError;
-
-    const TITLE: &str = "title";
-    const SIZE: &str = "size";
-
-    #[test]
-    fn test_top_collector_not_at_capacity() {
-        let mut schema_builder = Schema::builder();
-        let title = schema_builder.add_text_field(TITLE, TEXT);
-        let size = schema_builder.add_u64_field(SIZE, FAST);
-        let schema = schema_builder.build();
-        let (index, query) = index("beer", title, schema, |index_writer| {
-            index_writer.add_document(doc!(
-                title => "bottle of beer",
-                size => 12u64,
-            ));
-            index_writer.add_document(doc!(
-                title => "growler of beer",
-                size => 64u64,
-            ));
-            index_writer.add_document(doc!(
-                title => "pint of beer",
-                size => 16u64,
-            ));
-        });
-        let searcher = index.searcher();
-
-        let top_collector = TopDocs::with_limit(4).order_by_field(size);
-        let top_docs: Vec<(u64, DocAddress)> = searcher.search(&query, &top_collector).unwrap();
-        assert_eq!(
-            top_docs,
-            vec![
-                (64, DocAddress(0, 1)),
-                (16, DocAddress(0, 2)),
-                (12, DocAddress(0, 0))
-            ]
-        );
-    }
-
-    #[test]
-    #[should_panic]
-    fn test_field_does_not_exist() {
-        let mut schema_builder = Schema::builder();
-        let title = schema_builder.add_text_field(TITLE, TEXT);
-        let size = schema_builder.add_u64_field(SIZE, FAST);
-        let schema = schema_builder.build();
-        let (index, _) = index("beer", title, schema, |index_writer| {
-            index_writer.add_document(doc!(
-                title => "bottle of beer",
-                size => 12u64,
-            ));
-        });
-        let searcher = index.searcher();
-        let top_collector: TopDocsByField<u64> = TopDocs::with_limit(4).order_by_field(Field(2));
-        let segment_reader = searcher.segment_reader(0u32);
-        top_collector
-            .for_segment(0, segment_reader)
-            .expect("should panic");
-    }
-
-    #[test]
-    fn test_field_not_fast_field() {
-        let mut schema_builder = Schema::builder();
-        let title = schema_builder.add_text_field(TITLE, TEXT);
-        let size = schema_builder.add_u64_field(SIZE, IntOptions::default());
-        let schema = schema_builder.build();
-        let (index, _) = index("beer", title, schema, |index_writer| {
-            index_writer.add_document(doc!(
-                title => "bottle of beer",
-                size => 12u64,
-            ));
-        });
-        let searcher = index.searcher();
-        let segment = searcher.segment_reader(0);
-        let top_collector: TopDocsByField<u64> = TopDocs::with_limit(4).order_by_field(size);
-        assert_matches!(
-            top_collector
-                .for_segment(0, segment)
-                .map(|_| ())
-                .unwrap_err(),
-            TantivyError::FastFieldError(_)
-        );
-    }
-
-    fn index(
-        query: &str,
-        query_field: Field,
-        schema: Schema,
-        mut doc_adder: impl FnMut(&mut IndexWriter) -> (),
-    ) -> (Index, Box<Query>) {
-        let index = Index::create_in_ram(schema);
-
-        let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
-        doc_adder(&mut index_writer);
-        index_writer.commit().unwrap();
-        index.load_searchers().unwrap();
-
-        let query_parser = QueryParser::for_index(&index, vec![query_field]);
-        let query = query_parser.parse_query(query).unwrap();
-        (index, query)
-    }
-}

src/collector/top_score_collector.rs

@@ -1,71 +1,103 @@
 use super::Collector;
-use collector::top_collector::TopCollector;
-use collector::top_collector::TopSegmentCollector;
-use collector::SegmentCollector;
-use collector::TopDocsByField;
-use fastfield::FastValue;
-use schema::Field;
-use DocAddress;
-use DocId;
-use Result;
-use Score;
-use SegmentLocalId;
-use SegmentReader;
+use crate::collector::custom_score_top_collector::CustomScoreTopCollector;
+use crate::collector::top_collector::TopSegmentCollector;
+use crate::collector::top_collector::{ComparableDoc, TopCollector};
+use crate::collector::tweak_score_top_collector::TweakedScoreTopCollector;
+use crate::collector::{
+    CustomScorer, CustomSegmentScorer, ScoreSegmentTweaker, ScoreTweaker, SegmentCollector,
+};
+use crate::fastfield::FastFieldReader;
+use crate::query::Weight;
+use crate::schema::Field;
+use crate::DocAddress;
+use crate::DocId;
+use crate::Score;
+use crate::SegmentLocalId;
+use crate::SegmentReader;
+use std::collections::BinaryHeap;
+use std::fmt;
 
-/// The Top Score Collector keeps track of the K documents
+/// The `TopDocs` collector keeps track of the top `K` documents
 /// sorted by their score.
 ///
 /// The implementation is based on a `BinaryHeap`.
 /// The theorical complexity for collecting the top `K` out of `n` documents
 /// is `O(n log K)`.
 ///
+/// This collector guarantees a stable sorting in case of a tie on the
+/// document score. As such, it is suitable to implement pagination.
+///
 /// ```rust
-/// #[macro_use]
-/// extern crate tantivy;
-/// use tantivy::DocAddress;
-/// use tantivy::schema::{Schema, TEXT};
-/// use tantivy::{Index, Result};
 /// use tantivy::collector::TopDocs;
 /// use tantivy::query::QueryParser;
+/// use tantivy::schema::{Schema, TEXT};
+/// use tantivy::{doc, DocAddress, Index};
 ///
-/// # fn main() { example().unwrap(); }
-/// fn example() -> Result<()> {
-///     let mut schema_builder = Schema::builder();
-///     let title = schema_builder.add_text_field("title", TEXT);
-///     let schema = schema_builder.build();
-///     let index = Index::create_in_ram(schema);
-///     {
-///         let mut index_writer = index.writer_with_num_threads(1, 3_000_000)?;
-///         index_writer.add_document(doc!(
-///             title => "The Name of the Wind",
-///         ));
-///         index_writer.add_document(doc!(
-///             title => "The Diary of Muadib",
-///         ));
-///         index_writer.add_document(doc!(
-///             title => "A Dairy Cow",
-///         ));
-///         index_writer.add_document(doc!(
-///             title => "The Diary of a Young Girl",
-///         ));
-///         index_writer.commit().unwrap();
-///     }
+/// let mut schema_builder = Schema::builder();
+/// let title = schema_builder.add_text_field("title", TEXT);
+/// let schema = schema_builder.build();
+/// let index = Index::create_in_ram(schema);
 ///
-///     index.load_searchers()?;
-///     let searcher = index.searcher();
+/// let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+/// index_writer.add_document(doc!(title => "The Name of the Wind"));
+/// index_writer.add_document(doc!(title => "The Diary of Muadib"));
+/// index_writer.add_document(doc!(title => "A Dairy Cow"));
+/// index_writer.add_document(doc!(title => "The Diary of a Young Girl"));
+/// assert!(index_writer.commit().is_ok());
 ///
-///     let query_parser = QueryParser::for_index(&index, vec![title]);
-///     let query = query_parser.parse_query("diary")?;
-///     let top_docs = searcher.search(&query, &TopDocs::with_limit(2))?;
+/// let reader = index.reader().unwrap();
+/// let searcher = reader.searcher();
 ///
-///     assert_eq!(&top_docs[0], &(0.7261542, DocAddress(0, 1)));
-///     assert_eq!(&top_docs[1], &(0.6099695, DocAddress(0, 3)));
+/// let query_parser = QueryParser::for_index(&index, vec![title]);
+/// let query = query_parser.parse_query("diary").unwrap();
+/// let top_docs = searcher.search(&query, &TopDocs::with_limit(2)).unwrap();
 ///
-///     Ok(())
-/// }
+/// assert_eq!(&top_docs[0], &(0.7261542, DocAddress(0, 1)));
+/// assert_eq!(&top_docs[1], &(0.6099695, DocAddress(0, 3)));
 /// ```
 pub struct TopDocs(TopCollector<Score>);
 
+impl fmt::Debug for TopDocs {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        write!(
+            f,
+            "TopDocs(limit={}, offset={})",
+            self.0.limit, self.0.offset
+        )
+    }
+}
+
+struct ScorerByFastFieldReader {
+    ff_reader: FastFieldReader<u64>,
+}
+
+impl CustomSegmentScorer<u64> for ScorerByFastFieldReader {
+    fn score(&mut self, doc: DocId) -> u64 {
+        self.ff_reader.get_u64(u64::from(doc))
+    }
+}
+
+struct ScorerByField {
+    field: Field,
+}
+
+impl CustomScorer<u64> for ScorerByField {
+    type Child = ScorerByFastFieldReader;
+
+    fn segment_scorer(&self, segment_reader: &SegmentReader) -> crate::Result<Self::Child> {
+        let ff_reader = segment_reader
+            .fast_fields()
+            .u64(self.field)
+            .ok_or_else(|| {
+                crate::TantivyError::SchemaError(format!(
+                    "Field requested ({:?}) is not a i64/u64 fast field.",
+                    self.field
+                ))
+            })?;
+        Ok(ScorerByFastFieldReader { ff_reader })
+    }
+}
+
 impl TopDocs {
     /// Creates a top score collector, with a number of documents equal to "limit".
     ///
@@ -75,15 +107,340 @@ impl TopDocs {
         TopDocs(TopCollector::with_limit(limit))
     }
 
+    /// Skip the first "offset" documents when collecting.
+    ///
+    /// This is equivalent to `OFFSET` in MySQL or PostgreSQL and `start` in
+    /// Lucene's TopDocsCollector.
+    ///
+    /// ```rust
+    /// use tantivy::collector::TopDocs;
+    /// use tantivy::query::QueryParser;
+    /// use tantivy::schema::{Schema, TEXT};
+    /// use tantivy::{doc, DocAddress, Index};
+    ///
+    /// let mut schema_builder = Schema::builder();
+    /// let title = schema_builder.add_text_field("title", TEXT);
+    /// let schema = schema_builder.build();
+    /// let index = Index::create_in_ram(schema);
+    ///
+    /// let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+    /// index_writer.add_document(doc!(title => "The Name of the Wind"));
+    /// index_writer.add_document(doc!(title => "The Diary of Muadib"));
+    /// index_writer.add_document(doc!(title => "A Dairy Cow"));
+    /// index_writer.add_document(doc!(title => "The Diary of a Young Girl"));
+    /// index_writer.add_document(doc!(title => "The Diary of Lena Mukhina"));
+    /// assert!(index_writer.commit().is_ok());
+    ///
+    /// let reader = index.reader().unwrap();
+    /// let searcher = reader.searcher();
+    ///
+    /// let query_parser = QueryParser::for_index(&index, vec![title]);
+    /// let query = query_parser.parse_query("diary").unwrap();
+    /// let top_docs = searcher.search(&query, &TopDocs::with_limit(2).and_offset(1)).unwrap();
+    ///
+    /// assert_eq!(top_docs.len(), 2);
+    /// assert_eq!(&top_docs[0], &(0.5204813, DocAddress(0, 4)));
+    /// assert_eq!(&top_docs[1], &(0.4793185, DocAddress(0, 3)));
+    /// ```
+    pub fn and_offset(self, offset: usize) -> TopDocs {
+        TopDocs(self.0.and_offset(offset))
+    }
+
     /// Set top-K to rank documents by a given fast field.
     ///
-    /// (By default, `TopDocs` collects the top-K documents sorted by
-    /// the similarity score.)
-    pub fn order_by_field<T: PartialOrd + FastValue + Clone>(
+    /// ```rust
+    /// # use tantivy::schema::{Schema, FAST, TEXT};
+    /// # use tantivy::{doc, Index, DocAddress};
+    /// # use tantivy::query::{Query, QueryParser};
+    /// use tantivy::Searcher;
+    /// use tantivy::collector::TopDocs;
+    /// use tantivy::schema::Field;
+    ///
+    /// # fn main() -> tantivy::Result<()> {
+    /// #   let mut schema_builder = Schema::builder();
+    /// #   let title = schema_builder.add_text_field("title", TEXT);
+    /// #   let rating = schema_builder.add_u64_field("rating", FAST);
+    /// #   let schema = schema_builder.build();
+    /// #  
+    /// #   let index = Index::create_in_ram(schema);
+    /// #   let mut index_writer = index.writer_with_num_threads(1, 3_000_000)?;
+    /// #   index_writer.add_document(doc!(title => "The Name of the Wind", rating => 92u64));
+    /// #   index_writer.add_document(doc!(title => "The Diary of Muadib", rating => 97u64));
+    /// #   index_writer.add_document(doc!(title => "A Dairy Cow", rating => 63u64));
+    /// #   index_writer.add_document(doc!(title => "The Diary of a Young Girl", rating => 80u64));
+    /// #   assert!(index_writer.commit().is_ok());
+    /// #   let reader = index.reader().unwrap();
+    /// #   let query = QueryParser::for_index(&index, vec![title]).parse_query("diary")?;
+    /// #   let top_docs = docs_sorted_by_rating(&reader.searcher(), &query, rating)?;
+    /// #   assert_eq!(top_docs,
+    /// #            vec![(97u64, DocAddress(0u32, 1)),
+    /// #                 (80u64, DocAddress(0u32, 3))]);
+    /// #   Ok(())
+    /// # }
+    ///
+    ///
+    /// /// Searches the document matching the given query, and
+    /// /// collects the top 10 documents, order by the u64-`field`
+    /// /// given in argument.
+    /// ///
+    /// /// `field` is required to be a FAST field.
+    /// fn docs_sorted_by_rating(searcher: &Searcher,
+    ///                          query: &dyn Query,
+    ///                          sort_by_field: Field)
+    ///     -> tantivy::Result<Vec<(u64, DocAddress)>> {
+    ///
+    ///     // This is where we build our topdocs collector
+    ///     //
+    ///     // Note the generics parameter that needs to match the
+    ///     // type `sort_by_field`.
+    ///     let top_docs_by_rating = TopDocs
+    ///                 ::with_limit(10)
+    ///                  .order_by_u64_field(sort_by_field);
+    ///     
+    ///     // ... and here are our documents. Note this is a simple vec.
+    ///     // The `u64` in the pair is the value of our fast field for
+    ///     // each documents.
+    ///     //
+    ///     // The vec is sorted decreasingly by `sort_by_field`, and has a
+    ///     // length of 10, or less if not enough documents matched the
+    ///     // query.
+    ///     let resulting_docs: Vec<(u64, DocAddress)> =
+    ///          searcher.search(query, &top_docs_by_rating)?;
+    ///     
+    ///     Ok(resulting_docs)
+    /// }
+    /// ```
+    ///
+    /// # Panics
+    ///
+    /// May panic if the field requested is not a fast field.
+    ///
+    pub fn order_by_u64_field(
         self,
         field: Field,
-    ) -> TopDocsByField<T> {
-        TopDocsByField::new(field, self.0.limit())
+    ) -> impl Collector<Fruit = Vec<(u64, DocAddress)>> {
+        self.custom_score(ScorerByField { field })
+    }
+
+    /// Ranks the documents using a custom score.
+    ///
+    /// This method offers a convenient way to tweak or replace
+    /// the documents score. As suggested by the prototype you can
+    /// manually define your own [`ScoreTweaker`](./trait.ScoreTweaker.html)
+    /// and pass it as an argument, but there is a much simpler way to
+    /// tweak your score: you can use a closure as in the following
+    /// example.
+    ///
+    /// # Example
+    ///
+    /// Typically, you will want to rely on one or more fast fields,
+    /// to alter the original relevance `Score`.
+    ///
+    /// For instance, in the following, we assume that we are implementing
+    /// an e-commerce website that has a fast field called `popularity`
+    /// that rates whether a product is typically often bought by users.
+    ///
+    /// In the following example will will tweak our ranking a bit by
+    /// boosting popular products a notch.
+    ///  
+    /// In more serious application, this tweaking could involved running a
+    /// learning-to-rank model over various features
+    ///
+    /// ```rust
+    /// # use tantivy::schema::{Schema, FAST, TEXT};
+    /// # use tantivy::{doc, Index, DocAddress, DocId, Score};
+    /// # use tantivy::query::QueryParser;
+    /// use tantivy::SegmentReader;
+    /// use tantivy::collector::TopDocs;
+    /// use tantivy::schema::Field;
+    ///
+    /// fn create_schema() -> Schema {
+    ///    let mut schema_builder = Schema::builder();
+    ///    schema_builder.add_text_field("product_name", TEXT);
+    ///    schema_builder.add_u64_field("popularity", FAST);
+    ///    schema_builder.build()
+    /// }
+    ///
+    /// fn create_index() -> tantivy::Result<Index> {
+    ///   let schema = create_schema();
+    ///   let index = Index::create_in_ram(schema);
+    ///   let mut index_writer = index.writer_with_num_threads(1, 3_000_000)?;
+    ///   let product_name = index.schema().get_field("product_name").unwrap();
+    ///   let popularity: Field = index.schema().get_field("popularity").unwrap();
+    ///   index_writer.add_document(doc!(product_name => "The Diary of Muadib", popularity => 1u64));
+    ///   index_writer.add_document(doc!(product_name => "A Dairy Cow", popularity => 10u64));
+    ///   index_writer.add_document(doc!(product_name => "The Diary of a Young Girl", popularity => 15u64));
+    ///   index_writer.commit()?;
+    ///   Ok(index)
+    /// }
+    ///
+    /// let index = create_index().unwrap();
+    /// let product_name = index.schema().get_field("product_name").unwrap();
+    /// let popularity: Field = index.schema().get_field("popularity").unwrap();
+    ///
+    /// let user_query_str = "diary";
+    /// let query_parser = QueryParser::for_index(&index, vec![product_name]);
+    /// let query = query_parser.parse_query(user_query_str).unwrap();
+    ///
+    /// // This is where we build our collector with our custom score.
+    /// let top_docs_by_custom_score = TopDocs
+    ///         ::with_limit(10)
+    ///          .tweak_score(move |segment_reader: &SegmentReader| {
+    ///             // The argument is a function that returns our scoring
+    ///             // function.
+    ///             //
+    ///             // The point of this "mother" function is to gather all
+    ///             // of the segment level information we need for scoring.
+    ///             // Typically, fast_fields.
+    ///             //
+    ///             // In our case, we will get a reader for the popularity
+    ///             // fast field.
+    ///             let popularity_reader =
+    ///                 segment_reader.fast_fields().u64(popularity).unwrap();
+    ///
+    ///             // We can now define our actual scoring function
+    ///             move |doc: DocId, original_score: Score| {
+    ///                 let popularity: u64 = popularity_reader.get(doc);
+    ///                 // Well.. For the sake of the example we use a simple logarithm
+    ///                 // function.
+    ///                 let popularity_boost_score = ((2u64 + popularity) as f32).log2();
+    ///                 popularity_boost_score * original_score
+    ///             }
+    ///           });
+    /// let reader = index.reader().unwrap();
+    /// let searcher = reader.searcher();
+    /// // ... and here are our documents. Note this is a simple vec.
+    /// // The `Score` in the pair is our tweaked score.
+    /// let resulting_docs: Vec<(Score, DocAddress)> =
+    ///      searcher.search(&query, &top_docs_by_custom_score).unwrap();
+    /// ```
+    ///
+    /// # See also
+    /// [custom_score(...)](#method.custom_score).
+    pub fn tweak_score<TScore, TScoreSegmentTweaker, TScoreTweaker>(
+        self,
+        score_tweaker: TScoreTweaker,
+    ) -> impl Collector<Fruit = Vec<(TScore, DocAddress)>>
+    where
+        TScore: 'static + Send + Sync + Clone + PartialOrd,
+        TScoreSegmentTweaker: ScoreSegmentTweaker<TScore> + 'static,
+        TScoreTweaker: ScoreTweaker<TScore, Child = TScoreSegmentTweaker>,
+    {
+        TweakedScoreTopCollector::new(score_tweaker, self.0.into_tscore())
+    }
+
+    /// Ranks the documents using a custom score.
+    ///
+    /// This method offers a convenient way to use a different score.
+    ///
+    /// As suggested by the prototype you can manually define your
+    /// own [`CustomScorer`](./trait.CustomScorer.html)
+    /// and pass it as an argument, but there is a much simpler way to
+    /// tweak your score: you can use a closure as in the following
+    /// example.
+    ///
+    /// # Limitation
+    ///
+    /// This method only makes it possible to compute the score from a given
+    /// `DocId`, fastfield values for the doc and any information you could
+    /// have precomputed beforehands. It does not make it possible for instance
+    /// to compute something like TfIdf as it does not have access to the list of query
+    /// terms present in the document, nor the term frequencies for the different terms.
+    ///
+    /// It can be used if your search engine relies on a learning-to-rank model for instance,
+    /// which does not rely on the term frequencies or positions as features.
+    ///
+    /// # Example
+    ///
+    /// ```rust
+    /// # use tantivy::schema::{Schema, FAST, TEXT};
+    /// # use tantivy::{doc, Index, DocAddress, DocId};
+    /// # use tantivy::query::QueryParser;
+    /// use tantivy::SegmentReader;
+    /// use tantivy::collector::TopDocs;
+    /// use tantivy::schema::Field;
+    ///
+    /// # fn create_schema() -> Schema {
+    /// #    let mut schema_builder = Schema::builder();
+    /// #    schema_builder.add_text_field("product_name", TEXT);
+    /// #    schema_builder.add_u64_field("popularity", FAST);
+    /// #    schema_builder.add_u64_field("boosted", FAST);
+    /// #    schema_builder.build()
+    /// # }
+    /// #
+    /// # fn main() -> tantivy::Result<()> {
+    /// #   let schema = create_schema();
+    /// #   let index = Index::create_in_ram(schema);
+    /// #   let mut index_writer = index.writer_with_num_threads(1, 3_000_000)?;
+    /// #   let product_name = index.schema().get_field("product_name").unwrap();
+    /// #   
+    /// let popularity: Field = index.schema().get_field("popularity").unwrap();
+    /// let boosted: Field = index.schema().get_field("boosted").unwrap();
+    /// #   index_writer.add_document(doc!(boosted=>1u64, product_name => "The Diary of Muadib", popularity => 1u64));
+    /// #   index_writer.add_document(doc!(boosted=>0u64, product_name => "A Dairy Cow", popularity => 10u64));
+    /// #   index_writer.add_document(doc!(boosted=>0u64, product_name => "The Diary of a Young Girl", popularity => 15u64));
+    /// #   index_writer.commit()?;
+    /// // ...
+    /// # let user_query = "diary";
+    /// # let query = QueryParser::for_index(&index, vec![product_name]).parse_query(user_query)?;
+    ///
+    /// // This is where we build our collector with our custom score.
+    /// let top_docs_by_custom_score = TopDocs
+    ///         ::with_limit(10)
+    ///          .custom_score(move |segment_reader: &SegmentReader| {
+    ///             // The argument is a function that returns our scoring
+    ///             // function.
+    ///             //
+    ///             // The point of this "mother" function is to gather all
+    ///             // of the segment level information we need for scoring.
+    ///             // Typically, fast_fields.
+    ///             //
+    ///             // In our case, we will get a reader for the popularity
+    ///             // fast field and a boosted field.
+    ///             //
+    ///             // We want to get boosted items score, and when we get
+    ///             // a tie, return the item with the highest popularity.
+    ///             //
+    ///             // Note that this is implemented by using a `(u64, u64)`
+    ///             // as a score.
+    ///             let popularity_reader =
+    ///                 segment_reader.fast_fields().u64(popularity).unwrap();
+    ///             let boosted_reader =
+    ///                 segment_reader.fast_fields().u64(boosted).unwrap();
+    ///    
+    ///             // We can now define our actual scoring function
+    ///             move |doc: DocId| {
+    ///                 let popularity: u64 = popularity_reader.get(doc);
+    ///                 let boosted: u64 = boosted_reader.get(doc);
+    ///                 // Score do not have to be `f64` in tantivy.
+    ///                 // Here we return a couple to get lexicographical order
+    ///                 // for free.
+    ///                 (boosted, popularity)
+    ///             }
+    ///           });
+    /// # let reader = index.reader()?;
+    /// # let searcher = reader.searcher();
+    /// // ... and here are our documents. Note this is a simple vec.
+    /// // The `Score` in the pair is our tweaked score.
+    /// let resulting_docs: Vec<((u64, u64), DocAddress)> =
+    ///      searcher.search(&*query, &top_docs_by_custom_score)?;
+    ///
+    /// # Ok(())
+    /// # }
+    /// ```
+    ///
+    /// # See also
+    /// [tweak_score(...)](#method.tweak_score).
+    pub fn custom_score<TScore, TCustomSegmentScorer, TCustomScorer>(
+        self,
+        custom_score: TCustomScorer,
+    ) -> impl Collector<Fruit = Vec<(TScore, DocAddress)>>
+    where
+        TScore: 'static + Send + Sync + Clone + PartialOrd,
+        TCustomSegmentScorer: CustomSegmentScorer<TScore> + 'static,
+        TCustomScorer: CustomScorer<TScore, Child = TCustomSegmentScorer>,
+    {
+        CustomScoreTopCollector::new(custom_score, self.0.into_tscore())
     }
 }
 
@@ -96,7 +453,7 @@ impl Collector for TopDocs {
         &self,
         segment_local_id: SegmentLocalId,
         reader: &SegmentReader,
-    ) -> Result<Self::Child> {
+    ) -> crate::Result<Self::Child> {
         let collector = self.0.for_segment(segment_local_id, reader)?;
         Ok(TopScoreSegmentCollector(collector))
     }
@@ -105,9 +462,70 @@ impl Collector for TopDocs {
         true
     }
 
-    fn merge_fruits(&self, child_fruits: Vec<Vec<(Score, DocAddress)>>) -> Result<Self::Fruit> {
+    fn merge_fruits(
+        &self,
+        child_fruits: Vec<Vec<(Score, DocAddress)>>,
+    ) -> crate::Result<Self::Fruit> {
         self.0.merge_fruits(child_fruits)
     }
+
+    fn collect_segment(
+        &self,
+        weight: &dyn Weight,
+        segment_ord: u32,
+        reader: &SegmentReader,
+    ) -> crate::Result<<Self::Child as SegmentCollector>::Fruit> {
+        let heap_len = self.0.limit + self.0.offset;
+        let mut heap: BinaryHeap<ComparableDoc<Score, DocId>> = BinaryHeap::with_capacity(heap_len);
+
+        if let Some(delete_bitset) = reader.delete_bitset() {
+            let mut threshold = f32::MIN;
+            weight.for_each_pruning(threshold, reader, &mut |doc, score| {
+                if delete_bitset.is_deleted(doc) {
+                    return threshold;
+                }
+                let heap_item = ComparableDoc {
+                    feature: score,
+                    doc,
+                };
+                if heap.len() < heap_len {
+                    heap.push(heap_item);
+                    if heap.len() == heap_len {
+                        threshold = heap.peek().map(|el| el.feature).unwrap_or(f32::MIN);
+                    }
+                    return threshold;
+                }
+                *heap.peek_mut().unwrap() = heap_item;
+                threshold = heap.peek().map(|el| el.feature).unwrap_or(std::f32::MIN);
+                threshold
+            })?;
+        } else {
+            weight.for_each_pruning(f32::MIN, reader, &mut |doc, score| {
+                let heap_item = ComparableDoc {
+                    feature: score,
+                    doc,
+                };
+                if heap.len() < heap_len {
+                    heap.push(heap_item);
+                    // TODO the threshold is suboptimal for heap.len == heap_len
+                    if heap.len() == heap_len {
+                        return heap.peek().map(|el| el.feature).unwrap_or(f32::MIN);
+                    } else {
+                        return f32::MIN;
+                    }
+                }
+                *heap.peek_mut().unwrap() = heap_item;
+                heap.peek().map(|el| el.feature).unwrap_or(std::f32::MIN)
+            })?;
+        }
+
+        let fruit = heap
+            .into_sorted_vec()
+            .into_iter()
+            .map(|cid| (cid.feature, DocAddress(segment_ord, cid.doc)))
+            .collect();
+        Ok(fruit)
+    }
 }
 
 /// Segment Collector associated to `TopDocs`.
@@ -117,7 +535,7 @@ impl SegmentCollector for TopScoreSegmentCollector {
     type Fruit = Vec<(Score, DocAddress)>;
 
     fn collect(&mut self, doc: DocId, score: Score) {
-        self.0.collect(doc, score)
+        self.0.collect(doc, score);
     }
 
     fn harvest(self) -> Vec<(Score, DocAddress)> {
@@ -128,12 +546,13 @@ impl SegmentCollector for TopScoreSegmentCollector {
 #[cfg(test)]
 mod tests {
     use super::TopDocs;
-    use query::QueryParser;
-    use schema::Schema;
-    use schema::TEXT;
-    use DocAddress;
-    use Index;
-    use Score;
+    use crate::collector::Collector;
+    use crate::query::{AllQuery, Query, QueryParser};
+    use crate::schema::{Field, Schema, FAST, STORED, TEXT};
+    use crate::Index;
+    use crate::IndexWriter;
+    use crate::Score;
+    use crate::{DocAddress, DocId, SegmentReader};
 
     fn make_index() -> Index {
         let mut schema_builder = Schema::builder();
@@ -142,13 +561,12 @@ mod tests {
         let index = Index::create_in_ram(schema);
         {
             // writing the segment
-            let mut index_writer = index.writer_with_num_threads(1, 40_000_000).unwrap();
+            let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
             index_writer.add_document(doc!(text_field=>"Hello happy tax payer."));
             index_writer.add_document(doc!(text_field=>"Droopy says hello happy tax payer"));
             index_writer.add_document(doc!(text_field=>"I like Droopy"));
             assert!(index_writer.commit().is_ok());
         }
-        index.load_searchers().unwrap();
         index
     }
 
@@ -159,6 +577,8 @@ mod tests {
         let query_parser = QueryParser::for_index(&index, vec![field]);
         let text_query = query_parser.parse_query("droopy tax").unwrap();
         let score_docs: Vec<(Score, DocAddress)> = index
+            .reader()
+            .unwrap()
             .searcher()
             .search(&text_query, &TopDocs::with_limit(4))
             .unwrap();
@@ -172,6 +592,21 @@ mod tests {
         );
     }
 
+    #[test]
+    fn test_top_collector_not_at_capacity_with_offset() {
+        let index = make_index();
+        let field = index.schema().get_field("text").unwrap();
+        let query_parser = QueryParser::for_index(&index, vec![field]);
+        let text_query = query_parser.parse_query("droopy tax").unwrap();
+        let score_docs: Vec<(Score, DocAddress)> = index
+            .reader()
+            .unwrap()
+            .searcher()
+            .search(&text_query, &TopDocs::with_limit(4).and_offset(2))
+            .unwrap();
+        assert_eq!(score_docs, vec![(0.48527452, DocAddress(0, 0))]);
+    }
+
     #[test]
     fn test_top_collector_at_capacity() {
         let index = make_index();
@@ -179,6 +614,8 @@ mod tests {
         let query_parser = QueryParser::for_index(&index, vec![field]);
         let text_query = query_parser.parse_query("droopy tax").unwrap();
         let score_docs: Vec<(Score, DocAddress)> = index
+            .reader()
+            .unwrap()
             .searcher()
             .search(&text_query, &TopDocs::with_limit(2))
             .unwrap();
@@ -191,10 +628,197 @@ mod tests {
         );
     }
 
+    #[test]
+    fn test_top_collector_at_capacity_with_offset() {
+        let index = make_index();
+        let field = index.schema().get_field("text").unwrap();
+        let query_parser = QueryParser::for_index(&index, vec![field]);
+        let text_query = query_parser.parse_query("droopy tax").unwrap();
+        let score_docs: Vec<(Score, DocAddress)> = index
+            .reader()
+            .unwrap()
+            .searcher()
+            .search(&text_query, &TopDocs::with_limit(2).and_offset(1))
+            .unwrap();
+        assert_eq!(
+            score_docs,
+            vec![
+                (0.5376842, DocAddress(0u32, 2)),
+                (0.48527452, DocAddress(0, 0))
+            ]
+        );
+    }
+
+    #[test]
+    fn test_top_collector_stable_sorting() {
+        let index = make_index();
+
+        // using AllQuery to get a constant score
+        let searcher = index.reader().unwrap().searcher();
+
+        let page_1 = searcher.search(&AllQuery, &TopDocs::with_limit(2)).unwrap();
+
+        let page_2 = searcher.search(&AllQuery, &TopDocs::with_limit(3)).unwrap();
+
+        // precondition for the test to be meaningful: we did get documents
+        // with the same score
+        assert!(page_1.iter().all(|result| result.0 == page_1[0].0));
+        assert!(page_2.iter().all(|result| result.0 == page_2[0].0));
+
+        // sanity check since we're relying on make_index()
+        assert_eq!(page_1.len(), 2);
+        assert_eq!(page_2.len(), 3);
+
+        assert_eq!(page_1, &page_2[..page_1.len()]);
+    }
+
     #[test]
     #[should_panic]
     fn test_top_0() {
         TopDocs::with_limit(0);
     }
 
+    const TITLE: &str = "title";
+    const SIZE: &str = "size";
+
+    #[test]
+    fn test_top_field_collector_not_at_capacity() {
+        let mut schema_builder = Schema::builder();
+        let title = schema_builder.add_text_field(TITLE, TEXT);
+        let size = schema_builder.add_u64_field(SIZE, FAST);
+        let schema = schema_builder.build();
+        let (index, query) = index("beer", title, schema, |index_writer| {
+            index_writer.add_document(doc!(
+                title => "bottle of beer",
+                size => 12u64,
+            ));
+            index_writer.add_document(doc!(
+                title => "growler of beer",
+                size => 64u64,
+            ));
+            index_writer.add_document(doc!(
+                title => "pint of beer",
+                size => 16u64,
+            ));
+        });
+        let searcher = index.reader().unwrap().searcher();
+
+        let top_collector = TopDocs::with_limit(4).order_by_u64_field(size);
+        let top_docs: Vec<(u64, DocAddress)> = searcher.search(&query, &top_collector).unwrap();
+        assert_eq!(
+            top_docs,
+            vec![
+                (64, DocAddress(0, 1)),
+                (16, DocAddress(0, 2)),
+                (12, DocAddress(0, 0))
+            ]
+        );
+    }
+
+    #[test]
+    #[should_panic]
+    fn test_field_does_not_exist() {
+        let mut schema_builder = Schema::builder();
+        let title = schema_builder.add_text_field(TITLE, TEXT);
+        let size = schema_builder.add_u64_field(SIZE, FAST);
+        let schema = schema_builder.build();
+        let (index, _) = index("beer", title, schema, |index_writer| {
+            index_writer.add_document(doc!(
+                title => "bottle of beer",
+                size => 12u64,
+            ));
+        });
+        let searcher = index.reader().unwrap().searcher();
+        let top_collector = TopDocs::with_limit(4).order_by_u64_field(Field::from_field_id(2));
+        let segment_reader = searcher.segment_reader(0u32);
+        top_collector
+            .for_segment(0, segment_reader)
+            .expect("should panic");
+    }
+
+    #[test]
+    fn test_field_not_fast_field() {
+        let mut schema_builder = Schema::builder();
+        let title = schema_builder.add_text_field(TITLE, TEXT);
+        let size = schema_builder.add_u64_field(SIZE, STORED);
+        let schema = schema_builder.build();
+        let (index, _) = index("beer", title, schema, |index_writer| {
+            index_writer.add_document(doc!(
+                title => "bottle of beer",
+                size => 12u64,
+            ));
+        });
+        let searcher = index.reader().unwrap().searcher();
+        let segment = searcher.segment_reader(0);
+        let top_collector = TopDocs::with_limit(4).order_by_u64_field(size);
+        let err = top_collector.for_segment(0, segment);
+        if let Err(crate::TantivyError::SchemaError(msg)) = err {
+            assert_eq!(
+                msg,
+                "Field requested (Field(1)) is not a i64/u64 fast field."
+            );
+        } else {
+            assert!(false);
+        }
+    }
+
+    #[test]
+    fn test_tweak_score_top_collector_with_offset() {
+        let index = make_index();
+        let field = index.schema().get_field("text").unwrap();
+        let query_parser = QueryParser::for_index(&index, vec![field]);
+        let text_query = query_parser.parse_query("droopy tax").unwrap();
+        let collector = TopDocs::with_limit(2).and_offset(1).tweak_score(
+            move |_segment_reader: &SegmentReader| move |doc: DocId, _original_score: Score| doc,
+        );
+        let score_docs: Vec<(u32, DocAddress)> = index
+            .reader()
+            .unwrap()
+            .searcher()
+            .search(&text_query, &collector)
+            .unwrap();
+
+        assert_eq!(
+            score_docs,
+            vec![(1, DocAddress(0, 1)), (0, DocAddress(0, 0)),]
+        );
+    }
+
+    #[test]
+    fn test_custom_score_top_collector_with_offset() {
+        let index = make_index();
+        let field = index.schema().get_field("text").unwrap();
+        let query_parser = QueryParser::for_index(&index, vec![field]);
+        let text_query = query_parser.parse_query("droopy tax").unwrap();
+        let collector = TopDocs::with_limit(2)
+            .and_offset(1)
+            .custom_score(move |_segment_reader: &SegmentReader| move |doc: DocId| doc);
+        let score_docs: Vec<(u32, DocAddress)> = index
+            .reader()
+            .unwrap()
+            .searcher()
+            .search(&text_query, &collector)
+            .unwrap();
+
+        assert_eq!(
+            score_docs,
+            vec![(1, DocAddress(0, 1)), (0, DocAddress(0, 0)),]
+        );
+    }
+
+    fn index(
+        query: &str,
+        query_field: Field,
+        schema: Schema,
+        mut doc_adder: impl FnMut(&mut IndexWriter) -> (),
+    ) -> (Index, Box<dyn Query>) {
+        let index = Index::create_in_ram(schema);
+
+        let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+        doc_adder(&mut index_writer);
+        index_writer.commit().unwrap();
+        let query_parser = QueryParser::for_index(&index, vec![query_field]);
+        let query = query_parser.parse_query(query).unwrap();
+        (index, query)
+    }
 }

src/collector/tweak_score_top_collector.rs

@@ -0,0 +1,129 @@
+use crate::collector::top_collector::{TopCollector, TopSegmentCollector};
+use crate::collector::{Collector, SegmentCollector};
+use crate::DocAddress;
+use crate::{DocId, Result, Score, SegmentReader};
+
+pub(crate) struct TweakedScoreTopCollector<TScoreTweaker, TScore = Score> {
+    score_tweaker: TScoreTweaker,
+    collector: TopCollector<TScore>,
+}
+
+impl<TScoreTweaker, TScore> TweakedScoreTopCollector<TScoreTweaker, TScore>
+where
+    TScore: Clone + PartialOrd,
+{
+    pub fn new(
+        score_tweaker: TScoreTweaker,
+        collector: TopCollector<TScore>,
+    ) -> TweakedScoreTopCollector<TScoreTweaker, TScore> {
+        TweakedScoreTopCollector {
+            score_tweaker,
+            collector,
+        }
+    }
+}
+
+/// A `ScoreSegmentTweaker` makes it possible to modify the default score
+/// for a given document belonging to a specific segment.
+///
+/// It is the segment local version of the [`ScoreTweaker`](./trait.ScoreTweaker.html).
+pub trait ScoreSegmentTweaker<TScore>: 'static {
+    /// Tweak the given `score` for the document `doc`.
+    fn score(&mut self, doc: DocId, score: Score) -> TScore;
+}
+
+/// `ScoreTweaker` makes it possible to tweak the score
+/// emitted  by the scorer into another one.
+///
+/// The `ScoreTweaker` itself does not make much of the computation itself.
+/// Instead, it helps constructing `Self::Child` instances that will compute
+/// the score at a segment scale.
+pub trait ScoreTweaker<TScore>: Sync {
+    /// Type of the associated [`ScoreSegmentTweaker`](./trait.ScoreSegmentTweaker.html).
+    type Child: ScoreSegmentTweaker<TScore>;
+
+    /// Builds a child tweaker for a specific segment. The child scorer is associated to
+    /// a specific segment.
+    fn segment_tweaker(&self, segment_reader: &SegmentReader) -> Result<Self::Child>;
+}
+
+impl<TScoreTweaker, TScore> Collector for TweakedScoreTopCollector<TScoreTweaker, TScore>
+where
+    TScoreTweaker: ScoreTweaker<TScore>,
+    TScore: 'static + PartialOrd + Clone + Send + Sync,
+{
+    type Fruit = Vec<(TScore, DocAddress)>;
+
+    type Child = TopTweakedScoreSegmentCollector<TScoreTweaker::Child, TScore>;
+
+    fn for_segment(
+        &self,
+        segment_local_id: u32,
+        segment_reader: &SegmentReader,
+    ) -> Result<Self::Child> {
+        let segment_scorer = self.score_tweaker.segment_tweaker(segment_reader)?;
+        let segment_collector = self
+            .collector
+            .for_segment(segment_local_id, segment_reader)?;
+        Ok(TopTweakedScoreSegmentCollector {
+            segment_collector,
+            segment_scorer,
+        })
+    }
+
+    fn requires_scoring(&self) -> bool {
+        true
+    }
+
+    fn merge_fruits(&self, segment_fruits: Vec<Self::Fruit>) -> Result<Self::Fruit> {
+        self.collector.merge_fruits(segment_fruits)
+    }
+}
+
+pub struct TopTweakedScoreSegmentCollector<TSegmentScoreTweaker, TScore>
+where
+    TScore: 'static + PartialOrd + Clone + Send + Sync + Sized,
+    TSegmentScoreTweaker: ScoreSegmentTweaker<TScore>,
+{
+    segment_collector: TopSegmentCollector<TScore>,
+    segment_scorer: TSegmentScoreTweaker,
+}
+
+impl<TSegmentScoreTweaker, TScore> SegmentCollector
+    for TopTweakedScoreSegmentCollector<TSegmentScoreTweaker, TScore>
+where
+    TScore: 'static + PartialOrd + Clone + Send + Sync,
+    TSegmentScoreTweaker: 'static + ScoreSegmentTweaker<TScore>,
+{
+    type Fruit = Vec<(TScore, DocAddress)>;
+
+    fn collect(&mut self, doc: DocId, score: Score) {
+        let score = self.segment_scorer.score(doc, score);
+        self.segment_collector.collect(doc, score);
+    }
+
+    fn harvest(self) -> Vec<(TScore, DocAddress)> {
+        self.segment_collector.harvest()
+    }
+}
+
+impl<F, TScore, TSegmentScoreTweaker> ScoreTweaker<TScore> for F
+where
+    F: 'static + Send + Sync + Fn(&SegmentReader) -> TSegmentScoreTweaker,
+    TSegmentScoreTweaker: ScoreSegmentTweaker<TScore>,
+{
+    type Child = TSegmentScoreTweaker;
+
+    fn segment_tweaker(&self, segment_reader: &SegmentReader) -> Result<Self::Child> {
+        Ok((self)(segment_reader))
+    }
+}
+
+impl<F, TScore> ScoreSegmentTweaker<TScore> for F
+where
+    F: 'static + FnMut(DocId, Score) -> TScore,
+{
+    fn score(&mut self, doc: DocId, score: Score) -> TScore {
+        (self)(doc, score)
+    }
+}

src/common/bitpacker.rs

@@ -1,9 +1,6 @@
-use common::serialize::BinarySerializable;
+use byteorder::{ByteOrder, LittleEndian, WriteBytesExt};
 use std::io;
-use std::io::Write;
-use std::mem;
 use std::ops::Deref;
-use std::ptr;
 
 pub(crate) struct BitPacker {
     mini_buffer: u64,
@@ -18,7 +15,7 @@ impl BitPacker {
         }
     }
 
-    pub fn write<TWrite: Write>(
+    pub fn write<TWrite: io::Write>(
         &mut self,
         val: u64,
         num_bits: u8,
@@ -28,14 +25,14 @@ impl BitPacker {
         let num_bits = num_bits as usize;
         if self.mini_buffer_written + num_bits > 64 {
             self.mini_buffer |= val_u64.wrapping_shl(self.mini_buffer_written as u32);
-            self.mini_buffer.serialize(output)?;
+            output.write_u64::<LittleEndian>(self.mini_buffer)?;
             self.mini_buffer = val_u64.wrapping_shr((64 - self.mini_buffer_written) as u32);
             self.mini_buffer_written = self.mini_buffer_written + num_bits - 64;
         } else {
             self.mini_buffer |= val_u64 << self.mini_buffer_written;
             self.mini_buffer_written += num_bits;
             if self.mini_buffer_written == 64 {
-                self.mini_buffer.serialize(output)?;
+                output.write_u64::<LittleEndian>(self.mini_buffer)?;
                 self.mini_buffer_written = 0;
                 self.mini_buffer = 0u64;
             }
@@ -43,17 +40,18 @@ impl BitPacker {
         Ok(())
     }
 
-    pub fn flush<TWrite: Write>(&mut self, output: &mut TWrite) -> io::Result<()> {
+    pub fn flush<TWrite: io::Write>(&mut self, output: &mut TWrite) -> io::Result<()> {
         if self.mini_buffer_written > 0 {
             let num_bytes = (self.mini_buffer_written + 7) / 8;
-            let arr: [u8; 8] = unsafe { mem::transmute::<u64, [u8; 8]>(self.mini_buffer.to_le()) };
+            let mut arr: [u8; 8] = [0u8; 8];
+            LittleEndian::write_u64(&mut arr, self.mini_buffer);
             output.write_all(&arr[..num_bytes])?;
             self.mini_buffer_written = 0;
         }
         Ok(())
     }
 
-    pub fn close<TWrite: Write>(&mut self, output: &mut TWrite) -> io::Result<()> {
+    pub fn close<TWrite: io::Write>(&mut self, output: &mut TWrite) -> io::Result<()> {
         self.flush(output)?;
         // Padding the write file to simplify reads.
         output.write_all(&[0u8; 7])?;
@@ -66,7 +64,7 @@ pub struct BitUnpacker<Data>
 where
     Data: Deref<Target = [u8]>,
 {
-    num_bits: usize,
+    num_bits: u64,
     mask: u64,
     data: Data,
 }
@@ -82,13 +80,13 @@ where
             (1u64 << num_bits) - 1u64
         };
         BitUnpacker {
-            num_bits: num_bits as usize,
+            num_bits: u64::from(num_bits),
             mask,
             data,
         }
     }
 
-    pub fn get(&self, idx: usize) -> u64 {
+    pub fn get(&self, idx: u64) -> u64 {
         if self.num_bits == 0 {
             return 0u64;
         }
@@ -99,42 +97,13 @@ where
         let addr = addr_in_bits >> 3;
         let bit_shift = addr_in_bits & 7;
         debug_assert!(
-            addr + 8 <= data.len(),
+            addr + 8 <= data.len() as u64,
             "The fast field field should have been padded with 7 bytes."
         );
-        #[cfg_attr(feature = "cargo-clippy", allow(clippy::cast_ptr_alignment))]
-        let val_unshifted_unmasked: u64 =
-            u64::from_le(unsafe { ptr::read_unaligned(data[addr..].as_ptr() as *const u64) });
+        let val_unshifted_unmasked: u64 = LittleEndian::read_u64(&data[(addr as usize)..]);
         let val_shifted = (val_unshifted_unmasked >> bit_shift) as u64;
         val_shifted & mask
     }
-
-    /// Reads a range of values from the fast field.
-    ///
-    /// The range of values read is from
-    /// `[start..start + output.len()[`
-    pub fn get_range(&self, start: u32, output: &mut [u64]) {
-        if self.num_bits == 0 {
-            for val in output.iter_mut() {
-                *val = 0u64;
-            }
-        } else {
-            let data: &[u8] = &*self.data;
-            let num_bits = self.num_bits;
-            let mask = self.mask;
-            let mut addr_in_bits = (start as usize) * num_bits;
-            for output_val in output.iter_mut() {
-                let addr = addr_in_bits >> 3;
-                let bit_shift = addr_in_bits & 7;
-                #[cfg_attr(feature = "cargo-clippy", allow(clippy::cast_ptr_alignment))]
-                let val_unshifted_unmasked: u64 =
-                    unsafe { ptr::read_unaligned(data[addr..].as_ptr() as *const u64) };
-                let val_shifted = (val_unshifted_unmasked >> bit_shift) as u64;
-                *output_val = val_shifted & mask;
-                addr_in_bits += num_bits;
-            }
-        }
-    }
 }
 
 #[cfg(test)]
@@ -160,7 +129,7 @@ mod test {
     fn test_bitpacker_util(len: usize, num_bits: u8) {
         let (bitunpacker, vals) = create_fastfield_bitpacker(len, num_bits);
         for (i, val) in vals.iter().enumerate() {
-            assert_eq!(bitunpacker.get(i), *val);
+            assert_eq!(bitunpacker.get(i as u64), *val);
         }
     }
 
@@ -172,17 +141,4 @@ mod test {
         test_bitpacker_util(6, 14);
         test_bitpacker_util(1000, 14);
     }
-
-    #[test]
-    fn test_bitpacker_range() {
-        let (bitunpacker, vals) = create_fastfield_bitpacker(100_000, 12);
-        let buffer_len = 100;
-        let mut buffer = vec![0u64; buffer_len];
-        for start in vec![0, 10, 20, 100, 1_000] {
-            bitunpacker.get_range(start as u32, &mut buffer[..]);
-            for i in 0..buffer_len {
-                assert_eq!(buffer[i], vals[start + i]);
-            }
-        }
-    }
 }

src/common/bitset.rs

@@ -5,7 +5,7 @@ use std::u64;
 pub(crate) struct TinySet(u64);
 
 impl fmt::Debug for TinySet {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         self.into_iter().collect::<Vec<u32>>().fmt(f)
     }
 }
@@ -33,6 +33,10 @@ impl TinySet {
         TinySet(0u64)
     }
 
+    pub fn clear(&mut self) {
+        self.0 = 0u64;
+    }
+
     /// Returns the complement of the set in `[0, 64[`.
     fn complement(self) -> TinySet {
         TinySet(!self.0)
@@ -43,6 +47,11 @@ impl TinySet {
         !self.intersect(TinySet::singleton(el)).is_empty()
     }
 
+    /// Returns the number of elements in the TinySet.
+    pub fn len(self) -> u32 {
+        self.0.count_ones()
+    }
+
     /// Returns the intersection of `self` and `other`
     pub fn intersect(self, other: TinySet) -> TinySet {
         TinySet(self.0 & other.0)
@@ -109,22 +118,12 @@ impl TinySet {
     pub fn range_greater_or_equal(from_included: u32) -> TinySet {
         TinySet::range_lower(from_included).complement()
     }
-
-    pub fn clear(&mut self) {
-        self.0 = 0u64;
-    }
-
-    pub fn len(self) -> u32 {
-        self.0.count_ones()
-    }
 }
 
 #[derive(Clone)]
 pub struct BitSet {
     tinysets: Box<[TinySet]>,
-    len: usize, //< Technically it should be u32, but we
-    // count multiple inserts.
-    // `usize` guards us from overflow.
+    len: usize,
     max_value: u32,
 }
 
@@ -204,12 +203,12 @@ mod tests {
 
     use super::BitSet;
     use super::TinySet;
-    use docset::DocSet;
-    use query::BitSetDocSet;
+    use crate::docset::{DocSet, TERMINATED};
+    use crate::query::BitSetDocSet;
+    use crate::tests;
+    use crate::tests::generate_nonunique_unsorted;
     use std::collections::BTreeSet;
     use std::collections::HashSet;
-    use tests;
-    use tests::generate_nonunique_unsorted;
 
     #[test]
     fn test_tiny_set() {
@@ -278,11 +277,13 @@ mod tests {
         }
         assert_eq!(btreeset.len(), bitset.len());
         let mut bitset_docset = BitSetDocSet::from(bitset);
+        let mut remaining = true;
         for el in btreeset.into_iter() {
-            bitset_docset.advance();
+            assert!(remaining);
             assert_eq!(bitset_docset.doc(), el);
+            remaining = bitset_docset.advance() != TERMINATED;
         }
-        assert!(!bitset_docset.advance());
+        assert!(!remaining);
     }
 
     #[test]

src/common/composite_file.rs

@@ -1,11 +1,11 @@
-use common::BinarySerializable;
-use common::CountingWriter;
-use common::VInt;
-use directory::ReadOnlySource;
-use directory::WritePtr;
-use schema::Field;
-use space_usage::FieldUsage;
-use space_usage::PerFieldSpaceUsage;
+use crate::common::BinarySerializable;
+use crate::common::CountingWriter;
+use crate::common::VInt;
+use crate::directory::ReadOnlySource;
+use crate::directory::{TerminatingWrite, WritePtr};
+use crate::schema::Field;
+use crate::space_usage::FieldUsage;
+use crate::space_usage::PerFieldSpaceUsage;
 use std::collections::HashMap;
 use std::io::Write;
 use std::io::{self, Read};
@@ -39,10 +39,10 @@ impl BinarySerializable for FileAddr {
 /// A `CompositeWrite` is used to write a `CompositeFile`.
 pub struct CompositeWrite<W = WritePtr> {
     write: CountingWriter<W>,
-    offsets: HashMap<FileAddr, usize>,
+    offsets: HashMap<FileAddr, u64>,
 }
 
-impl<W: Write> CompositeWrite<W> {
+impl<W: TerminatingWrite + Write> CompositeWrite<W> {
     /// Crate a new API writer that writes a composite file
     /// in a given write.
     pub fn wrap(w: W) -> CompositeWrite<W> {
@@ -91,8 +91,7 @@ impl<W: Write> CompositeWrite<W> {
 
         let footer_len = (self.write.written_bytes() - footer_offset) as u32;
         footer_len.serialize(&mut self.write)?;
-        self.write.flush()?;
-        Ok(())
+        self.write.terminate()
     }
 }
 
@@ -185,10 +184,10 @@ impl CompositeFile {
 mod test {
 
     use super::{CompositeFile, CompositeWrite};
-    use common::BinarySerializable;
-    use common::VInt;
-    use directory::{Directory, RAMDirectory};
-    use schema::Field;
+    use crate::common::BinarySerializable;
+    use crate::common::VInt;
+    use crate::directory::{Directory, RAMDirectory};
+    use crate::schema::Field;
     use std::io::Write;
     use std::path::Path;
 
@@ -200,13 +199,13 @@ mod test {
             let w = directory.open_write(path).unwrap();
             let mut composite_write = CompositeWrite::wrap(w);
             {
-                let mut write_0 = composite_write.for_field(Field(0u32));
+                let mut write_0 = composite_write.for_field(Field::from_field_id(0u32));
                 VInt(32431123u64).serialize(&mut write_0).unwrap();
                 write_0.flush().unwrap();
             }
 
             {
-                let mut write_4 = composite_write.for_field(Field(4u32));
+                let mut write_4 = composite_write.for_field(Field::from_field_id(4u32));
                 VInt(2).serialize(&mut write_4).unwrap();
                 write_4.flush().unwrap();
             }
@@ -216,14 +215,18 @@ mod test {
             let r = directory.open_read(path).unwrap();
             let composite_file = CompositeFile::open(&r).unwrap();
             {
-                let file0 = composite_file.open_read(Field(0u32)).unwrap();
+                let file0 = composite_file
+                    .open_read(Field::from_field_id(0u32))
+                    .unwrap();
                 let mut file0_buf = file0.as_slice();
                 let payload_0 = VInt::deserialize(&mut file0_buf).unwrap().0;
                 assert_eq!(file0_buf.len(), 0);
                 assert_eq!(payload_0, 32431123u64);
             }
             {
-                let file4 = composite_file.open_read(Field(4u32)).unwrap();
+                let file4 = composite_file
+                    .open_read(Field::from_field_id(4u32))
+                    .unwrap();
                 let mut file4_buf = file4.as_slice();
                 let payload_4 = VInt::deserialize(&mut file4_buf).unwrap().0;
                 assert_eq!(file4_buf.len(), 0);
@@ -231,5 +234,4 @@ mod test {
             }
         }
     }
-
 }

src/common/counting_writer.rs

@@ -1,9 +1,11 @@
+use crate::directory::AntiCallToken;
+use crate::directory::TerminatingWrite;
 use std::io;
 use std::io::Write;
 
 pub struct CountingWriter<W> {
     underlying: W,
-    written_bytes: usize,
+    written_bytes: u64,
 }
 
 impl<W: Write> CountingWriter<W> {
@@ -14,11 +16,11 @@ impl<W: Write> CountingWriter<W> {
         }
     }
 
-    pub fn written_bytes(&self) -> usize {
+    pub fn written_bytes(&self) -> u64 {
         self.written_bytes
     }
 
-    pub fn finish(mut self) -> io::Result<(W, usize)> {
+    pub fn finish(mut self) -> io::Result<(W, u64)> {
         self.flush()?;
         Ok((self.underlying, self.written_bytes))
     }
@@ -27,15 +29,28 @@ impl<W: Write> CountingWriter<W> {
 impl<W: Write> Write for CountingWriter<W> {
     fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
         let written_size = self.underlying.write(buf)?;
-        self.written_bytes += written_size;
+        self.written_bytes += written_size as u64;
         Ok(written_size)
     }
 
+    fn write_all(&mut self, buf: &[u8]) -> io::Result<()> {
+        self.underlying.write_all(buf)?;
+        self.written_bytes += buf.len() as u64;
+        Ok(())
+    }
+
     fn flush(&mut self) -> io::Result<()> {
         self.underlying.flush()
     }
 }
 
+impl<W: TerminatingWrite> TerminatingWrite for CountingWriter<W> {
+    fn terminate_ref(&mut self, token: AntiCallToken) -> io::Result<()> {
+        self.flush()?;
+        self.underlying.terminate_ref(token)
+    }
+}
+
 #[cfg(test)]
 mod test {
 
@@ -48,8 +63,8 @@ mod test {
         let mut counting_writer = CountingWriter::wrap(buffer);
         let bytes = (0u8..10u8).collect::<Vec<u8>>();
         counting_writer.write_all(&bytes).unwrap();
-        let (w, len): (Vec<u8>, usize) = counting_writer.finish().unwrap();
-        assert_eq!(len, 10);
+        let (w, len): (Vec<u8>, u64) = counting_writer.finish().unwrap();
+        assert_eq!(len, 10u64);
         assert_eq!(w.len(), 10);
     }
 }

src/common/mod.rs

@@ -10,10 +10,26 @@ pub(crate) use self::bitset::TinySet;
 pub(crate) use self::composite_file::{CompositeFile, CompositeWrite};
 pub use self::counting_writer::CountingWriter;
 pub use self::serialize::{BinarySerializable, FixedSize};
-pub use self::vint::VInt;
+pub use self::vint::{read_u32_vint, serialize_vint_u32, write_u32_vint, VInt};
 pub use byteorder::LittleEndian as Endianness;
 
-use std::io;
+/// Segment's max doc must be `< MAX_DOC_LIMIT`.
+///
+/// We do not allow segments with more than
+pub const MAX_DOC_LIMIT: u32 = 1 << 31;
+
+pub fn minmax<I, T>(mut vals: I) -> Option<(T, T)>
+where
+    I: Iterator<Item = T>,
+    T: Copy + Ord,
+{
+    if let Some(first_el) = vals.next() {
+        return Some(vals.fold((first_el, first_el), |(min_val, max_val), el| {
+            (min_val.min(el), max_val.max(el))
+        }));
+    }
+    None
+}
 
 /// Computes the number of bits that will be used for bitpacking.
 ///
@@ -52,11 +68,6 @@ pub(crate) fn is_power_of_2(n: usize) -> bool {
     (n > 0) && (n & (n - 1) == 0)
 }
 
-/// Create a default io error given a string.
-pub(crate) fn make_io_err(msg: String) -> io::Error {
-    io::Error::new(io::ErrorKind::Other, msg)
-}
-
 /// Has length trait
 pub trait HasLen {
     /// Return length
@@ -101,16 +112,54 @@ pub fn u64_to_i64(val: u64) -> i64 {
     (val ^ HIGHEST_BIT) as i64
 }
 
+/// Maps a `f64` to `u64`
+///
+/// For simplicity, tantivy internally handles `f64` as `u64`.
+/// The mapping is defined by this function.
+///
+/// Maps `f64` to `u64` so that lexical order is preserved.
+///
+/// This is more suited than simply casting (`val as u64`)
+/// which would truncate the result
+///
+/// # See also
+/// The [reverse mapping is `u64_to_f64`](./fn.u64_to_f64.html).
+#[inline(always)]
+pub fn f64_to_u64(val: f64) -> u64 {
+    let bits = val.to_bits();
+    if val.is_sign_positive() {
+        bits ^ HIGHEST_BIT
+    } else {
+        !bits
+    }
+}
+
+/// Reverse the mapping given by [`i64_to_u64`](./fn.i64_to_u64.html).
+#[inline(always)]
+pub fn u64_to_f64(val: u64) -> f64 {
+    f64::from_bits(if val & HIGHEST_BIT != 0 {
+        val ^ HIGHEST_BIT
+    } else {
+        !val
+    })
+}
+
 #[cfg(test)]
 pub(crate) mod test {
 
+    pub use super::minmax;
     pub use super::serialize::test::fixed_size_test;
-    use super::{compute_num_bits, i64_to_u64, u64_to_i64};
+    use super::{compute_num_bits, f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64};
+    use std::f64;
 
     fn test_i64_converter_helper(val: i64) {
         assert_eq!(u64_to_i64(i64_to_u64(val)), val);
     }
 
+    fn test_f64_converter_helper(val: f64) {
+        assert_eq!(u64_to_f64(f64_to_u64(val)), val);
+    }
+
     #[test]
     fn test_i64_converter() {
         assert_eq!(i64_to_u64(i64::min_value()), u64::min_value());
@@ -123,6 +172,29 @@ pub(crate) mod test {
         }
     }
 
+    #[test]
+    fn test_f64_converter() {
+        test_f64_converter_helper(f64::INFINITY);
+        test_f64_converter_helper(f64::NEG_INFINITY);
+        test_f64_converter_helper(0.0);
+        test_f64_converter_helper(-0.0);
+        test_f64_converter_helper(1.0);
+        test_f64_converter_helper(-1.0);
+    }
+
+    #[test]
+    fn test_f64_order() {
+        assert!(!(f64_to_u64(f64::NEG_INFINITY)..f64_to_u64(f64::INFINITY))
+            .contains(&f64_to_u64(f64::NAN))); //nan is not a number
+        assert!(f64_to_u64(1.5) > f64_to_u64(1.0)); //same exponent, different mantissa
+        assert!(f64_to_u64(2.0) > f64_to_u64(1.0)); //same mantissa, different exponent
+        assert!(f64_to_u64(2.0) > f64_to_u64(1.5)); //different exponent and mantissa
+        assert!(f64_to_u64(1.0) > f64_to_u64(-1.0)); // pos > neg
+        assert!(f64_to_u64(-1.5) < f64_to_u64(-1.0));
+        assert!(f64_to_u64(-2.0) < f64_to_u64(1.0));
+        assert!(f64_to_u64(-2.0) < f64_to_u64(-1.5));
+    }
+
     #[test]
     fn test_compute_num_bits() {
         assert_eq!(compute_num_bits(1), 1u8);
@@ -134,4 +206,28 @@ pub(crate) mod test {
         assert_eq!(compute_num_bits(256), 9u8);
         assert_eq!(compute_num_bits(5_000_000_000), 33u8);
     }
+
+    #[test]
+    fn test_max_doc() {
+        // this is the first time I write a unit test for a constant.
+        assert!(((super::MAX_DOC_LIMIT - 1) as i32) >= 0);
+        assert!((super::MAX_DOC_LIMIT as i32) < 0);
+    }
+
+    #[test]
+    fn test_minmax_empty() {
+        let vals: Vec<u32> = vec![];
+        assert_eq!(minmax(vals.into_iter()), None);
+    }
+
+    #[test]
+    fn test_minmax_one() {
+        assert_eq!(minmax(vec![1].into_iter()), Some((1, 1)));
+    }
+
+    #[test]
+    fn test_minmax_two() {
+        assert_eq!(minmax(vec![1, 2].into_iter()), Some((1, 2)));
+        assert_eq!(minmax(vec![2, 1].into_iter()), Some((1, 2)));
+    }
 }

src/common/serialize.rs

@@ -1,6 +1,6 @@
+use crate::common::Endianness;
+use crate::common::VInt;
 use byteorder::{ReadBytesExt, WriteBytesExt};
-use common::Endianness;
-use common::VInt;
 use std::fmt;
 use std::io;
 use std::io::Read;
@@ -102,6 +102,19 @@ impl FixedSize for i64 {
     const SIZE_IN_BYTES: usize = 8;
 }
 
+impl BinarySerializable for f64 {
+    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
+        writer.write_f64::<Endianness>(*self)
+    }
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
+        reader.read_f64::<Endianness>()
+    }
+}
+
+impl FixedSize for f64 {
+    const SIZE_IN_BYTES: usize = 8;
+}
+
 impl BinarySerializable for u8 {
     fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
         writer.write_u8(*self)
@@ -136,7 +149,7 @@ impl BinarySerializable for String {
 pub mod test {
 
     use super::*;
-    use common::VInt;
+    use crate::common::VInt;
 
     pub fn fixed_size_test<O: BinarySerializable + FixedSize + Default>() {
         let mut buffer = Vec::new();
@@ -172,6 +185,11 @@ pub mod test {
         fixed_size_test::<i64>();
     }
 
+    #[test]
+    fn test_serialize_f64() {
+        fixed_size_test::<f64>();
+    }
+
     #[test]
     fn test_serialize_u64() {
         fixed_size_test::<u64>();
@@ -181,10 +199,7 @@ pub mod test {
     fn test_serialize_string() {
         assert_eq!(serialize_test(String::from("")), 1);
         assert_eq!(serialize_test(String::from("ぽよぽよ")), 1 + 3 * 4);
-        assert_eq!(
-            serialize_test(String::from("富士さん見える。")),
-            1 + 3 * 8
-        );
+        assert_eq!(serialize_test(String::from("富士さん見える。")), 1 + 3 * 8);
     }
 
     #[test]

src/common/vint.rs

@@ -1,4 +1,5 @@
 use super::BinarySerializable;
+use byteorder::{ByteOrder, LittleEndian};
 use std::io;
 use std::io::Read;
 use std::io::Write;
@@ -9,6 +10,100 @@ pub struct VInt(pub u64);
 
 const STOP_BIT: u8 = 128;
 
+pub fn serialize_vint_u32(val: u32) -> (u64, usize) {
+    const START_2: u64 = 1 << 7;
+    const START_3: u64 = 1 << 14;
+    const START_4: u64 = 1 << 21;
+    const START_5: u64 = 1 << 28;
+
+    const STOP_1: u64 = START_2 - 1;
+    const STOP_2: u64 = START_3 - 1;
+    const STOP_3: u64 = START_4 - 1;
+    const STOP_4: u64 = START_5 - 1;
+
+    const MASK_1: u64 = 127;
+    const MASK_2: u64 = MASK_1 << 7;
+    const MASK_3: u64 = MASK_2 << 7;
+    const MASK_4: u64 = MASK_3 << 7;
+    const MASK_5: u64 = MASK_4 << 7;
+
+    let val = u64::from(val);
+    const STOP_BIT: u64 = 128u64;
+    match val {
+        0..=STOP_1 => (val | STOP_BIT, 1),
+        START_2..=STOP_2 => (
+            (val & MASK_1) | ((val & MASK_2) << 1) | (STOP_BIT << (8)),
+            2,
+        ),
+        START_3..=STOP_3 => (
+            (val & MASK_1) | ((val & MASK_2) << 1) | ((val & MASK_3) << 2) | (STOP_BIT << (8 * 2)),
+            3,
+        ),
+        START_4..=STOP_4 => (
+            (val & MASK_1)
+                | ((val & MASK_2) << 1)
+                | ((val & MASK_3) << 2)
+                | ((val & MASK_4) << 3)
+                | (STOP_BIT << (8 * 3)),
+            4,
+        ),
+        _ => (
+            (val & MASK_1)
+                | ((val & MASK_2) << 1)
+                | ((val & MASK_3) << 2)
+                | ((val & MASK_4) << 3)
+                | ((val & MASK_5) << 4)
+                | (STOP_BIT << (8 * 4)),
+            5,
+        ),
+    }
+}
+
+/// Returns the number of bytes covered by a
+/// serialized vint `u32`.
+///
+/// Expects a buffer data that starts
+/// by the serialized `vint`, scans at most 5 bytes ahead until
+/// it finds the vint final byte.
+///
+/// # May Panic
+/// If the payload does not start by a valid `vint`
+fn vint_len(data: &[u8]) -> usize {
+    for (i, &val) in data.iter().enumerate().take(5) {
+        if val >= STOP_BIT {
+            return i + 1;
+        }
+    }
+    panic!("Corrupted data. Invalid VInt 32");
+}
+
+/// Reads a vint `u32` from a buffer, and
+/// consumes its payload data.
+///
+/// # Panics
+///
+/// If the buffer does not start by a valid
+/// vint payload
+pub fn read_u32_vint(data: &mut &[u8]) -> u32 {
+    let vlen = vint_len(*data);
+    let mut result = 0u32;
+    let mut shift = 0u64;
+    for &b in &data[..vlen] {
+        result |= u32::from(b & 127u8) << shift;
+        shift += 7;
+    }
+    *data = &data[vlen..];
+    result
+}
+
+/// Write a `u32` as a vint payload.
+pub fn write_u32_vint<W: io::Write>(val: u32, writer: &mut W) -> io::Result<()> {
+    let (val, num_bytes) = serialize_vint_u32(val);
+    let mut buffer = [0u8; 8];
+    LittleEndian::write_u64(&mut buffer, val);
+    writer.write_all(&buffer[..num_bytes])
+}
+
 impl VInt {
     pub fn val(&self) -> u64 {
         self.0
@@ -24,7 +119,7 @@ impl VInt {
         output.extend(&buffer[0..num_bytes]);
     }
 
-    fn serialize_into(&self, buffer: &mut [u8; 10]) -> usize {
+    pub fn serialize_into(&self, buffer: &mut [u8; 10]) -> usize {
         let mut remaining = self.0;
         for (i, b) in buffer.iter_mut().enumerate() {
             let next_byte: u8 = (remaining % 128u64) as u8;
@@ -64,7 +159,7 @@ impl BinarySerializable for VInt {
                     return Err(io::Error::new(
                         io::ErrorKind::InvalidData,
                         "Reach end of buffer while reading VInt",
-                    ))
+                    ));
                 }
             }
         }
@@ -74,8 +169,10 @@ impl BinarySerializable for VInt {
 #[cfg(test)]
 mod tests {
 
+    use super::serialize_vint_u32;
     use super::VInt;
-    use common::BinarySerializable;
+    use crate::common::BinarySerializable;
+    use byteorder::{ByteOrder, LittleEndian};
 
     fn aux_test_vint(val: u64) {
         let mut v = [14u8; 10];
@@ -108,4 +205,28 @@ mod tests {
         }
         aux_test_vint(10);
     }
+
+    fn aux_test_serialize_vint_u32(val: u32) {
+        let mut buffer = [0u8; 10];
+        let mut buffer2 = [0u8; 10];
+        let len_vint = VInt(val as u64).serialize_into(&mut buffer);
+        let (vint, len) = serialize_vint_u32(val);
+        assert_eq!(len, len_vint, "len wrong for val {}", val);
+        LittleEndian::write_u64(&mut buffer2, vint);
+        assert_eq!(&buffer[..len], &buffer2[..len], "array wrong for {}", val);
+    }
+
+    #[test]
+    fn test_vint_u32() {
+        aux_test_serialize_vint_u32(0);
+        aux_test_serialize_vint_u32(1);
+        aux_test_serialize_vint_u32(5);
+        for i in 1..3 {
+            let power_of_128 = 1u32 << (7 * i);
+            aux_test_serialize_vint_u32(power_of_128 - 1u32);
+            aux_test_serialize_vint_u32(power_of_128);
+            aux_test_serialize_vint_u32(power_of_128 + 1u32);
+        }
+        aux_test_serialize_vint_u32(u32::max_value());
+    }
 }

src/core/executor.rs

@@ -1,6 +1,5 @@
 use crossbeam::channel;
-use scoped_pool::{Pool, ThreadConfig};
-use Result;
+use rayon::{ThreadPool, ThreadPoolBuilder};
 
 /// Search executor whether search request are single thread or multithread.
 ///
@@ -10,8 +9,10 @@ use Result;
 /// API of a dependency, knowing it might conflict with a different version
 /// used by the client. Second, we may stop using rayon in the future.
 pub enum Executor {
+    /// Single thread variant of an Executor
     SingleThread,
-    ThreadPool(Pool),
+    /// Thread pool variant of an Executor
+    ThreadPool(ThreadPool),
 }
 
 impl Executor {
@@ -20,37 +21,39 @@ impl Executor {
         Executor::SingleThread
     }
 
-    // Creates an Executor that dispatches the tasks in a thread pool.
-    pub fn multi_thread(num_threads: usize, prefix: &'static str) -> Executor {
-        let thread_config = ThreadConfig::new().prefix(prefix);
-        let pool = Pool::with_thread_config(num_threads, thread_config);
-        Executor::ThreadPool(pool)
+    /// Creates an Executor that dispatches the tasks in a thread pool.
+    pub fn multi_thread(num_threads: usize, prefix: &'static str) -> crate::Result<Executor> {
+        let pool = ThreadPoolBuilder::new()
+            .num_threads(num_threads)
+            .thread_name(move |num| format!("{}{}", prefix, num))
+            .build()?;
+        Ok(Executor::ThreadPool(pool))
     }
 
-    // Perform a map in the thread pool.
-    //
-    // Regardless of the executor (`SingleThread` or `ThreadPool`), panics in the task
-    // will propagate to the caller.
+    /// Perform a map in the thread pool.
+    ///
+    /// Regardless of the executor (`SingleThread` or `ThreadPool`), panics in the task
+    /// will propagate to the caller.
     pub fn map<
         A: Send,
         R: Send,
         AIterator: Iterator<Item = A>,
-        F: Sized + Sync + Fn(A) -> Result<R>,
+        F: Sized + Sync + Fn(A) -> crate::Result<R>,
     >(
         &self,
         f: F,
         args: AIterator,
-    ) -> Result<Vec<R>> {
+    ) -> crate::Result<Vec<R>> {
         match self {
-            Executor::SingleThread => args.map(f).collect::<Result<_>>(),
+            Executor::SingleThread => args.map(f).collect::<crate::Result<_>>(),
             Executor::ThreadPool(pool) => {
                 let args_with_indices: Vec<(usize, A)> = args.enumerate().collect();
                 let num_fruits = args_with_indices.len();
                 let fruit_receiver = {
                     let (fruit_sender, fruit_receiver) = channel::unbounded();
-                    pool.scoped(|scope| {
+                    pool.scope(|scope| {
                         for arg_with_idx in args_with_indices {
-                            scope.execute(|| {
+                            scope.spawn(|_| {
                                 let (idx, arg) = arg_with_idx;
                                 let fruit = f(arg);
                                 if let Err(err) = fruit_sender.send((idx, fruit)) {
@@ -64,17 +67,18 @@ impl Executor {
                     // This is important as it makes it possible for the fruit_receiver iteration to
                     // terminate.
                 };
-                let mut results = Vec::with_capacity(num_fruits);
-                unsafe { results.set_len(num_fruits) };
-                let mut num_items = 0;
+                // This is lame, but safe.
+                let mut results_with_position = Vec::with_capacity(num_fruits);
                 for (pos, fruit_res) in fruit_receiver {
-                    results[pos] = fruit_res?;
-                    num_items += 1;
+                    let fruit = fruit_res?;
+                    results_with_position.push((pos, fruit));
                 }
-                // this checks ensures that we filled of this
-                // uninitialized memory.
-                assert_eq!(num_items, results.len());
-                Ok(results)
+                results_with_position.sort_by_key(|(pos, _)| *pos);
+                assert_eq!(results_with_position.len(), num_fruits);
+                Ok(results_with_position
+                    .into_iter()
+                    .map(|(_, fruit)| fruit)
+                    .collect::<Vec<_>>())
             }
         }
     }
@@ -94,19 +98,22 @@ mod tests {
                     panic!("panic should propagate");
                 },
                 vec![0].into_iter(),
-            ).unwrap();
+            )
+            .unwrap();
     }
 
     #[test]
     #[should_panic] //< unfortunately the panic message is not propagated
     fn test_panic_propagates_multi_thread() {
         let _result: Vec<usize> = Executor::multi_thread(1, "search-test")
+            .unwrap()
             .map(
                 |_| {
                     panic!("panic should propagate");
                 },
                 vec![0].into_iter(),
-            ).unwrap();
+            )
+            .unwrap();
     }
 
     #[test]
@@ -120,15 +127,15 @@ mod tests {
         }
     }
 
-}
-
-#[test]
-fn test_map_multithread() {
-    let result: Vec<usize> = Executor::multi_thread(3, "search-test")
-        .map(|i| Ok(i * 2), 0..10)
-        .unwrap();
-    assert_eq!(result.len(), 10);
-    for i in 0..10 {
-        assert_eq!(result[i], i * 2);
+    #[test]
+    fn test_map_multithread() {
+        let result: Vec<usize> = Executor::multi_thread(3, "search-test")
+            .unwrap()
+            .map(|i| Ok(i * 2), 0..10)
+            .unwrap();
+        assert_eq!(result.len(), 10);
+        for i in 0..10 {
+            assert_eq!(result[i], i * 2);
+        }
     }
 }

src/core/index.rs

@@ -1,53 +1,59 @@
-use super::pool::LeasedItem;
-use super::pool::Pool;
-use super::segment::create_segment;
 use super::segment::Segment;
-use core::searcher::Searcher;
-use core::Executor;
-use core::IndexMeta;
-use core::SegmentId;
-use core::SegmentMeta;
-use core::SegmentReader;
-use core::META_FILEPATH;
-use directory::ManagedDirectory;
+use crate::core::Executor;
+use crate::core::IndexMeta;
+use crate::core::SegmentId;
+use crate::core::SegmentMeta;
+use crate::core::SegmentMetaInventory;
+use crate::core::META_FILEPATH;
+use crate::directory::ManagedDirectory;
 #[cfg(feature = "mmap")]
-use directory::MmapDirectory;
-use directory::{Directory, RAMDirectory};
-use error::TantivyError;
-use indexer::index_writer::open_index_writer;
-use indexer::index_writer::HEAP_SIZE_MIN;
-use indexer::segment_updater::save_new_metas;
-use indexer::LockType;
-use num_cpus;
-use schema::Field;
-use schema::FieldType;
-use schema::Schema;
-use serde_json;
+use crate::directory::MmapDirectory;
+use crate::directory::INDEX_WRITER_LOCK;
+use crate::directory::{Directory, RAMDirectory};
+use crate::error::DataCorruption;
+use crate::error::TantivyError;
+use crate::indexer::index_writer::HEAP_SIZE_MIN;
+use crate::indexer::segment_updater::save_new_metas;
+use crate::reader::IndexReader;
+use crate::reader::IndexReaderBuilder;
+use crate::schema::Field;
+use crate::schema::FieldType;
+use crate::schema::Schema;
+use crate::tokenizer::{TextAnalyzer, TokenizerManager};
+use crate::IndexWriter;
 use std::borrow::BorrowMut;
+use std::collections::HashSet;
 use std::fmt;
-use std::path::Path;
-use std::sync::atomic::{AtomicUsize, Ordering};
-use std::sync::Arc;
-use tokenizer::BoxedTokenizer;
-use tokenizer::TokenizerManager;
-use IndexWriter;
-use Result;
 
-fn load_metas(directory: &Directory) -> Result<IndexMeta> {
+#[cfg(feature = "mmap")]
+use std::path::Path;
+use std::path::PathBuf;
+use std::sync::Arc;
+
+fn load_metas(
+    directory: &dyn Directory,
+    inventory: &SegmentMetaInventory,
+) -> crate::Result<IndexMeta> {
     let meta_data = directory.atomic_read(&META_FILEPATH)?;
     let meta_string = String::from_utf8_lossy(&meta_data);
-    serde_json::from_str(&meta_string)
-        .map_err(|_| TantivyError::CorruptedFile(META_FILEPATH.clone()))
+    IndexMeta::deserialize(&meta_string, &inventory)
+        .map_err(|e| {
+            DataCorruption::new(
+                META_FILEPATH.to_path_buf(),
+                format!("Meta file cannot be deserialized. {:?}.", e),
+            )
+        })
+        .map_err(From::from)
 }
 
 /// Search Index
+#[derive(Clone)]
 pub struct Index {
     directory: ManagedDirectory,
     schema: Schema,
-    num_searchers: Arc<AtomicUsize>,
-    searcher_pool: Arc<Pool<Searcher>>,
     executor: Arc<Executor>,
     tokenizers: TokenizerManager,
+    inventory: SegmentMetaInventory,
 }
 
 impl Index {
@@ -68,15 +74,16 @@ impl Index {
 
     /// Replace the default single thread search executor pool
     /// by a thread pool with a given number of threads.
-    pub fn set_multithread_executor(&mut self, num_threads: usize) {
-        self.executor = Arc::new(Executor::multi_thread(num_threads, "thrd-tantivy-search-"));
+    pub fn set_multithread_executor(&mut self, num_threads: usize) -> crate::Result<()> {
+        self.executor = Arc::new(Executor::multi_thread(num_threads, "thrd-tantivy-search-")?);
+        Ok(())
     }
 
     /// Replace the default single thread search executor pool
     /// by a thread pool with a given number of threads.
-    pub fn set_default_multithread_executor(&mut self) {
+    pub fn set_default_multithread_executor(&mut self) -> crate::Result<()> {
         let default_num_threads = num_cpus::get();
-        self.set_multithread_executor(default_num_threads);
+        self.set_multithread_executor(default_num_threads)
     }
 
     /// Creates a new index using the `RAMDirectory`.
@@ -93,29 +100,29 @@ impl Index {
     ///
     /// If a previous index was in this directory, then its meta file will be destroyed.
     #[cfg(feature = "mmap")]
-    pub fn create_in_dir<P: AsRef<Path>>(directory_path: P, schema: Schema) -> Result<Index> {
+    pub fn create_in_dir<P: AsRef<Path>>(
+        directory_path: P,
+        schema: Schema,
+    ) -> crate::Result<Index> {
         let mmap_directory = MmapDirectory::open(directory_path)?;
         if Index::exists(&mmap_directory) {
             return Err(TantivyError::IndexAlreadyExists);
         }
-
         Index::create(mmap_directory, schema)
     }
 
     /// Opens or creates a new index in the provided directory
-    #[cfg(feature = "mmap")]
-    pub fn open_or_create<Dir: Directory>(dir: Dir, schema: Schema) -> Result<Index> {
-        if Index::exists(&dir) {
-            let index = Index::open(dir)?;
-            if index.schema() == schema {
-                Ok(index)
-            } else {
-                Err(TantivyError::SchemaError(
-                    "An index exists but the schema does not match.".to_string(),
-                ))
-            }
+    pub fn open_or_create<Dir: Directory>(dir: Dir, schema: Schema) -> crate::Result<Index> {
+        if !Index::exists(&dir) {
+            return Index::create(dir, schema);
+        }
+        let index = Index::open(dir)?;
+        if index.schema() == schema {
+            Ok(index)
         } else {
-            Index::create(dir, schema)
+            Err(TantivyError::SchemaError(
+                "An index exists but the schema does not match.".to_string(),
+            ))
         }
     }
 
@@ -128,40 +135,40 @@ impl Index {
     /// The temp directory is only used for testing the `MmapDirectory`.
     /// For other unit tests, prefer the `RAMDirectory`, see: `create_in_ram`.
     #[cfg(feature = "mmap")]
-    pub fn create_from_tempdir(schema: Schema) -> Result<Index> {
+    pub fn create_from_tempdir(schema: Schema) -> crate::Result<Index> {
         let mmap_directory = MmapDirectory::create_from_tempdir()?;
         Index::create(mmap_directory, schema)
     }
 
     /// Creates a new index given an implementation of the trait `Directory`
-    pub fn create<Dir: Directory>(dir: Dir, schema: Schema) -> Result<Index> {
-        let directory = ManagedDirectory::new(dir)?;
+    pub fn create<Dir: Directory>(dir: Dir, schema: Schema) -> crate::Result<Index> {
+        let directory = ManagedDirectory::wrap(dir)?;
         Index::from_directory(directory, schema)
     }
 
     /// Create a new index from a directory.
     ///
     /// This will overwrite existing meta.json
-    fn from_directory(mut directory: ManagedDirectory, schema: Schema) -> Result<Index> {
-        save_new_metas(schema.clone(), 0, directory.borrow_mut())?;
+    fn from_directory(mut directory: ManagedDirectory, schema: Schema) -> crate::Result<Index> {
+        save_new_metas(schema.clone(), directory.borrow_mut())?;
         let metas = IndexMeta::with_schema(schema);
-        Index::create_from_metas(directory, &metas)
+        Index::create_from_metas(directory, &metas, SegmentMetaInventory::default())
     }
 
     /// Creates a new index given a directory and an `IndexMeta`.
-    fn create_from_metas(directory: ManagedDirectory, metas: &IndexMeta) -> Result<Index> {
+    fn create_from_metas(
+        directory: ManagedDirectory,
+        metas: &IndexMeta,
+        inventory: SegmentMetaInventory,
+    ) -> crate::Result<Index> {
         let schema = metas.schema.clone();
-        let n_cpus = num_cpus::get();
-        let index = Index {
+        Ok(Index {
             directory,
             schema,
-            num_searchers: Arc::new(AtomicUsize::new(n_cpus)),
-            searcher_pool: Arc::new(Pool::new()),
             tokenizers: TokenizerManager::default(),
             executor: Arc::new(Executor::single_thread()),
-        };
-        index.load_searchers()?;
-        Ok(index)
+            inventory,
+        })
     }
 
     /// Accessor for the tokenizer manager.
@@ -170,11 +177,11 @@ impl Index {
     }
 
     /// Helper to access the tokenizer associated to a specific field.
-    pub fn tokenizer_for_field(&self, field: Field) -> Result<Box<BoxedTokenizer>> {
+    pub fn tokenizer_for_field(&self, field: Field) -> crate::Result<TextAnalyzer> {
         let field_entry = self.schema.get_field_entry(field);
         let field_type = field_entry.field_type();
         let tokenizer_manager: &TokenizerManager = self.tokenizers();
-        let tokenizer_name_opt: Option<Box<BoxedTokenizer>> = match field_type {
+        let tokenizer_name_opt: Option<TextAnalyzer> = match field_type {
             FieldType::Str(text_options) => text_options
                 .get_indexing_options()
                 .map(|text_indexing_options| text_indexing_options.tokenizer().to_string())
@@ -190,23 +197,58 @@ impl Index {
         }
     }
 
+    /// Create a default `IndexReader` for the given index.
+    ///
+    /// See [`Index.reader_builder()`](#method.reader_builder).
+    pub fn reader(&self) -> crate::Result<IndexReader> {
+        self.reader_builder().try_into()
+    }
+
+    /// Create a `IndexReader` for the given index.
+    ///
+    /// Most project should create at most one reader for a given index.
+    /// This method is typically called only once per `Index` instance,
+    /// over the lifetime of most problem.
+    pub fn reader_builder(&self) -> IndexReaderBuilder {
+        IndexReaderBuilder::new(self.clone())
+    }
+
     /// Opens a new directory from an index path.
     #[cfg(feature = "mmap")]
-    pub fn open_in_dir<P: AsRef<Path>>(directory_path: P) -> Result<Index> {
+    pub fn open_in_dir<P: AsRef<Path>>(directory_path: P) -> crate::Result<Index> {
         let mmap_directory = MmapDirectory::open(directory_path)?;
         Index::open(mmap_directory)
     }
 
+    /// Returns the list of the segment metas tracked by the index.
+    ///
+    /// Such segments can of course be part of the index,
+    /// but also they could be segments being currently built or in the middle of a merge
+    /// operation.
+    pub fn list_all_segment_metas(&self) -> Vec<SegmentMeta> {
+        self.inventory.all()
+    }
+
+    /// Creates a new segment_meta (Advanced user only).
+    ///
+    /// As long as the `SegmentMeta` lives, the files associated with the
+    /// `SegmentMeta` are guaranteed to not be garbage collected, regardless of
+    /// whether the segment is recorded as part of the index or not.
+    pub fn new_segment_meta(&self, segment_id: SegmentId, max_doc: u32) -> SegmentMeta {
+        self.inventory.new_segment_meta(segment_id, max_doc)
+    }
+
     /// Open the index using the provided directory
-    pub fn open<D: Directory>(directory: D) -> Result<Index> {
-        let directory = ManagedDirectory::new(directory)?;
-        let metas = load_metas(&directory)?;
-        Index::create_from_metas(directory, &metas)
+    pub fn open<D: Directory>(directory: D) -> crate::Result<Index> {
+        let directory = ManagedDirectory::wrap(directory)?;
+        let inventory = SegmentMetaInventory::default();
+        let metas = load_metas(&directory, &inventory)?;
+        Index::create_from_metas(directory, &metas, inventory)
     }
 
     /// Reads the index meta file from the directory.
-    pub fn load_metas(&self) -> Result<IndexMeta> {
-        load_metas(self.directory())
+    pub fn load_metas(&self) -> crate::Result<IndexMeta> {
+        load_metas(self.directory(), &self.inventory)
     }
 
     /// Open a new index writer. Attempts to acquire a lockfile.
@@ -225,17 +267,32 @@ impl Index {
     /// Each thread will receive a budget of  `overall_heap_size_in_bytes / num_threads`.
     ///
     /// # Errors
-    /// If the lockfile already exists, returns `Error::FileAlreadyExists`.
+    /// If the lockfile already exists, returns `Error::DirectoryLockBusy` or an `Error::IOError`.
+    ///
     /// # Panics
     /// If the heap size per thread is too small, panics.
     pub fn writer_with_num_threads(
         &self,
         num_threads: usize,
         overall_heap_size_in_bytes: usize,
-    ) -> Result<IndexWriter> {
-        let directory_lock = LockType::IndexWriterLock.acquire_lock(&self.directory)?;
+    ) -> crate::Result<IndexWriter> {
+        let directory_lock = self
+            .directory
+            .acquire_lock(&INDEX_WRITER_LOCK)
+            .map_err(|err| {
+                TantivyError::LockFailure(
+                    err,
+                    Some(
+                        "Failed to acquire index lock. If you are using\
+                         a regular directory, this means there is already an \
+                         `IndexWriter` working on this `Directory`, in this process \
+                         or in a different process."
+                            .to_string(),
+                    ),
+                )
+            })?;
         let heap_size_in_bytes_per_thread = overall_heap_size_in_bytes / num_threads;
-        open_index_writer(
+        IndexWriter::new(
             self,
             num_threads,
             heap_size_in_bytes_per_thread,
@@ -253,7 +310,7 @@ impl Index {
     /// If the lockfile already exists, returns `Error::FileAlreadyExists`.
     /// # Panics
     /// If the heap size per thread is too small, panics.
-    pub fn writer(&self, overall_heap_size_in_bytes: usize) -> Result<IndexWriter> {
+    pub fn writer(&self, overall_heap_size_in_bytes: usize) -> crate::Result<IndexWriter> {
         let mut num_threads = num_cpus::get();
         let heap_size_in_bytes_per_thread = overall_heap_size_in_bytes / num_threads;
         if heap_size_in_bytes_per_thread < HEAP_SIZE_MIN {
@@ -270,7 +327,7 @@ impl Index {
     }
 
     /// Returns the list of segments that are searchable
-    pub fn searchable_segments(&self) -> Result<Vec<Segment>> {
+    pub fn searchable_segments(&self) -> crate::Result<Vec<Segment>> {
         Ok(self
             .searchable_segment_metas()?
             .into_iter()
@@ -280,12 +337,14 @@ impl Index {
 
     #[doc(hidden)]
     pub fn segment(&self, segment_meta: SegmentMeta) -> Segment {
-        create_segment(self.clone(), segment_meta)
+        Segment::for_index(self.clone(), segment_meta)
     }
 
     /// Creates a new segment.
     pub fn new_segment(&self) -> Segment {
-        let segment_meta = SegmentMeta::new(SegmentId::generate_random(), 0);
+        let segment_meta = self
+            .inventory
+            .new_segment_meta(SegmentId::generate_random(), 0);
         self.segment(segment_meta)
     }
 
@@ -301,96 +360,44 @@ impl Index {
 
     /// Reads the meta.json and returns the list of
     /// `SegmentMeta` from the last commit.
-    pub fn searchable_segment_metas(&self) -> Result<Vec<SegmentMeta>> {
+    pub fn searchable_segment_metas(&self) -> crate::Result<Vec<SegmentMeta>> {
         Ok(self.load_metas()?.segments)
     }
 
     /// Returns the list of segment ids that are searchable.
-    pub fn searchable_segment_ids(&self) -> Result<Vec<SegmentId>> {
+    pub fn searchable_segment_ids(&self) -> crate::Result<Vec<SegmentId>> {
         Ok(self
             .searchable_segment_metas()?
             .iter()
-            .map(|segment_meta| segment_meta.id())
+            .map(SegmentMeta::id)
             .collect())
     }
 
-    /// Sets the number of searchers to use
-    ///
-    /// Only works after the next call to `load_searchers`
-    pub fn set_num_searchers(&mut self, num_searchers: usize) {
-        self.num_searchers.store(num_searchers, Ordering::Release);
-    }
-
-    /// Update searchers so that they reflect the state of the last
-    /// `.commit()`.
-    ///
-    /// If indexing happens in the same process as searching,
-    /// you most likely want to call `.load_searchers()` right after each
-    /// successful call to `.commit()`.
-    ///
-    /// If indexing and searching happen in different processes, the way to
-    /// get the freshest `index` at all time, is to watch `meta.json` and
-    /// call `load_searchers` whenever a changes happen.
-    pub fn load_searchers(&self) -> Result<()> {
-        let _meta_lock = LockType::MetaLock.acquire_lock(self.directory())?;
-        let searchable_segments = self.searchable_segments()?;
-        let segment_readers: Vec<SegmentReader> = searchable_segments
-            .iter()
-            .map(SegmentReader::open)
-            .collect::<Result<_>>()?;
-        let schema = self.schema();
-        let num_searchers: usize = self.num_searchers.load(Ordering::Acquire);
-        let searchers = (0..num_searchers)
-            .map(|_| Searcher::new(schema.clone(), self.clone(), segment_readers.clone()))
-            .collect();
-        self.searcher_pool.publish_new_generation(searchers);
-        Ok(())
-    }
-
-    /// Returns a searcher
-    ///
-    /// This method should be called every single time a search
-    /// query is performed.
-    /// The searchers are taken from a pool of `num_searchers` searchers.
-    /// If no searcher is available
-    /// this may block.
-    ///
-    /// The same searcher must be used for a given query, as it ensures
-    /// the use of a consistent segment set.
-    pub fn searcher(&self) -> LeasedItem<Searcher> {
-        self.searcher_pool.acquire()
+    /// Returns the set of corrupted files
+    pub fn validate_checksum(&self) -> crate::Result<HashSet<PathBuf>> {
+        self.directory.list_damaged().map_err(Into::into)
     }
 }
 
 impl fmt::Debug for Index {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         write!(f, "Index({:?})", self.directory)
     }
 }
 
-impl Clone for Index {
-    fn clone(&self) -> Index {
-        Index {
-            directory: self.directory.clone(),
-            schema: self.schema.clone(),
-            num_searchers: Arc::clone(&self.num_searchers),
-            searcher_pool: Arc::clone(&self.searcher_pool),
-            tokenizers: self.tokenizers.clone(),
-            executor: self.executor.clone(),
-        }
-    }
-}
-
 #[cfg(test)]
 mod tests {
-    use directory::RAMDirectory;
-    use schema::{Schema, INT_INDEXED, TEXT};
-    use Index;
+    use crate::directory::RAMDirectory;
+    use crate::schema::Field;
+    use crate::schema::{Schema, INDEXED, TEXT};
+    use crate::IndexReader;
+    use crate::ReloadPolicy;
+    use crate::{Directory, Index};
 
     #[test]
     fn test_indexer_for_field() {
         let mut schema_builder = Schema::builder();
-        let num_likes_field = schema_builder.add_u64_field("num_likes", INT_INDEXED);
+        let num_likes_field = schema_builder.add_u64_field("num_likes", INDEXED);
         let body_field = schema_builder.add_text_field("body", TEXT);
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
@@ -448,7 +455,148 @@ mod tests {
 
     fn throw_away_schema() -> Schema {
         let mut schema_builder = Schema::builder();
-        let _ = schema_builder.add_u64_field("num_likes", INT_INDEXED);
+        let _ = schema_builder.add_u64_field("num_likes", INDEXED);
         schema_builder.build()
     }
+
+    #[test]
+    fn test_index_on_commit_reload_policy() {
+        let schema = throw_away_schema();
+        let field = schema.get_field("num_likes").unwrap();
+        let index = Index::create_in_ram(schema);
+        let reader = index
+            .reader_builder()
+            .reload_policy(ReloadPolicy::OnCommit)
+            .try_into()
+            .unwrap();
+        assert_eq!(reader.searcher().num_docs(), 0);
+        test_index_on_commit_reload_policy_aux(field, &index, &reader);
+    }
+
+    #[cfg(feature = "mmap")]
+    mod mmap_specific {
+
+        use super::*;
+        use crate::Directory;
+        use std::path::PathBuf;
+        use tempfile::TempDir;
+
+        #[test]
+        fn test_index_on_commit_reload_policy_mmap() {
+            let schema = throw_away_schema();
+            let field = schema.get_field("num_likes").unwrap();
+            let tempdir = TempDir::new().unwrap();
+            let tempdir_path = PathBuf::from(tempdir.path());
+            let index = Index::create_in_dir(&tempdir_path, schema).unwrap();
+            let reader = index
+                .reader_builder()
+                .reload_policy(ReloadPolicy::OnCommit)
+                .try_into()
+                .unwrap();
+            assert_eq!(reader.searcher().num_docs(), 0);
+            test_index_on_commit_reload_policy_aux(field, &index, &reader);
+        }
+
+        #[test]
+        fn test_index_manual_policy_mmap() {
+            let schema = throw_away_schema();
+            let field = schema.get_field("num_likes").unwrap();
+            let mut index = Index::create_from_tempdir(schema).unwrap();
+            let mut writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+            writer.commit().unwrap();
+            let reader = index
+                .reader_builder()
+                .reload_policy(ReloadPolicy::Manual)
+                .try_into()
+                .unwrap();
+            assert_eq!(reader.searcher().num_docs(), 0);
+            writer.add_document(doc!(field=>1u64));
+            let (sender, receiver) = crossbeam::channel::unbounded();
+            let _handle = index.directory_mut().watch(Box::new(move || {
+                let _ = sender.send(());
+            }));
+            writer.commit().unwrap();
+            assert!(receiver.recv().is_ok());
+            assert_eq!(reader.searcher().num_docs(), 0);
+            reader.reload().unwrap();
+            assert_eq!(reader.searcher().num_docs(), 1);
+        }
+
+        #[test]
+        fn test_index_on_commit_reload_policy_different_directories() {
+            let schema = throw_away_schema();
+            let field = schema.get_field("num_likes").unwrap();
+            let tempdir = TempDir::new().unwrap();
+            let tempdir_path = PathBuf::from(tempdir.path());
+            let write_index = Index::create_in_dir(&tempdir_path, schema).unwrap();
+            let read_index = Index::open_in_dir(&tempdir_path).unwrap();
+            let reader = read_index
+                .reader_builder()
+                .reload_policy(ReloadPolicy::OnCommit)
+                .try_into()
+                .unwrap();
+            assert_eq!(reader.searcher().num_docs(), 0);
+            test_index_on_commit_reload_policy_aux(field, &write_index, &reader);
+        }
+    }
+
+    fn test_index_on_commit_reload_policy_aux(field: Field, index: &Index, reader: &IndexReader) {
+        let mut reader_index = reader.index();
+        let (sender, receiver) = crossbeam::channel::unbounded();
+        let _watch_handle = reader_index.directory_mut().watch(Box::new(move || {
+            let _ = sender.send(());
+        }));
+        let mut writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+        assert_eq!(reader.searcher().num_docs(), 0);
+        writer.add_document(doc!(field=>1u64));
+        writer.commit().unwrap();
+        assert!(receiver.recv().is_ok());
+        assert_eq!(reader.searcher().num_docs(), 1);
+        writer.add_document(doc!(field=>2u64));
+        writer.commit().unwrap();
+        assert!(receiver.recv().is_ok());
+        assert_eq!(reader.searcher().num_docs(), 2);
+    }
+
+    // This test will not pass on windows, because windows
+    // prevent deleting files that are MMapped.
+    #[cfg(not(target_os = "windows"))]
+    #[test]
+    fn garbage_collect_works_as_intended() {
+        let directory = RAMDirectory::create();
+        let schema = throw_away_schema();
+        let field = schema.get_field("num_likes").unwrap();
+        let index = Index::create(directory.clone(), schema).unwrap();
+
+        let mut writer = index.writer_with_num_threads(8, 24_000_000).unwrap();
+        for i in 0u64..8_000u64 {
+            writer.add_document(doc!(field => i));
+        }
+        let (sender, receiver) = crossbeam::channel::unbounded();
+        let _handle = directory.watch(Box::new(move || {
+            let _ = sender.send(());
+        }));
+        writer.commit().unwrap();
+        let mem_right_after_commit = directory.total_mem_usage();
+        assert!(receiver.recv().is_ok());
+        let reader = index
+            .reader_builder()
+            .reload_policy(ReloadPolicy::Manual)
+            .try_into()
+            .unwrap();
+
+        assert_eq!(reader.searcher().num_docs(), 8_000);
+        writer.wait_merging_threads().unwrap();
+        let mem_right_after_merge_finished = directory.total_mem_usage();
+
+        reader.reload().unwrap();
+        let searcher = reader.searcher();
+        assert_eq!(searcher.num_docs(), 8_000);
+        assert!(
+            mem_right_after_merge_finished < mem_right_after_commit,
+            "(mem after merge){} is expected < (mem before merge){}",
+            mem_right_after_merge_finished,
+            mem_right_after_commit
+        );
+    }
 }

src/core/index_meta.rs

@@ -1,7 +1,198 @@
-use core::SegmentMeta;
-use schema::Schema;
-use serde_json;
+use super::SegmentComponent;
+use crate::core::SegmentId;
+use crate::schema::Schema;
+use crate::Opstamp;
+use census::{Inventory, TrackedObject};
+use serde::{Deserialize, Serialize};
+use std::collections::HashSet;
 use std::fmt;
+use std::path::PathBuf;
+
+#[derive(Clone, Debug, Serialize, Deserialize)]
+struct DeleteMeta {
+    num_deleted_docs: u32,
+    opstamp: Opstamp,
+}
+
+#[derive(Clone, Default)]
+pub struct SegmentMetaInventory {
+    inventory: Inventory<InnerSegmentMeta>,
+}
+
+impl SegmentMetaInventory {
+    /// Lists all living `SegmentMeta` object at the time of the call.
+    pub fn all(&self) -> Vec<SegmentMeta> {
+        self.inventory
+            .list()
+            .into_iter()
+            .map(SegmentMeta::from)
+            .collect::<Vec<_>>()
+    }
+
+    pub fn new_segment_meta(&self, segment_id: SegmentId, max_doc: u32) -> SegmentMeta {
+        let inner = InnerSegmentMeta {
+            segment_id,
+            max_doc,
+            deletes: None,
+        };
+        SegmentMeta::from(self.inventory.track(inner))
+    }
+}
+
+/// `SegmentMeta` contains simple meta information about a segment.
+///
+/// For instance the number of docs it contains,
+/// how many are deleted, etc.
+#[derive(Clone)]
+pub struct SegmentMeta {
+    tracked: TrackedObject<InnerSegmentMeta>,
+}
+
+impl fmt::Debug for SegmentMeta {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
+        self.tracked.fmt(f)
+    }
+}
+
+impl serde::Serialize for SegmentMeta {
+    fn serialize<S>(
+        &self,
+        serializer: S,
+    ) -> Result<<S as serde::Serializer>::Ok, <S as serde::Serializer>::Error>
+    where
+        S: serde::Serializer,
+    {
+        self.tracked.serialize(serializer)
+    }
+}
+
+impl From<TrackedObject<InnerSegmentMeta>> for SegmentMeta {
+    fn from(tracked: TrackedObject<InnerSegmentMeta>) -> SegmentMeta {
+        SegmentMeta { tracked }
+    }
+}
+
+impl SegmentMeta {
+    // Creates a new `SegmentMeta` object.
+
+    /// Returns the segment id.
+    pub fn id(&self) -> SegmentId {
+        self.tracked.segment_id
+    }
+
+    /// Returns the number of deleted documents.
+    pub fn num_deleted_docs(&self) -> u32 {
+        self.tracked
+            .deletes
+            .as_ref()
+            .map(|delete_meta| delete_meta.num_deleted_docs)
+            .unwrap_or(0u32)
+    }
+
+    /// Returns the list of files that
+    /// are required for the segment meta.
+    ///
+    /// This is useful as the way tantivy removes files
+    /// is by removing all files that have been created by tantivy
+    /// and are not used by any segment anymore.
+    pub fn list_files(&self) -> HashSet<PathBuf> {
+        SegmentComponent::iterator()
+            .map(|component| self.relative_path(*component))
+            .collect::<HashSet<PathBuf>>()
+    }
+
+    /// Returns the relative path of a component of our segment.
+    ///
+    /// It just joins the segment id with the extension
+    /// associated to a segment component.
+    pub fn relative_path(&self, component: SegmentComponent) -> PathBuf {
+        let mut path = self.id().uuid_string();
+        path.push_str(&*match component {
+            SegmentComponent::POSTINGS => ".idx".to_string(),
+            SegmentComponent::POSITIONS => ".pos".to_string(),
+            SegmentComponent::POSITIONSSKIP => ".posidx".to_string(),
+            SegmentComponent::TERMS => ".term".to_string(),
+            SegmentComponent::STORE => ".store".to_string(),
+            SegmentComponent::FASTFIELDS => ".fast".to_string(),
+            SegmentComponent::FIELDNORMS => ".fieldnorm".to_string(),
+            SegmentComponent::DELETE => format!(".{}.del", self.delete_opstamp().unwrap_or(0)),
+        });
+        PathBuf::from(path)
+    }
+
+    /// Return the highest doc id + 1
+    ///
+    /// If there are no deletes, then num_docs = max_docs
+    /// and all the doc ids contains in this segment
+    /// are exactly (0..max_doc).
+    pub fn max_doc(&self) -> u32 {
+        self.tracked.max_doc
+    }
+
+    /// Return the number of documents in the segment.
+    pub fn num_docs(&self) -> u32 {
+        self.max_doc() - self.num_deleted_docs()
+    }
+
+    /// Returns the `Opstamp` of the last delete operation
+    /// taken in account in this segment.
+    pub fn delete_opstamp(&self) -> Option<Opstamp> {
+        self.tracked
+            .deletes
+            .as_ref()
+            .map(|delete_meta| delete_meta.opstamp)
+    }
+
+    /// Returns true iff the segment meta contains
+    /// delete information.
+    pub fn has_deletes(&self) -> bool {
+        self.num_deleted_docs() > 0
+    }
+
+    /// Updates the max_doc value from the `SegmentMeta`.
+    ///
+    /// This method is only used when updating `max_doc` from 0
+    /// as we finalize a fresh new segment.
+    pub(crate) fn with_max_doc(self, max_doc: u32) -> SegmentMeta {
+        assert_eq!(self.tracked.max_doc, 0);
+        assert!(self.tracked.deletes.is_none());
+        let tracked = self.tracked.map(move |inner_meta| InnerSegmentMeta {
+            segment_id: inner_meta.segment_id,
+            max_doc,
+            deletes: None,
+        });
+        SegmentMeta { tracked }
+    }
+
+    #[doc(hidden)]
+    pub fn with_delete_meta(self, num_deleted_docs: u32, opstamp: Opstamp) -> SegmentMeta {
+        let delete_meta = DeleteMeta {
+            num_deleted_docs,
+            opstamp,
+        };
+        let tracked = self.tracked.map(move |inner_meta| InnerSegmentMeta {
+            segment_id: inner_meta.segment_id,
+            max_doc: inner_meta.max_doc,
+            deletes: Some(delete_meta),
+        });
+        SegmentMeta { tracked }
+    }
+}
+
+#[derive(Debug, Clone, Serialize, Deserialize)]
+struct InnerSegmentMeta {
+    segment_id: SegmentId,
+    max_doc: u32,
+    deletes: Option<DeleteMeta>,
+}
+
+impl InnerSegmentMeta {
+    pub fn track(self, inventory: &SegmentMetaInventory) -> SegmentMeta {
+        SegmentMeta {
+            tracked: inventory.inventory.track(self),
+        }
+    }
+}
 
 /// Meta information about the `Index`.
 ///
@@ -11,16 +202,53 @@ use std::fmt;
 /// * the index `docstamp`
 /// * the schema
 ///
-#[derive(Clone, Serialize, Deserialize)]
+#[derive(Clone, Serialize)]
 pub struct IndexMeta {
+    /// List of `SegmentMeta` informations associated to each finalized segment of the index.
     pub segments: Vec<SegmentMeta>,
+    /// Index `Schema`
     pub schema: Schema,
-    pub opstamp: u64,
+    /// Opstamp associated to the last `commit` operation.
+    pub opstamp: Opstamp,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    /// Payload associated to the last commit.
+    ///
+    /// Upon commit, clients can optionally add a small `Striing` payload to their commit
+    /// to help identify this commit.
+    /// This payload is entirely unused by tantivy.
+    pub payload: Option<String>,
+}
+
+#[derive(Deserialize)]
+struct UntrackedIndexMeta {
+    pub segments: Vec<InnerSegmentMeta>,
+    pub schema: Schema,
+    pub opstamp: Opstamp,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub payload: Option<String>,
 }
 
+impl UntrackedIndexMeta {
+    pub fn track(self, inventory: &SegmentMetaInventory) -> IndexMeta {
+        IndexMeta {
+            segments: self
+                .segments
+                .into_iter()
+                .map(|inner_seg_meta| inner_seg_meta.track(inventory))
+                .collect::<Vec<SegmentMeta>>(),
+            schema: self.schema,
+            opstamp: self.opstamp,
+            payload: self.payload,
+        }
+    }
+}
+
 impl IndexMeta {
+    /// Create an `IndexMeta` object representing a brand new `Index`
+    /// with the given index.
+    ///
+    /// This new index does not contains any segments.
+    /// Opstamp will the value `0u64`.
     pub fn with_schema(schema: Schema) -> IndexMeta {
         IndexMeta {
             segments: vec![],
@@ -29,10 +257,18 @@ impl IndexMeta {
             payload: None,
         }
     }
+
+    pub(crate) fn deserialize(
+        meta_json: &str,
+        inventory: &SegmentMetaInventory,
+    ) -> serde_json::Result<IndexMeta> {
+        let untracked_meta_json: UntrackedIndexMeta = serde_json::from_str(meta_json)?;
+        Ok(untracked_meta_json.track(inventory))
+    }
 }
 
 impl fmt::Debug for IndexMeta {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         write!(
             f,
             "{}",
@@ -46,7 +282,7 @@ impl fmt::Debug for IndexMeta {
 mod tests {
 
     use super::IndexMeta;
-    use schema::{Schema, TEXT};
+    use crate::schema::{Schema, TEXT};
     use serde_json;
 
     #[test]
@@ -63,6 +299,9 @@ mod tests {
             payload: None,
         };
         let json = serde_json::ser::to_string(&index_metas).expect("serialization failed");
-        assert_eq!(json, r#"{"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","tokenizer":"default"},"stored":false}}],"opstamp":0}"#);
+        assert_eq!(
+            json,
+            r#"{"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","tokenizer":"default"},"stored":false}}],"opstamp":0}"#
+        );
     }
 }

src/core/inverted_index_reader.rs

@@ -1,13 +1,12 @@
-use common::BinarySerializable;
-use directory::ReadOnlySource;
-use owned_read::OwnedRead;
-use positions::PositionReader;
-use postings::TermInfo;
-use postings::{BlockSegmentPostings, SegmentPostings};
-use schema::FieldType;
-use schema::IndexRecordOption;
-use schema::Term;
-use termdict::TermDictionary;
+use crate::common::BinarySerializable;
+use crate::directory::ReadOnlySource;
+use crate::positions::PositionReader;
+use crate::postings::TermInfo;
+use crate::postings::{BlockSegmentPostings, SegmentPostings};
+use crate::schema::FieldType;
+use crate::schema::IndexRecordOption;
+use crate::schema::Term;
+use crate::termdict::TermDictionary;
 
 /// The inverted index reader is in charge of accessing
 /// the inverted index associated to a specific field.
@@ -32,10 +31,7 @@ pub struct InvertedIndexReader {
 }
 
 impl InvertedIndexReader {
-    #[cfg_attr(
-        feature = "cargo-clippy",
-        allow(clippy::needless_pass_by_value)
-    )] // for symetry
+    #[cfg_attr(feature = "cargo-clippy", allow(clippy::needless_pass_by_value))] // for symmetry
     pub(crate) fn new(
         termdict: TermDictionary,
         postings_source: ReadOnlySource,
@@ -63,7 +59,7 @@ impl InvertedIndexReader {
             .get_index_record_option()
             .unwrap_or(IndexRecordOption::Basic);
         InvertedIndexReader {
-            termdict: TermDictionary::empty(&field_type),
+            termdict: TermDictionary::empty(),
             postings_source: ReadOnlySource::empty(),
             positions_source: ReadOnlySource::empty(),
             positions_idx_source: ReadOnlySource::empty(),
@@ -100,8 +96,7 @@ impl InvertedIndexReader {
         let offset = term_info.postings_offset as usize;
         let end_source = self.postings_source.len();
         let postings_slice = self.postings_source.slice(offset, end_source);
-        let postings_reader = OwnedRead::new(postings_slice);
-        block_postings.reset(term_info.doc_freq, postings_reader);
+        block_postings.reset(term_info.doc_freq, postings_slice);
     }
 
     /// Returns a block postings given a `Term`.
@@ -130,7 +125,7 @@ impl InvertedIndexReader {
         let postings_data = self.postings_source.slice_from(offset);
         BlockSegmentPostings::from_data(
             term_info.doc_freq,
-            OwnedRead::new(postings_data),
+            postings_data,
             self.record_option,
             requested_option,
         )

src/core/mod.rs

@@ -2,37 +2,33 @@ mod executor;
 pub mod index;
 mod index_meta;
 mod inverted_index_reader;
-mod pool;
 pub mod searcher;
 mod segment;
 mod segment_component;
 mod segment_id;
-mod segment_meta;
 mod segment_reader;
 
 pub use self::executor::Executor;
 pub use self::index::Index;
-pub use self::index_meta::IndexMeta;
+pub use self::index_meta::{IndexMeta, SegmentMeta, SegmentMetaInventory};
 pub use self::inverted_index_reader::InvertedIndexReader;
 pub use self::searcher::Searcher;
 pub use self::segment::Segment;
 pub use self::segment::SerializableSegment;
 pub use self::segment_component::SegmentComponent;
 pub use self::segment_id::SegmentId;
-pub use self::segment_meta::SegmentMeta;
 pub use self::segment_reader::SegmentReader;
 
-use std::path::PathBuf;
+use once_cell::sync::Lazy;
+use std::path::Path;
 
-lazy_static! {
-    /// The meta file contains all the information about the list of segments and the schema
-    /// of the index.
-    pub static ref META_FILEPATH: PathBuf = PathBuf::from("meta.json");
+/// The meta file contains all the information about the list of segments and the schema
+/// of the index.
+pub static META_FILEPATH: Lazy<&'static Path> = Lazy::new(|| Path::new("meta.json"));
 
-    /// The managed file contains a list of files that were created by the tantivy
-    /// and will therefore be garbage collected when they are deemed useless by tantivy.
-    ///
-    /// Removing this file is safe, but will prevent the garbage collection of all of the file that
-    /// are currently in the directory
-    pub static ref MANAGED_FILEPATH: PathBuf = PathBuf::from(".managed.json");
-}
+/// The managed file contains a list of files that were created by the tantivy
+/// and will therefore be garbage collected when they are deemed useless by tantivy.
+///
+/// Removing this file is safe, but will prevent the garbage collection of all of the file that
+/// are currently in the directory
+pub static MANAGED_FILEPATH: Lazy<&'static Path> = Lazy::new(|| Path::new(".managed.json"));

src/core/pool.rs

@@ -1,136 +0,0 @@
-use crossbeam::queue::MsQueue;
-use std::mem;
-use std::ops::{Deref, DerefMut};
-use std::sync::atomic::AtomicUsize;
-use std::sync::atomic::Ordering;
-use std::sync::Arc;
-
-pub struct GenerationItem<T> {
-    generation: usize,
-    item: T,
-}
-
-pub struct Pool<T> {
-    queue: Arc<MsQueue<GenerationItem<T>>>,
-    freshest_generation: AtomicUsize,
-    next_generation: AtomicUsize,
-}
-
-impl<T> Pool<T> {
-    pub fn new() -> Pool<T> {
-        let queue = Arc::new(MsQueue::new());
-        Pool {
-            queue,
-            freshest_generation: AtomicUsize::default(),
-            next_generation: AtomicUsize::default(),
-        }
-    }
-
-    pub fn publish_new_generation(&self, items: Vec<T>) {
-        let next_generation = self.next_generation.fetch_add(1, Ordering::SeqCst) + 1;
-        for item in items {
-            let gen_item = GenerationItem {
-                item,
-                generation: next_generation,
-            };
-            self.queue.push(gen_item);
-        }
-        self.advertise_generation(next_generation);
-    }
-
-    /// At the exit of this method,
-    /// - freshest_generation has a value greater or equal than generation
-    /// - freshest_generation has a value that has been advertised
-    /// - freshest_generation has)
-    fn advertise_generation(&self, generation: usize) {
-        // not optimal at all but the easiest to read proof.
-        loop {
-            let former_generation = self.freshest_generation.load(Ordering::Acquire);
-            if former_generation >= generation {
-                break;
-            }
-            self.freshest_generation.compare_and_swap(
-                former_generation,
-                generation,
-                Ordering::SeqCst,
-            );
-        }
-    }
-
-    fn generation(&self) -> usize {
-        self.freshest_generation.load(Ordering::Acquire)
-    }
-
-    pub fn acquire(&self) -> LeasedItem<T> {
-        let generation = self.generation();
-        loop {
-            let gen_item = self.queue.pop();
-            if gen_item.generation >= generation {
-                return LeasedItem {
-                    gen_item: Some(gen_item),
-                    recycle_queue: Arc::clone(&self.queue),
-                };
-            } else {
-                // this searcher is obsolete,
-                // removing it from the pool.
-            }
-        }
-    }
-}
-
-pub struct LeasedItem<T> {
-    gen_item: Option<GenerationItem<T>>,
-    recycle_queue: Arc<MsQueue<GenerationItem<T>>>,
-}
-
-impl<T> Deref for LeasedItem<T> {
-    type Target = T;
-
-    fn deref(&self) -> &T {
-        &self
-            .gen_item
-            .as_ref()
-            .expect("Unwrapping a leased item should never fail")
-            .item // unwrap is safe here
-    }
-}
-
-impl<T> DerefMut for LeasedItem<T> {
-    fn deref_mut(&mut self) -> &mut T {
-        &mut self
-            .gen_item
-            .as_mut()
-            .expect("Unwrapping a mut leased item should never fail")
-            .item // unwrap is safe here
-    }
-}
-
-impl<T> Drop for LeasedItem<T> {
-    fn drop(&mut self) {
-        let gen_item: GenerationItem<T> = mem::replace(&mut self.gen_item, None)
-            .expect("Unwrapping a leased item should never fail");
-        self.recycle_queue.push(gen_item);
-    }
-}
-
-#[cfg(test)]
-mod tests {
-
-    use super::Pool;
-    use std::iter;
-
-    #[test]
-    fn test_pool() {
-        let items10: Vec<usize> = iter::repeat(10).take(10).collect();
-        let pool = Pool::new();
-        pool.publish_new_generation(items10);
-        for _ in 0..20 {
-            assert_eq!(*pool.acquire(), 10);
-        }
-        let items11: Vec<usize> = iter::repeat(11).take(10).collect();
-        pool.publish_new_generation(items11);
-        for _ in 0..20 {
-            assert_eq!(*pool.acquire(), 11);
-        }
-    }
-}

src/core/searcher.rs

@@ -1,42 +1,18 @@
-use collector::Collector;
-use collector::SegmentCollector;
-use core::Executor;
-use core::InvertedIndexReader;
-use core::SegmentReader;
-use query::Query;
-use query::Scorer;
-use query::Weight;
-use schema::Document;
-use schema::Schema;
-use schema::{Field, Term};
-use space_usage::SearcherSpaceUsage;
+use crate::collector::Collector;
+use crate::core::Executor;
+use crate::core::InvertedIndexReader;
+use crate::core::SegmentReader;
+use crate::query::Query;
+use crate::schema::Document;
+use crate::schema::Schema;
+use crate::schema::{Field, Term};
+use crate::space_usage::SearcherSpaceUsage;
+use crate::store::StoreReader;
+use crate::termdict::TermMerger;
+use crate::DocAddress;
+use crate::Index;
 use std::fmt;
 use std::sync::Arc;
-use store::StoreReader;
-use termdict::TermMerger;
-use DocAddress;
-use Index;
-use Result;
-
-fn collect_segment<C: Collector>(
-    collector: &C,
-    weight: &Weight,
-    segment_ord: u32,
-    segment_reader: &SegmentReader,
-) -> Result<C::Fruit> {
-    let mut scorer = weight.scorer(segment_reader)?;
-    let mut segment_collector = collector.for_segment(segment_ord as u32, segment_reader)?;
-    if let Some(delete_bitset) = segment_reader.delete_bitset() {
-        scorer.for_each(&mut |doc, score| {
-            if !delete_bitset.is_deleted(doc) {
-                segment_collector.collect(doc, score);
-            }
-        });
-    } else {
-        scorer.for_each(&mut |doc, score| segment_collector.collect(doc, score));
-    }
-    Ok(segment_collector.harvest())
-}
 
 /// Holds a list of `SegmentReader`s ready for search.
 ///
@@ -59,7 +35,7 @@ impl Searcher {
     ) -> Searcher {
         let store_readers = segment_readers
             .iter()
-            .map(|segment_reader| segment_reader.get_store_reader())
+            .map(SegmentReader::get_store_reader)
             .collect();
         Searcher {
             schema,
@@ -78,7 +54,7 @@ impl Searcher {
     ///
     /// The searcher uses the segment ordinal to route the
     /// the request to the right `Segment`.
-    pub fn doc(&self, doc_address: DocAddress) -> Result<Document> {
+    pub fn doc(&self, doc_address: DocAddress) -> crate::Result<Document> {
         let DocAddress(segment_local_id, doc_id) = doc_address;
         let store_reader = &self.store_readers[segment_local_id as usize];
         store_reader.get(doc_id)
@@ -104,7 +80,8 @@ impl Searcher {
             .iter()
             .map(|segment_reader| {
                 u64::from(segment_reader.inverted_index(term.field()).doc_freq(term))
-            }).sum::<u64>()
+            })
+            .sum::<u64>()
     }
 
     /// Return the list of segment readers
@@ -131,7 +108,11 @@ impl Searcher {
     ///
     ///  Finally, the Collector merges each of the child collectors into itself for result usability
     ///  by the caller.
-    pub fn search<C: Collector>(&self, query: &Query, collector: &C) -> Result<C::Fruit> {
+    pub fn search<C: Collector>(
+        &self,
+        query: &dyn Query,
+        collector: &C,
+    ) -> crate::Result<C::Fruit> {
         let executor = self.index.search_executor();
         self.search_with_executor(query, collector, executor)
     }
@@ -150,21 +131,16 @@ impl Searcher {
     /// hurt it. It will however, decrease the average response time.
     pub fn search_with_executor<C: Collector>(
         &self,
-        query: &Query,
+        query: &dyn Query,
         collector: &C,
         executor: &Executor,
-    ) -> Result<C::Fruit> {
+    ) -> crate::Result<C::Fruit> {
         let scoring_enabled = collector.requires_scoring();
         let weight = query.weight(self, scoring_enabled)?;
         let segment_readers = self.segment_readers();
         let fruits = executor.map(
             |(segment_ord, segment_reader)| {
-                collect_segment(
-                    collector,
-                    weight.as_ref(),
-                    segment_ord as u32,
-                    segment_reader,
-                )
+                collector.collect_segment(weight.as_ref(), segment_ord as u32, segment_reader)
             },
             segment_readers.iter().enumerate(),
         )?;
@@ -202,7 +178,7 @@ impl FieldSearcher {
 
     /// Returns a Stream over all of the sorted unique terms of
     /// for the given field.
-    pub fn terms(&self) -> TermMerger {
+    pub fn terms(&self) -> TermMerger<'_> {
         let term_streamers: Vec<_> = self
             .inv_index_readers
             .iter()
@@ -213,11 +189,11 @@ impl FieldSearcher {
 }
 
 impl fmt::Debug for Searcher {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         let segment_ids = self
             .segment_readers
             .iter()
-            .map(|segment_reader| segment_reader.segment_id())
+            .map(SegmentReader::segment_id)
             .collect::<Vec<_>>();
         write!(f, "Searcher({:?})", segment_ids)
     }

src/core/segment.rs

@@ -1,16 +1,15 @@
 use super::SegmentComponent;
-use core::Index;
-use core::SegmentId;
-use core::SegmentMeta;
-use directory::error::{OpenReadError, OpenWriteError};
-use directory::Directory;
-use directory::{ReadOnlySource, WritePtr};
-use indexer::segment_serializer::SegmentSerializer;
-use schema::Schema;
+use crate::core::Index;
+use crate::core::SegmentId;
+use crate::core::SegmentMeta;
+use crate::directory::error::{OpenReadError, OpenWriteError};
+use crate::directory::Directory;
+use crate::directory::{ReadOnlySource, WritePtr};
+use crate::indexer::segment_serializer::SegmentSerializer;
+use crate::schema::Schema;
+use crate::Opstamp;
 use std::fmt;
 use std::path::PathBuf;
-use std::result;
-use Result;
 
 /// A segment is a piece of the index.
 #[derive(Clone)]
@@ -20,20 +19,17 @@ pub struct Segment {
 }
 
 impl fmt::Debug for Segment {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         write!(f, "Segment({:?})", self.id().uuid_string())
     }
 }
 
-/// Creates a new segment given an `Index` and a `SegmentId`
-///
-/// The function is here to make it private outside `tantivy`.
-/// #[doc(hidden)]
-pub fn create_segment(index: Index, meta: SegmentMeta) -> Segment {
-    Segment { index, meta }
-}
-
 impl Segment {
+    /// Creates a new segment given an `Index` and a `SegmentId`
+    pub(crate) fn for_index(index: Index, meta: SegmentMeta) -> Segment {
+        Segment { index, meta }
+    }
+
     /// Returns the index the segment belongs to.
     pub fn index(&self) -> &Index {
         &self.index
@@ -49,8 +45,19 @@ impl Segment {
         &self.meta
     }
 
+    /// Updates the max_doc value from the `SegmentMeta`.
+    ///
+    /// This method is only used when updating `max_doc` from 0
+    /// as we finalize a fresh new segment.
+    pub(crate) fn with_max_doc(self, max_doc: u32) -> Segment {
+        Segment {
+            index: self.index,
+            meta: self.meta.with_max_doc(max_doc),
+        }
+    }
+
     #[doc(hidden)]
-    pub fn with_delete_meta(self, num_deleted_docs: u32, opstamp: u64) -> Segment {
+    pub fn with_delete_meta(self, num_deleted_docs: u32, opstamp: Opstamp) -> Segment {
         Segment {
             index: self.index,
             meta: self.meta.with_delete_meta(num_deleted_docs, opstamp),
@@ -71,20 +78,14 @@ impl Segment {
     }
 
     /// Open one of the component file for a *regular* read.
-    pub fn open_read(
-        &self,
-        component: SegmentComponent,
-    ) -> result::Result<ReadOnlySource, OpenReadError> {
+    pub fn open_read(&self, component: SegmentComponent) -> Result<ReadOnlySource, OpenReadError> {
         let path = self.relative_path(component);
         let source = self.index.directory().open_read(&path)?;
         Ok(source)
     }
 
     /// Open one of the component file for *regular* write.
-    pub fn open_write(
-        &mut self,
-        component: SegmentComponent,
-    ) -> result::Result<WritePtr, OpenWriteError> {
+    pub fn open_write(&mut self, component: SegmentComponent) -> Result<WritePtr, OpenWriteError> {
         let path = self.relative_path(component);
         let write = self.index.directory_mut().open_write(&path)?;
         Ok(write)
@@ -97,5 +98,5 @@ pub trait SerializableSegment {
     ///
     /// # Returns
     /// The number of documents in the segment.
-    fn write(&self, serializer: SegmentSerializer) -> Result<u32>;
+    fn write(&self, serializer: SegmentSerializer) -> crate::Result<u32>;
 }

src/core/segment_component.rs

@@ -41,6 +41,6 @@ impl SegmentComponent {
             SegmentComponent::STORE,
             SegmentComponent::DELETE,
         ];
-        SEGMENT_COMPONENTS.into_iter()
+        SEGMENT_COMPONENTS.iter()
     }
 }

src/core/segment_id.rs

@@ -2,6 +2,11 @@ use std::cmp::{Ord, Ordering};
 use std::fmt;
 use uuid::Uuid;
 
+#[cfg(test)]
+use once_cell::sync::Lazy;
+use serde::{Deserialize, Serialize};
+use std::error::Error;
+use std::str::FromStr;
 #[cfg(test)]
 use std::sync::atomic;
 
@@ -17,10 +22,10 @@ use std::sync::atomic;
 pub struct SegmentId(Uuid);
 
 #[cfg(test)]
-lazy_static! {
-    static ref AUTO_INC_COUNTER: atomic::AtomicUsize = atomic::AtomicUsize::default();
-    static ref EMPTY_ARR: [u8; 8] = [0u8; 8];
-}
+static AUTO_INC_COUNTER: Lazy<atomic::AtomicUsize> = Lazy::new(|| atomic::AtomicUsize::default());
+
+#[cfg(test)]
+const ZERO_ARRAY: [u8; 8] = [0u8; 8];
 
 // During tests, we generate the segment id in a autoincrement manner
 // for consistency of segment id between run.
@@ -30,7 +35,7 @@ lazy_static! {
 #[cfg(test)]
 fn create_uuid() -> Uuid {
     let new_auto_inc_id = (*AUTO_INC_COUNTER).fetch_add(1, atomic::Ordering::SeqCst);
-    Uuid::from_fields(new_auto_inc_id as u32, 0, 0, &*EMPTY_ARR).unwrap()
+    Uuid::from_fields(new_auto_inc_id as u32, 0, 0, &ZERO_ARRAY).unwrap()
 }
 
 #[cfg(not(test))]
@@ -50,19 +55,55 @@ impl SegmentId {
     /// and the rest is random.
     ///
     /// Picking the first 8 chars is ok to identify
-    /// segments in a display message.
+    /// segments in a display message (e.g. a5c4dfcb).
     pub fn short_uuid_string(&self) -> String {
         (&self.0.to_simple_ref().to_string()[..8]).to_string()
     }
 
     /// Returns a segment uuid string.
+    ///
+    /// It consists in 32 lowercase hexadecimal chars
+    /// (e.g. a5c4dfcbdfe645089129e308e26d5523)
     pub fn uuid_string(&self) -> String {
         self.0.to_simple_ref().to_string()
     }
+
+    /// Build a `SegmentId` string from the full uuid string.
+    ///
+    /// E.g. "a5c4dfcbdfe645089129e308e26d5523"
+    pub fn from_uuid_string(uuid_string: &str) -> Result<SegmentId, SegmentIdParseError> {
+        FromStr::from_str(uuid_string)
+    }
+}
+
+/// Error type used when parsing a `SegmentId` from a string fails.
+pub struct SegmentIdParseError(uuid::Error);
+
+impl Error for SegmentIdParseError {}
+
+impl fmt::Debug for SegmentIdParseError {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        self.0.fmt(f)
+    }
+}
+
+impl fmt::Display for SegmentIdParseError {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        self.0.fmt(f)
+    }
+}
+
+impl FromStr for SegmentId {
+    type Err = SegmentIdParseError;
+
+    fn from_str(uuid_string: &str) -> Result<Self, SegmentIdParseError> {
+        let uuid = Uuid::parse_str(uuid_string).map_err(SegmentIdParseError)?;
+        Ok(SegmentId(uuid))
+    }
 }
 
 impl fmt::Debug for SegmentId {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         write!(f, "Seg({:?})", self.short_uuid_string())
     }
 }
@@ -78,3 +119,18 @@ impl Ord for SegmentId {
         self.0.as_bytes().cmp(other.0.as_bytes())
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::SegmentId;
+
+    #[test]
+    fn test_to_uuid_string() {
+        let full_uuid = "a5c4dfcbdfe645089129e308e26d5523";
+        let segment_id = SegmentId::from_uuid_string(full_uuid).unwrap();
+        assert_eq!(segment_id.uuid_string(), full_uuid);
+        assert_eq!(segment_id.short_uuid_string(), "a5c4dfcb");
+        // one extra char
+        assert!(SegmentId::from_uuid_string("a5c4dfcbdfe645089129e308e26d5523b").is_err());
+    }
+}

src/core/segment_meta.rs

@@ -1,174 +0,0 @@
-use super::SegmentComponent;
-use census::{Inventory, TrackedObject};
-use core::SegmentId;
-use serde;
-use std::collections::HashSet;
-use std::fmt;
-use std::path::PathBuf;
-
-lazy_static! {
-    static ref INVENTORY: Inventory<InnerSegmentMeta> = { Inventory::new() };
-}
-
-#[derive(Clone, Debug, Serialize, Deserialize)]
-struct DeleteMeta {
-    num_deleted_docs: u32,
-    opstamp: u64,
-}
-
-/// `SegmentMeta` contains simple meta information about a segment.
-///
-/// For instance the number of docs it contains,
-/// how many are deleted, etc.
-#[derive(Clone)]
-pub struct SegmentMeta {
-    tracked: TrackedObject<InnerSegmentMeta>,
-}
-
-impl fmt::Debug for SegmentMeta {
-    fn fmt(&self, f: &mut fmt::Formatter) -> Result<(), fmt::Error> {
-        self.tracked.fmt(f)
-    }
-}
-
-impl serde::Serialize for SegmentMeta {
-    fn serialize<S>(
-        &self,
-        serializer: S,
-    ) -> Result<<S as serde::Serializer>::Ok, <S as serde::Serializer>::Error>
-    where
-        S: serde::Serializer,
-    {
-        self.tracked.serialize(serializer)
-    }
-}
-
-impl<'a> serde::Deserialize<'a> for SegmentMeta {
-    fn deserialize<D>(deserializer: D) -> Result<Self, <D as serde::Deserializer<'a>>::Error>
-    where
-        D: serde::Deserializer<'a>,
-    {
-        let inner = InnerSegmentMeta::deserialize(deserializer)?;
-        let tracked = INVENTORY.track(inner);
-        Ok(SegmentMeta { tracked })
-    }
-}
-
-impl SegmentMeta {
-    /// Lists all living `SegmentMeta` object at the time of the call.
-    pub fn all() -> Vec<SegmentMeta> {
-        INVENTORY
-            .list()
-            .into_iter()
-            .map(|inner| SegmentMeta { tracked: inner })
-            .collect::<Vec<_>>()
-    }
-
-    /// Creates a new `SegmentMeta` object.
-    #[doc(hidden)]
-    pub fn new(segment_id: SegmentId, max_doc: u32) -> SegmentMeta {
-        let inner = InnerSegmentMeta {
-            segment_id,
-            max_doc,
-            deletes: None,
-        };
-        SegmentMeta {
-            tracked: INVENTORY.track(inner),
-        }
-    }
-
-    /// Returns the segment id.
-    pub fn id(&self) -> SegmentId {
-        self.tracked.segment_id
-    }
-
-    /// Returns the number of deleted documents.
-    pub fn num_deleted_docs(&self) -> u32 {
-        self.tracked
-            .deletes
-            .as_ref()
-            .map(|delete_meta| delete_meta.num_deleted_docs)
-            .unwrap_or(0u32)
-    }
-
-    /// Returns the list of files that
-    /// are required for the segment meta.
-    ///
-    /// This is useful as the way tantivy removes files
-    /// is by removing all files that have been created by tantivy
-    /// and are not used by any segment anymore.
-    pub fn list_files(&self) -> HashSet<PathBuf> {
-        SegmentComponent::iterator()
-            .map(|component| self.relative_path(*component))
-            .collect::<HashSet<PathBuf>>()
-    }
-
-    /// Returns the relative path of a component of our segment.
-    ///
-    /// It just joins the segment id with the extension
-    /// associated to a segment component.
-    pub fn relative_path(&self, component: SegmentComponent) -> PathBuf {
-        let mut path = self.id().uuid_string();
-        path.push_str(&*match component {
-            SegmentComponent::POSTINGS => ".idx".to_string(),
-            SegmentComponent::POSITIONS => ".pos".to_string(),
-            SegmentComponent::POSITIONSSKIP => ".posidx".to_string(),
-            SegmentComponent::TERMS => ".term".to_string(),
-            SegmentComponent::STORE => ".store".to_string(),
-            SegmentComponent::FASTFIELDS => ".fast".to_string(),
-            SegmentComponent::FIELDNORMS => ".fieldnorm".to_string(),
-            SegmentComponent::DELETE => format!(".{}.del", self.delete_opstamp().unwrap_or(0)),
-        });
-        PathBuf::from(path)
-    }
-
-    /// Return the highest doc id + 1
-    ///
-    /// If there are no deletes, then num_docs = max_docs
-    /// and all the doc ids contains in this segment
-    /// are exactly (0..max_doc).
-    pub fn max_doc(&self) -> u32 {
-        self.tracked.max_doc
-    }
-
-    /// Return the number of documents in the segment.
-    pub fn num_docs(&self) -> u32 {
-        self.max_doc() - self.num_deleted_docs()
-    }
-
-    /// Returns the opstamp of the last delete operation
-    /// taken in account in this segment.
-    pub fn delete_opstamp(&self) -> Option<u64> {
-        self.tracked
-            .deletes
-            .as_ref()
-            .map(|delete_meta| delete_meta.opstamp)
-    }
-
-    /// Returns true iff the segment meta contains
-    /// delete information.
-    pub fn has_deletes(&self) -> bool {
-        self.num_deleted_docs() > 0
-    }
-
-    #[doc(hidden)]
-    pub fn with_delete_meta(self, num_deleted_docs: u32, opstamp: u64) -> SegmentMeta {
-        let delete_meta = DeleteMeta {
-            num_deleted_docs,
-            opstamp,
-        };
-        let tracked = self.tracked.map(move |inner_meta| InnerSegmentMeta {
-            segment_id: inner_meta.segment_id,
-            max_doc: inner_meta.max_doc,
-            deletes: Some(delete_meta),
-        });
-        SegmentMeta { tracked }
-    }
-}
-
-#[derive(Clone, Debug, Serialize, Deserialize)]
-struct InnerSegmentMeta {
-    segment_id: SegmentId,
-    max_doc: u32,
-    deletes: Option<DeleteMeta>,
-}

src/core/segment_reader.rs

@@ -1,30 +1,26 @@
-use common::CompositeFile;
-use common::HasLen;
-use core::InvertedIndexReader;
-use core::Segment;
-use core::SegmentComponent;
-use core::SegmentId;
-use directory::ReadOnlySource;
-use error::TantivyError;
-use fastfield::DeleteBitSet;
-use fastfield::FacetReader;
-use fastfield::FastFieldReader;
-use fastfield::{self, FastFieldNotAvailableError};
-use fastfield::{BytesFastFieldReader, FastValue, MultiValueIntFastFieldReader};
-use fieldnorm::FieldNormReader;
-use schema::Cardinality;
-use schema::Field;
-use schema::FieldType;
-use schema::Schema;
-use space_usage::SegmentSpaceUsage;
+use crate::common::CompositeFile;
+use crate::common::HasLen;
+use crate::core::InvertedIndexReader;
+use crate::core::Segment;
+use crate::core::SegmentComponent;
+use crate::core::SegmentId;
+use crate::directory::ReadOnlySource;
+use crate::fastfield::DeleteBitSet;
+use crate::fastfield::FacetReader;
+use crate::fastfield::FastFieldReaders;
+use crate::fieldnorm::FieldNormReader;
+use crate::schema::Field;
+use crate::schema::FieldType;
+use crate::schema::Schema;
+use crate::space_usage::SegmentSpaceUsage;
+use crate::store::StoreReader;
+use crate::termdict::TermDictionary;
+use crate::DocId;
+use fail::fail_point;
 use std::collections::HashMap;
 use std::fmt;
 use std::sync::Arc;
 use std::sync::RwLock;
-use store::StoreReader;
-use termdict::TermDictionary;
-use DocId;
-use Result;
 
 /// Entry point to access all of the datastructures of the `Segment`
 ///
@@ -51,7 +47,7 @@ pub struct SegmentReader {
     postings_composite: CompositeFile,
     positions_composite: CompositeFile,
     positions_idx_composite: CompositeFile,
-    fast_fields_composite: CompositeFile,
+    fast_fields_readers: Arc<FastFieldReaders>,
     fieldnorms_composite: CompositeFile,
 
     store_source: ReadOnlySource,
@@ -105,93 +101,21 @@ impl SegmentReader {
     ///
     /// # Panics
     /// May panic if the index is corrupted.
-    pub fn fast_field_reader<Item: FastValue>(
-        &self,
-        field: Field,
-    ) -> fastfield::Result<FastFieldReader<Item>> {
-        let field_entry = self.schema.get_field_entry(field);
-        if Item::fast_field_cardinality(field_entry.field_type()) == Some(Cardinality::SingleValue)
-        {
-            self.fast_fields_composite
-                .open_read(field)
-                .ok_or_else(|| FastFieldNotAvailableError::new(field_entry))
-                .map(FastFieldReader::open)
-        } else {
-            Err(FastFieldNotAvailableError::new(field_entry))
-        }
-    }
-
-    pub(crate) fn fast_field_reader_with_idx<Item: FastValue>(
-        &self,
-        field: Field,
-        idx: usize,
-    ) -> fastfield::Result<FastFieldReader<Item>> {
-        if let Some(ff_source) = self.fast_fields_composite.open_read_with_idx(field, idx) {
-            Ok(FastFieldReader::open(ff_source))
-        } else {
-            let field_entry = self.schema.get_field_entry(field);
-            Err(FastFieldNotAvailableError::new(field_entry))
-        }
-    }
-
-    /// Accessor to the `MultiValueIntFastFieldReader` associated to a given `Field`.
-    /// May panick if the field is not a multivalued fastfield of the type `Item`.
-    pub fn multi_fast_field_reader<Item: FastValue>(
-        &self,
-        field: Field,
-    ) -> fastfield::Result<MultiValueIntFastFieldReader<Item>> {
-        let field_entry = self.schema.get_field_entry(field);
-        if Item::fast_field_cardinality(field_entry.field_type()) == Some(Cardinality::MultiValues)
-        {
-            let idx_reader = self.fast_field_reader_with_idx(field, 0)?;
-            let vals_reader = self.fast_field_reader_with_idx(field, 1)?;
-            Ok(MultiValueIntFastFieldReader::open(idx_reader, vals_reader))
-        } else {
-            Err(FastFieldNotAvailableError::new(field_entry))
-        }
-    }
-
-    /// Accessor to the `BytesFastFieldReader` associated to a given `Field`.
-    pub fn bytes_fast_field_reader(&self, field: Field) -> fastfield::Result<BytesFastFieldReader> {
-        let field_entry = self.schema.get_field_entry(field);
-        match *field_entry.field_type() {
-            FieldType::Bytes => {}
-            _ => return Err(FastFieldNotAvailableError::new(field_entry)),
-        }
-        let idx_reader = self
-            .fast_fields_composite
-            .open_read_with_idx(field, 0)
-            .ok_or_else(|| FastFieldNotAvailableError::new(field_entry))
-            .map(FastFieldReader::open)?;
-        let values = self
-            .fast_fields_composite
-            .open_read_with_idx(field, 1)
-            .ok_or_else(|| FastFieldNotAvailableError::new(field_entry))?;
-        Ok(BytesFastFieldReader::open(idx_reader, values))
+    pub fn fast_fields(&self) -> &FastFieldReaders {
+        &self.fast_fields_readers
     }
 
     /// Accessor to the `FacetReader` associated to a given `Field`.
-    pub fn facet_reader(&self, field: Field) -> Result<FacetReader> {
+    pub fn facet_reader(&self, field: Field) -> Option<FacetReader> {
         let field_entry = self.schema.get_field_entry(field);
         if field_entry.field_type() != &FieldType::HierarchicalFacet {
-            return Err(TantivyError::InvalidArgument(format!(
-                "The field {:?} is not a \
-                 hierarchical facet.",
-                field_entry
-            )));
+            return None;
         }
-        let term_ords_reader = self.multi_fast_field_reader(field)?;
-        let termdict_source = self.termdict_composite.open_read(field).ok_or_else(|| {
-            TantivyError::InvalidArgument(format!(
-                "The field \"{}\" is a hierarchical \
-                 but this segment does not seem to have the field term \
-                 dictionary.",
-                field_entry.name()
-            ))
-        })?;
+        let term_ords_reader = self.fast_fields().u64s(field)?;
+        let termdict_source = self.termdict_composite.open_read(field)?;
         let termdict = TermDictionary::from_source(&termdict_source);
         let facet_reader = FacetReader::new(term_ords_reader, termdict);
-        Ok(facet_reader)
+        Some(facet_reader)
     }
 
     /// Accessor to the segment's `Field norms`'s reader.
@@ -220,7 +144,7 @@ impl SegmentReader {
     }
 
     /// Open a new segment for reading.
-    pub fn open(segment: &Segment) -> Result<SegmentReader> {
+    pub fn open(segment: &Segment) -> crate::Result<SegmentReader> {
         let termdict_source = segment.open_read(SegmentComponent::TERMS)?;
         let termdict_composite = CompositeFile::open(&termdict_source)?;
 
@@ -247,8 +171,12 @@ impl SegmentReader {
             }
         };
 
+        let schema = segment.schema();
+
         let fast_fields_data = segment.open_read(SegmentComponent::FASTFIELDS)?;
         let fast_fields_composite = CompositeFile::open(&fast_fields_data)?;
+        let fast_field_readers =
+            Arc::new(FastFieldReaders::load_all(&schema, &fast_fields_composite)?);
 
         let fieldnorms_data = segment.open_read(SegmentComponent::FIELDNORMS)?;
         let fieldnorms_composite = CompositeFile::open(&fieldnorms_data)?;
@@ -260,14 +188,13 @@ impl SegmentReader {
             None
         };
 
-        let schema = segment.schema();
         Ok(SegmentReader {
             inv_idx_reader_cache: Arc::new(RwLock::new(HashMap::new())),
             max_doc: segment.meta().max_doc(),
             num_docs: segment.meta().num_docs(),
             termdict_composite,
             postings_composite,
-            fast_fields_composite,
+            fast_fields_readers: fast_field_readers,
             fieldnorms_composite,
             segment_id: segment.id(),
             store_source,
@@ -316,10 +243,9 @@ impl SegmentReader {
 
         let postings_source = postings_source_opt.unwrap();
 
-        let termdict_source = self
-            .termdict_composite
-            .open_read(field)
-            .expect("Failed to open field term dictionary in composite file. Is the field indexed");
+        let termdict_source = self.termdict_composite.open_read(field).expect(
+            "Failed to open field term dictionary in composite file. Is the field indexed?",
+        );
 
         let positions_source = self
             .positions_composite
@@ -369,7 +295,7 @@ impl SegmentReader {
     }
 
     /// Returns an iterator that will iterate over the alive document ids
-    pub fn doc_ids_alive(&self) -> SegmentReaderAliveDocsIterator {
+    pub fn doc_ids_alive(&self) -> SegmentReaderAliveDocsIterator<'_> {
         SegmentReaderAliveDocsIterator::new(&self)
     }
 
@@ -381,19 +307,19 @@ impl SegmentReader {
             self.postings_composite.space_usage(),
             self.positions_composite.space_usage(),
             self.positions_idx_composite.space_usage(),
-            self.fast_fields_composite.space_usage(),
+            self.fast_fields_readers.space_usage(),
             self.fieldnorms_composite.space_usage(),
             self.get_store_reader().space_usage(),
             self.delete_bitset_opt
                 .as_ref()
-                .map(|x| x.space_usage())
+                .map(DeleteBitSet::space_usage)
                 .unwrap_or(0),
         )
     }
 }
 
 impl fmt::Debug for SegmentReader {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         write!(f, "SegmentReader({:?})", self.segment_id)
     }
 }
@@ -446,9 +372,9 @@ impl<'a> Iterator for SegmentReaderAliveDocsIterator<'a> {
 
 #[cfg(test)]
 mod test {
-    use core::Index;
-    use schema::{Schema, Term, STORED, TEXT};
-    use DocId;
+    use crate::core::Index;
+    use crate::schema::{Schema, Term, STORED, TEXT};
+    use crate::DocId;
 
     #[test]
     fn test_alive_docs_iterator() {
@@ -477,9 +403,7 @@ mod test {
             // ok, now we should have a deleted doc
             index_writer2.commit().unwrap();
         }
-
-        index.load_searchers().unwrap();
-        let searcher = index.searcher();
+        let searcher = index.reader().unwrap().searcher();
         let docs: Vec<DocId> = searcher.segment_reader(0).doc_ids_alive().collect();
         assert_eq!(vec![0u32, 2u32], docs);
     }

src/directory/directory.rs

@@ -1,11 +1,104 @@
-use directory::error::{DeleteError, OpenReadError, OpenWriteError};
-use directory::{ReadOnlySource, WritePtr};
+use crate::directory::directory_lock::Lock;
+use crate::directory::error::LockError;
+use crate::directory::error::{DeleteError, OpenReadError, OpenWriteError};
+use crate::directory::WatchCallback;
+use crate::directory::WatchHandle;
+use crate::directory::{ReadOnlySource, WritePtr};
 use std::fmt;
 use std::io;
+use std::io::Write;
 use std::marker::Send;
 use std::marker::Sync;
 use std::path::Path;
+use std::path::PathBuf;
 use std::result;
+use std::thread;
+use std::time::Duration;
+
+/// Retry the logic of acquiring locks is pretty simple.
+/// We just retry `n` times after a given `duratio`, both
+/// depending on the type of lock.
+struct RetryPolicy {
+    num_retries: usize,
+    wait_in_ms: u64,
+}
+
+impl RetryPolicy {
+    fn no_retry() -> RetryPolicy {
+        RetryPolicy {
+            num_retries: 0,
+            wait_in_ms: 0,
+        }
+    }
+
+    fn wait_and_retry(&mut self) -> bool {
+        if self.num_retries == 0 {
+            false
+        } else {
+            self.num_retries -= 1;
+            let wait_duration = Duration::from_millis(self.wait_in_ms);
+            thread::sleep(wait_duration);
+            true
+        }
+    }
+}
+
+/// The `DirectoryLock` is an object that represents a file lock.
+/// See  [`LockType`](struct.LockType.html)
+///
+/// It is transparently associated to a lock file, that gets deleted
+/// on `Drop.` The lock is released automatically on `Drop`.
+pub struct DirectoryLock(Box<dyn Send + Sync + 'static>);
+
+struct DirectoryLockGuard {
+    directory: Box<dyn Directory>,
+    path: PathBuf,
+}
+
+impl<T: Send + Sync + 'static> From<Box<T>> for DirectoryLock {
+    fn from(underlying: Box<T>) -> Self {
+        DirectoryLock(underlying)
+    }
+}
+
+impl Drop for DirectoryLockGuard {
+    fn drop(&mut self) {
+        if let Err(e) = self.directory.delete(&*self.path) {
+            error!("Failed to remove the lock file. {:?}", e);
+        }
+    }
+}
+
+enum TryAcquireLockError {
+    FileExists,
+    IOError(io::Error),
+}
+
+fn try_acquire_lock(
+    filepath: &Path,
+    directory: &mut dyn Directory,
+) -> Result<DirectoryLock, TryAcquireLockError> {
+    let mut write = directory.open_write(filepath).map_err(|e| match e {
+        OpenWriteError::FileAlreadyExists(_) => TryAcquireLockError::FileExists,
+        OpenWriteError::IOError(io_error) => TryAcquireLockError::IOError(io_error.into()),
+    })?;
+    write.flush().map_err(TryAcquireLockError::IOError)?;
+    Ok(DirectoryLock::from(Box::new(DirectoryLockGuard {
+        directory: directory.box_clone(),
+        path: filepath.to_owned(),
+    })))
+}
+
+fn retry_policy(is_blocking: bool) -> RetryPolicy {
+    if is_blocking {
+        RetryPolicy {
+            num_retries: 100,
+            wait_in_ms: 100,
+        }
+    } else {
+        RetryPolicy::no_retry()
+    }
+}
 
 /// Write-once read many (WORM) abstraction for where
 /// tantivy's data should be stored.
@@ -25,6 +118,8 @@ pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
     ///
     /// Specifically, subsequent writes or flushes should
     /// have no effect on the returned `ReadOnlySource` object.
+    ///
+    /// You should only use this to read files create with [Directory::open_write].
     fn open_read(&self, path: &Path) -> result::Result<ReadOnlySource, OpenReadError>;
 
     /// Removes a file
@@ -64,6 +159,8 @@ pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
     /// atomic_write.
     ///
     /// This should only be used for small files.
+    ///
+    /// You should only use this to read files create with [Directory::atomic_write].
     fn atomic_read(&self, path: &Path) -> Result<Vec<u8>, OpenReadError>;
 
     /// Atomically replace the content of a file with data.
@@ -73,19 +170,58 @@ pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
     ///
     /// The file may or may not previously exist.
     fn atomic_write(&mut self, path: &Path, data: &[u8]) -> io::Result<()>;
+
+    /// Acquire a lock in the given directory.
+    ///
+    /// The method is blocking or not depending on the `Lock` object.
+    fn acquire_lock(&self, lock: &Lock) -> Result<DirectoryLock, LockError> {
+        let mut box_directory = self.box_clone();
+        let mut retry_policy = retry_policy(lock.is_blocking);
+        loop {
+            match try_acquire_lock(&lock.filepath, &mut *box_directory) {
+                Ok(result) => {
+                    return Ok(result);
+                }
+                Err(TryAcquireLockError::FileExists) => {
+                    if !retry_policy.wait_and_retry() {
+                        return Err(LockError::LockBusy);
+                    }
+                }
+                Err(TryAcquireLockError::IOError(io_error)) => {
+                    return Err(LockError::IOError(io_error));
+                }
+            }
+        }
+    }
+
+    /// Registers a callback that will be called whenever a change on the `meta.json`
+    /// using the `atomic_write` API is detected.
+    ///
+    /// The behavior when using `.watch()` on a file using [Directory::open_write] is, on the other
+    /// hand, undefined.
+    ///
+    /// The file will be watched for the lifetime of the returned `WatchHandle`. The caller is
+    /// required to keep it.
+    /// It does not override previous callbacks. When the file is modified, all callback that are
+    /// registered (and whose `WatchHandle` is still alive) are triggered.
+    ///
+    /// Internally, tantivy only uses this API to detect new commits to implement the
+    /// `OnCommit` `ReloadPolicy`. Not implementing watch in a `Directory` only prevents the
+    /// `OnCommit` `ReloadPolicy` to work properly.
+    fn watch(&self, watch_callback: WatchCallback) -> crate::Result<WatchHandle>;
 }
 
 /// DirectoryClone
 pub trait DirectoryClone {
     /// Clones the directory and boxes the clone
-    fn box_clone(&self) -> Box<Directory>;
+    fn box_clone(&self) -> Box<dyn Directory>;
 }
 
 impl<T> DirectoryClone for T
 where
     T: 'static + Directory + Clone,
 {
-    fn box_clone(&self) -> Box<Directory> {
+    fn box_clone(&self) -> Box<dyn Directory> {
         Box::new(self.clone())
     }
 }

src/directory/directory_lock.rs

@@ -0,0 +1,55 @@
+use once_cell::sync::Lazy;
+use std::path::PathBuf;
+
+/// A directory lock.
+///
+/// A lock is associated to a specific path and some
+/// [`LockParams`](./enum.LockParams.html).
+/// Tantivy itself uses only two locks but client application
+/// can use the directory facility to define their own locks.
+/// - [INDEX_WRITER_LOCK](./struct.INDEX_WRITER_LOCK.html)
+/// - [META_LOCK](./struct.META_LOCK.html)
+///
+/// Check out these locks documentation for more information.
+///
+#[derive(Debug)]
+pub struct Lock {
+    /// The lock needs to be associated with its own file `path`.
+    /// Depending on the platform, the lock might rely on the creation
+    /// and deletion of this filepath.
+    pub filepath: PathBuf,
+    /// `lock_params` describes whether acquiring the lock is meant
+    /// to be a blocking operation or a non-blocking.
+    ///
+    /// Acquiring a blocking lock blocks until the lock is
+    /// available.
+    /// Acquiring a blocking lock returns rapidly, either successfully
+    /// or with an error signifying that someone is already holding
+    /// the lock.
+    pub is_blocking: bool,
+}
+
+/// Only one process should be able to write tantivy's index at a time.
+/// This lock file, when present, is in charge of preventing other processes to open an IndexWriter.
+///
+/// If the process is killed and this file remains, it is safe to remove it manually.
+///
+/// Failing to acquire this lock usually means a misuse of tantivy's API,
+/// (creating more than one instance of the `IndexWriter`), are a spurious
+/// lock file remaining after a crash. In the latter case, removing the file after
+/// checking no process running tantivy is running is safe.
+pub static INDEX_WRITER_LOCK: Lazy<Lock> = Lazy::new(|| Lock {
+    filepath: PathBuf::from(".tantivy-writer.lock"),
+    is_blocking: false,
+});
+/// The meta lock file is here to protect the segment files being opened by
+/// `IndexReader::reload()` from being garbage collected.
+/// It makes it possible for another process to safely consume
+/// our index in-writing. Ideally, we may have prefered `RWLock` semantics
+/// here, but it is difficult to achieve on Windows.
+///
+/// Opening segment readers is a very fast process.
+pub static META_LOCK: Lazy<Lock> = Lazy::new(|| Lock {
+    filepath: PathBuf::from(".tantivy-meta.lock"),
+    is_blocking: true,
+});

src/directory/error.rs

@@ -1,8 +1,25 @@
+use crate::Version;
 use std::error::Error as StdError;
 use std::fmt;
 use std::io;
 use std::path::PathBuf;
 
+/// Error while trying to acquire a directory lock.
+#[derive(Debug, Fail)]
+pub enum LockError {
+    /// Failed to acquired a lock as it is already held by another
+    /// client.
+    /// - In the context of a blocking lock, this means the lock was not released within some `timeout` period.
+    /// - In the context of a non-blocking lock, this means the lock was busy at the moment of the call.
+    #[fail(
+        display = "Could not acquire lock as it is already held, possibly by a different process."
+    )]
+    LockBusy,
+    /// Trying to acquire a lock failed with an `IOError`
+    #[fail(display = "Failed to acquire the lock due to an io:Error.")]
+    IOError(io::Error),
+}
+
 /// General IO error with an optional path to the offending file.
 #[derive(Debug)]
 pub struct IOError {
@@ -10,8 +27,14 @@ pub struct IOError {
     err: io::Error,
 }
 
+impl Into<io::Error> for IOError {
+    fn into(self) -> io::Error {
+        self.err
+    }
+}
+
 impl fmt::Display for IOError {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         match self.path {
             Some(ref path) => write!(f, "io error occurred on path '{:?}': '{}'", path, self.err),
             None => write!(f, "io error occurred: '{}'", self.err),
@@ -24,7 +47,7 @@ impl StdError for IOError {
         "io error occurred"
     }
 
-    fn cause(&self) -> Option<&StdError> {
+    fn cause(&self) -> Option<&dyn StdError> {
         Some(&self.err)
     }
 }
@@ -51,10 +74,18 @@ pub enum OpenDirectoryError {
     DoesNotExist(PathBuf),
     /// The path exists but is not a directory.
     NotADirectory(PathBuf),
+    /// IoError
+    IoError(io::Error),
+}
+
+impl From<io::Error> for OpenDirectoryError {
+    fn from(io_err: io::Error) -> Self {
+        OpenDirectoryError::IoError(io_err)
+    }
 }
 
 impl fmt::Display for OpenDirectoryError {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         match *self {
             OpenDirectoryError::DoesNotExist(ref path) => {
                 write!(f, "the underlying directory '{:?}' does not exist", path)
@@ -62,6 +93,11 @@ impl fmt::Display for OpenDirectoryError {
             OpenDirectoryError::NotADirectory(ref path) => {
                 write!(f, "the path '{:?}' exists but is not a directory", path)
             }
+            OpenDirectoryError::IoError(ref err) => write!(
+                f,
+                "IOError while trying to open/create the directory. {:?}",
+                err
+            ),
         }
     }
 }
@@ -71,7 +107,7 @@ impl StdError for OpenDirectoryError {
         "error occurred while opening a directory"
     }
 
-    fn cause(&self) -> Option<&StdError> {
+    fn cause(&self) -> Option<&dyn StdError> {
         None
     }
 }
@@ -94,7 +130,7 @@ impl From<IOError> for OpenWriteError {
 }
 
 impl fmt::Display for OpenWriteError {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         match *self {
             OpenWriteError::FileAlreadyExists(ref path) => {
                 write!(f, "the file '{:?}' already exists", path)
@@ -113,7 +149,7 @@ impl StdError for OpenWriteError {
         "error occurred while opening a file for writing"
     }
 
-    fn cause(&self) -> Option<&StdError> {
+    fn cause(&self) -> Option<&dyn StdError> {
         match *self {
             OpenWriteError::FileAlreadyExists(_) => None,
             OpenWriteError::IOError(ref err) => Some(err),
@@ -121,6 +157,65 @@ impl StdError for OpenWriteError {
     }
 }
 
+/// Type of index incompatibility between the library and the index found on disk
+/// Used to catch and provide a hint to solve this incompatibility issue
+pub enum Incompatibility {
+    /// This library cannot decompress the index found on disk
+    CompressionMismatch {
+        /// Compression algorithm used by the current version of tantivy
+        library_compression_format: String,
+        /// Compression algorithm that was used to serialise the index
+        index_compression_format: String,
+    },
+    /// The index format found on disk isn't supported by this version of the library
+    IndexMismatch {
+        /// Version used by the library
+        library_version: Version,
+        /// Version the index was built with
+        index_version: Version,
+    },
+}
+
+impl fmt::Debug for Incompatibility {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
+        match self {
+            Incompatibility::CompressionMismatch {
+                library_compression_format,
+                index_compression_format,
+            } => {
+                let err = format!(
+                    "Library was compiled with {:?} compression, index was compressed with {:?}",
+                    library_compression_format, index_compression_format
+                );
+                let advice = format!(
+                    "Change the feature flag to {:?} and rebuild the library",
+                    index_compression_format
+                );
+                write!(f, "{}. {}", err, advice)?;
+            }
+            Incompatibility::IndexMismatch {
+                library_version,
+                index_version,
+            } => {
+                let err = format!(
+                    "Library version: {}, index version: {}",
+                    library_version.index_format_version, index_version.index_format_version
+                );
+                // TODO make a more useful error message
+                // include the version range that supports this index_format_version
+                let advice = format!(
+                    "Change tantivy to a version compatible with index format {} (e.g. {}.{}.x) \
+                     and rebuild your project.",
+                    index_version.index_format_version, index_version.major, index_version.minor
+                );
+                write!(f, "{}. {}", err, advice)?;
+            }
+        }
+
+        Ok(())
+    }
+}
+
 /// Error that may occur when accessing a file read
 #[derive(Debug)]
 pub enum OpenReadError {
@@ -129,6 +224,8 @@ pub enum OpenReadError {
     /// Any kind of IO error that happens when
     /// interacting with the underlying IO device.
     IOError(IOError),
+    /// This library doesn't support the index version found on disk
+    IncompatibleIndex(Incompatibility),
 }
 
 impl From<IOError> for OpenReadError {
@@ -138,7 +235,7 @@ impl From<IOError> for OpenReadError {
 }
 
 impl fmt::Display for OpenReadError {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         match *self {
             OpenReadError::FileDoesNotExist(ref path) => {
                 write!(f, "the file '{:?}' does not exist", path)
@@ -148,19 +245,9 @@ impl fmt::Display for OpenReadError {
                 "an io error occurred while opening a file for reading: '{}'",
                 err
             ),
-        }
-    }
-}
-
-impl StdError for OpenReadError {
-    fn description(&self) -> &str {
-        "error occurred while opening a file for reading"
-    }
-
-    fn cause(&self) -> Option<&StdError> {
-        match *self {
-            OpenReadError::FileDoesNotExist(_) => None,
-            OpenReadError::IOError(ref err) => Some(err),
+            OpenReadError::IncompatibleIndex(ref footer) => {
+                write!(f, "Incompatible index format: {:?}", footer)
+            }
         }
     }
 }
@@ -181,8 +268,14 @@ impl From<IOError> for DeleteError {
     }
 }
 
+impl From<Incompatibility> for OpenReadError {
+    fn from(incompatibility: Incompatibility) -> Self {
+        OpenReadError::IncompatibleIndex(incompatibility)
+    }
+}
+
 impl fmt::Display for DeleteError {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         match *self {
             DeleteError::FileDoesNotExist(ref path) => {
                 write!(f, "the file '{:?}' does not exist", path)
@@ -199,7 +292,7 @@ impl StdError for DeleteError {
         "error occurred while deleting a file"
     }
 
-    fn cause(&self) -> Option<&StdError> {
+    fn cause(&self) -> Option<&dyn StdError> {
         match *self {
             DeleteError::FileDoesNotExist(_) => None,
             DeleteError::IOError(ref err) => Some(err),

src/directory/footer.rs

@@ -0,0 +1,369 @@
+use crate::common::{BinarySerializable, CountingWriter, FixedSize, VInt};
+use crate::directory::error::Incompatibility;
+use crate::directory::read_only_source::ReadOnlySource;
+use crate::directory::{AntiCallToken, TerminatingWrite};
+use crate::Version;
+use byteorder::{ByteOrder, LittleEndian, WriteBytesExt};
+use crc32fast::Hasher;
+use std::io;
+use std::io::Write;
+
+const FOOTER_MAX_LEN: usize = 10_000;
+
+type CrcHashU32 = u32;
+
+#[derive(Debug, Clone, PartialEq)]
+pub struct Footer {
+    pub version: Version,
+    pub meta: String,
+    pub versioned_footer: VersionedFooter,
+}
+
+/// Serialises the footer to a byte-array
+/// - versioned_footer_len : 4 bytes
+///-  versioned_footer: variable bytes
+/// - meta_len: 4 bytes
+/// - meta: variable bytes
+/// - version_len: 4 bytes
+/// - version json: variable bytes
+impl BinarySerializable for Footer {
+    fn serialize<W: io::Write>(&self, writer: &mut W) -> io::Result<()> {
+        BinarySerializable::serialize(&self.versioned_footer, writer)?;
+        BinarySerializable::serialize(&self.meta, writer)?;
+        let version_string =
+            serde_json::to_string(&self.version).map_err(|_err| io::ErrorKind::InvalidInput)?;
+        BinarySerializable::serialize(&version_string, writer)?;
+        Ok(())
+    }
+
+    fn deserialize<R: io::Read>(reader: &mut R) -> io::Result<Self> {
+        let versioned_footer = VersionedFooter::deserialize(reader)?;
+        let meta = String::deserialize(reader)?;
+        let version_json = String::deserialize(reader)?;
+        let version = serde_json::from_str(&version_json)?;
+        Ok(Footer {
+            version,
+            meta,
+            versioned_footer,
+        })
+    }
+}
+
+impl Footer {
+    pub fn new(versioned_footer: VersionedFooter) -> Self {
+        let version = crate::VERSION.clone();
+        let meta = version.to_string();
+        Footer {
+            version,
+            meta,
+            versioned_footer,
+        }
+    }
+
+    pub fn append_footer<W: io::Write>(&self, mut write: &mut W) -> io::Result<()> {
+        let mut counting_write = CountingWriter::wrap(&mut write);
+        self.serialize(&mut counting_write)?;
+        let written_len = counting_write.written_bytes();
+        write.write_u32::<LittleEndian>(written_len as u32)?;
+        Ok(())
+    }
+
+    pub fn extract_footer(source: ReadOnlySource) -> Result<(Footer, ReadOnlySource), io::Error> {
+        if source.len() < 4 {
+            return Err(io::Error::new(
+                io::ErrorKind::UnexpectedEof,
+                format!(
+                    "File corrupted. The file is smaller than 4 bytes (len={}).",
+                    source.len()
+                ),
+            ));
+        }
+        let (body_footer, footer_len_bytes) = source.split_from_end(u32::SIZE_IN_BYTES);
+        let footer_len = LittleEndian::read_u32(footer_len_bytes.as_slice()) as usize;
+        let body_len = body_footer.len() - footer_len;
+        let (body, footer_data) = body_footer.split(body_len);
+        let mut cursor = footer_data.as_slice();
+        let footer = Footer::deserialize(&mut cursor)?;
+        Ok((footer, body))
+    }
+
+    /// Confirms that the index will be read correctly by this version of tantivy
+    /// Has to be called after `extract_footer` to make sure it's not accessing uninitialised memory
+    pub fn is_compatible(&self) -> Result<(), Incompatibility> {
+        let library_version = crate::version();
+        match &self.versioned_footer {
+            VersionedFooter::V1 {
+                crc32: _crc,
+                store_compression: compression,
+            } => {
+                if &library_version.store_compression != compression {
+                    return Err(Incompatibility::CompressionMismatch {
+                        library_compression_format: library_version.store_compression.to_string(),
+                        index_compression_format: compression.to_string(),
+                    });
+                }
+                Ok(())
+            }
+            VersionedFooter::UnknownVersion => Err(Incompatibility::IndexMismatch {
+                library_version: library_version.clone(),
+                index_version: self.version.clone(),
+            }),
+        }
+    }
+}
+
+/// Footer that includes a crc32 hash that enables us to checksum files in the index
+#[derive(Debug, Clone, PartialEq)]
+pub enum VersionedFooter {
+    UnknownVersion,
+    V1 {
+        crc32: CrcHashU32,
+        store_compression: String,
+    },
+}
+
+impl BinarySerializable for VersionedFooter {
+    fn serialize<W: io::Write>(&self, writer: &mut W) -> io::Result<()> {
+        let mut buf = Vec::new();
+        match self {
+            VersionedFooter::V1 {
+                crc32,
+                store_compression: compression,
+            } => {
+                // Serializes a valid `VersionedFooter` or panics if the version is unknown
+                // [   version    |   crc_hash  | compression_mode ]
+                // [    0..4      |     4..8    |     variable     ]
+                BinarySerializable::serialize(&1u32, &mut buf)?;
+                BinarySerializable::serialize(crc32, &mut buf)?;
+                BinarySerializable::serialize(compression, &mut buf)?;
+            }
+            VersionedFooter::UnknownVersion => {
+                return Err(io::Error::new(
+                    io::ErrorKind::InvalidInput,
+                    "Cannot serialize an unknown versioned footer ",
+                ));
+            }
+        }
+        BinarySerializable::serialize(&VInt(buf.len() as u64), writer)?;
+        assert!(buf.len() <= FOOTER_MAX_LEN);
+        writer.write_all(&buf[..])?;
+        Ok(())
+    }
+
+    fn deserialize<R: io::Read>(reader: &mut R) -> io::Result<Self> {
+        let len = VInt::deserialize(reader)?.0 as usize;
+        if len > FOOTER_MAX_LEN {
+            return Err(io::Error::new(
+                io::ErrorKind::InvalidData,
+                format!(
+                    "Footer seems invalid as it suggests a footer len of {}. File is corrupted, \
+            or the index was created with a different & old version of tantivy.",
+                    len
+                ),
+            ));
+        }
+        let mut buf = vec![0u8; len];
+        reader.read_exact(&mut buf[..])?;
+        let mut cursor = &buf[..];
+        let version = u32::deserialize(&mut cursor)?;
+        if version == 1 {
+            let crc32 = u32::deserialize(&mut cursor)?;
+            let compression = String::deserialize(&mut cursor)?;
+            Ok(VersionedFooter::V1 {
+                crc32,
+                store_compression: compression,
+            })
+        } else {
+            Ok(VersionedFooter::UnknownVersion)
+        }
+    }
+}
+
+impl VersionedFooter {
+    pub fn crc(&self) -> Option<CrcHashU32> {
+        match self {
+            VersionedFooter::V1 { crc32, .. } => Some(*crc32),
+            VersionedFooter::UnknownVersion { .. } => None,
+        }
+    }
+}
+
+pub(crate) struct FooterProxy<W: TerminatingWrite> {
+    /// always Some except after terminate call
+    hasher: Option<Hasher>,
+    /// always Some except after terminate call
+    writer: Option<W>,
+}
+
+impl<W: TerminatingWrite> FooterProxy<W> {
+    pub fn new(writer: W) -> Self {
+        FooterProxy {
+            hasher: Some(Hasher::new()),
+            writer: Some(writer),
+        }
+    }
+}
+
+impl<W: TerminatingWrite> Write for FooterProxy<W> {
+    fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+        let count = self.writer.as_mut().unwrap().write(buf)?;
+        self.hasher.as_mut().unwrap().update(&buf[..count]);
+        Ok(count)
+    }
+
+    fn flush(&mut self) -> io::Result<()> {
+        self.writer.as_mut().unwrap().flush()
+    }
+}
+
+impl<W: TerminatingWrite> TerminatingWrite for FooterProxy<W> {
+    fn terminate_ref(&mut self, _: AntiCallToken) -> io::Result<()> {
+        let crc32 = self.hasher.take().unwrap().finalize();
+        let footer = Footer::new(VersionedFooter::V1 {
+            crc32,
+            store_compression: crate::store::COMPRESSION.to_string(),
+        });
+        let mut writer = self.writer.take().unwrap();
+        footer.append_footer(&mut writer)?;
+        writer.terminate()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+
+    use super::CrcHashU32;
+    use super::FooterProxy;
+    use crate::common::{BinarySerializable, VInt};
+    use crate::directory::footer::{Footer, VersionedFooter};
+    use crate::directory::TerminatingWrite;
+    use byteorder::{ByteOrder, LittleEndian};
+    use regex::Regex;
+    use std::io;
+
+    #[test]
+    fn test_versioned_footer() {
+        let mut vec = Vec::new();
+        let footer_proxy = FooterProxy::new(&mut vec);
+        assert!(footer_proxy.terminate().is_ok());
+        assert_eq!(vec.len(), 167);
+        let footer = Footer::deserialize(&mut &vec[..]).unwrap();
+        if let VersionedFooter::V1 {
+            crc32: _,
+            store_compression,
+        } = footer.versioned_footer
+        {
+            assert_eq!(store_compression, crate::store::COMPRESSION);
+        } else {
+            panic!("Versioned footer should be V1.");
+        }
+        assert_eq!(&footer.version, crate::version());
+    }
+
+    #[test]
+    fn test_serialize_deserialize_footer() {
+        let mut buffer = Vec::new();
+        let crc32 = 123456u32;
+        let footer: Footer = Footer::new(VersionedFooter::V1 {
+            crc32,
+            store_compression: "lz4".to_string(),
+        });
+        footer.serialize(&mut buffer).unwrap();
+        let footer_deser = Footer::deserialize(&mut &buffer[..]).unwrap();
+        assert_eq!(footer_deser, footer);
+    }
+
+    #[test]
+    fn footer_length() {
+        let crc32 = 1111111u32;
+        let versioned_footer = VersionedFooter::V1 {
+            crc32,
+            store_compression: "lz4".to_string(),
+        };
+        let mut buf = Vec::new();
+        versioned_footer.serialize(&mut buf).unwrap();
+        assert_eq!(buf.len(), 13);
+        let footer = Footer::new(versioned_footer);
+        let regex_ptn = Regex::new(
+            "tantivy v[0-9]{1,3}\\.[0-9]{1,3}\\.[0-9]{1,3}\\.{0,10}, index_format v[0-9]{1,5}",
+        )
+        .unwrap();
+        assert!(regex_ptn.is_match(&footer.meta));
+    }
+
+    #[test]
+    fn versioned_footer_from_bytes() {
+        let v_footer_bytes = vec![
+            // versionned footer length
+            12 | 128,
+            // index format version
+            1,
+            0,
+            0,
+            0,
+            // crc 32
+            12,
+            35,
+            89,
+            18,
+            // compression format
+            3 | 128,
+            b'l',
+            b'z',
+            b'4',
+        ];
+        let mut cursor = &v_footer_bytes[..];
+        let versioned_footer = VersionedFooter::deserialize(&mut cursor).unwrap();
+        assert!(cursor.is_empty());
+        let expected_crc: u32 = LittleEndian::read_u32(&v_footer_bytes[5..9]) as CrcHashU32;
+        let expected_versioned_footer: VersionedFooter = VersionedFooter::V1 {
+            crc32: expected_crc,
+            store_compression: "lz4".to_string(),
+        };
+        assert_eq!(versioned_footer, expected_versioned_footer);
+        let mut buffer = Vec::new();
+        assert!(versioned_footer.serialize(&mut buffer).is_ok());
+        assert_eq!(&v_footer_bytes[..], &buffer[..]);
+    }
+
+    #[test]
+    fn versioned_footer_panic() {
+        let v_footer_bytes = vec![6u8 | 128u8, 3u8, 0u8, 0u8, 1u8, 0u8, 0u8];
+        let mut b = &v_footer_bytes[..];
+        let versioned_footer = VersionedFooter::deserialize(&mut b).unwrap();
+        assert!(b.is_empty());
+        let expected_versioned_footer = VersionedFooter::UnknownVersion;
+        assert_eq!(versioned_footer, expected_versioned_footer);
+        let mut buf = Vec::new();
+        assert!(versioned_footer.serialize(&mut buf).is_err());
+    }
+
+    #[test]
+    #[cfg(not(feature = "lz4"))]
+    fn compression_mismatch() {
+        let crc32 = 1111111u32;
+        let versioned_footer = VersionedFooter::V1 {
+            crc32,
+            store_compression: "lz4".to_string(),
+        };
+        let footer = Footer::new(versioned_footer);
+        let res = footer.is_compatible();
+        assert!(res.is_err());
+    }
+
+    #[test]
+    fn test_deserialize_too_large_footer() {
+        let mut buf = vec![];
+        assert!(FooterProxy::new(&mut buf).terminate().is_ok());
+        let mut long_len_buf = [0u8; 10];
+        let num_bytes = VInt(super::FOOTER_MAX_LEN as u64 + 1u64).serialize_into(&mut long_len_buf);
+        buf[0..num_bytes].copy_from_slice(&long_len_buf[..num_bytes]);
+        let err = Footer::deserialize(&mut &buf[..]).unwrap_err();
+        assert_eq!(err.kind(), io::ErrorKind::InvalidData);
+        assert_eq!(
+            err.to_string(),
+            "Footer seems invalid as it suggests a footer len of 10001. File is corrupted, \
+            or the index was created with a different & old version of tantivy."
+        );
+    }
+}

src/directory/managed_directory.rs

@@ -1,9 +1,16 @@
-use core::MANAGED_FILEPATH;
-use directory::error::{DeleteError, IOError, OpenReadError, OpenWriteError};
-use directory::{ReadOnlySource, WritePtr};
-use error::TantivyError;
-use indexer::LockType;
-use serde_json;
+use crate::core::MANAGED_FILEPATH;
+use crate::directory::error::{DeleteError, IOError, LockError, OpenReadError, OpenWriteError};
+use crate::directory::footer::{Footer, FooterProxy};
+use crate::directory::DirectoryLock;
+use crate::directory::GarbageCollectionResult;
+use crate::directory::Lock;
+use crate::directory::META_LOCK;
+use crate::directory::{ReadOnlySource, WritePtr};
+use crate::directory::{WatchCallback, WatchHandle};
+use crate::error::DataCorruption;
+use crate::Directory;
+
+use crc32fast::Hasher;
 use std::collections::HashSet;
 use std::io;
 use std::io::Write;
@@ -11,8 +18,6 @@ use std::path::{Path, PathBuf};
 use std::result;
 use std::sync::RwLockWriteGuard;
 use std::sync::{Arc, RwLock};
-use Directory;
-use Result;
 
 /// Returns true iff the file is "managed".
 /// Non-managed file are not subject to garbage collection.
@@ -36,7 +41,7 @@ fn is_managed(path: &Path) -> bool {
 /// useful anymore.
 #[derive(Debug)]
 pub struct ManagedDirectory {
-    directory: Box<Directory>,
+    directory: Box<dyn Directory>,
     meta_informations: Arc<RwLock<MetaInformation>>,
 }
 
@@ -48,8 +53,8 @@ struct MetaInformation {
 /// Saves the file containing the list of existing files
 /// that were created by tantivy.
 fn save_managed_paths(
-    directory: &mut Directory,
-    wlock: &RwLockWriteGuard<MetaInformation>,
+    directory: &mut dyn Directory,
+    wlock: &RwLockWriteGuard<'_, MetaInformation>,
 ) -> io::Result<()> {
     let mut w = serde_json::to_vec(&wlock.managed_paths)?;
     writeln!(&mut w)?;
@@ -59,12 +64,17 @@ fn save_managed_paths(
 
 impl ManagedDirectory {
     /// Wraps a directory as managed directory.
-    pub fn new<Dir: Directory>(directory: Dir) -> Result<ManagedDirectory> {
+    pub fn wrap<Dir: Directory>(directory: Dir) -> crate::Result<ManagedDirectory> {
         match directory.atomic_read(&MANAGED_FILEPATH) {
             Ok(data) => {
                 let managed_files_json = String::from_utf8_lossy(&data);
                 let managed_files: HashSet<PathBuf> = serde_json::from_str(&managed_files_json)
-                    .map_err(|_| TantivyError::CorruptedFile(MANAGED_FILEPATH.clone()))?;
+                    .map_err(|e| {
+                        DataCorruption::new(
+                            MANAGED_FILEPATH.to_path_buf(),
+                            format!("Managed file cannot be deserialized: {:?}. ", e),
+                        )
+                    })?;
                 Ok(ManagedDirectory {
                     directory: Box::new(directory),
                     meta_informations: Arc::new(RwLock::new(MetaInformation {
@@ -77,6 +87,11 @@ impl ManagedDirectory {
                 meta_informations: Arc::default(),
             }),
             Err(OpenReadError::IOError(e)) => Err(From::from(e)),
+            Err(OpenReadError::IncompatibleIndex(incompatibility)) => {
+                // For the moment, this should never happen  `meta.json`
+                // do not have any footer and cannot detect incompatibility.
+                Err(crate::TantivyError::IncompatibleIndex(incompatibility))
+            }
         }
     }
 
@@ -87,11 +102,17 @@ impl ManagedDirectory {
     ///
     /// * `living_files` - List of files that are still used by the index.
     ///
+    /// The use a callback ensures that the list of living_files is computed
+    /// while we hold the lock on meta.
+    ///
     /// This method does not panick nor returns errors.
     /// If a file cannot be deleted (for permission reasons for instance)
     /// an error is simply logged, and the file remains in the list of managed
     /// files.
-    pub fn garbage_collect<L: FnOnce() -> HashSet<PathBuf>>(&mut self, get_living_files: L) {
+    pub fn garbage_collect<L: FnOnce() -> HashSet<PathBuf>>(
+        &mut self,
+        get_living_files: L,
+    ) -> crate::Result<GarbageCollectionResult> {
         info!("Garbage collect");
         let mut files_to_delete = vec![];
 
@@ -117,35 +138,42 @@ impl ManagedDirectory {
             // 2) writer change meta.json (for instance after a merge or a commit)
             // 3) gc kicks in.
             // 4) gc removes a file that was useful for process B, before process B opened it.
-            if let Ok(_meta_lock) = LockType::MetaLock.acquire_lock(self) {
-                let living_files = get_living_files();
-                for managed_path in &meta_informations_rlock.managed_paths {
-                    if !living_files.contains(managed_path) {
-                        files_to_delete.push(managed_path.clone());
+            match self.acquire_lock(&META_LOCK) {
+                Ok(_meta_lock) => {
+                    let living_files = get_living_files();
+                    for managed_path in &meta_informations_rlock.managed_paths {
+                        if !living_files.contains(managed_path) {
+                            files_to_delete.push(managed_path.clone());
+                        }
                     }
                 }
+                Err(err) => {
+                    error!("Failed to acquire lock for GC");
+                    return Err(crate::TantivyError::from(err));
+                }
             }
         }
 
+        let mut failed_to_delete_files = vec![];
         let mut deleted_files = vec![];
-        {
-            for file_to_delete in files_to_delete {
-                match self.delete(&file_to_delete) {
-                    Ok(_) => {
-                        info!("Deleted {:?}", file_to_delete);
-                        deleted_files.push(file_to_delete);
-                    }
-                    Err(file_error) => {
-                        match file_error {
-                            DeleteError::FileDoesNotExist(_) => {
-                                deleted_files.push(file_to_delete);
-                            }
-                            DeleteError::IOError(_) => {
-                                if !cfg!(target_os = "windows") {
-                                    // On windows, delete is expected to fail if the file
-                                    // is mmapped.
-                                    error!("Failed to delete {:?}", file_to_delete);
-                                }
+
+        for file_to_delete in files_to_delete {
+            match self.delete(&file_to_delete) {
+                Ok(_) => {
+                    info!("Deleted {:?}", file_to_delete);
+                    deleted_files.push(file_to_delete);
+                }
+                Err(file_error) => {
+                    match file_error {
+                        DeleteError::FileDoesNotExist(_) => {
+                            deleted_files.push(file_to_delete.clone());
+                        }
+                        DeleteError::IOError(_) => {
+                            failed_to_delete_files.push(file_to_delete.clone());
+                            if !cfg!(target_os = "windows") {
+                                // On windows, delete is expected to fail if the file
+                                // is mmapped.
+                                error!("Failed to delete {:?}", file_to_delete);
                             }
                         }
                     }
@@ -160,16 +188,17 @@ impl ManagedDirectory {
                 .meta_informations
                 .write()
                 .expect("Managed directory wlock poisoned (2).");
-            {
-                let managed_paths_write = &mut meta_informations_wlock.managed_paths;
-                for delete_file in &deleted_files {
-                    managed_paths_write.remove(delete_file);
-                }
-            }
-            if save_managed_paths(self.directory.as_mut(), &meta_informations_wlock).is_err() {
-                error!("Failed to save the list of managed files.");
+            let managed_paths_write = &mut meta_informations_wlock.managed_paths;
+            for delete_file in &deleted_files {
+                managed_paths_write.remove(delete_file);
             }
+            save_managed_paths(self.directory.as_mut(), &meta_informations_wlock)?;
         }
+
+        Ok(GarbageCollectionResult {
+            deleted_files,
+            failed_to_delete_files,
+        })
     }
 
     /// Registers a file as managed
@@ -198,17 +227,60 @@ impl ManagedDirectory {
         }
         Ok(())
     }
+
+    /// Verify checksum of a managed file
+    pub fn validate_checksum(&self, path: &Path) -> result::Result<bool, OpenReadError> {
+        let reader = self.directory.open_read(path)?;
+        let (footer, data) = Footer::extract_footer(reader)
+            .map_err(|err| IOError::with_path(path.to_path_buf(), err))?;
+        let mut hasher = Hasher::new();
+        hasher.update(data.as_slice());
+        let crc = hasher.finalize();
+        Ok(footer
+            .versioned_footer
+            .crc()
+            .map(|v| v == crc)
+            .unwrap_or(false))
+    }
+
+    /// List files for which checksum does not match content
+    pub fn list_damaged(&self) -> result::Result<HashSet<PathBuf>, OpenReadError> {
+        let mut hashset = HashSet::new();
+        let managed_paths = self
+            .meta_informations
+            .read()
+            .expect("Managed directory rlock poisoned in list damaged.")
+            .managed_paths
+            .clone();
+
+        for path in managed_paths.into_iter() {
+            if !self.validate_checksum(&path)? {
+                hashset.insert(path);
+            }
+        }
+        Ok(hashset)
+    }
 }
 
 impl Directory for ManagedDirectory {
     fn open_read(&self, path: &Path) -> result::Result<ReadOnlySource, OpenReadError> {
-        self.directory.open_read(path)
+        let read_only_source = self.directory.open_read(path)?;
+        let (footer, reader) = Footer::extract_footer(read_only_source)
+            .map_err(|err| IOError::with_path(path.to_path_buf(), err))?;
+        footer.is_compatible()?;
+        Ok(reader)
     }
 
     fn open_write(&mut self, path: &Path) -> result::Result<WritePtr, OpenWriteError> {
         self.register_file_as_managed(path)
             .map_err(|e| IOError::with_path(path.to_owned(), e))?;
-        self.directory.open_write(path)
+        Ok(io::BufWriter::new(Box::new(FooterProxy::new(
+            self.directory
+                .open_write(path)?
+                .into_inner()
+                .map_err(|_| ())
+                .expect("buffer should be empty"),
+        ))))
     }
 
     fn atomic_write(&mut self, path: &Path, data: &[u8]) -> io::Result<()> {
@@ -227,6 +299,14 @@ impl Directory for ManagedDirectory {
     fn exists(&self, path: &Path) -> bool {
         self.directory.exists(path)
     }
+
+    fn acquire_lock(&self, lock: &Lock) -> result::Result<DirectoryLock, LockError> {
+        self.directory.acquire_lock(lock)
+    }
+
+    fn watch(&self, watch_callback: WatchCallback) -> crate::Result<WatchHandle> {
+        self.directory.watch(watch_callback)
+    }
 }
 
 impl Clone for ManagedDirectory {
@@ -238,98 +318,118 @@ impl Clone for ManagedDirectory {
     }
 }
 
+#[cfg(feature = "mmap")]
 #[cfg(test)]
-mod tests {
+mod tests_mmap_specific {
 
-    use super::*;
-    #[cfg(feature = "mmap")]
-    use directory::MmapDirectory;
+    use crate::directory::{Directory, ManagedDirectory, MmapDirectory, TerminatingWrite};
+    use std::collections::HashSet;
+    use std::fs::OpenOptions;
     use std::io::Write;
-    use std::path::Path;
-    use tempdir::TempDir;
-
-    lazy_static! {
-        static ref TEST_PATH1: &'static Path = Path::new("some_path_for_test");
-        static ref TEST_PATH2: &'static Path = Path::new("some_path_for_test2");
-    }
+    use std::path::{Path, PathBuf};
+    use tempfile::TempDir;
 
     #[test]
-    #[cfg(feature = "mmap")]
     fn test_managed_directory() {
-        let tempdir = TempDir::new("index").unwrap();
+        let tempdir = TempDir::new().unwrap();
         let tempdir_path = PathBuf::from(tempdir.path());
+
+        let test_path1: &'static Path = Path::new("some_path_for_test");
+        let test_path2: &'static Path = Path::new("some_path_for_test_2");
         {
             let mmap_directory = MmapDirectory::open(&tempdir_path).unwrap();
-            let mut managed_directory = ManagedDirectory::new(mmap_directory).unwrap();
-            {
-                let mut write_file = managed_directory.open_write(*TEST_PATH1).unwrap();
-                write_file.flush().unwrap();
-            }
-            {
-                managed_directory
-                    .atomic_write(*TEST_PATH2, &vec![0u8, 1u8])
-                    .unwrap();
-            }
-            {
-                assert!(managed_directory.exists(*TEST_PATH1));
-                assert!(managed_directory.exists(*TEST_PATH2));
-            }
-            {
-                let living_files: HashSet<PathBuf> =
-                    [TEST_PATH1.to_owned()].into_iter().cloned().collect();
-                managed_directory.garbage_collect(|| living_files);
-            }
-            {
-                assert!(managed_directory.exists(*TEST_PATH1));
-                assert!(!managed_directory.exists(*TEST_PATH2));
-            }
+            let mut managed_directory = ManagedDirectory::wrap(mmap_directory).unwrap();
+            let write_file = managed_directory.open_write(test_path1).unwrap();
+            write_file.terminate().unwrap();
+            managed_directory
+                .atomic_write(test_path2, &[0u8, 1u8])
+                .unwrap();
+            assert!(managed_directory.exists(test_path1));
+            assert!(managed_directory.exists(test_path2));
+            let living_files: HashSet<PathBuf> = [test_path1.to_owned()].iter().cloned().collect();
+            assert!(managed_directory.garbage_collect(|| living_files).is_ok());
+            assert!(managed_directory.exists(test_path1));
+            assert!(!managed_directory.exists(test_path2));
         }
         {
             let mmap_directory = MmapDirectory::open(&tempdir_path).unwrap();
-            let mut managed_directory = ManagedDirectory::new(mmap_directory).unwrap();
-            {
-                assert!(managed_directory.exists(*TEST_PATH1));
-                assert!(!managed_directory.exists(*TEST_PATH2));
-            }
-            {
-                let living_files: HashSet<PathBuf> = HashSet::new();
-                managed_directory.garbage_collect(|| living_files);
-            }
-            {
-                assert!(!managed_directory.exists(*TEST_PATH1));
-                assert!(!managed_directory.exists(*TEST_PATH2));
-            }
+            let mut managed_directory = ManagedDirectory::wrap(mmap_directory).unwrap();
+            assert!(managed_directory.exists(test_path1));
+            assert!(!managed_directory.exists(test_path2));
+            let living_files: HashSet<PathBuf> = HashSet::new();
+            assert!(managed_directory.garbage_collect(|| living_files).is_ok());
+            assert!(!managed_directory.exists(test_path1));
+            assert!(!managed_directory.exists(test_path2));
         }
     }
 
     #[test]
-    #[cfg(feature = "mmap ")]
     fn test_managed_directory_gc_while_mmapped() {
-        let tempdir = TempDir::new("index").unwrap();
+        let test_path1: &'static Path = Path::new("some_path_for_test");
+
+        let tempdir = TempDir::new().unwrap();
         let tempdir_path = PathBuf::from(tempdir.path());
         let living_files = HashSet::new();
 
         let mmap_directory = MmapDirectory::open(&tempdir_path).unwrap();
-        let mut managed_directory = ManagedDirectory::new(mmap_directory).unwrap();
-        managed_directory
-            .atomic_write(*TEST_PATH1, &vec![0u8, 1u8])
-            .unwrap();
-        assert!(managed_directory.exists(*TEST_PATH1));
+        let mut managed_directory = ManagedDirectory::wrap(mmap_directory).unwrap();
+        let mut write = managed_directory.open_write(test_path1).unwrap();
+        write.write_all(&[0u8, 1u8]).unwrap();
+        write.terminate().unwrap();
+        assert!(managed_directory.exists(test_path1));
 
-        let _mmap_read = managed_directory.open_read(*TEST_PATH1).unwrap();
-        managed_directory.garbage_collect(|| living_files.clone());
+        let _mmap_read = managed_directory.open_read(test_path1).unwrap();
+        assert!(managed_directory
+            .garbage_collect(|| living_files.clone())
+            .is_ok());
         if cfg!(target_os = "windows") {
             // On Windows, gc should try and fail the file as it is mmapped.
-            assert!(managed_directory.exists(*TEST_PATH1));
+            assert!(managed_directory.exists(test_path1));
             // unmap should happen here.
             drop(_mmap_read);
             // The file should still be in the list of managed file and
             // eventually be deleted once mmap is released.
-            managed_directory.garbage_collect(|| living_files);
-            assert!(!managed_directory.exists(*TEST_PATH1));
+            assert!(managed_directory.garbage_collect(|| living_files).is_ok());
+            assert!(!managed_directory.exists(test_path1));
         } else {
-            assert!(!managed_directory.exists(*TEST_PATH1));
+            assert!(!managed_directory.exists(test_path1));
         }
     }
 
+    #[test]
+    fn test_checksum() {
+        let test_path1: &'static Path = Path::new("some_path_for_test");
+        let test_path2: &'static Path = Path::new("other_test_path");
+
+        let tempdir = TempDir::new().unwrap();
+        let tempdir_path = PathBuf::from(tempdir.path());
+
+        let mmap_directory = MmapDirectory::open(&tempdir_path).unwrap();
+        let mut managed_directory = ManagedDirectory::wrap(mmap_directory).unwrap();
+        let mut write = managed_directory.open_write(test_path1).unwrap();
+        write.write_all(&[0u8, 1u8]).unwrap();
+        write.terminate().unwrap();
+
+        let mut write = managed_directory.open_write(test_path2).unwrap();
+        write.write_all(&[3u8, 4u8, 5u8]).unwrap();
+        write.terminate().unwrap();
+
+        let read_source = managed_directory.open_read(test_path2).unwrap();
+        assert_eq!(read_source.as_slice(), &[3u8, 4u8, 5u8]);
+        assert!(managed_directory.list_damaged().unwrap().is_empty());
+
+        let mut corrupted_path = tempdir_path.clone();
+        corrupted_path.push(test_path2);
+        let mut file = OpenOptions::new()
+            .write(true)
+            .open(&corrupted_path)
+            .unwrap();
+        file.write_all(&[255u8]).unwrap();
+        file.flush().unwrap();
+        drop(file);
+
+        let damaged = managed_directory.list_damaged().unwrap();
+        assert_eq!(damaged.len(), 1);
+        assert!(damaged.contains(test_path2));
+    }
 }

src/directory/mmap_directory.rs

@@ -1,12 +1,24 @@
-use atomicwrites;
-use common::make_io_err;
-use directory::error::{DeleteError, IOError, OpenDirectoryError, OpenReadError, OpenWriteError};
-use directory::shared_vec_slice::SharedVecSlice;
-use directory::Directory;
-use directory::ReadOnlySource;
-use directory::WritePtr;
-use fst::raw::MmapReadOnly;
-use std::collections::hash_map::Entry as HashMapEntry;
+use crate::core::META_FILEPATH;
+use crate::directory::error::LockError;
+use crate::directory::error::{
+    DeleteError, IOError, OpenDirectoryError, OpenReadError, OpenWriteError,
+};
+use crate::directory::read_only_source::BoxedData;
+use crate::directory::AntiCallToken;
+use crate::directory::Directory;
+use crate::directory::DirectoryLock;
+use crate::directory::Lock;
+use crate::directory::ReadOnlySource;
+use crate::directory::WatchCallback;
+use crate::directory::WatchCallbackList;
+use crate::directory::WatchHandle;
+use crate::directory::{TerminatingWrite, WritePtr};
+use fs2::FileExt;
+use memmap::Mmap;
+use notify::RawEvent;
+use notify::RecursiveMode;
+use notify::Watcher;
+use serde::{Deserialize, Serialize};
 use std::collections::HashMap;
 use std::convert::From;
 use std::fmt;
@@ -16,14 +28,22 @@ use std::io::{self, Seek, SeekFrom};
 use std::io::{BufWriter, Read, Write};
 use std::path::{Path, PathBuf};
 use std::result;
+use std::sync::mpsc::{channel, Receiver, Sender};
 use std::sync::Arc;
+use std::sync::Mutex;
 use std::sync::RwLock;
-use tempdir::TempDir;
+use std::sync::Weak;
+use std::thread;
+use tempfile::TempDir;
+
+/// Create a default io error given a string.
+pub(crate) fn make_io_err(msg: String) -> io::Error {
+    io::Error::new(io::ErrorKind::Other, msg)
+}
 
 /// Returns None iff the file exists, can be read, but is empty (and hence
-/// cannot be mmapped).
-///
-fn open_mmap(full_path: &Path) -> result::Result<Option<MmapReadOnly>, OpenReadError> {
+/// cannot be mmapped)
+fn open_mmap(full_path: &Path) -> result::Result<Option<Mmap>, OpenReadError> {
     let file = File::open(full_path).map_err(|e| {
         if e.kind() == io::ErrorKind::NotFound {
             OpenReadError::FileDoesNotExist(full_path.to_owned())
@@ -42,7 +62,7 @@ fn open_mmap(full_path: &Path) -> result::Result<Option<MmapReadOnly>, OpenReadE
         return Ok(None);
     }
     unsafe {
-        MmapReadOnly::open(&file)
+        memmap::Mmap::map(&file)
             .map(Some)
             .map_err(|e| From::from(IOError::with_path(full_path.to_owned(), e)))
     }
@@ -65,7 +85,7 @@ pub struct CacheInfo {
 
 struct MmapCache {
     counters: CacheCounters,
-    cache: HashMap<PathBuf, MmapReadOnly>,
+    cache: HashMap<PathBuf, Weak<BoxedData>>,
 }
 
 impl Default for MmapCache {
@@ -78,12 +98,7 @@ impl Default for MmapCache {
 }
 
 impl MmapCache {
-    /// Removes a `MmapReadOnly` entry from the mmap cache.
-    fn discard_from_cache(&mut self, full_path: &Path) -> bool {
-        self.cache.remove(full_path).is_some()
-    }
-
-    fn get_info(&mut self) -> CacheInfo {
+    fn get_info(&self) -> CacheInfo {
         let paths: Vec<PathBuf> = self.cache.keys().cloned().collect();
         CacheInfo {
             counters: self.counters.clone(),
@@ -91,57 +106,175 @@ impl MmapCache {
         }
     }
 
-    fn get_mmap(&mut self, full_path: &Path) -> Result<Option<MmapReadOnly>, OpenReadError> {
-        Ok(match self.cache.entry(full_path.to_owned()) {
-            HashMapEntry::Occupied(occupied_entry) => {
-                let mmap = occupied_entry.get();
+    fn remove_weak_ref(&mut self) {
+        let keys_to_remove: Vec<PathBuf> = self
+            .cache
+            .iter()
+            .filter(|(_, mmap_weakref)| mmap_weakref.upgrade().is_none())
+            .map(|(key, _)| key.clone())
+            .collect();
+        for key in keys_to_remove {
+            self.cache.remove(&key);
+        }
+    }
+
+    // Returns None if the file exists but as a len of 0 (and hence is not mmappable).
+    fn get_mmap(&mut self, full_path: &Path) -> Result<Option<Arc<BoxedData>>, OpenReadError> {
+        if let Some(mmap_weak) = self.cache.get(full_path) {
+            if let Some(mmap_arc) = mmap_weak.upgrade() {
                 self.counters.hit += 1;
-                Some(mmap.clone())
+                return Ok(Some(mmap_arc));
             }
-            HashMapEntry::Vacant(vacant_entry) => {
-                self.counters.miss += 1;
-                if let Some(mmap) = open_mmap(full_path)? {
-                    vacant_entry.insert(mmap.clone());
-                    Some(mmap)
-                } else {
-                    None
+        }
+        self.cache.remove(full_path);
+        self.counters.miss += 1;
+        let mmap_opt = open_mmap(full_path)?;
+        Ok(mmap_opt.map(|mmap| {
+            let mmap_arc: Arc<BoxedData> = Arc::new(Box::new(mmap));
+            let mmap_weak = Arc::downgrade(&mmap_arc);
+            self.cache.insert(full_path.to_owned(), mmap_weak);
+            mmap_arc
+        }))
+    }
+}
+
+struct WatcherWrapper {
+    _watcher: Mutex<notify::RecommendedWatcher>,
+    watcher_router: Arc<WatchCallbackList>,
+}
+
+impl WatcherWrapper {
+    pub fn new(path: &Path) -> Result<Self, OpenDirectoryError> {
+        let (tx, watcher_recv): (Sender<RawEvent>, Receiver<RawEvent>) = channel();
+        // We need to initialize the
+        let watcher = notify::raw_watcher(tx)
+            .and_then(|mut watcher| {
+                watcher.watch(path, RecursiveMode::Recursive)?;
+                Ok(watcher)
+            })
+            .map_err(|err| match err {
+                notify::Error::PathNotFound => OpenDirectoryError::DoesNotExist(path.to_owned()),
+                _ => {
+                    panic!("Unknown error while starting watching directory {:?}", path);
                 }
-            }
+            })?;
+        let watcher_router: Arc<WatchCallbackList> = Default::default();
+        let watcher_router_clone = watcher_router.clone();
+        thread::Builder::new()
+            .name("meta-file-watch-thread".to_string())
+            .spawn(move || {
+                loop {
+                    match watcher_recv.recv().map(|evt| evt.path) {
+                        Ok(Some(changed_path)) => {
+                            // ... Actually subject to false positive.
+                            // We might want to be more accurate than this at one point.
+                            if let Some(filename) = changed_path.file_name() {
+                                if filename == *META_FILEPATH {
+                                    let _ = watcher_router_clone.broadcast();
+                                }
+                            }
+                        }
+                        Ok(None) => {
+                            // not an event we are interested in.
+                        }
+                        Err(_e) => {
+                            // the watch send channel was dropped
+                            break;
+                        }
+                    }
+                }
+            })?;
+        Ok(WatcherWrapper {
+            _watcher: Mutex::new(watcher),
+            watcher_router,
         })
     }
+
+    pub fn watch(&mut self, watch_callback: WatchCallback) -> WatchHandle {
+        self.watcher_router.subscribe(watch_callback)
+    }
 }
 
 /// Directory storing data in files, read via mmap.
 ///
 /// The Mmap object are cached to limit the
 /// system calls.
+///
+/// In the `MmapDirectory`, locks are implemented using the `fs2` crate definition of locks.
+///
+/// On MacOS & linux, it relies on `flock` (aka `BSD Lock`). These locks solve most of the
+/// problems related to POSIX Locks, but may their contract may not be respected on `NFS`
+/// depending on the implementation.
+///
+/// On Windows the semantics are again different.
 #[derive(Clone)]
 pub struct MmapDirectory {
+    inner: Arc<MmapDirectoryInner>,
+}
+
+struct MmapDirectoryInner {
     root_path: PathBuf,
-    mmap_cache: Arc<RwLock<MmapCache>>,
-    _temp_directory: Arc<Option<TempDir>>,
+    mmap_cache: RwLock<MmapCache>,
+    _temp_directory: Option<TempDir>,
+    watcher: RwLock<Option<WatcherWrapper>>,
+}
+
+impl MmapDirectoryInner {
+    fn new(root_path: PathBuf, temp_directory: Option<TempDir>) -> MmapDirectoryInner {
+        MmapDirectoryInner {
+            root_path,
+            mmap_cache: Default::default(),
+            _temp_directory: temp_directory,
+            watcher: RwLock::new(None),
+        }
+    }
+
+    fn watch(&self, watch_callback: WatchCallback) -> crate::Result<WatchHandle> {
+        // a lot of juggling here, to ensure we don't do anything that panics
+        // while the rwlock is held. That way we ensure that the rwlock cannot
+        // be poisoned.
+        //
+        // The downside is that we might create a watch wrapper that is not useful.
+        let need_initialization = self.watcher.read().unwrap().is_none();
+        if need_initialization {
+            let watch_wrapper = WatcherWrapper::new(&self.root_path)?;
+            let mut watch_wlock = self.watcher.write().unwrap();
+            // the watcher could have been initialized when we released the lock, and
+            // we do not want to lose the watched files that were set.
+            if watch_wlock.is_none() {
+                *watch_wlock = Some(watch_wrapper);
+            }
+        }
+        if let Some(watch_wrapper) = self.watcher.write().unwrap().as_mut() {
+            Ok(watch_wrapper.watch(watch_callback))
+        } else {
+            unreachable!("At this point, watch wrapper is supposed to be initialized");
+        }
+    }
 }
 
 impl fmt::Debug for MmapDirectory {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        write!(f, "MmapDirectory({:?})", self.root_path)
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "MmapDirectory({:?})", self.inner.root_path)
     }
 }
 
 impl MmapDirectory {
+    fn new(root_path: PathBuf, temp_directory: Option<TempDir>) -> MmapDirectory {
+        let inner = MmapDirectoryInner::new(root_path, temp_directory);
+        MmapDirectory {
+            inner: Arc::new(inner),
+        }
+    }
+
     /// Creates a new MmapDirectory in a temporary directory.
     ///
     /// This is mostly useful to test the MmapDirectory itself.
     /// For your unit tests, prefer the RAMDirectory.
-    pub fn create_from_tempdir() -> io::Result<MmapDirectory> {
-        let tempdir = TempDir::new("index")?;
+    pub fn create_from_tempdir() -> Result<MmapDirectory, OpenDirectoryError> {
+        let tempdir = TempDir::new().map_err(OpenDirectoryError::IoError)?;
         let tempdir_path = PathBuf::from(tempdir.path());
-        let directory = MmapDirectory {
-            root_path: tempdir_path,
-            mmap_cache: Arc::new(RwLock::new(MmapCache::default())),
-            _temp_directory: Arc::new(Some(tempdir)),
-        };
-        Ok(directory)
+        Ok(MmapDirectory::new(tempdir_path, Some(tempdir)))
     }
 
     /// Opens a MmapDirectory in a directory.
@@ -159,18 +292,14 @@ impl MmapDirectory {
                 directory_path,
             )))
         } else {
-            Ok(MmapDirectory {
-                root_path: PathBuf::from(directory_path),
-                mmap_cache: Arc::new(RwLock::new(MmapCache::default())),
-                _temp_directory: Arc::new(None),
-            })
+            Ok(MmapDirectory::new(PathBuf::from(directory_path), None))
         }
     }
 
     /// Joins a relative_path to the directory `root_path`
     /// to create a proper complete `filepath`.
     fn resolve_path(&self, relative_path: &Path) -> PathBuf {
-        self.root_path.join(relative_path)
+        self.inner.root_path.join(relative_path)
     }
 
     /// Sync the root directory.
@@ -188,14 +317,14 @@ impl MmapDirectory {
         #[cfg(windows)]
         {
             use std::os::windows::fs::OpenOptionsExt;
-            use winapi::winbase;
+            use winapi::um::winbase;
 
             open_opts
                 .write(true)
                 .custom_flags(winbase::FILE_FLAG_BACKUP_SEMANTICS);
         }
 
-        let fd = open_opts.open(&self.root_path)?;
+        let fd = open_opts.open(&self.inner.root_path)?;
         fd.sync_all()?;
         Ok(())
     }
@@ -205,14 +334,35 @@ impl MmapDirectory {
     ///
     /// The `MmapDirectory` embeds a `MmapDirectory`
     /// to avoid multiplying the `mmap` system calls.
-    pub fn get_cache_info(&mut self) -> CacheInfo {
-        self.mmap_cache
+    pub fn get_cache_info(&self) -> CacheInfo {
+        self.inner
+            .mmap_cache
             .write()
+            .expect("mmap cache lock is poisoned")
+            .remove_weak_ref();
+        self.inner
+            .mmap_cache
+            .read()
             .expect("Mmap cache lock is poisoned.")
             .get_info()
     }
 }
 
+/// We rely on fs2 for file locking. On Windows & MacOS this
+/// uses BSD locks (`flock`). The lock is actually released when
+/// the `File` object is dropped and its associated file descriptor
+/// is closed.
+struct ReleaseLockFile {
+    _file: File,
+    path: PathBuf,
+}
+
+impl Drop for ReleaseLockFile {
+    fn drop(&mut self) {
+        debug!("Releasing lock {:?}", self.path);
+    }
+}
+
 /// This Write wraps a File, but has the specificity of
 /// call `sync_all` on flush.
 struct SafeFileWriter(File);
@@ -240,12 +390,18 @@ impl Seek for SafeFileWriter {
     }
 }
 
+impl TerminatingWrite for SafeFileWriter {
+    fn terminate_ref(&mut self, _: AntiCallToken) -> io::Result<()> {
+        self.flush()
+    }
+}
+
 impl Directory for MmapDirectory {
     fn open_read(&self, path: &Path) -> result::Result<ReadOnlySource, OpenReadError> {
         debug!("Open Read {:?}", path);
         let full_path = self.resolve_path(path);
 
-        let mut mmap_cache = self.mmap_cache.write().map_err(|_| {
+        let mut mmap_cache = self.inner.mmap_cache.write().map_err(|_| {
             let msg = format!(
                 "Failed to acquired write lock \
                  on mmap cache while reading {:?}",
@@ -253,11 +409,33 @@ impl Directory for MmapDirectory {
             );
             IOError::with_path(path.to_owned(), make_io_err(msg))
         })?;
-
         Ok(mmap_cache
             .get_mmap(&full_path)?
-            .map(ReadOnlySource::Mmap)
-            .unwrap_or_else(|| ReadOnlySource::Anonymous(SharedVecSlice::empty())))
+            .map(ReadOnlySource::from)
+            .unwrap_or_else(ReadOnlySource::empty))
+    }
+
+    /// Any entry associated to the path in the mmap will be
+    /// removed before the file is deleted.
+    fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
+        let full_path = self.resolve_path(path);
+        match fs::remove_file(&full_path) {
+            Ok(_) => self
+                .sync_directory()
+                .map_err(|e| IOError::with_path(path.to_owned(), e).into()),
+            Err(e) => {
+                if e.kind() == io::ErrorKind::NotFound {
+                    Err(DeleteError::FileDoesNotExist(path.to_owned()))
+                } else {
+                    Err(IOError::with_path(path.to_owned(), e).into())
+                }
+            }
+        }
+    }
+
+    fn exists(&self, path: &Path) -> bool {
+        let full_path = self.resolve_path(path);
+        full_path.exists()
     }
 
     fn open_write(&mut self, path: &Path) -> Result<WritePtr, OpenWriteError> {
@@ -290,44 +468,6 @@ impl Directory for MmapDirectory {
         Ok(BufWriter::new(Box::new(writer)))
     }
 
-    /// Any entry associated to the path in the mmap will be
-    /// removed before the file is deleted.
-    fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
-        debug!("Deleting file {:?}", path);
-        let full_path = self.resolve_path(path);
-        let mut mmap_cache = self.mmap_cache.write().map_err(|_| {
-            let msg = format!(
-                "Failed to acquired write lock \
-                 on mmap cache while deleting {:?}",
-                path
-            );
-            IOError::with_path(path.to_owned(), make_io_err(msg))
-        })?;
-        mmap_cache.discard_from_cache(path);
-
-        // Removing the entry in the MMap cache.
-        // The munmap will appear on Drop,
-        // when the last reference is gone.
-        mmap_cache.cache.remove(&full_path);
-        match fs::remove_file(&full_path) {
-            Ok(_) => self
-                .sync_directory()
-                .map_err(|e| IOError::with_path(path.to_owned(), e).into()),
-            Err(e) => {
-                if e.kind() == io::ErrorKind::NotFound {
-                    Err(DeleteError::FileDoesNotExist(path.to_owned()))
-                } else {
-                    Err(IOError::with_path(path.to_owned(), e).into())
-                }
-            }
-        }
-    }
-
-    fn exists(&self, path: &Path) -> bool {
-        let full_path = self.resolve_path(path);
-        full_path.exists()
-    }
-
     fn atomic_read(&self, path: &Path) -> Result<Vec<u8>, OpenReadError> {
         let full_path = self.resolve_path(path);
         let mut buffer = Vec::new();
@@ -354,6 +494,30 @@ impl Directory for MmapDirectory {
         meta_file.write(|f| f.write_all(data))?;
         Ok(())
     }
+
+    fn acquire_lock(&self, lock: &Lock) -> Result<DirectoryLock, LockError> {
+        let full_path = self.resolve_path(&lock.filepath);
+        // We make sure that the file exists.
+        let file: File = OpenOptions::new()
+            .write(true)
+            .create(true) //< if the file does not exist yet, create it.
+            .open(&full_path)
+            .map_err(LockError::IOError)?;
+        if lock.is_blocking {
+            file.lock_exclusive().map_err(LockError::IOError)?;
+        } else {
+            file.try_lock_exclusive().map_err(|_| LockError::LockBusy)?
+        }
+        // dropping the file handle will release the lock.
+        Ok(DirectoryLock::from(Box::new(ReleaseLockFile {
+            path: lock.filepath.clone(),
+            _file: file,
+        })))
+    }
+
+    fn watch(&self, watch_callback: WatchCallback) -> crate::Result<WatchHandle> {
+        self.inner.watch(watch_callback)
+    }
 }
 
 #[cfg(test)]
@@ -363,9 +527,15 @@ mod tests {
     // The following tests are specific to the MmapDirectory
 
     use super::*;
+    use crate::indexer::LogMergePolicy;
+    use crate::schema::{Schema, SchemaBuilder, TEXT};
+    use crate::Index;
+    use crate::ReloadPolicy;
+    use std::fs;
+    use std::sync::atomic::{AtomicUsize, Ordering};
 
     #[test]
-    fn test_open_non_existant_path() {
+    fn test_open_non_existent_path() {
         assert!(MmapDirectory::open(PathBuf::from("./nowhere")).is_err());
     }
 
@@ -387,7 +557,7 @@ mod tests {
 
     #[test]
     fn test_cache() {
-        let content = "abc".as_bytes();
+        let content = b"abc";
 
         // here we test if the cache releases
         // mmaps correctly.
@@ -403,26 +573,117 @@ mod tests {
                 w.flush().unwrap();
             }
         }
-        {
-            for (i, path) in paths.iter().enumerate() {
-                let _r = mmap_directory.open_read(path).unwrap();
-                assert_eq!(mmap_directory.get_cache_info().mmapped.len(), i + 1);
-            }
-            for path in paths.iter() {
-                let _r = mmap_directory.open_read(path).unwrap();
-                assert_eq!(mmap_directory.get_cache_info().mmapped.len(), num_paths);
-            }
-            for (i, path) in paths.iter().enumerate() {
-                mmap_directory.delete(path).unwrap();
-                assert_eq!(
-                    mmap_directory.get_cache_info().mmapped.len(),
-                    num_paths - i - 1
-                );
-            }
+
+        let mut keep = vec![];
+        for (i, path) in paths.iter().enumerate() {
+            keep.push(mmap_directory.open_read(path).unwrap());
+            assert_eq!(mmap_directory.get_cache_info().mmapped.len(), i + 1);
+        }
+        assert_eq!(mmap_directory.get_cache_info().counters.hit, 0);
+        assert_eq!(mmap_directory.get_cache_info().counters.miss, 10);
+        assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 10);
+        for path in paths.iter() {
+            let _r = mmap_directory.open_read(path).unwrap();
+            assert_eq!(mmap_directory.get_cache_info().mmapped.len(), num_paths);
         }
         assert_eq!(mmap_directory.get_cache_info().counters.hit, 10);
         assert_eq!(mmap_directory.get_cache_info().counters.miss, 10);
+        assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 10);
+
+        for path in paths.iter() {
+            let _r = mmap_directory.open_read(path).unwrap();
+            assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 10);
+        }
+
+        assert_eq!(mmap_directory.get_cache_info().counters.hit, 20);
+        assert_eq!(mmap_directory.get_cache_info().counters.miss, 10);
+        assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 10);
+        drop(keep);
+        for path in paths.iter() {
+            let _r = mmap_directory.open_read(path).unwrap();
+            assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 1);
+        }
+        assert_eq!(mmap_directory.get_cache_info().counters.hit, 20);
+        assert_eq!(mmap_directory.get_cache_info().counters.miss, 20);
+        assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 0);
+
+        for path in &paths {
+            mmap_directory.delete(path).unwrap();
+        }
+        assert_eq!(mmap_directory.get_cache_info().counters.hit, 20);
+        assert_eq!(mmap_directory.get_cache_info().counters.miss, 20);
+        assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 0);
+        for path in paths.iter() {
+            assert!(mmap_directory.open_read(path).is_err());
+        }
+        assert_eq!(mmap_directory.get_cache_info().counters.hit, 20);
+        assert_eq!(mmap_directory.get_cache_info().counters.miss, 30);
         assert_eq!(mmap_directory.get_cache_info().mmapped.len(), 0);
     }
 
+    #[test]
+    fn test_watch_wrapper() {
+        let counter: Arc<AtomicUsize> = Default::default();
+        let counter_clone = counter.clone();
+        let tmp_dir = tempfile::TempDir::new().unwrap();
+        let tmp_dirpath = tmp_dir.path().to_owned();
+        let mut watch_wrapper = WatcherWrapper::new(&tmp_dirpath).unwrap();
+        let tmp_file = tmp_dirpath.join(*META_FILEPATH);
+        let _handle = watch_wrapper.watch(Box::new(move || {
+            counter_clone.fetch_add(1, Ordering::SeqCst);
+        }));
+        let (sender, receiver) = crossbeam::channel::unbounded();
+        let _handle2 = watch_wrapper.watch(Box::new(move || {
+            let _ = sender.send(());
+        }));
+        assert_eq!(counter.load(Ordering::SeqCst), 0);
+        fs::write(&tmp_file, b"whateverwilldo").unwrap();
+        assert!(receiver.recv().is_ok());
+        assert!(counter.load(Ordering::SeqCst) >= 1);
+    }
+
+    #[test]
+    fn test_mmap_released() {
+        let mmap_directory = MmapDirectory::create_from_tempdir().unwrap();
+        let mut schema_builder: SchemaBuilder = Schema::builder();
+        let text_field = schema_builder.add_text_field("text", TEXT);
+        let schema = schema_builder.build();
+
+        {
+            let index = Index::create(mmap_directory.clone(), schema).unwrap();
+
+            let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+            let mut log_merge_policy = LogMergePolicy::default();
+            log_merge_policy.set_min_merge_size(3);
+            index_writer.set_merge_policy(Box::new(log_merge_policy));
+            for _num_commits in 0..10 {
+                for _ in 0..10 {
+                    index_writer.add_document(doc!(text_field=>"abc"));
+                }
+                index_writer.commit().unwrap();
+            }
+
+            let reader = index
+                .reader_builder()
+                .reload_policy(ReloadPolicy::Manual)
+                .try_into()
+                .unwrap();
+
+            for _ in 0..4 {
+                index_writer.add_document(doc!(text_field=>"abc"));
+                index_writer.commit().unwrap();
+                reader.reload().unwrap();
+            }
+            index_writer.wait_merging_threads().unwrap();
+
+            reader.reload().unwrap();
+            let num_segments = reader.searcher().segment_readers().len();
+            assert!(num_segments <= 4);
+            assert_eq!(
+                num_segments * 7,
+                mmap_directory.get_cache_info().mmapped.len()
+            );
+        }
+        assert!(mmap_directory.get_cache_info().mmapped.is_empty());
+    }
 }

src/directory/mod.rs

@@ -8,158 +8,89 @@ WORM directory abstraction.
 mod mmap_directory;
 
 mod directory;
+mod directory_lock;
+mod footer;
 mod managed_directory;
 mod ram_directory;
 mod read_only_source;
-mod shared_vec_slice;
+mod watch_event_router;
 
 /// Errors specific to the directory module.
 pub mod error;
 
-use std::io::{BufWriter, Seek, Write};
-
+pub use self::directory::DirectoryLock;
 pub use self::directory::{Directory, DirectoryClone};
+pub use self::directory_lock::{Lock, INDEX_WRITER_LOCK, META_LOCK};
 pub use self::ram_directory::RAMDirectory;
 pub use self::read_only_source::ReadOnlySource;
+pub use self::watch_event_router::{WatchCallback, WatchCallbackList, WatchHandle};
+use std::io::{self, BufWriter, Write};
+use std::path::PathBuf;
+/// Outcome of the Garbage collection
+pub struct GarbageCollectionResult {
+    /// List of files that were deleted in this cycle
+    pub deleted_files: Vec<PathBuf>,
+    /// List of files that were schedule to be deleted in this cycle,
+    /// but deletion did not work. This typically happens on windows,
+    /// as deleting a memory mapped file is forbidden.
+    ///
+    /// If a searcher is still held, a file cannot be deleted.
+    /// This is not considered a bug, the file will simply be deleted
+    /// in the next GC.
+    pub failed_to_delete_files: Vec<PathBuf>,
+}
 
 #[cfg(feature = "mmap")]
 pub use self::mmap_directory::MmapDirectory;
 
-pub(crate) use self::managed_directory::ManagedDirectory;
+pub use self::managed_directory::ManagedDirectory;
 
-/// Synonym of Seek + Write
-pub trait SeekableWrite: Seek + Write {}
-impl<T: Seek + Write> SeekableWrite for T {}
+/// Struct used to prevent from calling [`terminate_ref`](trait.TerminatingWrite#method.terminate_ref) directly
+///
+/// The point is that while the type is public, it cannot be built by anyone
+/// outside of this module.
+pub struct AntiCallToken(());
+
+/// Trait used to indicate when no more write need to be done on a writer
+pub trait TerminatingWrite: Write {
+    /// Indicate that the writer will no longer be used. Internally call terminate_ref.
+    fn terminate(mut self) -> io::Result<()>
+    where
+        Self: Sized,
+    {
+        self.terminate_ref(AntiCallToken(()))
+    }
+
+    /// You should implement this function to define custom behavior.
+    /// This function should flush any buffer it may hold.
+    fn terminate_ref(&mut self, _: AntiCallToken) -> io::Result<()>;
+}
+
+impl<W: TerminatingWrite + ?Sized> TerminatingWrite for Box<W> {
+    fn terminate_ref(&mut self, token: AntiCallToken) -> io::Result<()> {
+        self.as_mut().terminate_ref(token)
+    }
+}
+
+impl<W: TerminatingWrite> TerminatingWrite for BufWriter<W> {
+    fn terminate_ref(&mut self, a: AntiCallToken) -> io::Result<()> {
+        self.flush()?;
+        self.get_mut().terminate_ref(a)
+    }
+}
+
+#[cfg(test)]
+impl<'a> TerminatingWrite for &'a mut Vec<u8> {
+    fn terminate_ref(&mut self, _a: AntiCallToken) -> io::Result<()> {
+        self.flush()
+    }
+}
 
 /// Write object for Directory.
 ///
 /// `WritePtr` are required to implement both Write
 /// and Seek.
-pub type WritePtr = BufWriter<Box<SeekableWrite>>;
+pub type WritePtr = BufWriter<Box<dyn TerminatingWrite>>;
 
 #[cfg(test)]
-mod tests {
-
-    use super::*;
-    use std::io::{Seek, SeekFrom, Write};
-    use std::path::Path;
-
-    lazy_static! {
-        static ref TEST_PATH: &'static Path = Path::new("some_path_for_test");
-    }
-
-    #[test]
-    fn test_ram_directory() {
-        let mut ram_directory = RAMDirectory::create();
-        test_directory(&mut ram_directory);
-    }
-
-    #[test]
-    #[cfg(feature = "mmap")]
-    fn test_mmap_directory() {
-        let mut mmap_directory = MmapDirectory::create_from_tempdir().unwrap();
-        test_directory(&mut mmap_directory);
-    }
-
-    #[test]
-    #[should_panic]
-    fn ram_directory_panics_if_flush_forgotten() {
-        let mut ram_directory = RAMDirectory::create();
-        let mut write_file = ram_directory.open_write(*TEST_PATH).unwrap();
-        assert!(write_file.write_all(&[4]).is_ok());
-    }
-
-    fn test_simple(directory: &mut Directory) {
-        {
-            {
-                let mut write_file = directory.open_write(*TEST_PATH).unwrap();
-                assert!(directory.exists(*TEST_PATH));
-                write_file.write_all(&[4]).unwrap();
-                write_file.write_all(&[3]).unwrap();
-                write_file.write_all(&[7, 3, 5]).unwrap();
-                write_file.flush().unwrap();
-            }
-            let read_file = directory.open_read(*TEST_PATH).unwrap();
-            let data: &[u8] = &*read_file;
-            assert_eq!(data, &[4u8, 3u8, 7u8, 3u8, 5u8]);
-        }
-
-        assert!(directory.delete(*TEST_PATH).is_ok());
-        assert!(!directory.exists(*TEST_PATH));
-    }
-
-    fn test_seek(directory: &mut Directory) {
-        {
-            {
-                let mut write_file = directory.open_write(*TEST_PATH).unwrap();
-                write_file.write_all(&[4, 3, 7, 3, 5]).unwrap();
-                write_file.seek(SeekFrom::Start(0)).unwrap();
-                write_file.write_all(&[3, 1]).unwrap();
-                write_file.flush().unwrap();
-            }
-            let read_file = directory.open_read(*TEST_PATH).unwrap();
-            let data: &[u8] = &*read_file;
-            assert_eq!(data, &[3u8, 1u8, 7u8, 3u8, 5u8]);
-        }
-
-        assert!(directory.delete(*TEST_PATH).is_ok());
-    }
-
-    fn test_rewrite_forbidden(directory: &mut Directory) {
-        {
-            directory.open_write(*TEST_PATH).unwrap();
-            assert!(directory.exists(*TEST_PATH));
-        }
-        {
-            assert!(directory.open_write(*TEST_PATH).is_err());
-        }
-        assert!(directory.delete(*TEST_PATH).is_ok());
-    }
-
-    fn test_write_create_the_file(directory: &mut Directory) {
-        {
-            assert!(directory.open_read(*TEST_PATH).is_err());
-            let _w = directory.open_write(*TEST_PATH).unwrap();
-            assert!(directory.exists(*TEST_PATH));
-            assert!(directory.open_read(*TEST_PATH).is_ok());
-            assert!(directory.delete(*TEST_PATH).is_ok());
-        }
-    }
-
-    fn test_directory_delete(directory: &mut Directory) {
-        assert!(directory.open_read(*TEST_PATH).is_err());
-        let mut write_file = directory.open_write(*TEST_PATH).unwrap();
-        write_file.write_all(&[1, 2, 3, 4]).unwrap();
-        write_file.flush().unwrap();
-        {
-            let read_handle = directory.open_read(*TEST_PATH).unwrap();
-            {
-                assert_eq!(&*read_handle, &[1u8, 2u8, 3u8, 4u8]);
-
-                // Mapped files can't be deleted on Windows
-                if !cfg!(windows) {
-                    assert!(directory.delete(*TEST_PATH).is_ok());
-                    assert_eq!(&*read_handle, &[1u8, 2u8, 3u8, 4u8]);
-                }
-
-                assert!(directory.delete(Path::new("SomeOtherPath")).is_err());
-            }
-        }
-
-        if cfg!(windows) {
-            assert!(directory.delete(*TEST_PATH).is_ok());
-        }
-
-        assert!(directory.open_read(*TEST_PATH).is_err());
-        assert!(directory.delete(*TEST_PATH).is_err());
-    }
-
-    fn test_directory(directory: &mut Directory) {
-        test_simple(directory);
-        test_seek(directory);
-        test_rewrite_forbidden(directory);
-        test_write_create_the_file(directory);
-        test_directory_delete(directory);
-    }
-
-}
+mod tests;

src/directory/ram_directory.rs

@@ -1,8 +1,10 @@
-use super::shared_vec_slice::SharedVecSlice;
-use common::make_io_err;
-use directory::error::{DeleteError, IOError, OpenReadError, OpenWriteError};
-use directory::WritePtr;
-use directory::{Directory, ReadOnlySource};
+use crate::core::META_FILEPATH;
+use crate::directory::error::{DeleteError, OpenReadError, OpenWriteError};
+use crate::directory::AntiCallToken;
+use crate::directory::WatchCallbackList;
+use crate::directory::{Directory, ReadOnlySource, WatchCallback, WatchHandle};
+use crate::directory::{TerminatingWrite, WritePtr};
+use fail::fail_point;
 use std::collections::HashMap;
 use std::fmt;
 use std::io::{self, BufWriter, Cursor, Seek, SeekFrom, Write};
@@ -22,13 +24,13 @@ use std::sync::{Arc, RwLock};
 ///
 struct VecWriter {
     path: PathBuf,
-    shared_directory: InnerDirectory,
+    shared_directory: RAMDirectory,
     data: Cursor<Vec<u8>>,
     is_flushed: bool,
 }
 
 impl VecWriter {
-    fn new(path_buf: PathBuf, shared_directory: InnerDirectory) -> VecWriter {
+    fn new(path_buf: PathBuf, shared_directory: RAMDirectory) -> VecWriter {
         VecWriter {
             path: path_buf,
             data: Cursor::new(Vec::new()),
@@ -64,78 +66,59 @@ impl Write for VecWriter {
 
     fn flush(&mut self) -> io::Result<()> {
         self.is_flushed = true;
-        self.shared_directory
-            .write(self.path.clone(), self.data.get_ref())?;
+        let mut fs = self.shared_directory.fs.write().unwrap();
+        fs.write(self.path.clone(), self.data.get_ref());
         Ok(())
     }
 }
 
-#[derive(Clone)]
-struct InnerDirectory(Arc<RwLock<HashMap<PathBuf, Arc<Vec<u8>>>>>);
+impl TerminatingWrite for VecWriter {
+    fn terminate_ref(&mut self, _: AntiCallToken) -> io::Result<()> {
+        self.flush()
+    }
+}
+
+#[derive(Default)]
+struct InnerDirectory {
+    fs: HashMap<PathBuf, ReadOnlySource>,
+    watch_router: WatchCallbackList,
+}
 
 impl InnerDirectory {
-    fn new() -> InnerDirectory {
-        InnerDirectory(Arc::new(RwLock::new(HashMap::new())))
-    }
-
-    fn write(&self, path: PathBuf, data: &[u8]) -> io::Result<bool> {
-        let mut map = self.0.write().map_err(|_| {
-            make_io_err(format!(
-                "Failed to lock the directory, when trying to write {:?}",
-                path
-            ))
-        })?;
-        let prev_value = map.insert(path, Arc::new(Vec::from(data)));
-        Ok(prev_value.is_some())
+    fn write(&mut self, path: PathBuf, data: &[u8]) -> bool {
+        let data = ReadOnlySource::new(Vec::from(data));
+        self.fs.insert(path, data).is_some()
     }
 
     fn open_read(&self, path: &Path) -> Result<ReadOnlySource, OpenReadError> {
-        self.0
-            .read()
-            .map_err(|_| {
-                let msg = format!(
-                    "Failed to acquire read lock for the \
-                     directory when trying to read {:?}",
-                    path
-                );
-                let io_err = make_io_err(msg);
-                OpenReadError::IOError(IOError::with_path(path.to_owned(), io_err))
-            }).and_then(|readable_map| {
-                readable_map
-                    .get(path)
-                    .ok_or_else(|| OpenReadError::FileDoesNotExist(PathBuf::from(path)))
-                    .map(Arc::clone)
-                    .map(|data| ReadOnlySource::Anonymous(SharedVecSlice::new(data)))
-            })
+        self.fs
+            .get(path)
+            .ok_or_else(|| OpenReadError::FileDoesNotExist(PathBuf::from(path)))
+            .map(Clone::clone)
     }
 
-    fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
-        self.0
-            .write()
-            .map_err(|_| {
-                let msg = format!(
-                    "Failed to acquire write lock for the \
-                     directory when trying to delete {:?}",
-                    path
-                );
-                let io_err = make_io_err(msg);
-                DeleteError::IOError(IOError::with_path(path.to_owned(), io_err))
-            }).and_then(|mut writable_map| match writable_map.remove(path) {
-                Some(_) => Ok(()),
-                None => Err(DeleteError::FileDoesNotExist(PathBuf::from(path))),
-            })
+    fn delete(&mut self, path: &Path) -> result::Result<(), DeleteError> {
+        match self.fs.remove(path) {
+            Some(_) => Ok(()),
+            None => Err(DeleteError::FileDoesNotExist(PathBuf::from(path))),
+        }
     }
 
     fn exists(&self, path: &Path) -> bool {
-        self.0
-            .read()
-            .expect("Failed to get read lock directory.")
-            .contains_key(path)
+        self.fs.contains_key(path)
+    }
+
+    fn watch(&mut self, watch_handle: WatchCallback) -> WatchHandle {
+        self.watch_router.subscribe(watch_handle)
+    }
+
+    fn total_mem_usage(&self) -> usize {
+        self.fs.values().map(|f| f.len()).sum()
     }
 }
 
 impl fmt::Debug for RAMDirectory {
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         write!(f, "RAMDirectory")
     }
 }
@@ -145,33 +128,63 @@ impl fmt::Debug for RAMDirectory {
 /// It is mainly meant for unit testing.
 /// Writes are only made visible upon flushing.
 ///
-#[derive(Clone)]
+#[derive(Clone, Default)]
 pub struct RAMDirectory {
-    fs: InnerDirectory,
+    fs: Arc<RwLock<InnerDirectory>>,
 }
 
 impl RAMDirectory {
     /// Constructor
     pub fn create() -> RAMDirectory {
-        RAMDirectory {
-            fs: InnerDirectory::new(),
+        Self::default()
+    }
+
+    /// Returns the sum of the size of the different files
+    /// in the RAMDirectory.
+    pub fn total_mem_usage(&self) -> usize {
+        self.fs.read().unwrap().total_mem_usage()
+    }
+
+    /// Write a copy of all of the files saved in the RAMDirectory in the target `Directory`.
+    ///
+    /// Files are all written using the `Directory::write` meaning, even if they were
+    /// written using the `atomic_write` api.
+    ///
+    /// If an error is encounterred, files may be persisted partially.
+    pub fn persist(&self, dest: &mut dyn Directory) -> crate::Result<()> {
+        let wlock = self.fs.write().unwrap();
+        for (path, source) in wlock.fs.iter() {
+            let mut dest_wrt = dest.open_write(path)?;
+            dest_wrt.write_all(source.as_slice())?;
+            dest_wrt.terminate()?;
         }
+        Ok(())
     }
 }
 
 impl Directory for RAMDirectory {
     fn open_read(&self, path: &Path) -> result::Result<ReadOnlySource, OpenReadError> {
-        self.fs.open_read(path)
+        self.fs.read().unwrap().open_read(path)
+    }
+
+    fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
+        fail_point!("RAMDirectory::delete", |_| {
+            use crate::directory::error::IOError;
+            let io_error = IOError::from(io::Error::from(io::ErrorKind::Other));
+            Err(DeleteError::from(io_error))
+        });
+        self.fs.write().unwrap().delete(path)
+    }
+
+    fn exists(&self, path: &Path) -> bool {
+        self.fs.read().unwrap().exists(path)
     }
 
     fn open_write(&mut self, path: &Path) -> Result<WritePtr, OpenWriteError> {
+        let mut fs = self.fs.write().unwrap();
         let path_buf = PathBuf::from(path);
-        let vec_writer = VecWriter::new(path_buf.clone(), self.fs.clone());
-
-        let exists = self
-            .fs
-            .write(path_buf.clone(), &Vec::new())
-            .map_err(|err| IOError::with_path(path.to_owned(), err))?;
+        let vec_writer = VecWriter::new(path_buf.clone(), self.clone());
+        let exists = fs.write(path_buf.clone(), &[]);
         // force the creation of the file to mimic the MMap directory.
         if exists {
             Err(OpenWriteError::FileAlreadyExists(path_buf))
@@ -180,29 +193,55 @@ impl Directory for RAMDirectory {
         }
     }
 
-    fn delete(&self, path: &Path) -> result::Result<(), DeleteError> {
-        self.fs.delete(path)
-    }
-
-    fn exists(&self, path: &Path) -> bool {
-        self.fs.exists(path)
-    }
-
     fn atomic_read(&self, path: &Path) -> Result<Vec<u8>, OpenReadError> {
-        let read = self.open_read(path)?;
-        Ok(read.as_slice().to_owned())
+        Ok(self.open_read(path)?.as_slice().to_owned())
     }
 
     fn atomic_write(&mut self, path: &Path, data: &[u8]) -> io::Result<()> {
         fail_point!("RAMDirectory::atomic_write", |msg| Err(io::Error::new(
             io::ErrorKind::Other,
-            msg.unwrap_or("Undefined".to_string())
+            msg.unwrap_or_else(|| "Undefined".to_string())
         )));
         let path_buf = PathBuf::from(path);
-        let mut vec_writer = VecWriter::new(path_buf.clone(), self.fs.clone());
-        self.fs.write(path_buf, &Vec::new())?;
+
+        // Reserve the path to prevent calls to .write() to succeed.
+        self.fs.write().unwrap().write(path_buf.clone(), &[]);
+
+        let mut vec_writer = VecWriter::new(path_buf, self.clone());
         vec_writer.write_all(data)?;
         vec_writer.flush()?;
+        if path == Path::new(&*META_FILEPATH) {
+            let _ = self.fs.write().unwrap().watch_router.broadcast();
+        }
         Ok(())
     }
+
+    fn watch(&self, watch_callback: WatchCallback) -> crate::Result<WatchHandle> {
+        Ok(self.fs.write().unwrap().watch(watch_callback))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::RAMDirectory;
+    use crate::Directory;
+    use std::io::Write;
+    use std::path::Path;
+
+    #[test]
+    fn test_persist() {
+        let msg_atomic: &'static [u8] = b"atomic is the way";
+        let msg_seq: &'static [u8] = b"sequential is the way";
+        let path_atomic: &'static Path = Path::new("atomic");
+        let path_seq: &'static Path = Path::new("seq");
+        let mut directory = RAMDirectory::create();
+        assert!(directory.atomic_write(path_atomic, msg_atomic).is_ok());
+        let mut wrt = directory.open_write(path_seq).unwrap();
+        assert!(wrt.write_all(msg_seq).is_ok());
+        assert!(wrt.flush().is_ok());
+        let mut directory_copy = RAMDirectory::create();
+        assert!(directory.persist(&mut directory_copy).is_ok());
+        assert_eq!(directory_copy.atomic_read(path_atomic).unwrap(), msg_atomic);
+        assert_eq!(directory_copy.atomic_read(path_seq).unwrap(), msg_seq);
+    }
 }

src/directory/read_only_source.rs

@@ -1,9 +1,9 @@
-use super::shared_vec_slice::SharedVecSlice;
-use common::HasLen;
-#[cfg(feature = "mmap")]
-use fst::raw::MmapReadOnly;
+use crate::common::HasLen;
 use stable_deref_trait::{CloneStableDeref, StableDeref};
 use std::ops::Deref;
+use std::sync::Arc;
+
+pub type BoxedData = Box<dyn Deref<Target = [u8]> + Send + Sync + 'static>;
 
 /// Read object that represents files in tantivy.
 ///
@@ -11,12 +11,10 @@ use std::ops::Deref;
 /// the data in the form of a constant read-only `&[u8]`.
 /// Whatever happens to the directory file, the data
 /// hold by this object should never be altered or destroyed.
-pub enum ReadOnlySource {
-    /// Mmap source of data
-    #[cfg(feature = "mmap")]
-    Mmap(MmapReadOnly),
-    /// Wrapping a `Vec<u8>`
-    Anonymous(SharedVecSlice),
+pub struct ReadOnlySource {
+    data: Arc<BoxedData>,
+    start: usize,
+    stop: usize,
 }
 
 unsafe impl StableDeref for ReadOnlySource {}
@@ -30,19 +28,38 @@ impl Deref for ReadOnlySource {
     }
 }
 
+impl From<Arc<BoxedData>> for ReadOnlySource {
+    fn from(data: Arc<BoxedData>) -> Self {
+        let len = data.len();
+        ReadOnlySource {
+            data,
+            start: 0,
+            stop: len,
+        }
+    }
+}
+
 impl ReadOnlySource {
+    pub(crate) fn new<D>(data: D) -> ReadOnlySource
+    where
+        D: Deref<Target = [u8]> + Send + Sync + 'static,
+    {
+        let len = data.len();
+        ReadOnlySource {
+            data: Arc::new(Box::new(data)),
+            start: 0,
+            stop: len,
+        }
+    }
+
     /// Creates an empty ReadOnlySource
     pub fn empty() -> ReadOnlySource {
-        ReadOnlySource::Anonymous(SharedVecSlice::empty())
+        ReadOnlySource::new(&[][..])
     }
 
     /// Returns the data underlying the ReadOnlySource object.
     pub fn as_slice(&self) -> &[u8] {
-        match *self {
-            #[cfg(feature = "mmap")]
-            ReadOnlySource::Mmap(ref mmap_read_only) => mmap_read_only.as_slice(),
-            ReadOnlySource::Anonymous(ref shared_vec) => shared_vec.as_slice(),
-        }
+        &self.data[self.start..self.stop]
     }
 
     /// Splits into 2 `ReadOnlySource`, at the offset given
@@ -53,6 +70,12 @@ impl ReadOnlySource {
         (left, right)
     }
 
+    /// Splits into 2 `ReadOnlySource`, at the offset `end - right_len`.
+    pub fn split_from_end(self, right_len: usize) -> (ReadOnlySource, ReadOnlySource) {
+        let left_len = self.len() - right_len;
+        self.split(left_len)
+    }
+
     /// Creates a ReadOnlySource that is just a
     /// view over a slice of the data.
     ///
@@ -63,22 +86,18 @@ impl ReadOnlySource {
     /// worth of data in anonymous memory, and only a
     /// 1KB slice is remaining, the whole `500MBs`
     /// are retained in memory.
-    pub fn slice(&self, from_offset: usize, to_offset: usize) -> ReadOnlySource {
+    pub fn slice(&self, start: usize, stop: usize) -> ReadOnlySource {
         assert!(
-            from_offset <= to_offset,
+            start <= stop,
             "Requested negative slice [{}..{}]",
-            from_offset,
-            to_offset
+            start,
+            stop
         );
-        match *self {
-            #[cfg(feature = "mmap")]
-            ReadOnlySource::Mmap(ref mmap_read_only) => {
-                let sliced_mmap = mmap_read_only.range(from_offset, to_offset - from_offset);
-                ReadOnlySource::Mmap(sliced_mmap)
-            }
-            ReadOnlySource::Anonymous(ref shared_vec) => {
-                ReadOnlySource::Anonymous(shared_vec.slice(from_offset, to_offset))
-            }
+        assert!(stop <= self.len());
+        ReadOnlySource {
+            data: self.data.clone(),
+            start: self.start + start,
+            stop: self.start + stop,
         }
     }
 
@@ -87,8 +106,7 @@ impl ReadOnlySource {
     ///
     /// Equivalent to `.slice(from_offset, self.len())`
     pub fn slice_from(&self, from_offset: usize) -> ReadOnlySource {
-        let len = self.len();
-        self.slice(from_offset, len)
+        self.slice(from_offset, self.len())
     }
 
     /// Like `.slice(...)` but enforcing only the `to`
@@ -102,19 +120,18 @@ impl ReadOnlySource {
 
 impl HasLen for ReadOnlySource {
     fn len(&self) -> usize {
-        self.as_slice().len()
+        self.stop - self.start
     }
 }
 
 impl Clone for ReadOnlySource {
     fn clone(&self) -> Self {
-        self.slice(0, self.len())
+        self.slice_from(0)
     }
 }
 
 impl From<Vec<u8>> for ReadOnlySource {
     fn from(data: Vec<u8>) -> ReadOnlySource {
-        let shared_data = SharedVecSlice::from(data);
-        ReadOnlySource::Anonymous(shared_data)
+        ReadOnlySource::new(data)
     }
 }

src/directory/shared_vec_slice.rs

@@ -1,41 +0,0 @@
-use std::sync::Arc;
-
-#[derive(Clone)]
-pub struct SharedVecSlice {
-    pub data: Arc<Vec<u8>>,
-    pub start: usize,
-    pub len: usize,
-}
-
-impl SharedVecSlice {
-    pub fn empty() -> SharedVecSlice {
-        SharedVecSlice::new(Arc::new(Vec::new()))
-    }
-
-    pub fn new(data: Arc<Vec<u8>>) -> SharedVecSlice {
-        let data_len = data.len();
-        SharedVecSlice {
-            data,
-            start: 0,
-            len: data_len,
-        }
-    }
-
-    pub fn as_slice(&self) -> &[u8] {
-        &self.data[self.start..self.start + self.len]
-    }
-
-    pub fn slice(&self, from_offset: usize, to_offset: usize) -> SharedVecSlice {
-        SharedVecSlice {
-            data: Arc::clone(&self.data),
-            start: self.start + from_offset,
-            len: to_offset - from_offset,
-        }
-    }
-}
-
-impl From<Vec<u8>> for SharedVecSlice {
-    fn from(data: Vec<u8>) -> SharedVecSlice {
-        SharedVecSlice::new(Arc::new(data))
-    }
-}

src/directory/tests.rs

@@ -0,0 +1,293 @@
+use super::*;
+use futures::channel::oneshot;
+use futures::executor::block_on;
+use std::io::Write;
+use std::mem;
+use std::path::{Path, PathBuf};
+use std::sync::atomic::Ordering::SeqCst;
+use std::sync::atomic::{AtomicBool, AtomicUsize};
+use std::sync::Arc;
+use std::time::Duration;
+
+#[cfg(feature = "mmap")]
+mod mmap_directory_tests {
+    use crate::directory::MmapDirectory;
+
+    type DirectoryImpl = MmapDirectory;
+
+    fn make_directory() -> DirectoryImpl {
+        MmapDirectory::create_from_tempdir().unwrap()
+    }
+
+    #[test]
+    fn test_simple() {
+        let mut directory = make_directory();
+        super::test_simple(&mut directory);
+    }
+
+    #[test]
+    fn test_write_create_the_file() {
+        let mut directory = make_directory();
+        super::test_write_create_the_file(&mut directory);
+    }
+
+    #[test]
+    fn test_rewrite_forbidden() {
+        let mut directory = make_directory();
+        super::test_rewrite_forbidden(&mut directory);
+    }
+
+    #[test]
+    fn test_directory_delete() {
+        let mut directory = make_directory();
+        super::test_directory_delete(&mut directory);
+    }
+
+    #[test]
+    fn test_lock_non_blocking() {
+        let mut directory = make_directory();
+        super::test_lock_non_blocking(&mut directory);
+    }
+
+    #[test]
+    fn test_lock_blocking() {
+        let mut directory = make_directory();
+        super::test_lock_blocking(&mut directory);
+    }
+
+    #[test]
+    fn test_watch() {
+        let mut directory = make_directory();
+        super::test_watch(&mut directory);
+    }
+}
+
+mod ram_directory_tests {
+    use crate::directory::RAMDirectory;
+
+    type DirectoryImpl = RAMDirectory;
+
+    fn make_directory() -> DirectoryImpl {
+        RAMDirectory::default()
+    }
+
+    #[test]
+    fn test_simple() {
+        let mut directory = make_directory();
+        super::test_simple(&mut directory);
+    }
+
+    #[test]
+    fn test_write_create_the_file() {
+        let mut directory = make_directory();
+        super::test_write_create_the_file(&mut directory);
+    }
+
+    #[test]
+    fn test_rewrite_forbidden() {
+        let mut directory = make_directory();
+        super::test_rewrite_forbidden(&mut directory);
+    }
+
+    #[test]
+    fn test_directory_delete() {
+        let mut directory = make_directory();
+        super::test_directory_delete(&mut directory);
+    }
+
+    #[test]
+    fn test_lock_non_blocking() {
+        let mut directory = make_directory();
+        super::test_lock_non_blocking(&mut directory);
+    }
+
+    #[test]
+    fn test_lock_blocking() {
+        let mut directory = make_directory();
+        super::test_lock_blocking(&mut directory);
+    }
+
+    #[test]
+    fn test_watch() {
+        let mut directory = make_directory();
+        super::test_watch(&mut directory);
+    }
+}
+
+#[test]
+#[should_panic]
+fn ram_directory_panics_if_flush_forgotten() {
+    let test_path: &'static Path = Path::new("some_path_for_test");
+    let mut ram_directory = RAMDirectory::create();
+    let mut write_file = ram_directory.open_write(test_path).unwrap();
+    assert!(write_file.write_all(&[4]).is_ok());
+}
+
+fn test_simple(directory: &mut dyn Directory) {
+    let test_path: &'static Path = Path::new("some_path_for_test");
+    {
+        let mut write_file = directory.open_write(test_path).unwrap();
+        assert!(directory.exists(test_path));
+        write_file.write_all(&[4]).unwrap();
+        write_file.write_all(&[3]).unwrap();
+        write_file.write_all(&[7, 3, 5]).unwrap();
+        write_file.flush().unwrap();
+    }
+    {
+        let read_file = directory.open_read(test_path).unwrap();
+        let data: &[u8] = &*read_file;
+        assert_eq!(data, &[4u8, 3u8, 7u8, 3u8, 5u8]);
+    }
+    assert!(directory.delete(test_path).is_ok());
+    assert!(!directory.exists(test_path));
+}
+
+fn test_rewrite_forbidden(directory: &mut dyn Directory) {
+    let test_path: &'static Path = Path::new("some_path_for_test");
+    {
+        directory.open_write(test_path).unwrap();
+        assert!(directory.exists(test_path));
+    }
+    {
+        assert!(directory.open_write(test_path).is_err());
+    }
+    assert!(directory.delete(test_path).is_ok());
+}
+
+fn test_write_create_the_file(directory: &mut dyn Directory) {
+    let test_path: &'static Path = Path::new("some_path_for_test");
+    {
+        assert!(directory.open_read(test_path).is_err());
+        let _w = directory.open_write(test_path).unwrap();
+        assert!(directory.exists(test_path));
+        assert!(directory.open_read(test_path).is_ok());
+        assert!(directory.delete(test_path).is_ok());
+    }
+}
+
+fn test_directory_delete(directory: &mut dyn Directory) {
+    let test_path: &'static Path = Path::new("some_path_for_test");
+    assert!(directory.open_read(test_path).is_err());
+    let mut write_file = directory.open_write(&test_path).unwrap();
+    write_file.write_all(&[1, 2, 3, 4]).unwrap();
+    write_file.flush().unwrap();
+    {
+        let read_handle = directory.open_read(&test_path).unwrap();
+        assert_eq!(&*read_handle, &[1u8, 2u8, 3u8, 4u8]);
+        // Mapped files can't be deleted on Windows
+        if !cfg!(windows) {
+            assert!(directory.delete(&test_path).is_ok());
+            assert_eq!(&*read_handle, &[1u8, 2u8, 3u8, 4u8]);
+        }
+
+        assert!(directory.delete(Path::new("SomeOtherPath")).is_err());
+    }
+
+    if cfg!(windows) {
+        assert!(directory.delete(&test_path).is_ok());
+    }
+
+    assert!(directory.open_read(&test_path).is_err());
+    assert!(directory.delete(&test_path).is_err());
+}
+
+fn test_watch(directory: &mut dyn Directory) {
+    let num_progress: Arc<AtomicUsize> = Default::default();
+    let counter: Arc<AtomicUsize> = Default::default();
+    let counter_clone = counter.clone();
+    let (sender, receiver) = crossbeam::channel::unbounded();
+    let watch_callback = Box::new(move || {
+        counter_clone.fetch_add(1, SeqCst);
+    });
+    // This callback is used to synchronize watching in our unit test.
+    // We bind it to a variable because the callback is removed when that
+    // handle is dropped.
+    let watch_handle = directory.watch(watch_callback).unwrap();
+    let _progress_listener = directory
+        .watch(Box::new(move || {
+            let val = num_progress.fetch_add(1, SeqCst);
+            let _ = sender.send(val);
+        }))
+        .unwrap();
+
+    for i in 0..10 {
+        assert_eq!(i, counter.load(SeqCst));
+        assert!(directory
+            .atomic_write(Path::new("meta.json"), b"random_test_data_2")
+            .is_ok());
+        assert_eq!(receiver.recv_timeout(Duration::from_millis(500)), Ok(i));
+        assert_eq!(i + 1, counter.load(SeqCst));
+    }
+    mem::drop(watch_handle);
+    assert!(directory
+        .atomic_write(Path::new("meta.json"), b"random_test_data")
+        .is_ok());
+    assert!(receiver.recv_timeout(Duration::from_millis(500)).is_ok());
+    assert_eq!(10, counter.load(SeqCst));
+}
+
+fn test_lock_non_blocking(directory: &mut dyn Directory) {
+    {
+        let lock_a_res = directory.acquire_lock(&Lock {
+            filepath: PathBuf::from("a.lock"),
+            is_blocking: false,
+        });
+        assert!(lock_a_res.is_ok());
+        let lock_b_res = directory.acquire_lock(&Lock {
+            filepath: PathBuf::from("b.lock"),
+            is_blocking: false,
+        });
+        assert!(lock_b_res.is_ok());
+        let lock_a_res2 = directory.acquire_lock(&Lock {
+            filepath: PathBuf::from("a.lock"),
+            is_blocking: false,
+        });
+        assert!(lock_a_res2.is_err());
+    }
+    let lock_a_res = directory.acquire_lock(&Lock {
+        filepath: PathBuf::from("a.lock"),
+        is_blocking: false,
+    });
+    assert!(lock_a_res.is_ok());
+}
+
+fn test_lock_blocking(directory: &mut dyn Directory) {
+    let lock_a_res = directory.acquire_lock(&Lock {
+        filepath: PathBuf::from("a.lock"),
+        is_blocking: true,
+    });
+    assert!(lock_a_res.is_ok());
+    let in_thread = Arc::new(AtomicBool::default());
+    let in_thread_clone = in_thread.clone();
+    let (sender, receiver) = oneshot::channel();
+    std::thread::spawn(move || {
+        //< lock_a_res is sent to the thread.
+        in_thread_clone.store(true, SeqCst);
+        let _just_sync = block_on(receiver);
+        // explicitely droping lock_a_res. It would have been sufficient to just force it
+        // to be part of the move, but the intent seems clearer that way.
+        drop(lock_a_res);
+    });
+    {
+        // A non-blocking call should fail, as the thread is running and holding the lock.
+        let lock_a_res = directory.acquire_lock(&Lock {
+            filepath: PathBuf::from("a.lock"),
+            is_blocking: false,
+        });
+        assert!(lock_a_res.is_err());
+    }
+    let directory_clone = directory.box_clone();
+    let (sender2, receiver2) = oneshot::channel();
+    let join_handle = std::thread::spawn(move || {
+        assert!(sender2.send(()).is_ok());
+        let lock_a_res = directory_clone.acquire_lock(&Lock {
+            filepath: PathBuf::from("a.lock"),
+            is_blocking: true,
+        });
+        assert!(in_thread.load(SeqCst));
+        assert!(lock_a_res.is_ok());
+    });
+    assert!(block_on(receiver2).is_ok());
+    assert!(sender.send(()).is_ok());
+    assert!(join_handle.join().is_ok());
+}

src/directory/watch_event_router.rs

@@ -0,0 +1,170 @@
+use futures::channel::oneshot;
+use futures::{Future, TryFutureExt};
+use std::sync::Arc;
+use std::sync::RwLock;
+use std::sync::Weak;
+
+/// Type alias for callbacks registered when watching files of a `Directory`.
+pub type WatchCallback = Box<dyn Fn() -> () + Sync + Send>;
+
+/// Helper struct to implement the watch method in `Directory` implementations.
+///
+/// It registers callbacks (See `.subscribe(...)`) and
+/// calls them upon calls to `.broadcast(...)`.
+#[derive(Default)]
+pub struct WatchCallbackList {
+    router: RwLock<Vec<Weak<WatchCallback>>>,
+}
+
+/// Controls how long a directory should watch for a file change.
+///
+/// After all the clones of `WatchHandle` are dropped, the associated will not be called when a
+/// file change is detected.
+#[must_use = "This `WatchHandle` controls the lifetime of the watch and should therefore be used."]
+#[derive(Clone)]
+pub struct WatchHandle(Arc<WatchCallback>);
+
+impl WatchHandle {
+    /// Create a WatchHandle handle.
+    pub fn new(watch_callback: Arc<WatchCallback>) -> WatchHandle {
+        WatchHandle(watch_callback)
+    }
+}
+
+impl WatchCallbackList {
+    /// Suscribes a new callback and returns a handle that controls the lifetime of the callback.
+    pub fn subscribe(&self, watch_callback: WatchCallback) -> WatchHandle {
+        let watch_callback_arc = Arc::new(watch_callback);
+        let watch_callback_weak = Arc::downgrade(&watch_callback_arc);
+        self.router.write().unwrap().push(watch_callback_weak);
+        WatchHandle::new(watch_callback_arc)
+    }
+
+    fn list_callback(&self) -> Vec<Arc<WatchCallback>> {
+        let mut callbacks = vec![];
+        let mut router_wlock = self.router.write().unwrap();
+        let mut i = 0;
+        while i < router_wlock.len() {
+            if let Some(watch) = router_wlock[i].upgrade() {
+                callbacks.push(watch);
+                i += 1;
+            } else {
+                router_wlock.swap_remove(i);
+            }
+        }
+        callbacks
+    }
+
+    /// Triggers all callbacks
+    pub fn broadcast(&self) -> impl Future<Output = ()> {
+        let callbacks = self.list_callback();
+        let (sender, receiver) = oneshot::channel();
+        let result = receiver.unwrap_or_else(|_| ());
+        if callbacks.is_empty() {
+            let _ = sender.send(());
+            return result;
+        }
+        let spawn_res = std::thread::Builder::new()
+            .name("watch-callbacks".to_string())
+            .spawn(move || {
+                for callback in callbacks {
+                    callback();
+                }
+                let _ = sender.send(());
+            });
+        if let Err(err) = spawn_res {
+            error!(
+                "Failed to spawn thread to call watch callbacks. Cause: {:?}",
+                err
+            );
+        }
+        result
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use crate::directory::WatchCallbackList;
+    use futures::executor::block_on;
+    use std::mem;
+    use std::sync::atomic::{AtomicUsize, Ordering};
+    use std::sync::Arc;
+
+    #[test]
+    fn test_watch_event_router_simple() {
+        let watch_event_router = WatchCallbackList::default();
+        let counter: Arc<AtomicUsize> = Default::default();
+        let counter_clone = counter.clone();
+        let inc_callback = Box::new(move || {
+            counter_clone.fetch_add(1, Ordering::SeqCst);
+        });
+        block_on(watch_event_router.broadcast());
+        assert_eq!(0, counter.load(Ordering::SeqCst));
+        let handle_a = watch_event_router.subscribe(inc_callback);
+        assert_eq!(0, counter.load(Ordering::SeqCst));
+        block_on(watch_event_router.broadcast());
+        assert_eq!(1, counter.load(Ordering::SeqCst));
+        block_on(async {
+            (
+                watch_event_router.broadcast().await,
+                watch_event_router.broadcast().await,
+                watch_event_router.broadcast().await,
+            )
+        });
+        assert_eq!(4, counter.load(Ordering::SeqCst));
+        mem::drop(handle_a);
+        block_on(watch_event_router.broadcast());
+        assert_eq!(4, counter.load(Ordering::SeqCst));
+    }
+
+    #[test]
+    fn test_watch_event_router_multiple_callback_same_key() {
+        let watch_event_router = WatchCallbackList::default();
+        let counter: Arc<AtomicUsize> = Default::default();
+        let inc_callback = |inc: usize| {
+            let counter_clone = counter.clone();
+            Box::new(move || {
+                counter_clone.fetch_add(inc, Ordering::SeqCst);
+            })
+        };
+        let handle_a = watch_event_router.subscribe(inc_callback(1));
+        let handle_a2 = watch_event_router.subscribe(inc_callback(10));
+        assert_eq!(0, counter.load(Ordering::SeqCst));
+        block_on(async {
+            futures::join!(
+                watch_event_router.broadcast(),
+                watch_event_router.broadcast()
+            )
+        });
+        assert_eq!(22, counter.load(Ordering::SeqCst));
+        mem::drop(handle_a);
+        block_on(watch_event_router.broadcast());
+        assert_eq!(32, counter.load(Ordering::SeqCst));
+        mem::drop(handle_a2);
+        block_on(watch_event_router.broadcast());
+        block_on(watch_event_router.broadcast());
+        assert_eq!(32, counter.load(Ordering::SeqCst));
+    }
+
+    #[test]
+    fn test_watch_event_router_multiple_callback_different_key() {
+        let watch_event_router = WatchCallbackList::default();
+        let counter: Arc<AtomicUsize> = Default::default();
+        let counter_clone = counter.clone();
+        let inc_callback = Box::new(move || {
+            counter_clone.fetch_add(1, Ordering::SeqCst);
+        });
+        let handle_a = watch_event_router.subscribe(inc_callback);
+        assert_eq!(0, counter.load(Ordering::SeqCst));
+        block_on(async {
+            let future1 = watch_event_router.broadcast();
+            let future2 = watch_event_router.broadcast();
+            futures::join!(future1, future2)
+        });
+        assert_eq!(2, counter.load(Ordering::SeqCst));
+        mem::drop(handle_a);
+        let _ = watch_event_router.broadcast();
+        block_on(watch_event_router.broadcast());
+        assert_eq!(2, counter.load(Ordering::SeqCst));
+    }
+}

src/docset.rs

@@ -1,57 +1,47 @@
-use common::BitSet;
+use crate::fastfield::DeleteBitSet;
+use crate::DocId;
 use std::borrow::Borrow;
 use std::borrow::BorrowMut;
-use std::cmp::Ordering;
-use DocId;
 
-/// Expresses the outcome of a call to `DocSet`'s `.skip_next(...)`.
-#[derive(PartialEq, Eq, Debug)]
-pub enum SkipResult {
-    /// target was in the docset
-    Reached,
-    /// target was not in the docset, skipping stopped as a greater element was found
-    OverStep,
-    /// the docset was entirely consumed without finding the target, nor any
-    /// element greater than the target.
-    End,
-}
+/// Sentinel value returned when a DocSet has been entirely consumed.
+///
+/// This is not u32::MAX as one would have expected, due to the lack of SSE2 instructions
+/// to compare [u32; 4].
+pub const TERMINATED: DocId = std::i32::MAX as u32;
 
 /// Represents an iterable set of sorted doc ids.
 pub trait DocSet {
     /// Goes to the next element.
-    /// `.advance(...)` needs to be called a first time to point to the correct
-    /// element.
-    fn advance(&mut self) -> bool;
+    ///
+    /// The DocId of the next element is returned.
+    /// In other words we should always have :
+    /// ```ignore
+    /// let doc = docset.advance();
+    /// assert_eq!(doc, docset.doc());
+    /// ```
+    ///
+    /// If we reached the end of the DocSet, TERMINATED should be returned.
+    ///
+    /// Calling `.advance()` on a terminated DocSet should be supported, and TERMINATED should
+    /// be returned.
+    /// TODO Test existing docsets.
+    fn advance(&mut self) -> DocId;
 
-    /// After skipping, position the iterator in such a way that `.doc()`
-    /// will return a value greater than or equal to target.
+    /// Advances the DocSet forward until reaching the target, or going to the
+    /// lowest DocId greater than the target.
     ///
-    /// SkipResult expresses whether the `target value` was reached, overstepped,
-    /// or if the `DocSet` was entirely consumed without finding any value
-    /// greater or equal to the `target`.
+    /// If the end of the DocSet is reached, TERMINATED is returned.
     ///
-    /// WARNING: Calling skip always advances the docset.
-    /// More specifically, if the docset is already positionned on the target
-    /// skipping will advance to the next position and return SkipResult::Overstep.
+    /// Calling `.seek(target)` on a terminated DocSet is legal. Implementation
+    /// of DocSet should support it.
     ///
-    /// If `.skip_next()` oversteps, then the docset must be positionned correctly
-    /// on an existing document. In other words, `.doc()` should return the first document
-    /// greater than `DocId`.
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
-        if !self.advance() {
-            return SkipResult::End;
-        }
-        loop {
-            match self.doc().cmp(&target) {
-                Ordering::Less => {
-                    if !self.advance() {
-                        return SkipResult::End;
-                    }
-                }
-                Ordering::Equal => return SkipResult::Reached,
-                Ordering::Greater => return SkipResult::OverStep,
-            }
+    /// Calling `seek(TERMINATED)` is also legal and is the normal way to consume a DocSet.
+    fn seek(&mut self, target: DocId) -> DocId {
+        let mut doc = self.doc();
+        while doc < target {
+            doc = self.advance();
         }
+        doc
     }
 
     /// Fills a given mutable buffer with the next doc ids from the
@@ -70,51 +60,85 @@ pub trait DocSet {
     /// use case where batching. The normal way to
     /// go through the `DocId`'s is to call `.advance()`.
     fn fill_buffer(&mut self, buffer: &mut [DocId]) -> usize {
+        if self.doc() == TERMINATED {
+            return 0;
+        }
         for (i, buffer_val) in buffer.iter_mut().enumerate() {
-            if self.advance() {
-                *buffer_val = self.doc();
-            } else {
-                return i;
+            *buffer_val = self.doc();
+            if self.advance() == TERMINATED {
+                return i + 1;
             }
         }
         buffer.len()
     }
 
     /// Returns the current document
+    /// Right after creating a new DocSet, the docset points to the first document.
+    ///
+    /// If the DocSet is empty, .doc() should return `TERMINATED`.
     fn doc(&self) -> DocId;
 
     /// Returns a best-effort hint of the
     /// length of the docset.
     fn size_hint(&self) -> u32;
 
-    /// Appends all docs to a `bitset`.
-    fn append_to_bitset(&mut self, bitset: &mut BitSet) {
-        while self.advance() {
-            bitset.insert(self.doc());
+    /// Returns the number documents matching.
+    /// Calling this method consumes the `DocSet`.
+    fn count(&mut self, delete_bitset: &DeleteBitSet) -> u32 {
+        let mut count = 0u32;
+        let mut doc = self.doc();
+        while doc != TERMINATED {
+            if !delete_bitset.is_deleted(doc) {
+                count += 1u32;
+            }
+            doc = self.advance();
         }
+        count
     }
 
-    /// Returns the number documents matching.
-    ///
+    /// Returns the count of documents, deleted or not.
     /// Calling this method consumes the `DocSet`.
-    fn count(&mut self) -> u32 {
+    ///
+    /// Of course, the result is an upper bound of the result
+    /// given by `count()`.
+    fn count_including_deleted(&mut self) -> u32 {
         let mut count = 0u32;
-        while self.advance() {
+        let mut doc = self.doc();
+        while doc != TERMINATED {
             count += 1u32;
+            doc = self.advance();
         }
         count
     }
 }
 
+impl<'a> DocSet for &'a mut dyn DocSet {
+    fn advance(&mut self) -> u32 {
+        (**self).advance()
+    }
+
+    fn seek(&mut self, target: DocId) -> DocId {
+        (**self).seek(target)
+    }
+
+    fn doc(&self) -> u32 {
+        (**self).doc()
+    }
+
+    fn size_hint(&self) -> u32 {
+        (**self).size_hint()
+    }
+}
+
 impl<TDocSet: DocSet + ?Sized> DocSet for Box<TDocSet> {
-    fn advance(&mut self) -> bool {
+    fn advance(&mut self) -> DocId {
         let unboxed: &mut TDocSet = self.borrow_mut();
         unboxed.advance()
     }
 
-    fn skip_next(&mut self, target: DocId) -> SkipResult {
+    fn seek(&mut self, target: DocId) -> DocId {
         let unboxed: &mut TDocSet = self.borrow_mut();
-        unboxed.skip_next(target)
+        unboxed.seek(target)
     }
 
     fn doc(&self) -> DocId {
@@ -127,13 +151,13 @@ impl<TDocSet: DocSet + ?Sized> DocSet for Box<TDocSet> {
         unboxed.size_hint()
     }
 
-    fn count(&mut self) -> u32 {
+    fn count(&mut self, delete_bitset: &DeleteBitSet) -> u32 {
         let unboxed: &mut TDocSet = self.borrow_mut();
-        unboxed.count()
+        unboxed.count(delete_bitset)
     }
 
-    fn append_to_bitset(&mut self, bitset: &mut BitSet) {
+    fn count_including_deleted(&mut self) -> u32 {
         let unboxed: &mut TDocSet = self.borrow_mut();
-        unboxed.append_to_bitset(bitset);
+        unboxed.count_including_deleted()
     }
 }

src/error.rs

@@ -2,15 +2,47 @@
 
 use std::io;
 
-use directory::error::{IOError, OpenDirectoryError, OpenReadError, OpenWriteError};
-use fastfield::FastFieldNotAvailableError;
-use indexer::LockType;
-use query;
-use schema;
-use serde_json;
+use crate::directory::error::{IOError, OpenDirectoryError, OpenReadError, OpenWriteError};
+use crate::directory::error::{Incompatibility, LockError};
+use crate::fastfield::FastFieldNotAvailableError;
+use crate::query;
+use crate::schema;
+use std::fmt;
 use std::path::PathBuf;
 use std::sync::PoisonError;
 
+pub struct DataCorruption {
+    filepath: Option<PathBuf>,
+    comment: String,
+}
+
+impl DataCorruption {
+    pub fn new(filepath: PathBuf, comment: String) -> DataCorruption {
+        DataCorruption {
+            filepath: Some(filepath),
+            comment,
+        }
+    }
+
+    pub fn comment_only(comment: String) -> DataCorruption {
+        DataCorruption {
+            filepath: None,
+            comment,
+        }
+    }
+}
+
+impl fmt::Debug for DataCorruption {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
+        write!(f, "Data corruption: ")?;
+        if let Some(ref filepath) = &self.filepath {
+            write!(f, "(in file `{:?}`)", filepath)?;
+        }
+        write!(f, ": {}.", self.comment)?;
+        Ok(())
+    }
+}
+
 /// The library's failure based error enum
 #[derive(Debug, Fail)]
 pub enum TantivyError {
@@ -24,17 +56,14 @@ pub enum TantivyError {
     #[fail(display = "Index already exists")]
     IndexAlreadyExists,
     /// Failed to acquire file lock
-    #[fail(
-        display = "Failed to acquire Lockfile: {:?}. Possible causes: another IndexWriter instance or panic during previous lock drop.",
-        _0
-    )]
-    LockFailure(LockType),
+    #[fail(display = "Failed to acquire Lockfile: {:?}. {:?}", _0, _1)]
+    LockFailure(LockError, Option<String>),
     /// IO Error.
     #[fail(display = "An IO error occurred: '{}'", _0)]
     IOError(#[cause] IOError),
     /// Data corruption.
-    #[fail(display = "File contains corrupted data: '{:?}'", _0)]
-    CorruptedFile(PathBuf),
+    #[fail(display = "{:?}", _0)]
+    DataCorruption(DataCorruption),
     /// A thread holding the locked panicked and poisoned the lock.
     #[fail(display = "A thread holding the locked panicked and poisoned the lock")]
     Poisoned,
@@ -47,17 +76,29 @@ pub enum TantivyError {
     /// An Error appeared related to the schema.
     #[fail(display = "Schema error: '{}'", _0)]
     SchemaError(String),
-    /// Tried to access a fastfield reader for a field not configured accordingly.
-    #[fail(display = "Fast field not available: '{:?}'", _0)]
-    FastFieldError(#[cause] FastFieldNotAvailableError),
     /// System error. (e.g.: We failed spawning a new thread)
     #[fail(display = "System error.'{}'", _0)]
     SystemError(String),
+    /// Index incompatible with current version of tantivy
+    #[fail(display = "{:?}", _0)]
+    IncompatibleIndex(Incompatibility),
+}
+
+impl From<DataCorruption> for TantivyError {
+    fn from(data_corruption: DataCorruption) -> TantivyError {
+        TantivyError::DataCorruption(data_corruption)
+    }
 }
 
 impl From<FastFieldNotAvailableError> for TantivyError {
     fn from(fastfield_error: FastFieldNotAvailableError) -> TantivyError {
-        TantivyError::FastFieldError(fastfield_error)
+        TantivyError::SchemaError(format!("{}", fastfield_error))
+    }
+}
+
+impl From<LockError> for TantivyError {
+    fn from(lock_error: LockError) -> TantivyError {
+        TantivyError::LockFailure(lock_error, None)
     }
 }
 
@@ -90,6 +131,9 @@ impl From<OpenReadError> for TantivyError {
         match error {
             OpenReadError::FileDoesNotExist(filepath) => TantivyError::PathDoesNotExist(filepath),
             OpenReadError::IOError(io_error) => TantivyError::IOError(io_error),
+            OpenReadError::IncompatibleIndex(incompatibility) => {
+                TantivyError::IncompatibleIndex(incompatibility)
+            }
         }
     }
 }
@@ -120,6 +164,7 @@ impl From<OpenDirectoryError> for TantivyError {
             OpenDirectoryError::NotADirectory(directory_path) => {
                 TantivyError::InvalidArgument(format!("{:?} is not a directory", directory_path))
             }
+            OpenDirectoryError::IoError(err) => TantivyError::IOError(IOError::from(err)),
         }
     }
 }
@@ -130,3 +175,9 @@ impl From<serde_json::Error> for TantivyError {
         TantivyError::IOError(io_err.into())
     }
 }
+
+impl From<rayon::ThreadPoolBuildError> for TantivyError {
+    fn from(error: rayon::ThreadPoolBuildError) -> TantivyError {
+        TantivyError::SystemError(error.to_string())
+    }
+}

src/fastfield/bytes/mod.rs

@@ -6,8 +6,8 @@ pub use self::writer::BytesFastFieldWriter;
 
 #[cfg(test)]
 mod tests {
-    use schema::Schema;
-    use Index;
+    use crate::schema::Schema;
+    use crate::Index;
 
     #[test]
     fn test_bytes() {
@@ -22,17 +22,15 @@ mod tests {
         index_writer.add_document(doc!(field=>vec![1u8, 3, 5, 7, 9]));
         index_writer.add_document(doc!(field=>vec![0u8; 1000]));
         assert!(index_writer.commit().is_ok());
+        let searcher = index.reader().unwrap().searcher();
+        let segment_reader = searcher.segment_reader(0);
+        let bytes_reader = segment_reader.fast_fields().bytes(field).unwrap();
 
-        index.load_searchers().unwrap();
-        let searcher = index.searcher();
-        let reader = searcher.segment_reader(0);
-        let bytes_reader = reader.bytes_fast_field_reader(field).unwrap();
-
-        assert_eq!(bytes_reader.get_val(0), &[0u8, 1, 2, 3]);
-        assert!(bytes_reader.get_val(1).is_empty());
-        assert_eq!(bytes_reader.get_val(2), &[255u8]);
-        assert_eq!(bytes_reader.get_val(3), &[1u8, 3, 5, 7, 9]);
+        assert_eq!(bytes_reader.get_bytes(0), &[0u8, 1, 2, 3]);
+        assert!(bytes_reader.get_bytes(1).is_empty());
+        assert_eq!(bytes_reader.get_bytes(2), &[255u8]);
+        assert_eq!(bytes_reader.get_bytes(3), &[1u8, 3, 5, 7, 9]);
         let long = vec![0u8; 1000];
-        assert_eq!(bytes_reader.get_val(4), long.as_slice());
+        assert_eq!(bytes_reader.get_bytes(4), long.as_slice());
     }
 }

src/fastfield/bytes/reader.rs

@@ -1,8 +1,8 @@
 use owning_ref::OwningRef;
 
-use directory::ReadOnlySource;
-use fastfield::FastFieldReader;
-use DocId;
+use crate::directory::ReadOnlySource;
+use crate::fastfield::FastFieldReader;
+use crate::DocId;
 
 /// Reader for byte array fast fields
 ///
@@ -14,6 +14,7 @@ use DocId;
 ///
 /// Reading the value for a document is done by reading the start index for it,
 /// and the start index for the next document, and keeping the bytes in between.
+#[derive(Clone)]
 pub struct BytesFastFieldReader {
     idx_reader: FastFieldReader<u64>,
     values: OwningRef<ReadOnlySource, [u8]>,
@@ -28,10 +29,20 @@ impl BytesFastFieldReader {
         BytesFastFieldReader { idx_reader, values }
     }
 
-    /// Returns the bytes associated to the given `doc`
-    pub fn get_val(&self, doc: DocId) -> &[u8] {
+    fn range(&self, doc: DocId) -> (usize, usize) {
         let start = self.idx_reader.get(doc) as usize;
         let stop = self.idx_reader.get(doc + 1) as usize;
+        (start, stop)
+    }
+
+    /// Returns the bytes associated to the given `doc`
+    pub fn get_bytes(&self, doc: DocId) -> &[u8] {
+        let (start, stop) = self.range(doc);
         &self.values[start..stop]
     }
+
+    /// Returns the overall number of bytes in this bytes fast field.
+    pub fn total_num_bytes(&self) -> usize {
+        self.values.len()
+    }
 }

src/fastfield/bytes/writer.rs

@@ -1,8 +1,8 @@
 use std::io;
 
-use fastfield::serializer::FastFieldSerializer;
-use schema::{Document, Field, Value};
-use DocId;
+use crate::fastfield::serializer::FastFieldSerializer;
+use crate::schema::{Document, Field, Value};
+use crate::DocId;
 
 /// Writer for byte array (as in, any number of bytes per document) fast fields
 ///

src/fastfield/delete.rs

@@ -1,17 +1,19 @@
-use bit_set::BitSet;
-use common::HasLen;
-use directory::ReadOnlySource;
-use directory::WritePtr;
-use space_usage::ByteCount;
+use crate::common::{BitSet, HasLen};
+use crate::directory::ReadOnlySource;
+use crate::directory::WritePtr;
+use crate::space_usage::ByteCount;
+use crate::DocId;
 use std::io;
 use std::io::Write;
-use DocId;
 
 /// Write a delete `BitSet`
 ///
 /// where `delete_bitset` is the set of deleted `DocId`.
-pub fn write_delete_bitset(delete_bitset: &BitSet, writer: &mut WritePtr) -> io::Result<()> {
-    let max_doc = delete_bitset.capacity();
+pub fn write_delete_bitset(
+    delete_bitset: &BitSet,
+    max_doc: u32,
+    writer: &mut WritePtr,
+) -> io::Result<()> {
     let mut byte = 0u8;
     let mut shift = 0u8;
     for doc in 0..max_doc {
@@ -29,7 +31,7 @@ pub fn write_delete_bitset(delete_bitset: &BitSet, writer: &mut WritePtr) -> io:
     if max_doc % 8 > 0 {
         writer.write_all(&[byte])?;
     }
-    writer.flush()
+    Ok(())
 }
 
 /// Set of deleted `DocId`s.
@@ -53,16 +55,18 @@ impl DeleteBitSet {
         }
     }
 
-    /// Returns whether the document has been marked as deleted.
+    /// Returns true iff the document is still "alive". In other words, if it has not been deleted.
+    pub fn is_alive(&self, doc: DocId) -> bool {
+        !self.is_deleted(doc)
+    }
+
+    /// Returns true iff the document has been marked as deleted.
+    #[inline(always)]
     pub fn is_deleted(&self, doc: DocId) -> bool {
-        if self.len == 0 {
-            false
-        } else {
-            let byte_offset = doc / 8u32;
-            let b: u8 = (*self.data)[byte_offset as usize];
-            let shift = (doc & 7u32) as u8;
-            b & (1u8 << shift) != 0
-        }
+        let byte_offset = doc / 8u32;
+        let b: u8 = (*self.data)[byte_offset as usize];
+        let shift = (doc & 7u32) as u8;
+        b & (1u8 << shift) != 0
     }
 
     /// Summarize total space usage of this bitset.
@@ -80,44 +84,41 @@ impl HasLen for DeleteBitSet {
 #[cfg(test)]
 mod tests {
     use super::*;
-    use bit_set::BitSet;
-    use directory::*;
+    use crate::directory::*;
     use std::path::PathBuf;
 
-    fn test_delete_bitset_helper(bitset: &BitSet) {
+    fn test_delete_bitset_helper(bitset: &BitSet, max_doc: u32) {
         let test_path = PathBuf::from("test");
         let mut directory = RAMDirectory::create();
         {
             let mut writer = directory.open_write(&*test_path).unwrap();
-            write_delete_bitset(bitset, &mut writer).unwrap();
+            write_delete_bitset(bitset, max_doc, &mut writer).unwrap();
+            writer.terminate().unwrap();
         }
-        {
-            let source = directory.open_read(&test_path).unwrap();
-            let delete_bitset = DeleteBitSet::open(source);
-            let n = bitset.capacity();
-            for doc in 0..n {
-                assert_eq!(bitset.contains(doc), delete_bitset.is_deleted(doc as DocId));
-            }
-            assert_eq!(delete_bitset.len(), bitset.len());
+        let source = directory.open_read(&test_path).unwrap();
+        let delete_bitset = DeleteBitSet::open(source);
+        for doc in 0..max_doc {
+            assert_eq!(bitset.contains(doc), delete_bitset.is_deleted(doc as DocId));
         }
+        assert_eq!(delete_bitset.len(), bitset.len());
     }
 
     #[test]
     fn test_delete_bitset() {
         {
-            let mut bitset = BitSet::with_capacity(10);
+            let mut bitset = BitSet::with_max_value(10);
             bitset.insert(1);
             bitset.insert(9);
-            test_delete_bitset_helper(&bitset);
+            test_delete_bitset_helper(&bitset, 10);
         }
         {
-            let mut bitset = BitSet::with_capacity(8);
+            let mut bitset = BitSet::with_max_value(8);
             bitset.insert(1);
             bitset.insert(2);
             bitset.insert(3);
             bitset.insert(5);
             bitset.insert(7);
-            test_delete_bitset_helper(&bitset);
+            test_delete_bitset_helper(&bitset, 8);
         }
     }
 }

src/fastfield/error.rs

@@ -1,11 +1,11 @@
-use schema::FieldEntry;
+use crate::schema::FieldEntry;
 use std::result;
 
 /// `FastFieldNotAvailableError` is returned when the
 /// user requested for a fast field reader, and the field was not
 /// defined in the schema as a fast field.
 #[derive(Debug, Fail)]
-#[fail(display = "field not available: '{:?}'", field_name)]
+#[fail(display = "Fast field not available: '{:?}'", field_name)]
 pub struct FastFieldNotAvailableError {
     field_name: String,
 }

src/fastfield/facet_reader.rs

@@ -1,8 +1,9 @@
 use super::MultiValueIntFastFieldReader;
-use schema::Facet;
-use termdict::TermDictionary;
-use termdict::TermOrdinal;
-use DocId;
+use crate::schema::Facet;
+use crate::termdict::TermDictionary;
+use crate::termdict::TermOrdinal;
+use crate::DocId;
+use std::str;
 
 /// The facet reader makes it possible to access the list of
 /// facets associated to a given document in a specific
@@ -20,6 +21,7 @@ use DocId;
 pub struct FacetReader {
     term_ords: MultiValueIntFastFieldReader<u64>,
     term_dict: TermDictionary,
+    buffer: Vec<u8>,
 }
 
 impl FacetReader {
@@ -37,6 +39,7 @@ impl FacetReader {
         FacetReader {
             term_ords,
             term_dict,
+            buffer: vec![],
         }
     }
 
@@ -55,11 +58,18 @@ impl FacetReader {
     }
 
     /// Given a term ordinal returns the term associated to it.
-    pub fn facet_from_ord(&self, facet_ord: TermOrdinal, output: &mut Facet) {
+    pub fn facet_from_ord(
+        &mut self,
+        facet_ord: TermOrdinal,
+        output: &mut Facet,
+    ) -> Result<(), str::Utf8Error> {
         let found_term = self
             .term_dict
-            .ord_to_term(facet_ord as u64, output.inner_buffer_mut());
+            .ord_to_term(facet_ord as u64, &mut self.buffer);
         assert!(found_term, "Term ordinal {} no found.", facet_ord);
+        let facet_str = str::from_utf8(&self.buffer[..])?;
+        output.set_facet_str(facet_str);
+        Ok(())
     }
 
     /// Return the list of facet ordinals associated to a document.

src/fastfield/mod.rs

@@ -30,12 +30,14 @@ pub use self::error::{FastFieldNotAvailableError, Result};
 pub use self::facet_reader::FacetReader;
 pub use self::multivalued::{MultiValueIntFastFieldReader, MultiValueIntFastFieldWriter};
 pub use self::reader::FastFieldReader;
+pub use self::readers::FastFieldReaders;
 pub use self::serializer::FastFieldSerializer;
 pub use self::writer::{FastFieldsWriter, IntFastFieldWriter};
-use common;
-use schema::Cardinality;
-use schema::FieldType;
-use schema::Value;
+use crate::chrono::{NaiveDateTime, Utc};
+use crate::common;
+use crate::schema::Cardinality;
+use crate::schema::FieldType;
+use crate::schema::Value;
 
 mod bytes;
 mod delete;
@@ -43,11 +45,12 @@ mod error;
 mod facet_reader;
 mod multivalued;
 mod reader;
+mod readers;
 mod serializer;
 mod writer;
 
-/// Trait for types that are allowed for fast fields: (u64 or i64).
-pub trait FastValue: Default + Clone + Copy {
+/// Trait for types that are allowed for fast fields: (u64, i64 and f64).
+pub trait FastValue: Clone + Copy + Send + Sync + PartialOrd {
     /// Converts a value from u64
     ///
     /// Internally all fast field values are encoded as u64.
@@ -67,6 +70,12 @@ pub trait FastValue: Default + Clone + Copy {
     /// Cast value to `u64`.
     /// The value is just reinterpreted in memory.
     fn as_u64(&self) -> u64;
+
+    /// Build a default value. This default value is never used, so the value does not
+    /// really matter.
+    fn make_zero() -> Self {
+        Self::from_u64(0i64.to_u64())
+    }
 }
 
 impl FastValue for u64 {
@@ -78,10 +87,6 @@ impl FastValue for u64 {
         *self
     }
 
-    fn as_u64(&self) -> u64 {
-        *self
-    }
-
     fn fast_field_cardinality(field_type: &FieldType) -> Option<Cardinality> {
         match *field_type {
             FieldType::U64(ref integer_options) => integer_options.get_fastfield_cardinality(),
@@ -89,6 +94,10 @@ impl FastValue for u64 {
             _ => None,
         }
     }
+
+    fn as_u64(&self) -> u64 {
+        *self
+    }
 }
 
 impl FastValue for i64 {
@@ -112,11 +121,56 @@ impl FastValue for i64 {
     }
 }
 
+impl FastValue for f64 {
+    fn from_u64(val: u64) -> Self {
+        common::u64_to_f64(val)
+    }
+
+    fn to_u64(&self) -> u64 {
+        common::f64_to_u64(*self)
+    }
+
+    fn fast_field_cardinality(field_type: &FieldType) -> Option<Cardinality> {
+        match *field_type {
+            FieldType::F64(ref integer_options) => integer_options.get_fastfield_cardinality(),
+            _ => None,
+        }
+    }
+
+    fn as_u64(&self) -> u64 {
+        self.to_bits()
+    }
+}
+
+impl FastValue for crate::DateTime {
+    fn from_u64(timestamp_u64: u64) -> Self {
+        let timestamp_i64 = i64::from_u64(timestamp_u64);
+        crate::DateTime::from_utc(NaiveDateTime::from_timestamp(timestamp_i64, 0), Utc)
+    }
+
+    fn to_u64(&self) -> u64 {
+        self.timestamp().to_u64()
+    }
+
+    fn fast_field_cardinality(field_type: &FieldType) -> Option<Cardinality> {
+        match *field_type {
+            FieldType::Date(ref integer_options) => integer_options.get_fastfield_cardinality(),
+            _ => None,
+        }
+    }
+
+    fn as_u64(&self) -> u64 {
+        self.timestamp().as_u64()
+    }
+}
+
 fn value_to_u64(value: &Value) -> u64 {
     match *value {
         Value::U64(ref val) => *val,
         Value::I64(ref val) => common::i64_to_u64(*val),
-        _ => panic!("Expected a u64/i64 field, got {:?} ", value),
+        Value::F64(ref val) => common::f64_to_u64(*val),
+        Value::Date(ref datetime) => common::i64_to_u64(datetime.timestamp()),
+        _ => panic!("Expected a u64/i64/f64 field, got {:?} ", value),
     }
 }
 
@@ -124,27 +178,29 @@ fn value_to_u64(value: &Value) -> u64 {
 mod tests {
 
     use super::*;
-    use common::CompositeFile;
-    use directory::{Directory, RAMDirectory, WritePtr};
-    use fastfield::FastFieldReader;
+    use crate::common::CompositeFile;
+    use crate::directory::{Directory, RAMDirectory, WritePtr};
+    use crate::fastfield::FastFieldReader;
+    use crate::merge_policy::NoMergePolicy;
+    use crate::schema::Field;
+    use crate::schema::Schema;
+    use crate::schema::FAST;
+    use crate::schema::{Document, IntOptions};
+    use crate::{Index, SegmentId, SegmentReader};
+    use once_cell::sync::Lazy;
     use rand::prelude::SliceRandom;
     use rand::rngs::StdRng;
     use rand::SeedableRng;
-    use schema::Document;
-    use schema::Field;
-    use schema::Schema;
-    use schema::FAST;
     use std::collections::HashMap;
     use std::path::Path;
 
-    lazy_static! {
-        pub static ref SCHEMA: Schema = {
-            let mut schema_builder = Schema::builder();
-            schema_builder.add_u64_field("field", FAST);
-            schema_builder.build()
-        };
-        pub static ref FIELD: Field = { SCHEMA.get_field("field").unwrap() };
-    }
+    pub static SCHEMA: Lazy<Schema> = Lazy::new(|| {
+        let mut schema_builder = Schema::builder();
+        schema_builder.add_u64_field("field", FAST);
+        schema_builder.build()
+    });
+
+    pub static FIELD: Lazy<Field> = Lazy::new(|| SCHEMA.get_field("field").unwrap());
 
     #[test]
     pub fn test_fastfield() {
@@ -154,6 +210,12 @@ mod tests {
         assert_eq!(test_fastfield.get(2), 300);
     }
 
+    #[test]
+    pub fn test_fastfield_i64_u64() {
+        let datetime = crate::DateTime::from_utc(NaiveDateTime::from_timestamp(0i64, 0), Utc);
+        assert_eq!(i64::from_u64(datetime.to_u64()), 0i64);
+    }
+
     #[test]
     fn test_intfastfield_small() {
         let path = Path::new("test");
@@ -406,6 +468,92 @@ mod tests {
         }
     }
 
+    #[test]
+    fn test_merge_missing_date_fast_field() {
+        let mut schema_builder = Schema::builder();
+        let date_field = schema_builder.add_date_field("date", FAST);
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+        index_writer.set_merge_policy(Box::new(NoMergePolicy));
+        index_writer.add_document(doc!(date_field =>crate::chrono::prelude::Utc::now()));
+        index_writer.commit().unwrap();
+        index_writer.add_document(doc!());
+        index_writer.commit().unwrap();
+        let reader = index.reader().unwrap();
+        let segment_ids: Vec<SegmentId> = reader
+            .searcher()
+            .segment_readers()
+            .iter()
+            .map(SegmentReader::segment_id)
+            .collect();
+        assert_eq!(segment_ids.len(), 2);
+        let merge_future = index_writer.merge(&segment_ids[..]);
+        let merge_res = futures::executor::block_on(merge_future);
+        assert!(merge_res.is_ok());
+        assert!(reader.reload().is_ok());
+        assert_eq!(reader.searcher().segment_readers().len(), 1);
+    }
+
+    #[test]
+    fn test_default_datetime() {
+        assert_eq!(crate::DateTime::make_zero().timestamp(), 0i64);
+    }
+
+    #[test]
+    fn test_datefastfield() {
+        use crate::fastfield::FastValue;
+        let mut schema_builder = Schema::builder();
+        let date_field = schema_builder.add_date_field("date", FAST);
+        let multi_date_field = schema_builder.add_date_field(
+            "multi_date",
+            IntOptions::default().set_fast(Cardinality::MultiValues),
+        );
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+        index_writer.set_merge_policy(Box::new(NoMergePolicy));
+        index_writer.add_document(doc!(
+            date_field => crate::DateTime::from_u64(1i64.to_u64()),
+            multi_date_field => crate::DateTime::from_u64(2i64.to_u64()),
+            multi_date_field => crate::DateTime::from_u64(3i64.to_u64())
+        ));
+        index_writer.add_document(doc!(
+            date_field => crate::DateTime::from_u64(4i64.to_u64())
+        ));
+        index_writer.add_document(doc!(
+            multi_date_field => crate::DateTime::from_u64(5i64.to_u64()),
+            multi_date_field => crate::DateTime::from_u64(6i64.to_u64())
+        ));
+        index_writer.commit().unwrap();
+        let reader = index.reader().unwrap();
+        let searcher = reader.searcher();
+        assert_eq!(searcher.segment_readers().len(), 1);
+        let segment_reader = searcher.segment_reader(0);
+        let fast_fields = segment_reader.fast_fields();
+        let date_fast_field = fast_fields.date(date_field).unwrap();
+        let dates_fast_field = fast_fields.dates(multi_date_field).unwrap();
+        let mut dates = vec![];
+        {
+            assert_eq!(date_fast_field.get(0u32).timestamp(), 1i64);
+            dates_fast_field.get_vals(0u32, &mut dates);
+            assert_eq!(dates.len(), 2);
+            assert_eq!(dates[0].timestamp(), 2i64);
+            assert_eq!(dates[1].timestamp(), 3i64);
+        }
+        {
+            assert_eq!(date_fast_field.get(1u32).timestamp(), 4i64);
+            dates_fast_field.get_vals(1u32, &mut dates);
+            assert!(dates.is_empty());
+        }
+        {
+            assert_eq!(date_fast_field.get(2u32).timestamp(), 0i64);
+            dates_fast_field.get_vals(2u32, &mut dates);
+            assert_eq!(dates.len(), 2);
+            assert_eq!(dates[0].timestamp(), 5i64);
+            assert_eq!(dates[1].timestamp(), 6i64);
+        }
+    }
 }
 
 #[cfg(all(test, feature = "unstable"))]
@@ -413,9 +561,9 @@ mod bench {
     use super::tests::FIELD;
     use super::tests::{generate_permutation, SCHEMA};
     use super::*;
-    use common::CompositeFile;
-    use directory::{Directory, RAMDirectory, WritePtr};
-    use fastfield::FastFieldReader;
+    use crate::common::CompositeFile;
+    use crate::directory::{Directory, RAMDirectory, WritePtr};
+    use crate::fastfield::FastFieldReader;
     use std::collections::HashMap;
     use std::path::Path;
     use test::{self, Bencher};
@@ -513,5 +661,4 @@ mod bench {
             });
         }
     }
-
 }

src/fastfield/multivalued/mod.rs

@@ -7,10 +7,14 @@ pub use self::writer::MultiValueIntFastFieldWriter;
 #[cfg(test)]
 mod tests {
 
-    use schema::Cardinality;
-    use schema::IntOptions;
-    use schema::Schema;
-    use Index;
+    use crate::collector::TopDocs;
+    use crate::query::QueryParser;
+    use crate::schema::Cardinality;
+    use crate::schema::Facet;
+    use crate::schema::IntOptions;
+    use crate::schema::Schema;
+    use crate::Index;
+    use chrono::Duration;
 
     #[test]
     fn test_multivalued_u64() {
@@ -28,11 +32,10 @@ mod tests {
         index_writer.add_document(doc!(field=>5u64, field=>20u64,field=>1u64));
         assert!(index_writer.commit().is_ok());
 
-        index.load_searchers().unwrap();
-        let searcher = index.searcher();
-        let reader = searcher.segment_reader(0);
+        let searcher = index.reader().unwrap().searcher();
+        let segment_reader = searcher.segment_reader(0);
         let mut vals = Vec::new();
-        let multi_value_reader = reader.multi_fast_field_reader::<u64>(field).unwrap();
+        let multi_value_reader = segment_reader.fast_fields().u64s(field).unwrap();
         {
             multi_value_reader.get_vals(2, &mut vals);
             assert_eq!(&vals, &[4u64]);
@@ -47,6 +50,133 @@ mod tests {
         }
     }
 
+    #[test]
+    fn test_multivalued_date() {
+        let mut schema_builder = Schema::builder();
+        let date_field = schema_builder.add_date_field(
+            "multi_date_field",
+            IntOptions::default()
+                .set_fast(Cardinality::MultiValues)
+                .set_indexed()
+                .set_stored(),
+        );
+        let time_i =
+            schema_builder.add_i64_field("time_stamp_i", IntOptions::default().set_stored());
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+        let first_time_stamp = chrono::Utc::now();
+        index_writer.add_document(
+            doc!(date_field=>first_time_stamp, date_field=>first_time_stamp, time_i=>1i64),
+        );
+        index_writer.add_document(doc!(time_i=>0i64));
+        // add one second
+        index_writer
+            .add_document(doc!(date_field=>first_time_stamp + Duration::seconds(1), time_i=>2i64));
+        // add another second
+        let two_secs_ahead = first_time_stamp + Duration::seconds(2);
+        index_writer.add_document(doc!(date_field=>two_secs_ahead, date_field=>two_secs_ahead,date_field=>two_secs_ahead, time_i=>3i64));
+        assert!(index_writer.commit().is_ok());
+
+        let reader = index.reader().unwrap();
+        let searcher = reader.searcher();
+        let reader = searcher.segment_reader(0);
+        assert_eq!(reader.num_docs(), 4);
+
+        {
+            let parser = QueryParser::for_index(&index, vec![date_field]);
+            let query = parser
+                .parse_query(&format!("\"{}\"", first_time_stamp.to_rfc3339()).to_string())
+                .expect("could not parse query");
+            let results = searcher
+                .search(&query, &TopDocs::with_limit(5))
+                .expect("could not query index");
+
+            assert_eq!(results.len(), 1);
+            for (_score, doc_address) in results {
+                let retrieved_doc = searcher.doc(doc_address).expect("cannot fetch doc");
+                assert_eq!(
+                    retrieved_doc
+                        .get_first(date_field)
+                        .expect("cannot find value")
+                        .date_value()
+                        .timestamp(),
+                    first_time_stamp.timestamp()
+                );
+                assert_eq!(
+                    retrieved_doc
+                        .get_first(time_i)
+                        .expect("cannot find value")
+                        .i64_value(),
+                    1i64
+                );
+            }
+        }
+
+        {
+            let parser = QueryParser::for_index(&index, vec![date_field]);
+            let query = parser
+                .parse_query(&format!("\"{}\"", two_secs_ahead.to_rfc3339()).to_string())
+                .expect("could not parse query");
+            let results = searcher
+                .search(&query, &TopDocs::with_limit(5))
+                .expect("could not query index");
+
+            assert_eq!(results.len(), 1);
+
+            for (_score, doc_address) in results {
+                let retrieved_doc = searcher.doc(doc_address).expect("cannot fetch doc");
+                assert_eq!(
+                    retrieved_doc
+                        .get_first(date_field)
+                        .expect("cannot find value")
+                        .date_value()
+                        .timestamp(),
+                    two_secs_ahead.timestamp()
+                );
+                assert_eq!(
+                    retrieved_doc
+                        .get_first(time_i)
+                        .expect("cannot find value")
+                        .i64_value(),
+                    3i64
+                );
+            }
+        }
+
+        // TODO: support Date range queries
+        //        {
+        //            let parser = QueryParser::for_index(&index, vec![date_field]);
+        //            let range_q = format!("\"{}\"..\"{}\"",
+        //                                  (first_time_stamp + Duration::seconds(1)).to_rfc3339(),
+        //                                  (first_time_stamp + Duration::seconds(3)).to_rfc3339()
+        //            );
+        //            let query = parser.parse_query(&range_q)
+        //                .expect("could not parse query");
+        //            let results = searcher.search(&query, &TopDocs::with_limit(5))
+        //                .expect("could not query index");
+        //
+        //
+        //            assert_eq!(results.len(), 2);
+        //            for (i, doc_pair) in results.iter().enumerate() {
+        //                let retrieved_doc = searcher.doc(doc_pair.1).expect("cannot fetch doc");
+        //                let offset_sec = match i {
+        //                    0 => 1,
+        //                    1 => 3,
+        //                    _ => panic!("should not have more than 2 docs")
+        //                };
+        //                let time_i_val = match i {
+        //                    0 => 2,
+        //                    1 => 3,
+        //                    _ => panic!("should not have more than 2 docs")
+        //                };
+        //                assert_eq!(retrieved_doc.get_first(date_field).expect("cannot find value").date_value().timestamp(),
+        //                           (first_time_stamp + Duration::seconds(offset_sec)).timestamp());
+        //                assert_eq!(retrieved_doc.get_first(time_i).expect("cannot find value").i64_value(), time_i_val);
+        //            }
+        //        }
+    }
+
     #[test]
     fn test_multivalued_i64() {
         let mut schema_builder = Schema::builder();
@@ -63,11 +193,10 @@ mod tests {
         index_writer.add_document(doc!(field=> -5i64, field => -20i64, field=>1i64));
         assert!(index_writer.commit().is_ok());
 
-        index.load_searchers().unwrap();
-        let searcher = index.searcher();
-        let reader = searcher.segment_reader(0);
+        let searcher = index.reader().unwrap().searcher();
+        let segment_reader = searcher.segment_reader(0);
         let mut vals = Vec::new();
-        let multi_value_reader = reader.multi_fast_field_reader::<i64>(field).unwrap();
+        let multi_value_reader = segment_reader.fast_fields().i64s(field).unwrap();
         {
             multi_value_reader.get_vals(2, &mut vals);
             assert_eq!(&vals, &[-4i64]);
@@ -85,4 +214,17 @@ mod tests {
             assert_eq!(&vals, &[-5i64, -20i64, 1i64]);
         }
     }
+    #[test]
+    #[ignore]
+    fn test_many_facets() {
+        let mut schema_builder = Schema::builder();
+        let field = schema_builder.add_facet_field("facetfield");
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+        for i in 0..100_000 {
+            index_writer.add_document(doc!(field=> Facet::from(format!("/lang/{}", i).as_str())));
+        }
+        assert!(index_writer.commit().is_ok());
+    }
 }

src/fastfield/multivalued/reader.rs

@@ -1,5 +1,5 @@
-use fastfield::{FastFieldReader, FastValue};
-use DocId;
+use crate::fastfield::{FastFieldReader, FastValue};
+use crate::DocId;
 
 /// Reader for a multivalued `u64` fast field.
 ///
@@ -26,6 +26,13 @@ impl<Item: FastValue> MultiValueIntFastFieldReader<Item> {
         }
     }
 
+    pub(crate) fn into_u64s_reader(self) -> MultiValueIntFastFieldReader<u64> {
+        MultiValueIntFastFieldReader {
+            idx_reader: self.idx_reader,
+            vals_reader: self.vals_reader.into_u64_reader(),
+        }
+    }
+
     /// Returns `(start, stop)`, such that the values associated
     /// to the given document are `start..stop`.
     fn range(&self, doc: DocId) -> (u64, u64) {
@@ -38,16 +45,27 @@ impl<Item: FastValue> MultiValueIntFastFieldReader<Item> {
     pub fn get_vals(&self, doc: DocId, vals: &mut Vec<Item>) {
         let (start, stop) = self.range(doc);
         let len = (stop - start) as usize;
-        vals.resize(len, Item::default());
-        self.vals_reader.get_range(start as u32, &mut vals[..]);
+        vals.resize(len, Item::make_zero());
+        self.vals_reader.get_range_u64(start, &mut vals[..]);
+    }
+
+    /// Returns the number of values associated with the document `DocId`.
+    pub fn num_vals(&self, doc: DocId) -> usize {
+        let (start, stop) = self.range(doc);
+        (stop - start) as usize
+    }
+
+    /// Returns the overall number of values in this field  .
+    pub fn total_num_vals(&self) -> u64 {
+        self.idx_reader.max_value()
     }
 }
 
 #[cfg(test)]
 mod tests {
 
-    use core::Index;
-    use schema::{Document, Facet, Schema};
+    use crate::core::Index;
+    use crate::schema::{Facet, Schema};
 
     #[test]
     fn test_multifastfield_reader() {
@@ -58,44 +76,33 @@ mod tests {
         let mut index_writer = index
             .writer_with_num_threads(1, 30_000_000)
             .expect("Failed to create index writer.");
-        {
-            let mut doc = Document::new();
-            doc.add_facet(facet_field, "/category/cat2");
-            doc.add_facet(facet_field, "/category/cat1");
-            index_writer.add_document(doc);
-        }
-        {
-            let mut doc = Document::new();
-            doc.add_facet(facet_field, "/category/cat2");
-            index_writer.add_document(doc);
-        }
-        {
-            let mut doc = Document::new();
-            doc.add_facet(facet_field, "/category/cat3");
-            index_writer.add_document(doc);
-        }
+        index_writer.add_document(doc!(
+            facet_field => Facet::from("/category/cat2"),
+            facet_field => Facet::from("/category/cat1"),
+        ));
+        index_writer.add_document(doc!(facet_field => Facet::from("/category/cat2")));
+        index_writer.add_document(doc!(facet_field => Facet::from("/category/cat3")));
         index_writer.commit().expect("Commit failed");
-        index.load_searchers().expect("Reloading searchers");
-        let searcher = index.searcher();
+        let searcher = index.reader().unwrap().searcher();
         let segment_reader = searcher.segment_reader(0);
         let mut facet_reader = segment_reader.facet_reader(facet_field).unwrap();
 
         let mut facet = Facet::root();
         {
-            facet_reader.facet_from_ord(1, &mut facet);
+            facet_reader.facet_from_ord(1, &mut facet).unwrap();
             assert_eq!(facet, Facet::from("/category"));
         }
         {
-            facet_reader.facet_from_ord(2, &mut facet);
+            facet_reader.facet_from_ord(2, &mut facet).unwrap();
             assert_eq!(facet, Facet::from("/category/cat1"));
         }
         {
-            facet_reader.facet_from_ord(3, &mut facet);
+            facet_reader.facet_from_ord(3, &mut facet).unwrap();
             assert_eq!(format!("{}", facet), "/category/cat2");
             assert_eq!(facet, Facet::from("/category/cat2"));
         }
         {
-            facet_reader.facet_from_ord(4, &mut facet);
+            facet_reader.facet_from_ord(4, &mut facet).unwrap();
             assert_eq!(facet, Facet::from("/category/cat3"));
         }
 

src/fastfield/multivalued/writer.rs

@@ -1,13 +1,12 @@
-use fastfield::serializer::FastSingleFieldSerializer;
-use fastfield::value_to_u64;
-use fastfield::FastFieldSerializer;
-use itertools::Itertools;
-use postings::UnorderedTermId;
-use schema::{Document, Field};
-use std::collections::HashMap;
+use crate::fastfield::serializer::FastSingleFieldSerializer;
+use crate::fastfield::value_to_u64;
+use crate::fastfield::FastFieldSerializer;
+use crate::postings::UnorderedTermId;
+use crate::schema::{Document, Field};
+use crate::termdict::TermOrdinal;
+use crate::DocId;
+use fnv::FnvHashMap;
 use std::io;
-use termdict::TermOrdinal;
-use DocId;
 
 /// Writer for multi-valued (as in, more than one value per document)
 /// int fast field.
@@ -32,7 +31,7 @@ use DocId;
 /// term ids when the segment is getting serialized.
 pub struct MultiValueIntFastFieldWriter {
     field: Field,
-    vals: Vec<u64>,
+    vals: Vec<UnorderedTermId>,
     doc_index: Vec<u64>,
     is_facet: bool,
 }
@@ -102,7 +101,7 @@ impl MultiValueIntFastFieldWriter {
     pub fn serialize(
         &self,
         serializer: &mut FastFieldSerializer,
-        mapping_opt: Option<&HashMap<UnorderedTermId, TermOrdinal>>,
+        mapping_opt: Option<&FnvHashMap<UnorderedTermId, TermOrdinal>>,
     ) -> io::Result<()> {
         {
             // writing the offset index
@@ -116,7 +115,7 @@ impl MultiValueIntFastFieldWriter {
         }
         {
             // writing the values themselves.
-            let mut value_serializer: FastSingleFieldSerializer<_>;
+            let mut value_serializer: FastSingleFieldSerializer<'_, _>;
             match mapping_opt {
                 Some(mapping) => {
                     value_serializer = serializer.new_u64_fast_field_with_idx(
@@ -151,8 +150,8 @@ impl MultiValueIntFastFieldWriter {
                     }
                 }
                 None => {
-                    let val_min_max = self.vals.iter().cloned().minmax();
-                    let (val_min, val_max) = val_min_max.into_option().unwrap_or((0u64, 0u64));
+                    let val_min_max = crate::common::minmax(self.vals.iter().cloned());
+                    let (val_min, val_max) = val_min_max.unwrap_or((0u64, 0u64));
                     value_serializer =
                         serializer.new_u64_fast_field_with_idx(self.field, val_min, val_max, 1)?;
                     for &val in &self.vals {

src/fastfield/reader.rs

@@ -1,18 +1,18 @@
 use super::FastValue;
-use common::bitpacker::BitUnpacker;
-use common::compute_num_bits;
-use common::BinarySerializable;
-use common::CompositeFile;
-use directory::ReadOnlySource;
-use directory::{Directory, RAMDirectory, WritePtr};
-use fastfield::{FastFieldSerializer, FastFieldsWriter};
+use crate::common::bitpacker::BitUnpacker;
+use crate::common::compute_num_bits;
+use crate::common::BinarySerializable;
+use crate::common::CompositeFile;
+use crate::directory::ReadOnlySource;
+use crate::directory::{Directory, RAMDirectory, WritePtr};
+use crate::fastfield::{FastFieldSerializer, FastFieldsWriter};
+use crate::schema::Schema;
+use crate::schema::FAST;
+use crate::DocId;
 use owning_ref::OwningRef;
-use schema::Schema;
-use schema::FAST;
 use std::collections::HashMap;
 use std::marker::PhantomData;
 use std::path::Path;
-use DocId;
 
 /// Trait for accessing a fastfield.
 ///
@@ -50,6 +50,15 @@ impl<Item: FastValue> FastFieldReader<Item> {
         }
     }
 
+    pub(crate) fn into_u64_reader(self) -> FastFieldReader<u64> {
+        FastFieldReader {
+            bit_unpacker: self.bit_unpacker,
+            min_value_u64: self.min_value_u64,
+            max_value_u64: self.max_value_u64,
+            _phantom: PhantomData,
+        }
+    }
+
     /// Return the value associated to the given document.
     ///
     /// This accessor should return as fast as possible.
@@ -59,7 +68,29 @@ impl<Item: FastValue> FastFieldReader<Item> {
     /// May panic if `doc` is greater than the segment
     // `maxdoc`.
     pub fn get(&self, doc: DocId) -> Item {
-        Item::from_u64(self.min_value_u64 + self.bit_unpacker.get(doc as usize))
+        self.get_u64(u64::from(doc))
+    }
+
+    pub(crate) fn get_u64(&self, doc: u64) -> Item {
+        Item::from_u64(self.min_value_u64 + self.bit_unpacker.get(doc))
+    }
+
+    /// Internally `multivalued` also use SingleValue Fast fields.
+    /// It works as follows... A first column contains the list of start index
+    /// for each document, a second column contains the actual values.
+    ///
+    /// The values associated to a given doc, are then
+    ///  `second_column[first_column.get(doc)..first_column.get(doc+1)]`.
+    ///
+    /// Which means single value fast field reader can be indexed internally with
+    /// something different from a `DocId`. For this use case, we want to use `u64`
+    /// values.
+    ///
+    /// See `get_range` for an actual documentation about this method.
+    pub(crate) fn get_range_u64(&self, start: u64, output: &mut [Item]) {
+        for (i, out) in output.iter_mut().enumerate() {
+            *out = self.get_u64(start + (i as u64));
+        }
     }
 
     /// Fills an output buffer with the fast field values
@@ -75,16 +106,8 @@ impl<Item: FastValue> FastFieldReader<Item> {
     ///
     /// May panic if `start + output.len()` is greater than
     /// the segment's `maxdoc`.
-    ///
-    // TODO change start to `u64`.
-    // For multifastfield, start is an index in a second fastfield, not a `DocId`
-    pub fn get_range(&self, start: u32, output: &mut [Item]) {
-        // ok: Item is either `u64` or `i64`
-        let output_u64: &mut [u64] = unsafe { &mut *(output as *mut [Item] as *mut [u64]) };
-        self.bit_unpacker.get_range(start, output_u64);
-        for out in output_u64.iter_mut() {
-            *out = Item::from_u64(*out + self.min_value_u64).as_u64();
-        }
+    pub fn get_range(&self, start: DocId, output: &mut [Item]) {
+        self.get_range_u64(u64::from(start), output);
     }
 
     /// Returns the minimum value for this fast field.

src/fastfield/readers.rs

@@ -0,0 +1,270 @@
+use crate::common::CompositeFile;
+use crate::fastfield::BytesFastFieldReader;
+use crate::fastfield::MultiValueIntFastFieldReader;
+use crate::fastfield::{FastFieldNotAvailableError, FastFieldReader};
+use crate::schema::{Cardinality, Field, FieldType, Schema};
+use crate::space_usage::PerFieldSpaceUsage;
+use std::collections::HashMap;
+
+/// Provides access to all of the FastFieldReader.
+///
+/// Internally, `FastFieldReaders` have preloaded fast field readers,
+/// and just wraps several `HashMap`.
+pub struct FastFieldReaders {
+    fast_field_i64: HashMap<Field, FastFieldReader<i64>>,
+    fast_field_u64: HashMap<Field, FastFieldReader<u64>>,
+    fast_field_f64: HashMap<Field, FastFieldReader<f64>>,
+    fast_field_date: HashMap<Field, FastFieldReader<crate::DateTime>>,
+    fast_field_i64s: HashMap<Field, MultiValueIntFastFieldReader<i64>>,
+    fast_field_u64s: HashMap<Field, MultiValueIntFastFieldReader<u64>>,
+    fast_field_f64s: HashMap<Field, MultiValueIntFastFieldReader<f64>>,
+    fast_field_dates: HashMap<Field, MultiValueIntFastFieldReader<crate::DateTime>>,
+    fast_bytes: HashMap<Field, BytesFastFieldReader>,
+    fast_fields_composite: CompositeFile,
+}
+
+enum FastType {
+    I64,
+    U64,
+    F64,
+    Date,
+}
+
+fn type_and_cardinality(field_type: &FieldType) -> Option<(FastType, Cardinality)> {
+    match field_type {
+        FieldType::U64(options) => options
+            .get_fastfield_cardinality()
+            .map(|cardinality| (FastType::U64, cardinality)),
+        FieldType::I64(options) => options
+            .get_fastfield_cardinality()
+            .map(|cardinality| (FastType::I64, cardinality)),
+        FieldType::F64(options) => options
+            .get_fastfield_cardinality()
+            .map(|cardinality| (FastType::F64, cardinality)),
+        FieldType::Date(options) => options
+            .get_fastfield_cardinality()
+            .map(|cardinality| (FastType::Date, cardinality)),
+        FieldType::HierarchicalFacet => Some((FastType::U64, Cardinality::MultiValues)),
+        _ => None,
+    }
+}
+
+impl FastFieldReaders {
+    pub(crate) fn load_all(
+        schema: &Schema,
+        fast_fields_composite: &CompositeFile,
+    ) -> crate::Result<FastFieldReaders> {
+        let mut fast_field_readers = FastFieldReaders {
+            fast_field_i64: Default::default(),
+            fast_field_u64: Default::default(),
+            fast_field_f64: Default::default(),
+            fast_field_date: Default::default(),
+            fast_field_i64s: Default::default(),
+            fast_field_u64s: Default::default(),
+            fast_field_f64s: Default::default(),
+            fast_field_dates: Default::default(),
+            fast_bytes: Default::default(),
+            fast_fields_composite: fast_fields_composite.clone(),
+        };
+        for (field, field_entry) in schema.fields() {
+            let field_type = field_entry.field_type();
+            if field_type == &FieldType::Bytes {
+                let idx_reader = fast_fields_composite
+                    .open_read_with_idx(field, 0)
+                    .ok_or_else(|| FastFieldNotAvailableError::new(field_entry))
+                    .map(FastFieldReader::open)?;
+                let data = fast_fields_composite
+                    .open_read_with_idx(field, 1)
+                    .ok_or_else(|| FastFieldNotAvailableError::new(field_entry))?;
+                fast_field_readers
+                    .fast_bytes
+                    .insert(field, BytesFastFieldReader::open(idx_reader, data));
+            } else if let Some((fast_type, cardinality)) = type_and_cardinality(field_type) {
+                match cardinality {
+                    Cardinality::SingleValue => {
+                        if let Some(fast_field_data) = fast_fields_composite.open_read(field) {
+                            match fast_type {
+                                FastType::U64 => {
+                                    let fast_field_reader = FastFieldReader::open(fast_field_data);
+                                    fast_field_readers
+                                        .fast_field_u64
+                                        .insert(field, fast_field_reader);
+                                }
+                                FastType::I64 => {
+                                    fast_field_readers.fast_field_i64.insert(
+                                        field,
+                                        FastFieldReader::open(fast_field_data.clone()),
+                                    );
+                                }
+                                FastType::F64 => {
+                                    fast_field_readers.fast_field_f64.insert(
+                                        field,
+                                        FastFieldReader::open(fast_field_data.clone()),
+                                    );
+                                }
+                                FastType::Date => {
+                                    fast_field_readers.fast_field_date.insert(
+                                        field,
+                                        FastFieldReader::open(fast_field_data.clone()),
+                                    );
+                                }
+                            }
+                        } else {
+                            return Err(From::from(FastFieldNotAvailableError::new(field_entry)));
+                        }
+                    }
+                    Cardinality::MultiValues => {
+                        let idx_opt = fast_fields_composite.open_read_with_idx(field, 0);
+                        let data_opt = fast_fields_composite.open_read_with_idx(field, 1);
+                        if let (Some(fast_field_idx), Some(fast_field_data)) = (idx_opt, data_opt) {
+                            let idx_reader = FastFieldReader::open(fast_field_idx);
+                            match fast_type {
+                                FastType::I64 => {
+                                    let vals_reader = FastFieldReader::open(fast_field_data);
+                                    let multivalued_int_fast_field =
+                                        MultiValueIntFastFieldReader::open(idx_reader, vals_reader);
+                                    fast_field_readers
+                                        .fast_field_i64s
+                                        .insert(field, multivalued_int_fast_field);
+                                }
+                                FastType::U64 => {
+                                    let vals_reader = FastFieldReader::open(fast_field_data);
+                                    let multivalued_int_fast_field =
+                                        MultiValueIntFastFieldReader::open(idx_reader, vals_reader);
+                                    fast_field_readers
+                                        .fast_field_u64s
+                                        .insert(field, multivalued_int_fast_field);
+                                }
+                                FastType::F64 => {
+                                    let vals_reader = FastFieldReader::open(fast_field_data);
+                                    let multivalued_int_fast_field =
+                                        MultiValueIntFastFieldReader::open(idx_reader, vals_reader);
+                                    fast_field_readers
+                                        .fast_field_f64s
+                                        .insert(field, multivalued_int_fast_field);
+                                }
+                                FastType::Date => {
+                                    let vals_reader = FastFieldReader::open(fast_field_data);
+                                    let multivalued_int_fast_field =
+                                        MultiValueIntFastFieldReader::open(idx_reader, vals_reader);
+                                    fast_field_readers
+                                        .fast_field_dates
+                                        .insert(field, multivalued_int_fast_field);
+                                }
+                            }
+                        } else {
+                            return Err(From::from(FastFieldNotAvailableError::new(field_entry)));
+                        }
+                    }
+                }
+            }
+        }
+        Ok(fast_field_readers)
+    }
+
+    pub(crate) fn space_usage(&self) -> PerFieldSpaceUsage {
+        self.fast_fields_composite.space_usage()
+    }
+
+    /// Returns the `u64` fast field reader reader associated to `field`.
+    ///
+    /// If `field` is not a u64 fast field, this method returns `None`.
+    pub fn u64(&self, field: Field) -> Option<FastFieldReader<u64>> {
+        self.fast_field_u64.get(&field).cloned()
+    }
+
+    /// If the field is a u64-fast field return the associated reader.
+    /// If the field is a i64-fast field, return the associated u64 reader. Values are
+    /// mapped from i64 to u64 using a (well the, it is unique) monotonic mapping.    ///
+    ///
+    /// This method is useful when merging segment reader.
+    pub(crate) fn u64_lenient(&self, field: Field) -> Option<FastFieldReader<u64>> {
+        if let Some(u64_ff_reader) = self.u64(field) {
+            return Some(u64_ff_reader);
+        }
+        if let Some(i64_ff_reader) = self.i64(field) {
+            return Some(i64_ff_reader.into_u64_reader());
+        }
+        if let Some(f64_ff_reader) = self.f64(field) {
+            return Some(f64_ff_reader.into_u64_reader());
+        }
+        if let Some(date_ff_reader) = self.date(field) {
+            return Some(date_ff_reader.into_u64_reader());
+        }
+        None
+    }
+
+    /// Returns the `i64` fast field reader reader associated to `field`.
+    ///
+    /// If `field` is not a i64 fast field, this method returns `None`.
+    pub fn i64(&self, field: Field) -> Option<FastFieldReader<i64>> {
+        self.fast_field_i64.get(&field).cloned()
+    }
+
+    /// Returns the `i64` fast field reader reader associated to `field`.
+    ///
+    /// If `field` is not a i64 fast field, this method returns `None`.
+    pub fn date(&self, field: Field) -> Option<FastFieldReader<crate::DateTime>> {
+        self.fast_field_date.get(&field).cloned()
+    }
+
+    /// Returns the `f64` fast field reader reader associated to `field`.
+    ///
+    /// If `field` is not a f64 fast field, this method returns `None`.
+    pub fn f64(&self, field: Field) -> Option<FastFieldReader<f64>> {
+        self.fast_field_f64.get(&field).cloned()
+    }
+
+    /// Returns a `u64s` multi-valued fast field reader reader associated to `field`.
+    ///
+    /// If `field` is not a u64 multi-valued fast field, this method returns `None`.
+    pub fn u64s(&self, field: Field) -> Option<MultiValueIntFastFieldReader<u64>> {
+        self.fast_field_u64s.get(&field).cloned()
+    }
+
+    /// If the field is a u64s-fast field return the associated reader.
+    /// If the field is a i64s-fast field, return the associated u64s reader. Values are
+    /// mapped from i64 to u64 using a (well the, it is unique) monotonic mapping.
+    ///
+    /// This method is useful when merging segment reader.
+    pub(crate) fn u64s_lenient(&self, field: Field) -> Option<MultiValueIntFastFieldReader<u64>> {
+        if let Some(u64s_ff_reader) = self.u64s(field) {
+            return Some(u64s_ff_reader);
+        }
+        if let Some(i64s_ff_reader) = self.i64s(field) {
+            return Some(i64s_ff_reader.into_u64s_reader());
+        }
+        if let Some(f64s_ff_reader) = self.f64s(field) {
+            return Some(f64s_ff_reader.into_u64s_reader());
+        }
+        None
+    }
+
+    /// Returns a `i64s` multi-valued fast field reader reader associated to `field`.
+    ///
+    /// If `field` is not a i64 multi-valued fast field, this method returns `None`.
+    pub fn i64s(&self, field: Field) -> Option<MultiValueIntFastFieldReader<i64>> {
+        self.fast_field_i64s.get(&field).cloned()
+    }
+
+    /// Returns a `f64s` multi-valued fast field reader reader associated to `field`.
+    ///
+    /// If `field` is not a f64 multi-valued fast field, this method returns `None`.
+    pub fn f64s(&self, field: Field) -> Option<MultiValueIntFastFieldReader<f64>> {
+        self.fast_field_f64s.get(&field).cloned()
+    }
+
+    /// Returns a `crate::DateTime` multi-valued fast field reader reader associated to `field`.
+    ///
+    /// If `field` is not a `crate::DateTime` multi-valued fast field, this method returns `None`.
+    pub fn dates(&self, field: Field) -> Option<MultiValueIntFastFieldReader<crate::DateTime>> {
+        self.fast_field_dates.get(&field).cloned()
+    }
+
+    /// Returns the `bytes` fast field reader associated to `field`.
+    ///
+    /// If `field` is not a bytes fast field, returns `None`.
+    pub fn bytes(&self, field: Field) -> Option<BytesFastFieldReader> {
+        self.fast_bytes.get(&field).cloned()
+    }
+}

src/fastfield/serializer.rs

@@ -1,10 +1,10 @@
-use common::bitpacker::BitPacker;
-use common::compute_num_bits;
-use common::BinarySerializable;
-use common::CompositeWrite;
-use common::CountingWriter;
-use directory::WritePtr;
-use schema::Field;
+use crate::common::bitpacker::BitPacker;
+use crate::common::compute_num_bits;
+use crate::common::BinarySerializable;
+use crate::common::CompositeWrite;
+use crate::common::CountingWriter;
+use crate::directory::WritePtr;
+use crate::schema::Field;
 use std::io::{self, Write};
 
 /// `FastFieldSerializer` is in charge of serializing
@@ -45,7 +45,7 @@ impl FastFieldSerializer {
         field: Field,
         min_value: u64,
         max_value: u64,
-    ) -> io::Result<FastSingleFieldSerializer<CountingWriter<WritePtr>>> {
+    ) -> io::Result<FastSingleFieldSerializer<'_, CountingWriter<WritePtr>>> {
         self.new_u64_fast_field_with_idx(field, min_value, max_value, 0)
     }
 
@@ -56,7 +56,7 @@ impl FastFieldSerializer {
         min_value: u64,
         max_value: u64,
         idx: usize,
-    ) -> io::Result<FastSingleFieldSerializer<CountingWriter<WritePtr>>> {
+    ) -> io::Result<FastSingleFieldSerializer<'_, CountingWriter<WritePtr>>> {
         let field_write = self.composite_write.for_field_with_idx(field, idx);
         FastSingleFieldSerializer::open(field_write, min_value, max_value)
     }
@@ -66,7 +66,7 @@ impl FastFieldSerializer {
         &mut self,
         field: Field,
         idx: usize,
-    ) -> io::Result<FastBytesFieldSerializer<CountingWriter<WritePtr>>> {
+    ) -> io::Result<FastBytesFieldSerializer<'_, CountingWriter<WritePtr>>> {
         let field_write = self.composite_write.for_field_with_idx(field, idx);
         FastBytesFieldSerializer::open(field_write)
     }
@@ -79,7 +79,7 @@ impl FastFieldSerializer {
     }
 }
 
-pub struct FastSingleFieldSerializer<'a, W: Write + 'a> {
+pub struct FastSingleFieldSerializer<'a, W: Write> {
     bit_packer: BitPacker,
     write: &'a mut W,
     min_value: u64,
@@ -127,7 +127,7 @@ impl<'a, W: Write> FastSingleFieldSerializer<'a, W> {
     }
 }
 
-pub struct FastBytesFieldSerializer<'a, W: Write + 'a> {
+pub struct FastBytesFieldSerializer<'a, W: Write> {
     write: &'a mut W,
 }
 

src/fastfield/writer.rs

@@ -1,13 +1,14 @@
 use super::multivalued::MultiValueIntFastFieldWriter;
-use common;
-use common::BinarySerializable;
-use common::VInt;
-use fastfield::{BytesFastFieldWriter, FastFieldSerializer};
-use postings::UnorderedTermId;
-use schema::{Cardinality, Document, Field, FieldType, Schema};
+use crate::common;
+use crate::common::BinarySerializable;
+use crate::common::VInt;
+use crate::fastfield::{BytesFastFieldWriter, FastFieldSerializer};
+use crate::postings::UnorderedTermId;
+use crate::schema::{Cardinality, Document, Field, FieldEntry, FieldType, Schema};
+use crate::termdict::TermOrdinal;
+use fnv::FnvHashMap;
 use std::collections::HashMap;
 use std::io;
-use termdict::TermOrdinal;
 
 /// The fastfieldswriter regroup all of the fast field writers.
 pub struct FastFieldsWriter {
@@ -16,6 +17,14 @@ pub struct FastFieldsWriter {
     bytes_value_writers: Vec<BytesFastFieldWriter>,
 }
 
+fn fast_field_default_value(field_entry: &FieldEntry) -> u64 {
+    match *field_entry.field_type() {
+        FieldType::I64(_) | FieldType::Date(_) => common::i64_to_u64(0i64),
+        FieldType::F64(_) => common::f64_to_u64(0.0f64),
+        _ => 0u64,
+    }
+}
+
 impl FastFieldsWriter {
     /// Create all `FastFieldWriter` required by the schema.
     pub fn from_schema(schema: &Schema) -> FastFieldsWriter {
@@ -23,18 +32,16 @@ impl FastFieldsWriter {
         let mut multi_values_writers = Vec::new();
         let mut bytes_value_writers = Vec::new();
 
-        for (field_id, field_entry) in schema.fields().iter().enumerate() {
-            let field = Field(field_id as u32);
-            let default_value = if let FieldType::I64(_) = *field_entry.field_type() {
-                common::i64_to_u64(0i64)
-            } else {
-                0u64
-            };
+        for (field, field_entry) in schema.fields() {
             match *field_entry.field_type() {
-                FieldType::I64(ref int_options) | FieldType::U64(ref int_options) => {
+                FieldType::I64(ref int_options)
+                | FieldType::U64(ref int_options)
+                | FieldType::F64(ref int_options)
+                | FieldType::Date(ref int_options) => {
                     match int_options.get_fastfield_cardinality() {
                         Some(Cardinality::SingleValue) => {
                             let mut fast_field_writer = IntFastFieldWriter::new(field);
+                            let default_value = fast_field_default_value(field_entry);
                             fast_field_writer.set_val_if_missing(default_value);
                             single_value_writers.push(fast_field_writer);
                         }
@@ -114,7 +121,7 @@ impl FastFieldsWriter {
     pub fn serialize(
         &self,
         serializer: &mut FastFieldSerializer,
-        mapping: &HashMap<Field, HashMap<UnorderedTermId, TermOrdinal>>,
+        mapping: &HashMap<Field, FnvHashMap<UnorderedTermId, TermOrdinal>>,
     ) -> io::Result<()> {
         for field_writer in &self.single_value_writers {
             field_writer.serialize(serializer)?;
@@ -142,9 +149,9 @@ impl FastFieldsWriter {
 /// bitpacked and the number of bits required for bitpacking
 /// can only been known once we have seen all of the values.
 ///
-/// Both u64, and i64 use the same writer.
-/// i64 are just remapped to the `0..2^64 - 1`
-/// using `common::i64_to_u64`.
+/// Both u64, i64 and f64 use the same writer.
+/// i64 and f64 are just remapped to the `0..2^64 - 1`
+/// using `common::i64_to_u64` and `common::f64_to_u64`.
 pub struct IntFastFieldWriter {
     field: Field,
     vals: Vec<u8>,
@@ -203,8 +210,8 @@ impl IntFastFieldWriter {
     /// Extract the value associated to the fast field for
     /// this document.
     ///
-    /// i64 are remapped to u64 using the logic
-    /// in `common::i64_to_u64`.
+    /// i64 and f64 are remapped to u64 using the logic
+    /// in `common::i64_to_u64` and `common::f64_to_u64`.
     ///
     /// If the value is missing, then the default value is used
     /// instead.

src/fieldnorm/code.rs

@@ -10,28 +10,263 @@ pub fn fieldnorm_to_id(fieldnorm: u32) -> u8 {
         .unwrap_or_else(|idx| idx - 1) as u8
 }
 
-#[cfg_attr(feature = "cargo-clippy", allow(clippy::unreadable_literal))]
 pub const FIELD_NORMS_TABLE: [u32; 256] = [
-    0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25,
-    26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 42, 44, 46, 48, 50, 52, 54, 56, 60,
-    64, 68, 72, 76, 80, 84, 88, 96, 104, 112, 120, 128, 136, 144, 152, 168, 184, 200, 216, 232,
-    248, 264, 280, 312, 344, 376, 408, 440, 472, 504, 536, 600, 664, 728, 792, 856, 920, 984,
-    1_048, 1176, 1304, 1432, 1560, 1688, 1816, 1944, 2072, 2328, 2584, 2840, 3096, 3352, 3608,
-    3864, 4120, 4632, 5144, 5656, 6168, 6680, 7192, 7704, 8216, 9240, 10264, 11288, 12312, 13336,
-    14360, 15384, 16408, 18456, 20504, 22552, 24600, 26648, 28696, 30744, 32792, 36888, 40984,
-    45080, 49176, 53272, 57368, 61464, 65560, 73752, 81944, 90136, 98328, 106520, 114712, 122904,
-    131096, 147480, 163864, 180248, 196632, 213016, 229400, 245784, 262168, 294936, 327704, 360472,
-    393240, 426008, 458776, 491544, 524312, 589848, 655384, 720920, 786456, 851992, 917528, 983064,
-    1048600, 1179672, 1310744, 1441816, 1572888, 1703960, 1835032, 1966104, 2097176, 2359320,
-    2621464, 2883608, 3145752, 3407896, 3670040, 3932184, 4194328, 4718616, 5242904, 5767192,
-    6291480, 6815768, 7340056, 7864344, 8388632, 9437208, 10485784, 11534360, 12582936, 13631512,
-    14680088, 15728664, 16777240, 18874392, 20971544, 23068696, 25165848, 27263000, 29360152,
-    31457304, 33554456, 37748760, 41943064, 46137368, 50331672, 54525976, 58720280, 62914584,
-    67108888, 75497496, 83886104, 92274712, 100663320, 109051928, 117440536, 125829144, 134217752,
-    150994968, 167772184, 184549400, 201326616, 218103832, 234881048, 251658264, 268435480,
-    301989912, 335544344, 369098776, 402653208, 436207640, 469762072, 503316504, 536870936,
-    603979800, 671088664, 738197528, 805306392, 872415256, 939524120, 1006632984, 1073741848,
-    1207959576, 1342177304, 1476395032, 1610612760, 1744830488, 1879048216, 2013265944,
+    0,
+    1,
+    2,
+    3,
+    4,
+    5,
+    6,
+    7,
+    8,
+    9,
+    10,
+    11,
+    12,
+    13,
+    14,
+    15,
+    16,
+    17,
+    18,
+    19,
+    20,
+    21,
+    22,
+    23,
+    24,
+    25,
+    26,
+    27,
+    28,
+    29,
+    30,
+    31,
+    32,
+    33,
+    34,
+    35,
+    36,
+    37,
+    38,
+    39,
+    40,
+    42,
+    44,
+    46,
+    48,
+    50,
+    52,
+    54,
+    56,
+    60,
+    64,
+    68,
+    72,
+    76,
+    80,
+    84,
+    88,
+    96,
+    104,
+    112,
+    120,
+    128,
+    136,
+    144,
+    152,
+    168,
+    184,
+    200,
+    216,
+    232,
+    248,
+    264,
+    280,
+    312,
+    344,
+    376,
+    408,
+    440,
+    472,
+    504,
+    536,
+    600,
+    664,
+    728,
+    792,
+    856,
+    920,
+    984,
+    1_048,
+    1_176,
+    1_304,
+    1_432,
+    1_560,
+    1_688,
+    1_816,
+    1_944,
+    2_072,
+    2_328,
+    2_584,
+    2_840,
+    3_096,
+    3_352,
+    3_608,
+    3_864,
+    4_120,
+    4_632,
+    5_144,
+    5_656,
+    6_168,
+    6_680,
+    7_192,
+    7_704,
+    8_216,
+    9_240,
+    10_264,
+    11_288,
+    12_312,
+    13_336,
+    14_360,
+    15_384,
+    16_408,
+    18_456,
+    20_504,
+    22_552,
+    24_600,
+    26_648,
+    28_696,
+    30_744,
+    32_792,
+    36_888,
+    40_984,
+    45_080,
+    49_176,
+    53_272,
+    57_368,
+    61_464,
+    65_560,
+    73_752,
+    81_944,
+    90_136,
+    98_328,
+    106_520,
+    114_712,
+    122_904,
+    131_096,
+    147_480,
+    163_864,
+    180_248,
+    196_632,
+    213_016,
+    229_400,
+    245_784,
+    262_168,
+    294_936,
+    327_704,
+    360_472,
+    393_240,
+    426_008,
+    458_776,
+    491_544,
+    524_312,
+    589_848,
+    655_384,
+    720_920,
+    786_456,
+    851_992,
+    917_528,
+    983_064,
+    1_048_600,
+    1_179_672,
+    1_310_744,
+    1_441_816,
+    1_572_888,
+    1_703_960,
+    1_835_032,
+    1_966_104,
+    2_097_176,
+    2_359_320,
+    2_621_464,
+    2_883_608,
+    3_145_752,
+    3_407_896,
+    3_670_040,
+    3_932_184,
+    4_194_328,
+    4_718_616,
+    5_242_904,
+    5_767_192,
+    6_291_480,
+    6_815_768,
+    7_340_056,
+    7_864_344,
+    8_388_632,
+    9_437_208,
+    10_485_784,
+    11_534_360,
+    12_582_936,
+    13_631_512,
+    14_680_088,
+    15_728_664,
+    16_777_240,
+    18_874_392,
+    20_971_544,
+    23_068_696,
+    25_165_848,
+    27_263_000,
+    29_360_152,
+    31_457_304,
+    33_554_456,
+    37_748_760,
+    41_943_064,
+    46_137_368,
+    50_331_672,
+    54_525_976,
+    58_720_280,
+    62_914_584,
+    67_108_888,
+    75_497_496,
+    83_886_104,
+    92_274_712,
+    100_663_320,
+    109_051_928,
+    117_440_536,
+    125_829_144,
+    134_217_752,
+    150_994_968,
+    167_772_184,
+    184_549_400,
+    201_326_616,
+    218_103_832,
+    234_881_048,
+    251_658_264,
+    268_435_480,
+    301_989_912,
+    335_544_344,
+    369_098_776,
+    402_653_208,
+    436_207_640,
+    469_762_072,
+    503_316_504,
+    536_870_936,
+    603_979_800,
+    671_088_664,
+    738_197_528,
+    805_306_392,
+    872_415_256,
+    939_524_120,
+    1_006_632_984,
+    1_073_741_848,
+    1_207_959_576,
+    1_342_177_304,
+    1_476_395_032,
+    1_610_612_760,
+    1_744_830_488,
+    1_879_048_216,
+    2_013_265_944,
 ];
 
 #[cfg(test)]

src/fieldnorm/reader.rs

@@ -1,6 +1,6 @@
 use super::{fieldnorm_to_id, id_to_fieldnorm};
-use directory::ReadOnlySource;
-use DocId;
+use crate::directory::ReadOnlySource;
+use crate::DocId;
 
 /// Reads the fieldnorm associated to a document.
 /// The fieldnorm represents the length associated to

src/fieldnorm/serializer.rs

@@ -1,6 +1,6 @@
-use common::CompositeWrite;
-use directory::WritePtr;
-use schema::Field;
+use crate::common::CompositeWrite;
+use crate::directory::WritePtr;
+use crate::schema::Field;
 use std::io;
 use std::io::Write;
 

src/fieldnorm/writer.rs

@@ -1,9 +1,9 @@
-use DocId;
+use crate::DocId;
 
 use super::fieldnorm_to_id;
 use super::FieldNormsSerializer;
-use schema::Field;
-use schema::Schema;
+use crate::schema::Field;
+use crate::schema::Schema;
 use std::io;
 
 /// The `FieldNormsWriter` is in charge of tracking the fieldnorm byte
@@ -22,11 +22,14 @@ impl FieldNormsWriter {
     pub(crate) fn fields_with_fieldnorm(schema: &Schema) -> Vec<Field> {
         schema
             .fields()
-            .iter()
-            .enumerate()
-            .filter(|&(_, field_entry)| field_entry.is_indexed())
-            .map(|(field, _)| Field(field as u32))
-            .collect::<Vec<Field>>()
+            .filter_map(|(field, field_entry)| {
+                if field_entry.is_indexed() {
+                    Some(field)
+                } else {
+                    None
+                }
+            })
+            .collect::<Vec<_>>()
     }
 
     /// Initialize with state for tracking the field norm fields
@@ -35,7 +38,7 @@ impl FieldNormsWriter {
         let fields = FieldNormsWriter::fields_with_fieldnorm(schema);
         let max_field = fields
             .iter()
-            .map(|field| field.0)
+            .map(Field::field_id)
             .max()
             .map(|max_field_id| max_field_id as usize + 1)
             .unwrap_or(0);
@@ -50,8 +53,8 @@ impl FieldNormsWriter {
     ///
     /// Will extend with 0-bytes for documents that have not been seen.
     pub fn fill_up_to_max_doc(&mut self, max_doc: DocId) {
-        for &field in self.fields.iter() {
-            self.fieldnorms_buffer[field.0 as usize].resize(max_doc as usize, 0u8);
+        for field in self.fields.iter() {
+            self.fieldnorms_buffer[field.field_id() as usize].resize(max_doc as usize, 0u8);
         }
     }
 
@@ -64,7 +67,7 @@ impl FieldNormsWriter {
     /// * field     - the field being set
     /// * fieldnorm - the number of terms present in document `doc` in field `field`
     pub fn record(&mut self, doc: DocId, field: Field, fieldnorm: u32) {
-        let fieldnorm_buffer: &mut Vec<u8> = &mut self.fieldnorms_buffer[field.0 as usize];
+        let fieldnorm_buffer: &mut Vec<u8> = &mut self.fieldnorms_buffer[field.field_id() as usize];
         assert!(
             fieldnorm_buffer.len() <= doc as usize,
             "Cannot register a given fieldnorm twice"
@@ -77,7 +80,7 @@ impl FieldNormsWriter {
     /// Serialize the seen fieldnorm values to the serializer for all fields.
     pub fn serialize(&self, fieldnorms_serializer: &mut FieldNormsSerializer) -> io::Result<()> {
         for &field in self.fields.iter() {
-            let fieldnorm_values: &[u8] = &self.fieldnorms_buffer[field.0 as usize][..];
+            let fieldnorm_values: &[u8] = &self.fieldnorms_buffer[field.field_id() as usize][..];
             fieldnorms_serializer.serialize_field(field, fieldnorm_values)?;
         }
         Ok(())

src/functional_test.rs

@@ -1,10 +1,10 @@
 use rand::thread_rng;
 use std::collections::HashSet;
 
+use crate::schema::*;
+use crate::Index;
+use crate::Searcher;
 use rand::Rng;
-use schema::*;
-use Index;
-use Searcher;
 
 fn check_index_content(searcher: &Searcher, vals: &HashSet<u64>) {
     assert!(searcher.segment_readers().len() < 20);
@@ -13,15 +13,15 @@ fn check_index_content(searcher: &Searcher, vals: &HashSet<u64>) {
 
 #[test]
 #[ignore]
-#[cfg(feature = "mmap")]
 fn test_indexing() {
     let mut schema_builder = Schema::builder();
 
-    let id_field = schema_builder.add_u64_field("id", INT_INDEXED);
-    let multiples_field = schema_builder.add_u64_field("multiples", INT_INDEXED);
+    let id_field = schema_builder.add_u64_field("id", INDEXED);
+    let multiples_field = schema_builder.add_u64_field("multiples", INDEXED);
     let schema = schema_builder.build();
 
     let index = Index::create_from_tempdir(schema).unwrap();
+    let reader = index.reader().unwrap();
 
     let mut rng = thread_rng();
 
@@ -36,8 +36,8 @@ fn test_indexing() {
             index_writer.commit().expect("Commit failed");
             committed_docs.extend(&uncommitted_docs);
             uncommitted_docs.clear();
-            index.load_searchers().unwrap();
-            let searcher = index.searcher();
+            reader.reload().unwrap();
+            let searcher = reader.searcher();
             // check that everything is correct.
             check_index_content(&searcher, &committed_docs);
         } else {

src/indexer/delete_queue.rs

@@ -1,7 +1,8 @@
 use super::operation::DeleteOperation;
+use crate::Opstamp;
 use std::mem;
 use std::ops::DerefMut;
-use std::sync::{Arc, RwLock};
+use std::sync::{Arc, RwLock, Weak};
 
 // The DeleteQueue is similar in conceptually to a multiple
 // consumer single producer broadcast channel.
@@ -13,17 +14,18 @@ use std::sync::{Arc, RwLock};
 //
 // New consumer can be created in two ways
 // - calling `delete_queue.cursor()` returns a cursor, that
-//   will include all future delete operation (and no past operations).
+//   will include all future delete operation (and some or none
+//   of the past operations... The client is in charge of checking the opstamps.).
 // - cloning an existing cursor returns a new cursor, that
 //   is at the exact same position, and can now advance independently
 //   from the original cursor.
 #[derive(Default)]
 struct InnerDeleteQueue {
     writer: Vec<DeleteOperation>,
-    last_block: Option<Arc<Block>>,
+    last_block: Weak<Block>,
 }
 
-#[derive(Clone, Default)]
+#[derive(Clone)]
 pub struct DeleteQueue {
     inner: Arc<RwLock<InnerDeleteQueue>>,
 }
@@ -31,20 +33,31 @@ pub struct DeleteQueue {
 impl DeleteQueue {
     // Creates a new delete queue.
     pub fn new() -> DeleteQueue {
-        let delete_queue = DeleteQueue {
+        DeleteQueue {
             inner: Arc::default(),
-        };
-
-        let next_block = NextBlock::from(delete_queue.clone());
-        {
-            let mut delete_queue_wlock = delete_queue.inner.write().unwrap();
-            delete_queue_wlock.last_block = Some(Arc::new(Block {
-                operations: Arc::default(),
-                next: next_block,
-            }));
         }
+    }
 
-        delete_queue
+    fn get_last_block(&self) -> Arc<Block> {
+        {
+            // try get the last block with simply acquiring the read lock.
+            let rlock = self.inner.read().unwrap();
+            if let Some(block) = rlock.last_block.upgrade() {
+                return block;
+            }
+        }
+        // It failed. Let's double check after acquiring the write, as someone could have called
+        // `get_last_block` right after we released the rlock.
+        let mut wlock = self.inner.write().unwrap();
+        if let Some(block) = wlock.last_block.upgrade() {
+            return block;
+        }
+        let block = Arc::new(Block {
+            operations: Arc::default(),
+            next: NextBlock::from(self.clone()),
+        });
+        wlock.last_block = Arc::downgrade(&block);
+        block
     }
 
     // Creates a new cursor that makes it possible to
@@ -52,17 +65,7 @@ impl DeleteQueue {
     //
     // Past delete operations are not accessible.
     pub fn cursor(&self) -> DeleteCursor {
-        let last_block = self
-            .inner
-            .read()
-            .expect("Read lock poisoned when opening delete queue cursor")
-            .last_block
-            .clone()
-            .expect(
-                "Failed to unwrap last_block. This should never happen
-                as the Option<> is only here to make
-                initialization possible",
-            );
+        let last_block = self.get_last_block();
         let operations_len = last_block.operations.len();
         DeleteCursor {
             block: last_block,
@@ -98,23 +101,19 @@ impl DeleteQueue {
             .write()
             .expect("Failed to acquire write lock on delete queue writer");
 
-        let delete_operations;
-        {
-            let writer: &mut Vec<DeleteOperation> = &mut self_wlock.writer;
-            if writer.is_empty() {
-                return None;
-            }
-            delete_operations = mem::replace(writer, vec![]);
+        if self_wlock.writer.is_empty() {
+            return None;
         }
 
-        let next_block = NextBlock::from(self.clone());
-        {
-            self_wlock.last_block = Some(Arc::new(Block {
-                operations: Arc::new(delete_operations),
-                next: next_block,
-            }));
-        }
-        self_wlock.last_block.clone()
+        let delete_operations = mem::replace(&mut self_wlock.writer, vec![]);
+
+        let new_block = Arc::new(Block {
+            operations: Arc::new(delete_operations.into_boxed_slice()),
+            next: NextBlock::from(self.clone()),
+        });
+
+        self_wlock.last_block = Arc::downgrade(&new_block);
+        Some(new_block)
     }
 }
 
@@ -168,7 +167,7 @@ impl NextBlock {
 }
 
 struct Block {
-    operations: Arc<Vec<DeleteOperation>>,
+    operations: Arc<Box<[DeleteOperation]>>,
     next: NextBlock,
 }
 
@@ -184,18 +183,15 @@ impl DeleteCursor {
     ///   queue are consume and the next get will return None.
     /// - the next get will return the first operation with an
     /// `opstamp >= target_opstamp`.
-    pub fn skip_to(&mut self, target_opstamp: u64) {
+    pub fn skip_to(&mut self, target_opstamp: Opstamp) {
         // TODO Can be optimize as we work with block.
         while self.is_behind_opstamp(target_opstamp) {
             self.advance();
         }
     }
 
-    #[cfg_attr(
-        feature = "cargo-clippy",
-        allow(clippy::wrong_self_convention)
-    )]
-    fn is_behind_opstamp(&mut self, target_opstamp: u64) -> bool {
+    #[cfg_attr(feature = "cargo-clippy", allow(clippy::wrong_self_convention))]
+    fn is_behind_opstamp(&mut self, target_opstamp: Opstamp) -> bool {
         self.get()
             .map(|operation| operation.opstamp < target_opstamp)
             .unwrap_or(false)
@@ -252,14 +248,14 @@ impl DeleteCursor {
 mod tests {
 
     use super::{DeleteOperation, DeleteQueue};
-    use schema::{Field, Term};
+    use crate::schema::{Field, Term};
 
     #[test]
     fn test_deletequeue() {
         let delete_queue = DeleteQueue::new();
 
         let make_op = |i: usize| {
-            let field = Field(1u32);
+            let field = Field::from_field_id(1u32);
             DeleteOperation {
                 opstamp: i as u64,
                 term: Term::from_field_u64(field, i as u64),

src/indexer/directory_lock.rs

@@ -1,131 +0,0 @@
-use directory::error::OpenWriteError;
-use std::io::Write;
-use std::path::{Path, PathBuf};
-use std::thread;
-use std::time::Duration;
-use Directory;
-use TantivyError;
-
-#[derive(Debug, Clone, Copy)]
-pub enum LockType {
-    /// Only one process should be able to write tantivy's index at a time.
-    /// This lock file, when present, is in charge of preventing other processes to open an IndexWriter.
-    ///
-    /// If the process is killed and this file remains, it is safe to remove it manually.
-    ///
-    /// Failing to acquire this lock usually means a misuse of tantivy's API,
-    /// (creating more than one instance of the `IndexWriter`), are a spurious
-    /// lock file remaining after a crash. In the latter case, removing the file after
-    /// checking no process running tantivy is running is safe.
-    IndexWriterLock,
-    /// The meta lock file is here to protect the segment files being opened by
-    /// `.load_searchers()` from being garbage collected.
-    /// It makes it possible for another process to safely consume
-    /// our index in-writing. Ideally, we may have prefered `RWLock` semantics
-    /// here, but it is difficult to achieve on Windows.
-    ///
-    /// Opening segment readers is a very fast process.
-    /// Right now if the lock cannot be acquire on the first attempt, the logic
-    /// is very simplistic. We retry after `100ms` until we effectively
-    /// acquire the lock.
-    /// This lock should not have much contention in normal usage.
-    MetaLock,
-}
-
-/// Retry the logic of acquiring locks is pretty simple.
-/// We just retry `n` times after a given `duratio`, both
-/// depending on the type of lock.
-struct RetryPolicy {
-    num_retries: usize,
-    wait_in_ms: u64,
-}
-
-impl RetryPolicy {
-    fn no_retry() -> RetryPolicy {
-        RetryPolicy {
-            num_retries: 0,
-            wait_in_ms: 0,
-        }
-    }
-
-    fn wait_and_retry(&mut self) -> bool {
-        if self.num_retries == 0 {
-            false
-        } else {
-            self.num_retries -= 1;
-            let wait_duration = Duration::from_millis(self.wait_in_ms);
-            thread::sleep(wait_duration);
-            true
-        }
-    }
-}
-
-impl LockType {
-    fn retry_policy(self) -> RetryPolicy {
-        match self {
-            LockType::IndexWriterLock => RetryPolicy::no_retry(),
-            LockType::MetaLock => RetryPolicy {
-                num_retries: 100,
-                wait_in_ms: 100,
-            },
-        }
-    }
-
-    fn try_acquire_lock(self, directory: &mut Directory) -> Result<DirectoryLock, TantivyError> {
-        let path = self.filename();
-        let mut write = directory.open_write(path).map_err(|e| match e {
-            OpenWriteError::FileAlreadyExists(_) => TantivyError::LockFailure(self),
-            OpenWriteError::IOError(io_error) => TantivyError::IOError(io_error),
-        })?;
-        write.flush()?;
-        Ok(DirectoryLock {
-            directory: directory.box_clone(),
-            path: path.to_owned(),
-        })
-    }
-
-    /// Acquire a lock in the given directory.
-    pub fn acquire_lock(self, directory: &Directory) -> Result<DirectoryLock, TantivyError> {
-        let mut box_directory = directory.box_clone();
-        let mut retry_policy = self.retry_policy();
-        loop {
-            let lock_result = self.try_acquire_lock(&mut *box_directory);
-            match lock_result {
-                Ok(result) => {
-                    return Ok(result);
-                }
-                Err(TantivyError::LockFailure(ref filepath)) => {
-                    if !retry_policy.wait_and_retry() {
-                        return Err(TantivyError::LockFailure(filepath.to_owned()));
-                    }
-                }
-                Err(_) => {}
-            }
-        }
-    }
-
-    fn filename(&self) -> &Path {
-        match *self {
-            LockType::MetaLock => Path::new(".tantivy-meta.lock"),
-            LockType::IndexWriterLock => Path::new(".tantivy-indexer.lock"),
-        }
-    }
-}
-
-/// The `DirectoryLock` is an object that represents a file lock.
-/// See  [`LockType`](struct.LockType.html)
-///
-/// It is transparently associated to a lock file, that gets deleted
-/// on `Drop.` The lock is release automatically on `Drop`.
-pub struct DirectoryLock {
-    directory: Box<Directory>,
-    path: PathBuf,
-}
-
-impl Drop for DirectoryLock {
-    fn drop(&mut self) {
-        if let Err(e) = self.directory.delete(&*self.path) {
-            error!("Failed to remove the lock file. {:?}", e);
-        }
-    }
-}

src/indexer/doc_opstamp_mapping.rs

@@ -1,5 +1,5 @@
-use std::sync::Arc;
-use DocId;
+use crate::DocId;
+use crate::Opstamp;
 
 // Doc to opstamp is used to identify which
 // document should be deleted.
@@ -17,25 +17,25 @@ use DocId;
 // This mapping is (for the moment) stricly increasing
 // because of the way document id are allocated.
 #[derive(Clone)]
-pub enum DocToOpstampMapping {
-    WithMap(Arc<Vec<u64>>),
+pub enum DocToOpstampMapping<'a> {
+    WithMap(&'a [Opstamp]),
     None,
 }
 
-impl From<Vec<u64>> for DocToOpstampMapping {
-    fn from(opstamps: Vec<u64>) -> DocToOpstampMapping {
-        DocToOpstampMapping::WithMap(Arc::new(opstamps))
+impl<'a> From<&'a [u64]> for DocToOpstampMapping<'a> {
+    fn from(opstamps: &[Opstamp]) -> DocToOpstampMapping {
+        DocToOpstampMapping::WithMap(opstamps)
     }
 }
 
-impl DocToOpstampMapping {
+impl<'a> DocToOpstampMapping<'a> {
     /// Given an opstamp return the limit doc id L
     /// such that all doc id D such that
     // D >= L iff opstamp(D) >= than `target_opstamp`.
     //
     // The edge case opstamp = some doc opstamp is in practise
     // never called.
-    pub fn compute_doc_limit(&self, target_opstamp: u64) -> DocId {
+    pub fn compute_doc_limit(&self, target_opstamp: Opstamp) -> DocId {
         match *self {
             DocToOpstampMapping::WithMap(ref doc_opstamps) => {
                 match doc_opstamps.binary_search(&target_opstamp) {
@@ -64,17 +64,18 @@ mod tests {
     #[test]
     fn test_doc_to_opstamp_mapping_complex() {
         {
-            let doc_to_opstamp_mapping = DocToOpstampMapping::from(vec![]);
+            let doc_to_opstamp_mapping = DocToOpstampMapping::from(&[][..]);
             assert_eq!(doc_to_opstamp_mapping.compute_doc_limit(0u64), 0);
             assert_eq!(doc_to_opstamp_mapping.compute_doc_limit(2u64), 0);
         }
         {
-            let doc_to_opstamp_mapping = DocToOpstampMapping::from(vec![1u64]);
+            let doc_to_opstamp_mapping = DocToOpstampMapping::from(&[1u64][..]);
             assert_eq!(doc_to_opstamp_mapping.compute_doc_limit(0u64), 0);
             assert_eq!(doc_to_opstamp_mapping.compute_doc_limit(2u64), 1);
         }
         {
-            let doc_to_opstamp_mapping = DocToOpstampMapping::from(vec![1u64, 12u64, 17u64, 23u64]);
+            let doc_to_opstamp_mapping =
+                DocToOpstampMapping::from(&[1u64, 12u64, 17u64, 23u64][..]);
             assert_eq!(doc_to_opstamp_mapping.compute_doc_limit(0u64), 0);
             for i in 2u64..13u64 {
                 assert_eq!(doc_to_opstamp_mapping.compute_doc_limit(i), 1);

src/indexer/log_merge_policy.rs

@@ -1,17 +1,19 @@
 use super::merge_policy::{MergeCandidate, MergePolicy};
-use core::SegmentMeta;
+use crate::core::SegmentMeta;
 use std::cmp;
 use std::f64;
 
 const DEFAULT_LEVEL_LOG_SIZE: f64 = 0.75;
 const DEFAULT_MIN_LAYER_SIZE: u32 = 10_000;
 const DEFAULT_MIN_MERGE_SIZE: usize = 8;
+const DEFAULT_MAX_MERGE_SIZE: usize = 10_000_000;
 
 /// `LogMergePolicy` tries tries to merge segments that have a similar number of
 /// documents.
 #[derive(Debug, Clone)]
 pub struct LogMergePolicy {
     min_merge_size: usize,
+    max_merge_size: usize,
     min_layer_size: u32,
     level_log_size: f64,
 }
@@ -26,6 +28,12 @@ impl LogMergePolicy {
         self.min_merge_size = min_merge_size;
     }
 
+    /// Set the maximum number docs in a segment for it to be considered for
+    /// merging.
+    pub fn set_max_merge_size(&mut self, max_merge_size: usize) {
+        self.max_merge_size = max_merge_size;
+    }
+
     /// Set the minimum segment size under which all segment belong
     /// to the same level.
     pub fn set_min_layer_size(&mut self, min_layer_size: u32) {
@@ -46,39 +54,44 @@ impl LogMergePolicy {
 
 impl MergePolicy for LogMergePolicy {
     fn compute_merge_candidates(&self, segments: &[SegmentMeta]) -> Vec<MergeCandidate> {
-        if segments.is_empty() {
-            return Vec::new();
-        }
-
         let mut size_sorted_tuples = segments
             .iter()
-            .map(|x| x.num_docs())
+            .map(SegmentMeta::num_docs)
+            .filter(|s| s <= &(self.max_merge_size as u32))
             .enumerate()
             .collect::<Vec<(usize, u32)>>();
 
         size_sorted_tuples.sort_by(|x, y| y.1.cmp(&(x.1)));
 
+        if size_sorted_tuples.len() <= 1 {
+            return Vec::new();
+        }
+
         let size_sorted_log_tuples: Vec<_> = size_sorted_tuples
             .into_iter()
             .map(|(ind, num_docs)| (ind, f64::from(self.clip_min_size(num_docs)).log2()))
             .collect();
 
-        let (first_ind, first_score) = size_sorted_log_tuples[0];
-        let mut current_max_log_size = first_score;
-        let mut levels = vec![vec![first_ind]];
-        for &(ind, score) in (&size_sorted_log_tuples).iter().skip(1) {
-            if score < (current_max_log_size - self.level_log_size) {
-                current_max_log_size = score;
-                levels.push(Vec::new());
+        if let Some(&(first_ind, first_score)) = size_sorted_log_tuples.first() {
+            let mut current_max_log_size = first_score;
+            let mut levels = vec![vec![first_ind]];
+            for &(ind, score) in (&size_sorted_log_tuples).iter().skip(1) {
+                if score < (current_max_log_size - self.level_log_size) {
+                    current_max_log_size = score;
+                    levels.push(Vec::new());
+                }
+                levels.last_mut().unwrap().push(ind);
             }
-            levels.last_mut().unwrap().push(ind);
+            levels
+                .iter()
+                .filter(|level| level.len() >= self.min_merge_size)
+                .map(|ind_vec| {
+                    MergeCandidate(ind_vec.iter().map(|&ind| segments[ind].id()).collect())
+                })
+                .collect()
+        } else {
+            return vec![];
         }
-
-        levels
-            .iter()
-            .filter(|level| level.len() >= self.min_merge_size)
-            .map(|ind_vec| MergeCandidate(ind_vec.iter().map(|&ind| segments[ind].id()).collect()))
-            .collect()
     }
 }
 
@@ -86,6 +99,7 @@ impl Default for LogMergePolicy {
     fn default() -> LogMergePolicy {
         LogMergePolicy {
             min_merge_size: DEFAULT_MIN_MERGE_SIZE,
+            max_merge_size: DEFAULT_MAX_MERGE_SIZE,
             min_layer_size: DEFAULT_MIN_LAYER_SIZE,
             level_log_size: DEFAULT_LEVEL_LOG_SIZE,
         }
@@ -95,12 +109,16 @@ impl Default for LogMergePolicy {
 #[cfg(test)]
 mod tests {
     use super::*;
-    use core::{SegmentId, SegmentMeta};
-    use indexer::merge_policy::MergePolicy;
+    use crate::core::{SegmentId, SegmentMeta, SegmentMetaInventory};
+    use crate::indexer::merge_policy::MergePolicy;
+    use once_cell::sync::Lazy;
+
+    static INVENTORY: Lazy<SegmentMetaInventory> = Lazy::new(SegmentMetaInventory::default);
 
     fn test_merge_policy() -> LogMergePolicy {
         let mut log_merge_policy = LogMergePolicy::default();
         log_merge_policy.set_min_merge_size(3);
+        log_merge_policy.set_max_merge_size(100_000);
         log_merge_policy.set_min_layer_size(2);
         log_merge_policy
     }
@@ -113,7 +131,7 @@ mod tests {
     }
 
     fn create_random_segment_meta(num_docs: u32) -> SegmentMeta {
-        SegmentMeta::new(SegmentId::generate_random(), num_docs)
+        INVENTORY.new_segment_meta(SegmentId::generate_random(), num_docs)
     }
 
     #[test]
@@ -138,11 +156,11 @@ mod tests {
             create_random_segment_meta(10),
             create_random_segment_meta(10),
             create_random_segment_meta(10),
-            create_random_segment_meta(1000),
-            create_random_segment_meta(1000),
-            create_random_segment_meta(1000),
-            create_random_segment_meta(10000),
-            create_random_segment_meta(10000),
+            create_random_segment_meta(1_000),
+            create_random_segment_meta(1_000),
+            create_random_segment_meta(1_000),
+            create_random_segment_meta(10_000),
+            create_random_segment_meta(10_000),
             create_random_segment_meta(10),
             create_random_segment_meta(10),
             create_random_segment_meta(10),
@@ -165,6 +183,7 @@ mod tests {
         let result_list = test_merge_policy().compute_merge_candidates(&test_input);
         assert_eq!(result_list.len(), 2);
     }
+
     #[test]
     fn test_log_merge_policy_small_segments() {
         // segments under min_layer_size are merged together
@@ -179,4 +198,30 @@ mod tests {
         let result_list = test_merge_policy().compute_merge_candidates(&test_input);
         assert_eq!(result_list.len(), 1);
     }
+
+    #[test]
+    fn test_log_merge_policy_all_segments_too_large_to_merge() {
+        let eight_large_segments: Vec<SegmentMeta> =
+            std::iter::repeat_with(|| create_random_segment_meta(100_001))
+                .take(8)
+                .collect();
+        assert!(test_merge_policy()
+            .compute_merge_candidates(&eight_large_segments)
+            .is_empty());
+    }
+
+    #[test]
+    fn test_large_merge_segments() {
+        let test_input = vec![
+            create_random_segment_meta(1_000_000),
+            create_random_segment_meta(100_001),
+            create_random_segment_meta(100_000),
+            create_random_segment_meta(100_000),
+            create_random_segment_meta(100_000),
+        ];
+        let result_list = test_merge_policy().compute_merge_candidates(&test_input);
+        // Do not include large segments
+        assert_eq!(result_list.len(), 1);
+        assert_eq!(result_list[0].0.len(), 3)
+    }
 }

src/indexer/merge_operation.rs

@@ -0,0 +1,74 @@
+use crate::Opstamp;
+use crate::SegmentId;
+use census::{Inventory, TrackedObject};
+use std::collections::HashSet;
+use std::ops::Deref;
+
+#[derive(Default)]
+pub(crate) struct MergeOperationInventory(Inventory<InnerMergeOperation>);
+
+impl Deref for MergeOperationInventory {
+    type Target = Inventory<InnerMergeOperation>;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+impl MergeOperationInventory {
+    pub fn segment_in_merge(&self) -> HashSet<SegmentId> {
+        let mut segment_in_merge = HashSet::default();
+        for merge_op in self.list() {
+            for &segment_id in &merge_op.segment_ids {
+                segment_in_merge.insert(segment_id);
+            }
+        }
+        segment_in_merge
+    }
+}
+
+/// A `MergeOperation` has two roles.
+/// It carries all of the information required to describe a merge:
+/// - `target_opstamp` is the opstamp up to which we want to consume the
+/// delete queue and reflect their deletes.
+/// - `segment_ids` is the list of segment to be merged.
+///
+/// The second role is to ensure keep track of the fact that these
+/// segments are in merge and avoid starting a merge operation that
+/// may conflict with this one.
+///
+/// This works by tracking merge operations. When considering computing
+/// merge candidates, we simply list tracked merge operations and remove
+/// their segments from possible merge candidates.
+pub struct MergeOperation {
+    inner: TrackedObject<InnerMergeOperation>,
+}
+
+pub(crate) struct InnerMergeOperation {
+    target_opstamp: Opstamp,
+    segment_ids: Vec<SegmentId>,
+}
+
+impl MergeOperation {
+    pub(crate) fn new(
+        inventory: &MergeOperationInventory,
+        target_opstamp: Opstamp,
+        segment_ids: Vec<SegmentId>,
+    ) -> MergeOperation {
+        let inner_merge_operation = InnerMergeOperation {
+            target_opstamp,
+            segment_ids,
+        };
+        MergeOperation {
+            inner: inventory.track(inner_merge_operation),
+        }
+    }
+
+    pub fn target_opstamp(&self) -> Opstamp {
+        self.inner.target_opstamp
+    }
+
+    pub fn segment_ids(&self) -> &[SegmentId] {
+        &self.inner.segment_ids[..]
+    }
+}

src/indexer/merge_policy.rs

@@ -1,5 +1,5 @@
-use core::SegmentId;
-use core::SegmentMeta;
+use crate::core::SegmentId;
+use crate::core::SegmentMeta;
 use std::fmt::Debug;
 use std::marker;
 
@@ -11,7 +11,7 @@ pub struct MergeCandidate(pub Vec<SegmentId>);
 ///
 /// Every time a the list of segments changes, the segment updater
 /// asks the merge policy if some segments should be merged.
-pub trait MergePolicy: MergePolicyClone + marker::Send + marker::Sync + Debug {
+pub trait MergePolicy: marker::Send + marker::Sync + Debug {
     /// Given the list of segment metas, returns the list of merge candidates.
     ///
     /// This call happens on the segment updater thread, and will block
@@ -19,21 +19,6 @@ pub trait MergePolicy: MergePolicyClone + marker::Send + marker::Sync + Debug {
     fn compute_merge_candidates(&self, segments: &[SegmentMeta]) -> Vec<MergeCandidate>;
 }
 
-/// MergePolicyClone
-pub trait MergePolicyClone {
-    /// Returns a boxed clone of the MergePolicy.
-    fn box_clone(&self) -> Box<MergePolicy>;
-}
-
-impl<T> MergePolicyClone for T
-where
-    T: 'static + MergePolicy + Clone,
-{
-    fn box_clone(&self) -> Box<MergePolicy> {
-        Box::new(self.clone())
-    }
-}
-
 /// Never merge segments.
 #[derive(Debug, Clone)]
 pub struct NoMergePolicy;
@@ -54,8 +39,8 @@ impl MergePolicy for NoMergePolicy {
 pub mod tests {
 
     use super::*;
-    use core::SegmentId;
-    use core::SegmentMeta;
+    use crate::core::SegmentId;
+    use crate::core::SegmentMeta;
 
     /// `MergePolicy` useful for test purposes.
     ///

src/indexer/mod.rs

@@ -1,8 +1,9 @@
 pub mod delete_queue;
-mod directory_lock;
+
 mod doc_opstamp_mapping;
 pub mod index_writer;
 mod log_merge_policy;
+mod merge_operation;
 pub mod merge_policy;
 pub mod merger;
 pub mod operation;
@@ -15,17 +16,38 @@ pub mod segment_updater;
 mod segment_writer;
 mod stamper;
 
-pub(crate) use self::directory_lock::DirectoryLock;
-pub use self::directory_lock::LockType;
-
 pub use self::index_writer::IndexWriter;
 pub use self::log_merge_policy::LogMergePolicy;
+pub use self::merge_operation::MergeOperation;
 pub use self::merge_policy::{MergeCandidate, MergePolicy, NoMergePolicy};
 pub use self::prepared_commit::PreparedCommit;
-pub use self::segment_entry::{SegmentEntry, SegmentState};
+pub use self::segment_entry::SegmentEntry;
 pub use self::segment_manager::SegmentManager;
 pub use self::segment_serializer::SegmentSerializer;
 pub use self::segment_writer::SegmentWriter;
 
 /// Alias for the default merge policy, which is the `LogMergePolicy`.
 pub type DefaultMergePolicy = LogMergePolicy;
+
+#[cfg(test)]
+mod tests {
+    use crate::schema::{self, Schema};
+    use crate::{Index, Term};
+
+    #[test]
+    fn test_advance_delete_bug() {
+        let mut schema_builder = Schema::builder();
+        let text_field = schema_builder.add_text_field("text", schema::TEXT);
+        let index = Index::create_from_tempdir(schema_builder.build()).unwrap();
+        let mut index_writer = index.writer_with_num_threads(1, 3_000_000).unwrap();
+        // there must be one deleted document in the segment
+        index_writer.add_document(doc!(text_field=>"b"));
+        index_writer.delete_term(Term::from_field_text(text_field, "b"));
+        // we need enough data to trigger the bug (at least 32 documents)
+        for _ in 0..32 {
+            index_writer.add_document(doc!(text_field=>"c"));
+        }
+        index_writer.commit().unwrap();
+        index_writer.commit().unwrap();
+    }
+}