Preparing for release

Closes #1068

.github/ISSUE_TEMPLATE/actions.md

@@ -0,0 +1,13 @@
+---
+name: Actions
+about: Actions not directly related to producing code.
+
+---
+
+# Actions title
+
+Action description. 
+e.g. 
+- benchmark
+- investigate and report
+- etc.

.github/dependabot.yml

@@ -0,0 +1,15 @@
+version: 2
+updates:
+- package-ecosystem: cargo
+  directory: "/"
+  schedule:
+    interval: daily
+    time: "20:00"
+  open-pull-requests-limit: 10
+
+- package-ecosystem: "github-actions"
+  directory: "/"
+  schedule:
+    interval: daily
+    time: "20:00"
+  open-pull-requests-limit: 10

.github/workflows/coverage.yml

@@ -0,0 +1,25 @@
+name: Coverage
+
+on:
+  push:
+    branches: [ main ]
+  pull_request:
+    branches: [ main ]
+
+jobs:
+  coverage:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v3
+      - name: Install Rust
+        run: rustup toolchain install nightly --component llvm-tools-preview
+      - name: Install cargo-llvm-cov
+        run: curl -LsSf https://github.com/taiki-e/cargo-llvm-cov/releases/latest/download/cargo-llvm-cov-x86_64-unknown-linux-gnu.tar.gz | tar xzf - -C ~/.cargo/bin
+      - name: Generate code coverage
+        run: cargo llvm-cov --all-features --workspace --lcov --output-path lcov.info
+      - name: Upload coverage to Codecov
+        uses: codecov/codecov-action@v2
+        with:
+          token: ${{ secrets.CODECOV_TOKEN }} # not required for public repos
+          files: lcov.info
+          fail_ci_if_error: true

.github/workflows/long_running.yml

@@ -0,0 +1,24 @@
+name: Long running tests
+
+on:
+  push:
+    branches: [ main ]
+
+env:
+  CARGO_TERM_COLOR: always
+  NUM_FUNCTIONAL_TEST_ITERATIONS: 20000
+
+jobs:
+  functional_test_unsorted:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v3
+    - name: Run indexing_unsorted
+      run: cargo test indexing_unsorted -- --ignored
+  functional_test_sorted:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v3
+    - name: Run indexing_sorted
+      run: cargo test indexing_sorted -- --ignored
+

.github/workflows/test.yml

@@ -0,0 +1,49 @@
+name: Unit tests
+
+on:
+  push:
+    branches: [ main ]
+  pull_request:
+    branches: [ main ]
+
+env:
+  CARGO_TERM_COLOR: always
+
+jobs:
+  test:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v3
+    - name: Build
+      run: cargo build --verbose --workspace
+    - name: Install latest nightly to test also against unstable feature flag
+      uses: actions-rs/toolchain@v1
+      with:
+            toolchain: nightly
+            override: true
+            components: rustfmt
+
+    - name: Install latest nightly to test also against unstable feature flag
+      uses: actions-rs/toolchain@v1
+      with:
+            toolchain: stable
+            override: true
+            components: rustfmt, clippy
+
+    - name: Run tests
+      run: cargo +stable test --features mmap,brotli-compression,lz4-compression,snappy-compression,failpoints --verbose --workspace
+
+    - name: Run tests quickwit feature
+      run: cargo +stable test --features mmap,quickwit,failpoints --verbose --workspace
+
+    - name: Check Formatting
+      run: cargo +nightly fmt --all -- --check
+
+    - uses: actions-rs/clippy-check@v1
+      with:
+        toolchain: stable
+        token: ${{ secrets.GITHUB_TOKEN }}
+        args: --tests
+

.gitignore

@@ -1,4 +1,5 @@
 tantivy.iml
+.cargo
 proptest-regressions
 *.swp
 target

.travis.yml

@@ -1,92 +0,0 @@
-# Based on the "trust" template v0.1.2
-# https://github.com/japaric/trust/tree/v0.1.2
-
-dist: trusty
-language: rust
-services: docker
-sudo: required
-
-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=
-
-addons:
-  apt:
-    sources:
-      - ubuntu-toolchain-r-test
-      - kalakris-cmake
-    packages:
-      - gcc-4.8
-      - g++-4.8
-      - libcurl4-openssl-dev
-      - libelf-dev
-      - libdw-dev
-      - binutils-dev
-      - cmake
-
-matrix:
-  include:
-    # Android
-    - env: TARGET=aarch64-linux-android DISABLE_TESTS=1
-    #- env: TARGET=arm-linux-androideabi DISABLE_TESTS=1
-    #- env: TARGET=armv7-linux-androideabi DISABLE_TESTS=1
-    #- env: TARGET=i686-linux-android DISABLE_TESTS=1
-    #- env: TARGET=x86_64-linux-android DISABLE_TESTS=1
-
-    # Linux
-    #- env: TARGET=aarch64-unknown-linux-gnu
-    #- env: TARGET=i686-unknown-linux-gnu
-    - 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
-
-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
-  - cargo install cargo-update || echo "cargo-update already installed"
-  - cargo install cargo-travis || echo "cargo-travis already installed"
-
-script:
-  - bash ci/script.sh
-  - cargo fmt --all -- --check
-
-before_deploy:
-  - sh ci/before_deploy.sh
-
-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:
-#    # release tags
-#    - /^v\d+\.\d+\.\d+.*$/
-#    - master
-
-notifications:
-  email:
-    on_success: never

ARCHITECTURE.md

@@ -0,0 +1,295 @@
+# Tantivy
+
+## What is tantivy?
+
+Tantivy is a library that is meant to build search engines. Although it is by no means a port of Lucene, its architecture is strongly inspired by it. If you are familiar with Lucene, you may be struck by the overlapping vocabulary.
+This is not fortuitous.
+
+Tantivy's bread and butter is to address the problem of full-text search :
+
+Given a large set of textual documents, and a text query, return the K-most relevant documents in a very efficient way. To execute these queries rapidly, the tantivy needs to build an index beforehand. The relevance score implemented in the tantivy is not configurable. Tantivy uses the same score as the default similarity used in Lucene / Elasticsearch, called [BM25](https://en.wikipedia.org/wiki/Okapi_BM25).
+
+But tantivy's scope does not stop there. Numerous features are required to power rich-search applications. For instance, one may want to:
+- compute the count of documents matching a query in the different section of an e-commerce website,
+- display an average price per meter square for a real estate search engine,
+- take into account historical user data to rank documents in a specific way,
+- or even use tantivy to power an OLAP database.
+
+A more abstract description of the problem space tantivy is trying to address is the following.
+
+Ingest a large set of documents, create an index that makes it possible to
+rapidly select all documents matching a given predicate (also known as a query) and
+collect some information about them ([See collector](#collector-define-what-to-do-with-matched-documents)).
+
+Roughly speaking the design is following these guiding principles:
+- Search should be O(1) in memory.
+- Indexing should be O(1) in memory. (In practice it is just sublinear)
+- Search should be as fast as possible
+
+This comes at the cost of the dynamicity of the index: while it is possible to add, and delete documents from our corpus, the tantivy is designed to handle these updates in large batches.
+
+## [core/](src/core): Index, segments, searchers.
+
+Core contains all of the high-level code to make it possible to create an index, add documents, delete documents and commit.
+
+This is both the most high-level part of tantivy, the least performance-sensitive one, the seemingly most mundane code... And paradoxically the most complicated part.
+
+### Index and Segments...
+
+A tantivy index is a collection of smaller independent immutable segments. 
+Each segment contains its own independent set of data structures.
+
+A segment is identified by a segment id that is in fact a UUID.
+The file of a segment has the format
+
+ ```segment-id . ext ```
+
+The extension signals which data structure (or [`SegmentComponent`](src/core/segment_component.rs)) is stored in the file.
+
+A small `meta.json` file is in charge of keeping track of the list of segments, as well as the schema.
+
+On commit, one segment per indexing thread is written to disk, and the `meta.json` is then updated atomically.
+
+For a better idea of how indexing works, you may read the [following blog post](https://fulmicoton.com/posts/behold-tantivy-part2/).
+
+
+### Deletes
+
+Deletes happen by deleting a "term". Tantivy does not offer any notion of primary id, so it is up to the user to use a field in their schema as if it was a primary id, and delete the associated term if they want to delete only one specific document.
+
+On commit, tantivy will find all of the segments with documents matching this existing term and create a [tombstone file](src/fastfield/delete.rs) that represents the bitset of the document that are deleted.
+Like all segment files, this file is immutable. Because it is possible to have more than one tombstone file at a given instant, the tombstone filename has the format ``` segment_id . commit_opstamp . del```.
+
+An opstamp is simply an incremental id that identifies any operation applied to the index. For instance, performing a commit or adding a document.
+
+
+### DocId
+
+Within a segment, all documents are identified by a DocId that ranges within `[0, max_doc)`.
+where `max_doc` is the number of documents in the segment, (deleted or not). Having such a compact `DocId` space is key to the compression of our data structures.
+
+The DocIds are simply allocated in the order documents are added to the index.
+
+### Merges
+
+In separate threads, tantivy's index writer search for opportunities to merge segments.
+The point of segment merge is to:
+- eventually get rid of tombstoned documents
+- reduce the otherwise ever-growing number of segments.
+
+Indeed, while having several segments instead of one does not hurt search too much, having hundreds can have a measurable impact on the search performance.
+
+### Searcher
+
+The user of the library usually does not need to know about the existence of Segments.
+Searching is done through an object called a [`Searcher`](src/core/searcher.rs), that captures a
+snapshot of the index at one point of time, by holding a list of [SegmentReader](src/core/segment_reader.rs).
+
+In other words, regardless of commits, file garbage collection, or segment merge that might happen, as long as the user holds and reuse the same [Searcher](src/core/searcher.rs), search will happen on an immutable snapshot of the index.
+
+## [directory/](src/directory): Where should the data be stored?
+
+Tantivy, like Lucene, abstracts the place where the data should be stored in a key-trait
+called [`Directory`](src/directory/directory.rs).
+Contrary to Lucene however, "files" are quite different from some kind of `io::Read` object.
+Check out [`src/directory/directory.rs`](src/directory/directory.rs) trait for more details.
+
+Tantivy ships two main directory implementation: the `MMapDirectory` and the `RAMDirectory`,
+but users can extend tantivy with their own implementation.
+
+## [schema/](src/schema): What are documents?
+
+Tantivy's document follows a very strict schema, decided before building any index.
+
+The schema defines all of the fields that the indexes [`Document`](src/schema/document.rs) may and should contain, their types (`text`, `i64`, `u64`, `Date`, ...) as well as how it should be indexed / represented in tantivy.
+
+Depending on the type of the field, you can decide to
+- put it in the docstore
+- store it as a fast field
+- index it
+
+Practically, tantivy will push values associated with this type to up to 3 respective
+data structures.
+
+*Limitations*
+
+As of today, tantivy's schema imposes a 1:1 relationship between a field that is being ingested and a field represented in the search index. In sophisticated search application, it is fairly common to want to index a field twice using different tokenizers, or to index the concatenation of several fields together into one field.
+
+This is not something tantivy supports, and it is up to the user to duplicate field / concatenate fields before feeding them to tantivy.
+
+## General information about these data structures.
+
+All data structures in tantivy, have:
+- a writer
+- a serializer
+- a reader
+
+The writer builds an in-memory representation of a batch of documents. This representation is not searchable. It is just meant as an intermediary mutable representation, to which we can sequentially add
+the document of a batch. At the end of the batch (or if a memory limit is reached), this representation
+is then converted into an on-disk immutable representation, that is extremely compact.
+This conversion is done by the serializer.
+
+Finally, the reader is in charge of offering an API to read on this on-disk read-only representation.
+In tantivy, readers are designed to require very little anonymous memory. The data is read straight from an mmapped file, and loading an index is as fast as mmapping its files.
+
+## [store/](src/store): Here is my DocId, Gimme my document!
+
+The docstore is a row-oriented storage that, for each document, stores a subset of the fields
+that are marked as stored in the schema. The docstore is compressed using a general-purpose algorithm
+like LZ4.
+
+**Useful for**
+
+In search engines, it is often used to display search results.
+Once the top 10 documents have been identified, we fetch them from the store, and display them or their snippet on the search result page (aka SERP).
+
+**Not useful for**
+
+Fetching a document from the store is typically a "slow" operation. It usually consists in
+- searching into a compact tree-like data structure to find the position of the right block.
+- decompressing a small block
+- returning the document from this block.
+
+It is NOT meant to be called for every document matching a query.
+
+As a rule of thumb, if you hit the docstore more than 100 times per search query, you are probably misusing tantivy.
+
+
+## [fastfield/](src/fastfield): Here is my DocId, Gimme my value!
+
+Fast fields are stored in a column-oriented storage that allows for random access.
+The only compression applied is bitpacking. The column comes with two meta data.
+The minimum value in the column and the number of bits per doc.
+
+Fetching a value for a `DocId` is then as simple as computing
+
+```
+min_value + fetch_bits(num_bits * doc_id..num_bits * (doc_id+1))
+```
+
+This operation just requires one memory fetch.
+Because, DocSets are scanned through in order (DocId are iterated in a sorted manner) which
+also help locality.
+
+In Lucene's jargon, fast fields are called DocValues.
+
+**Useful for**
+
+They are typically integer values that are useful to either rank or compute aggregate over
+all of the documents matching a query (aka [DocSet](src/docset.rs)).
+
+For instance, one could define a function to combine upvotes with tantivy's internal relevancy score.
+This can be done by fetching a fast field during scoring.
+One could also compute the mean price of the items matching a query in an e-commerce website.
+This can be done by fetching a fast field in a collector.
+Finally one could decide to post-filter a docset to remove docset with a price within a specific range.
+If the ratio of filtered out documents is not too low, an efficient way to do this is to fetch the price and apply the filter on the collector side.
+
+Aside from integer values, it is also possible to store an actual byte payload.
+For advanced search engine, it is possible to store all of the features required for learning-to-rank in a byte payload, access it during search, and apply the learning-to-rank model.
+
+Finally facets are a specific kind of fast field, and the associated source code is in [`fastfield/facet_reader.rs`](src/fastfield/facet_reader.rs).
+
+# The inverted search index.
+
+The inverted index is the core part of full-text search.
+When presented a new document with the text field "Hello, happy tax payer!", tantivy breaks it into a list of so-called tokens. In addition to just splitting these strings into tokens, it might also do different kinds of operations like dropping the punctuation, converting the character to lowercase, apply stemming, etc. Tantivy makes it possible to configure the operations to be applied in the schema (tokenizer/ is the place where these operations are implemented).
+
+For instance, the default tokenizer of tantivy would break our text into: `[hello, happy, tax, payer]`.
+The document will therefore be registered in the inverted index as containing the terms
+`[text:hello, text:happy, text:tax, text:payer]`.
+
+The role of the inverted index is, when given a term, gives us in return a very fast iterator over the sorted doc ids that match the term.
+
+Such an iterator is called a posting list. In addition to giving us `DocId`, they can also give us optionally the number of occurrence of the term for each document, also called term frequency or TF.
+
+These iterators being sorted by DocId, one can create an iterator over the document containing `text:tax AND text:payer`, `(text:tax AND text:payer) OR (text:contribuable)` or any boolean expression.
+
+In order to represent the function
+```Term ⟶ Posting```
+
+The inverted index actually consists of two data structures chained together.
+
+- [Term](src/schema/term.rs) ⟶ [TermInfo](src/postings/term_info.rs) is addressed by the term dictionary.
+- [TermInfo](src/postings/term_info.rs) ⟶ [Posting](src/postings/postings.rs) is addressed by the posting lists.
+
+Where [TermInfo](src/postings/term_info.rs) is an object containing some meta data about a term.
+
+
+## [termdict/](src/termdict): Here is a term, give me the [TermInfo](src/postings/term_info.rs)!
+
+Tantivy's term dictionary is mainly in charge of supplying the function
+
+[Term](src/schema/term.rs) ⟶ [TermInfo](src/postings/term_info.rs)
+
+It is itself broken into two parts.
+- [Term](src/schema/term.rs) ⟶ [TermOrdinal](src/termdict/mod.rs) is addressed by a finite state transducer, implemented by the fst crate.
+- [TermOrdinal](src/termdict/mod.rs) ⟶ [TermInfo](src/postings/term_info.rs) is addressed by the term info store.
+
+
+## [postings/](src/postings): Iterate over documents... very fast!
+
+A posting list makes it possible to store a sorted list of doc ids and for each doc store
+a term frequency as well.
+
+The posting lists are stored in a separate file. The [TermInfo](src/postings/term_info.rs) contains an offset into that file and a number of documents for the given posting list. Both are required and sufficient to read the posting list.
+
+The posting list is organized in block of 128 documents.
+One block of doc ids is followed by one block of term frequencies.
+
+The doc ids are delta encoded and bitpacked.
+The term frequencies are bitpacked.
+
+Because the number of docs is rarely a multiple of 128, the last block may contain an arbitrary number of docs between 1 and 127 documents. We then use variable int encoding instead of bitpacking.
+
+## [positions/](src/positions): Where are my terms within the documents?
+
+Phrase queries make it possible to search for documents containing a specific sequence of terms.
+For instance, when the phrase query "the art of war" does not match "the war of art".
+To make it possible, it is possible to specify in the schema that a field should store positions in addition to being indexed.
+
+The token positions of all of the terms are then stored in a separate file with the extension `.pos`.
+The [TermInfo](src/postings/term_info.rs) gives an offset (expressed in position this time) in this file. As we iterate throught the docset,
+we advance the position reader by the number of term frequencies of the current document.
+
+## [fieldnorms/](src/fieldnorms): Here is my doc, how many tokens in this field?
+
+The [BM25](https://en.wikipedia.org/wiki/Okapi_BM25) formula also requires to know the number of tokens stored in a specific field for a given document. We store this information on one byte per document in the fieldnorm.
+The fieldnorm is therefore compressed. Values up to 40 are encoded unchanged.
+
+
+## [tokenizer/](src/tokenizer): How should we process text?
+
+Text processing is key to a good search experience.
+Splits or normalize your text too much, and the search results will have a less precision and a higher recall.
+Do not normalize, or under split your text, you will end up with a higher precision and a lesser recall.
+
+Text processing can be configured by selecting an off-the-shelf [`Tokenizer`](./src/tokenizer/tokenizer.rs) or implementing your own to first split the text into tokens, and then chain different [`TokenFilter`](src/tokenizer/tokenizer.rs)'s to it.
+
+Tantivy's comes with few tokenizers, but external crates are offering advanced tokenizers, such as [Lindera](https://crates.io/crates/lindera) for Japanese.
+
+
+## [query/](src/query): Define and compose queries
+
+The [Query](src/query/query.rs) trait defines what a query is.
+Due to the necessity for some queries to compute some statistics over the entire index, and because the
+index is composed of several `SegmentReader`, the path from transforming a `Query` to an iterator over documents is slightly convoluted, but fundamentally, this is what a Query is.
+
+The iterator over a document comes with some scoring function. The resulting trait is called a
+[Scorer](src/query/scorer.rs) and is specific to a segment.
+
+Different queries can be combined using the [BooleanQuery](src/query/boolean_query/).
+Tantivy comes with different types of queries and can be extended by implementing
+the `Query`, `Weight`, and `Scorer` traits.
+
+## [collector](src/collector): Define what to do with matched documents
+
+Collectors define how to aggregate the documents matching a query, in the broadest sense possible.
+The search will push matched documents one by one, calling their
+`fn collect(doc: DocId, score: Score);` method.
+
+Users may implement their own collectors by implementing the [Collector](src/collector/mod.rs) trait.
+
+## [query-grammar](query-grammar): Defines the grammar of the query parser
+
+While the [QueryParser](src/query/query_parser/query_parser.rs) struct is located in the `query/` directory, the actual parser combinator used to convert user queries into an AST is in an external crate called `query-grammar`. This part was externalized to lighten the work of the compiler.

CHANGELOG.md

@@ -1,3 +1,68 @@
+Tantivy 0.17
+================================
+- LogMergePolicy now triggers merges if the ratio of deleted documents reaches a threshold (@shikhar @fulmicoton) [#115](https://github.com/quickwit-oss/tantivy/issues/115)
+- Adds a searcher Warmer API (@shikhar @fulmicoton)
+- Change to non-strict schema. Ignore fields in data which are not defined in schema. Previously this returned an error. #1211
+- Facets are necessarily indexed. Existing index with indexed facets should work out of the box. Index without facets that are marked with index: false should be broken (but they were already broken in a sense). (@fulmicoton) #1195 .
+- Bugfix that could in theory impact durability in theory on some filesystems [#1224](https://github.com/quickwit-oss/tantivy/issues/1224)
+- Schema now offers not indexing fieldnorms (@lpouget) [#922](https://github.com/quickwit-oss/tantivy/issues/922)
+- Reduce the number of fsync calls [#1225](https://github.com/quickwit-oss/tantivy/issues/1225)
+- Fix opening bytes index with dynamic codec (@PSeitz) [#1278](https://github.com/quickwit-oss/tantivy/issues/1278)
+- Added an aggregation collector compatible with Elasticsearch (@PSeitz)
+- Added a JSON schema type @fulmicoton [#1251](https://github.com/quickwit-oss/tantivy/issues/1251)
+- Added support for slop in phrase queries @halvorboe [#1068](https://github.com/quickwit-oss/tantivy/issues/1068)
+
+Tantivy 0.16.2
+================================
+- Bugfix in FuzzyTermQuery. (tranposition_cost_one was not doing anything)
+
+Tantivy 0.16.1
+========================
+- Major Bugfix on multivalued fastfield.  #1151
+- Demux operation (@PSeitz)
+
+Tantivy 0.16.0
+=========================
+- Bugfix in the filesum check. (@evanxg852000) #1127
+- Bugfix in positions when the index is sorted by a field. (@appaquet) #1125
+
+Tantivy 0.15.3
+=========================
+- Major bugfix. Deleting documents was broken when the index was sorted by a field. (@appaquet, @fulmicoton) #1101
+
+
+Tantivy 0.15.2
+========================
+- Major bugfix. DocStore still panics when a deleted doc is at the beginning of a block. (@appaquet) #1088
+
+Tantivy 0.15.1
+=========================
+- Major bugfix. DocStore panics when first block is deleted. (@appaquet) #1077
+
+Tantivy 0.15.0
+=========================
+- API Changes. Using Range instead of (start, end) in the API and internals (`FileSlice`, `OwnedBytes`, `Snippets`, ...)
+  This change is breaking but migration is trivial.
+- Added an Histogram collector. (@fulmicoton) #994
+- Added support for Option<TCollector>.  (@fulmicoton)
+- DocAddress is now a struct (@scampi) #987
+- Bugfix consistent tie break handling in facet's topk (@hardikpnsp) #357
+- Date field support for range queries (@rihardsk) #516
+- Added lz4-flex as the default compression scheme in tantivy (@PSeitz) #1009
+- Renamed a lot of symbols to avoid all uppercasing on acronyms, as per new clippy recommendation. For instance, RAMDirectory -> RamDirectory. (@fulmicoton)
+- Simplified positions index format (@fulmicoton) #1022
+- Moved bitpacking to bitpacker subcrate and add BlockedBitpacker, which bitpacks blocks of 128 elements (@PSeitz) #1030
+- Added support for more-like-this query in tantivy (@evanxg852000) #1011
+- Added support for sorting an index, e.g presorting documents in an index by a timestamp field. This can heavily improve performance for certain scenarios, by utilizing the sorted data (Top-n optimizations)(@PSeitz). #1026
+- Add iterator over documents in doc store (@PSeitz). #1044
+- Fix log merge policy (@PSeitz). #1043
+- Add detection to avoid small doc store blocks on merge (@PSeitz). #1054
+- Make doc store compression dynamic (@PSeitz). #1060
+- Switch to json for footer version handling (@PSeitz). #1060
+- Updated TermMerger implementation to rely on the union feature of the FST (@scampi) #469
+- Add boolean marking whether position is required in the query_terms API call (@fulmicoton). #1070
+
+
 Tantivy 0.14.0
 =========================
 - Remove dependency to atomicwrites #833 .Implemented by @fulmicoton upon suggestion and research from @asafigan).
@@ -12,6 +77,7 @@ Tantivy 0.14.0
 - Simplified the encoding of the skip reader struct. BlockWAND max tf is now encoded over a single byte. (@fulmicoton)
 - `FilterCollector` now supports all Fast Field value types (@barrotsteindev)
 - FastField are not all loaded when opening the segment reader. (@fulmicoton)
+- Added an API to merge segments, see `tantivy::merge_segments` #1005. (@evanxg852000)
 
 This version breaks compatibility and requires users to reindex everything.
 
@@ -66,7 +132,7 @@ Tantivy 0.12.0
 ## 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/main/examples/custom_tokenizer.rs
+minor changes. Check https://github.com/quickwit-oss/tantivy/blob/main/examples/custom_tokenizer.rs
 to check for some code sample.
 
 Tantivy 0.11.3

Cargo.toml

@@ -1,66 +1,79 @@
 [package]
 name = "tantivy"
-version = "0.14.0"
+version = "0.17.0"
 authors = ["Paul Masurel <paul.masurel@gmail.com>"]
 license = "MIT"
 categories = ["database-implementations", "data-structures"]
 description = """Search engine library"""
 documentation = "https://docs.rs/tantivy/"
-homepage = "https://github.com/tantivy-search/tantivy"
-repository = "https://github.com/tantivy-search/tantivy"
+homepage = "https://github.com/quickwit-oss/tantivy"
+repository = "https://github.com/quickwit-oss/tantivy"
 readme = "README.md"
 keywords = ["search", "information", "retrieval"]
 edition = "2018"
 
 [dependencies]
 base64 = "0.13"
-byteorder = "1"
-crc32fast = "1"
-once_cell = "1"
-regex ={version = "1", default-features = false, features = ["std"]}
+byteorder = "1.4.3"
+crc32fast = "1.2.1"
+once_cell = "1.7.2"
+regex ={ version = "1.5.4", default-features = false, features = ["std"] }
 tantivy-fst = "0.3"
-memmap = {version = "0.7", optional=true}
-lz4 = {version="1", optional=true}
-brotli = {version="3.3.0", optional=true}
-snap = "1"
-tempfile = {version="3", optional=true}
-log = "0.4"
-serde = {version="1", features=["derive"]}
-serde_json = "1"
-num_cpus = "1"
-fs2={version="0.4", optional=true}
+memmap2 = {version = "0.5", optional=true}
+lz4_flex = { version = "0.9", default-features = false, features = ["checked-decode"], optional = true }
+brotli = { version = "3.3", optional = true }
+snap = { version = "1.0.5", optional = true }
+tempfile = { version = "3.2", optional = true }
+log = "0.4.14"
+serde = { version = "1.0.126", features = ["derive"] }
+serde_json = "1.0.64"
+num_cpus = "1.13"
+fs2={ version = "0.4.3", optional = true }
 levenshtein_automata = "0.2"
-uuid = { version = "0.8", features = ["v4", "serde"] }
-crossbeam = "0.8"
-futures = {version = "0.3",  features=["thread-pool"] }
-tantivy-query-grammar = { version="0.14.0", path="./query-grammar" }
-stable_deref_trait = "1"
-rust-stemmers = "1"
-downcast-rs = "1"
-bitpacking = {version="0.8", default-features = false, features=["bitpacker4x"]}
+uuid = { version = "0.8.2", features = ["v4", "serde"] }
+crossbeam = "0.8.1"
+futures = { version = "0.3.15", features = ["thread-pool"] }
+tantivy-query-grammar = { version="0.15.0", path="./query-grammar" }
+tantivy-bitpacker = { version="0.1", path="./bitpacker" }
+common = { version = "0.2", path = "./common/", package = "tantivy-common" }
+fastfield_codecs = { version="0.1", path="./fastfield_codecs", default-features = false }
+ownedbytes = { version="0.2", path="./ownedbytes" }
+stable_deref_trait = "1.2"
+rust-stemmers = "1.2"
+downcast-rs = "1.2"
+bitpacking = { version = "0.8.4", default-features = false, features = ["bitpacker4x"] }
 census = "0.4"
-fnv = "1"
-thiserror = "1.0"
-htmlescape = "0.3"
-fail = "0.4"
+fnv = "1.0.7"
+thiserror = "1.0.24"
+htmlescape = "0.3.1"
+fail = "0.5"
 murmurhash32 = "0.2"
-chrono = "0.4"
-smallvec = "1"
-rayon = "1"
-lru = "0.6"
+chrono = "0.4.19"
+smallvec = "1.6.1"
+rayon = "1.5"
+lru = "0.7.0"
+fastdivide = "0.4"
+itertools = "0.10.0"
+measure_time = "0.8.0"
+pretty_assertions = "1.1.0"
+serde_cbor = {version="0.11", optional=true}
+async-trait = "0.1"
 
 [target.'cfg(windows)'.dependencies]
-winapi = "0.3"
+winapi = "0.3.9"
 
 [dev-dependencies]
-rand = "0.8"
-maplit = "1"
+rand = "0.8.3"
+maplit = "1.0.2"
 matches = "0.1.8"
-proptest = "0.10"
-criterion = "0.3"
+proptest = "1.0"
+criterion = "0.3.5"
+test-log = "0.2.8"
+env_logger = "0.9.0"
+pprof = {version= "0.6", features=["flamegraph", "criterion"]}
 
 [dev-dependencies.fail]
-version = "0.4"
+version = "0.5"
 features = ["failpoints"]
 
 [profile.release]
@@ -73,19 +86,20 @@ debug-assertions = true
 overflow-checks = true
 
 [features]
-default = ["mmap"]
-mmap = ["fs2", "tempfile", "memmap"]
+default = ["mmap", "lz4-compression" ]
+mmap = ["fs2", "tempfile", "memmap2"]
+
 brotli-compression = ["brotli"]
-lz4-compression = ["lz4"]
+lz4-compression = ["lz4_flex"]
+snappy-compression = ["snap"]
+
 failpoints = ["fail/failpoints"]
 unstable = [] # useful for benches.
-wasm-bindgen = ["uuid/wasm-bindgen"]
+
+quickwit = ["serde_cbor"]
 
 [workspace]
-members = ["query-grammar"]
-
-[badges]
-travis-ci = { repository = "tantivy-search/tantivy" }
+members = ["query-grammar", "bitpacker", "common", "fastfield_codecs", "ownedbytes"]
 
 # Following the "fail" crate best practises, we isolate
 # tests that define specific behavior in fail check points
@@ -102,3 +116,8 @@ required-features = ["fail/failpoints"]
 [[bench]]
 name = "analyzer"
 harness = false
+
+[[bench]]
+name = "index-bench"
+harness = false
+

Makefile

@@ -1,3 +1,6 @@
 test:
 	echo "Run test only... No examples."
 	cargo test --tests --lib
+
+fmt:
+	cargo +nightly fmt --all

README.md

@@ -1,25 +1,12 @@
-
-[![Build Status](https://travis-ci.org/tantivy-search/tantivy.svg?branch=main)](https://travis-ci.org/tantivy-search/tantivy)
-[![codecov](https://codecov.io/gh/tantivy-search/tantivy/branch/main/graph/badge.svg)](https://codecov.io/gh/tantivy-search/tantivy)
-[![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)
+[![Docs](https://docs.rs/tantivy/badge.svg)](https://docs.rs/crate/tantivy/)
+[![Build Status](https://github.com/quickwit-oss/tantivy/actions/workflows/test.yml/badge.svg)](https://github.com/quickwit-oss/tantivy/actions/workflows/test.yml)
+[![codecov](https://codecov.io/gh/quickwit-oss/tantivy/branch/main/graph/badge.svg)](https://codecov.io/gh/quickwit-oss/tantivy)
+[![Join the chat at https://discord.gg/MT27AG5EVE](https://shields.io/discord/908281611840282624?label=chat%20on%20discord)](https://discord.gg/MT27AG5EVE)
 [![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/main?svg=true)](https://ci.appveyor.com/project/fulmicoton/tantivy/branch/main)
 [![Crates.io](https://img.shields.io/crates/v/tantivy.svg)](https://crates.io/crates/tantivy)
 
 ![Tantivy](https://tantivy-search.github.io/logo/tantivy-logo.png)
 
-[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/0)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/0)
-[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/1)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/1)
-[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/2)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/2)
-[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/3)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/3)
-[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/4)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/4)
-[![](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/images/5)](https://sourcerer.io/fame/fulmicoton/tantivy-search/tantivy/links/5)
-[![](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.
 
 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
@@ -30,7 +17,7 @@ Tantivy is, in fact, strongly inspired by Lucene's design.
 
 # Benchmark
 
-The following [benchmark](https://tantivy-search.github.io/bench/) break downs 
+The following [benchmark](https://tantivy-search.github.io/bench/) break downs
 performance for different type of queries / collection.
 
 Your mileage WILL vary depending on the nature of queries and their load.
@@ -38,7 +25,7 @@ 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))
+- 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), [Vaporetto](https://crates.io/crates/vaporetto_tantivy), and [tantivy-tokenizer-tiny-segmenter](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)
@@ -60,7 +47,7 @@ Your mileage WILL vary depending on the nature of queries and their load.
 ## Non-features
 
 - 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, 
+library upon which one could build a distributed search. Serializable/mergeable collector state for instance,
 are within the scope of Tantivy.
 
 
@@ -69,22 +56,21 @@ are within the scope of Tantivy.
 Tantivy works on stable Rust (>= 1.27) and supports Linux, MacOS, and Windows.
 
 - [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,
+- [tantivy-cli and its tutorial](https://github.com/quickwit-oss/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/)
 
 # How can I support this project?
 
-There are many ways to support this project. 
+There are many ways to support this project.
 
-- Use Tantivy and tell us about your experience on [Gitter](https://gitter.im/tantivy-search/tantivy) or by email (paul.masurel@gmail.com)
+- Use Tantivy and tell us about your experience on [Discord](https://discord.gg/MT27AG5EVE) 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))
+- Contribute code (you can join [our Discord server](https://discord.gg/MT27AG5EVE))
 - Talk about Tantivy around you
-- [![Become a patron](https://c5.patreon.com/external/logo/become_a_patron_button.png)](https://www.patreon.com/fulmicoton)
 
 # Contributing code
 
@@ -96,7 +82,7 @@ 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
+    git clone https://github.com/quickwit-oss/tantivy.git
     cd tantivy
     cargo build
 ```

appveyor.yml

@@ -18,5 +18,6 @@ install:
 build: false
 
 test_script:
-  - REM SET RUST_LOG=tantivy,test & cargo test --all --verbose --no-default-features --features mmap
+  - REM SET RUST_LOG=tantivy,test & cargo test --all --verbose --no-default-features --features lz4-compression --features mmap
+  - REM SET RUST_LOG=tantivy,test & cargo test test_store --verbose --no-default-features --features lz4-compression --features snappy-compression --features brotli-compression --features mmap
   - REM SET RUST_BACKTRACE=1 & cargo build --examples

benches/analyzer.rs

@@ -1,7 +1,7 @@
 use criterion::{criterion_group, criterion_main, Criterion};
 use tantivy::tokenizer::TokenizerManager;
 
-const ALICE_TXT: &'static str = include_str!("alice.txt");
+const ALICE_TXT: &str = include_str!("alice.txt");
 
 pub fn criterion_benchmark(c: &mut Criterion) {
     let tokenizer_manager = TokenizerManager::default();

benches/index-bench.rs

@@ -0,0 +1,121 @@
+use criterion::{criterion_group, criterion_main, Criterion};
+use pprof::criterion::{Output, PProfProfiler};
+use tantivy::schema::{INDEXED, STORED, STRING, TEXT};
+use tantivy::Index;
+
+const HDFS_LOGS: &str = include_str!("hdfs.json");
+const NUM_REPEATS: usize = 2;
+
+pub fn hdfs_index_benchmark(c: &mut Criterion) {
+    let schema = {
+        let mut schema_builder = tantivy::schema::SchemaBuilder::new();
+        schema_builder.add_u64_field("timestamp", INDEXED);
+        schema_builder.add_text_field("body", TEXT);
+        schema_builder.add_text_field("severity", STRING);
+        schema_builder.build()
+    };
+    let schema_with_store = {
+        let mut schema_builder = tantivy::schema::SchemaBuilder::new();
+        schema_builder.add_u64_field("timestamp", INDEXED | STORED);
+        schema_builder.add_text_field("body", TEXT | STORED);
+        schema_builder.add_text_field("severity", STRING | STORED);
+        schema_builder.build()
+    };
+    let dynamic_schema = {
+        let mut schema_builder = tantivy::schema::SchemaBuilder::new();
+        schema_builder.add_json_field("json", TEXT);
+        schema_builder.build()
+    };
+
+    let mut group = c.benchmark_group("index-hdfs");
+    group.sample_size(20);
+    group.bench_function("index-hdfs-no-commit", |b| {
+        b.iter(|| {
+            let index = Index::create_in_ram(schema.clone());
+            let index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
+            for _ in 0..NUM_REPEATS {
+                for doc_json in HDFS_LOGS.trim().split("\n") {
+                    let doc = schema.parse_document(doc_json).unwrap();
+                    index_writer.add_document(doc).unwrap();
+                }
+            }
+        })
+    });
+    group.bench_function("index-hdfs-with-commit", |b| {
+        b.iter(|| {
+            let index = Index::create_in_ram(schema.clone());
+            let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
+            for _ in 0..NUM_REPEATS {
+                for doc_json in HDFS_LOGS.trim().split("\n") {
+                    let doc = schema.parse_document(doc_json).unwrap();
+                    index_writer.add_document(doc).unwrap();
+                }
+            }
+            index_writer.commit().unwrap();
+        })
+    });
+    group.bench_function("index-hdfs-no-commit-with-docstore", |b| {
+        b.iter(|| {
+            let index = Index::create_in_ram(schema_with_store.clone());
+            let index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
+            for _ in 0..NUM_REPEATS {
+                for doc_json in HDFS_LOGS.trim().split("\n") {
+                    let doc = schema.parse_document(doc_json).unwrap();
+                    index_writer.add_document(doc).unwrap();
+                }
+            }
+        })
+    });
+    group.bench_function("index-hdfs-with-commit-with-docstore", |b| {
+        b.iter(|| {
+            let index = Index::create_in_ram(schema_with_store.clone());
+            let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
+            for _ in 0..NUM_REPEATS {
+                for doc_json in HDFS_LOGS.trim().split("\n") {
+                    let doc = schema.parse_document(doc_json).unwrap();
+                    index_writer.add_document(doc).unwrap();
+                }
+            }
+            index_writer.commit().unwrap();
+        })
+    });
+    group.bench_function("index-hdfs-no-commit-json-without-docstore", |b| {
+        b.iter(|| {
+            let index = Index::create_in_ram(dynamic_schema.clone());
+            let json_field = dynamic_schema.get_field("json").unwrap();
+            let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
+            for _ in 0..NUM_REPEATS {
+                for doc_json in HDFS_LOGS.trim().split("\n") {
+                    let json_val: serde_json::Map<String, serde_json::Value> =
+                        serde_json::from_str(doc_json).unwrap();
+                    let doc = tantivy::doc!(json_field=>json_val);
+                    index_writer.add_document(doc).unwrap();
+                }
+            }
+            index_writer.commit().unwrap();
+        })
+    });
+    group.bench_function("index-hdfs-with-commit-json-without-docstore", |b| {
+        b.iter(|| {
+            let index = Index::create_in_ram(dynamic_schema.clone());
+            let json_field = dynamic_schema.get_field("json").unwrap();
+            let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
+            for _ in 0..NUM_REPEATS {
+                for doc_json in HDFS_LOGS.trim().split("\n") {
+                    let json_val: serde_json::Map<String, serde_json::Value> =
+                        serde_json::from_str(doc_json).unwrap();
+                    let doc = tantivy::doc!(json_field=>json_val);
+                    index_writer.add_document(doc).unwrap();
+                }
+            }
+            index_writer.commit().unwrap();
+        })
+    });
+}
+
+criterion_group! {
+    name = benches;
+    config = Criterion::default().with_profiler(PProfProfiler::new(100, Output::Flamegraph(None)));
+    targets = hdfs_index_benchmark
+}
+criterion_main!(benches);

bitpacker/Cargo.toml

@@ -0,0 +1,15 @@
+[package]
+name = "tantivy-bitpacker"
+version = "0.1.1"
+edition = "2018"
+authors = ["Paul Masurel <paul.masurel@gmail.com>"]
+license = "MIT"
+categories = []
+description = """Tantivy-sub crate: bitpacking"""
+repository = "https://github.com/quickwit-oss/tantivy"
+keywords = []
+
+
+# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
+
+[dependencies]

bitpacker/benches/bench.rs

@@ -0,0 +1,35 @@
+#![feature(test)]
+
+extern crate test;
+
+#[cfg(test)]
+mod tests {
+    use tantivy_bitpacker::BlockedBitpacker;
+    use test::Bencher;
+
+    #[bench]
+    fn bench_blockedbitp_read(b: &mut Bencher) {
+        let mut blocked_bitpacker = BlockedBitpacker::new();
+        for val in 0..=21500 {
+            blocked_bitpacker.add(val * val);
+        }
+        b.iter(|| {
+            let mut out = 0;
+            for val in 0..=21500 {
+                out = blocked_bitpacker.get(val);
+            }
+            out
+        });
+    }
+
+    #[bench]
+    fn bench_blockedbitp_create(b: &mut Bencher) {
+        b.iter(|| {
+            let mut blocked_bitpacker = BlockedBitpacker::new();
+            for val in 0..=21500 {
+                blocked_bitpacker.add(val * val);
+            }
+            blocked_bitpacker
+        });
+    }
+}

bitpacker/src/bitpacker.rs

@@ -1,13 +1,15 @@
-use byteorder::{ByteOrder, LittleEndian, WriteBytesExt};
+use std::convert::TryInto;
 use std::io;
 
-use crate::directory::OwnedBytes;
-
-pub(crate) struct BitPacker {
+pub struct BitPacker {
     mini_buffer: u64,
     mini_buffer_written: usize,
 }
-
+impl Default for BitPacker {
+    fn default() -> Self {
+        BitPacker::new()
+    }
+}
 impl BitPacker {
     pub fn new() -> BitPacker {
         BitPacker {
@@ -16,6 +18,7 @@ impl BitPacker {
         }
     }
 
+    #[inline]
     pub fn write<TWrite: io::Write>(
         &mut self,
         val: u64,
@@ -26,14 +29,14 @@ 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);
-            output.write_u64::<LittleEndian>(self.mini_buffer)?;
+            output.write_all(self.mini_buffer.to_le_bytes().as_ref())?;
             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 {
-                output.write_u64::<LittleEndian>(self.mini_buffer)?;
+                output.write_all(self.mini_buffer.to_le_bytes().as_ref())?;
                 self.mini_buffer_written = 0;
                 self.mini_buffer = 0u64;
             }
@@ -44,10 +47,10 @@ impl BitPacker {
     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 mut arr: [u8; 8] = [0u8; 8];
-            LittleEndian::write_u64(&mut arr, self.mini_buffer);
-            output.write_all(&arr[..num_bytes])?;
+            let bytes = self.mini_buffer.to_le_bytes();
+            output.write_all(&bytes[..num_bytes])?;
             self.mini_buffer_written = 0;
+            self.mini_buffer = 0;
         }
         Ok(())
     }
@@ -60,15 +63,14 @@ impl BitPacker {
     }
 }
 
-#[derive(Clone)]
+#[derive(Clone, Debug, Default)]
 pub struct BitUnpacker {
     num_bits: u64,
     mask: u64,
-    data: OwnedBytes,
 }
 
 impl BitUnpacker {
-    pub fn new(data: OwnedBytes, num_bits: u8) -> BitUnpacker {
+    pub fn new(num_bits: u8) -> BitUnpacker {
         let mask: u64 = if num_bits == 64 {
             !0u64
         } else {
@@ -77,15 +79,14 @@ impl BitUnpacker {
         BitUnpacker {
             num_bits: u64::from(num_bits),
             mask,
-            data,
         }
     }
 
-    pub fn get(&self, idx: u64) -> u64 {
+    #[inline]
+    pub fn get(&self, idx: u64, data: &[u8]) -> u64 {
         if self.num_bits == 0 {
             return 0u64;
         }
-        let data: &[u8] = self.data.as_slice();
         let num_bits = self.num_bits;
         let mask = self.mask;
         let addr_in_bits = idx * num_bits;
@@ -95,7 +96,10 @@ impl BitUnpacker {
             addr + 8 <= data.len() as u64,
             "The fast field field should have been padded with 7 bytes."
         );
-        let val_unshifted_unmasked: u64 = LittleEndian::read_u64(&data[(addr as usize)..]);
+        let bytes: [u8; 8] = (&data[(addr as usize)..(addr as usize) + 8])
+            .try_into()
+            .unwrap();
+        let val_unshifted_unmasked: u64 = u64::from_le_bytes(bytes);
         let val_shifted = (val_unshifted_unmasked >> bit_shift) as u64;
         val_shifted & mask
     }
@@ -104,9 +108,8 @@ impl BitUnpacker {
 #[cfg(test)]
 mod test {
     use super::{BitPacker, BitUnpacker};
-    use crate::directory::OwnedBytes;
 
-    fn create_fastfield_bitpacker(len: usize, num_bits: u8) -> (BitUnpacker, Vec<u64>) {
+    fn create_fastfield_bitpacker(len: usize, num_bits: u8) -> (BitUnpacker, Vec<u64>, Vec<u8>) {
         let mut data = Vec::new();
         let mut bitpacker = BitPacker::new();
         let max_val: u64 = (1u64 << num_bits as u64) - 1u64;
@@ -118,14 +121,14 @@ mod test {
         }
         bitpacker.close(&mut data).unwrap();
         assert_eq!(data.len(), ((num_bits as usize) * len + 7) / 8 + 7);
-        let bitunpacker = BitUnpacker::new(OwnedBytes::new(data), num_bits);
-        (bitunpacker, vals)
+        let bitunpacker = BitUnpacker::new(num_bits);
+        (bitunpacker, vals, data)
     }
 
     fn test_bitpacker_util(len: usize, num_bits: u8) {
-        let (bitunpacker, vals) = create_fastfield_bitpacker(len, num_bits);
+        let (bitunpacker, vals, data) = create_fastfield_bitpacker(len, num_bits);
         for (i, val) in vals.iter().enumerate() {
-            assert_eq!(bitunpacker.get(i as u64), *val);
+            assert_eq!(bitunpacker.get(i as u64, &data), *val);
         }
     }
 

bitpacker/src/blocked_bitpacker.rs

@@ -0,0 +1,177 @@
+use super::bitpacker::BitPacker;
+use super::compute_num_bits;
+use crate::{minmax, BitUnpacker};
+
+const BLOCK_SIZE: usize = 128;
+
+/// `BlockedBitpacker` compresses data in blocks of
+/// 128 elements, while keeping an index on it
+#[derive(Debug, Clone)]
+pub struct BlockedBitpacker {
+    // bitpacked blocks
+    compressed_blocks: Vec<u8>,
+    // uncompressed data, collected until BLOCK_SIZE
+    buffer: Vec<u64>,
+    offset_and_bits: Vec<BlockedBitpackerEntryMetaData>,
+}
+impl Default for BlockedBitpacker {
+    fn default() -> Self {
+        BlockedBitpacker::new()
+    }
+}
+
+/// `BlockedBitpackerEntryMetaData` encodes the
+/// offset and bit_width into a u64 bit field
+///
+/// This saves some space, since 7byte is more
+/// than enough and also keeps the access fast
+/// because of alignment
+#[derive(Debug, Clone, Default)]
+struct BlockedBitpackerEntryMetaData {
+    encoded: u64,
+    base_value: u64,
+}
+
+impl BlockedBitpackerEntryMetaData {
+    fn new(offset: u64, num_bits: u8, base_value: u64) -> Self {
+        let encoded = offset | (num_bits as u64) << (64 - 8);
+        Self {
+            encoded,
+            base_value,
+        }
+    }
+    fn offset(&self) -> u64 {
+        (self.encoded << 8) >> 8
+    }
+    fn num_bits(&self) -> u8 {
+        (self.encoded >> 56) as u8
+    }
+    fn base_value(&self) -> u64 {
+        self.base_value
+    }
+}
+
+#[test]
+fn metadata_test() {
+    let meta = BlockedBitpackerEntryMetaData::new(50000, 6, 40000);
+    assert_eq!(meta.offset(), 50000);
+    assert_eq!(meta.num_bits(), 6);
+}
+
+impl BlockedBitpacker {
+    pub fn new() -> Self {
+        let mut compressed_blocks = vec![];
+        compressed_blocks.resize(8, 0);
+        Self {
+            compressed_blocks,
+            buffer: vec![],
+            offset_and_bits: vec![],
+        }
+    }
+
+    /// The memory used (inclusive childs)
+    pub fn mem_usage(&self) -> usize {
+        std::mem::size_of::<BlockedBitpacker>()
+            + self.compressed_blocks.capacity()
+            + self.offset_and_bits.capacity()
+                * std::mem::size_of_val(&self.offset_and_bits.get(0).cloned().unwrap_or_default())
+            + self.buffer.capacity()
+                * std::mem::size_of_val(&self.buffer.get(0).cloned().unwrap_or_default())
+    }
+
+    #[inline]
+    pub fn add(&mut self, val: u64) {
+        self.buffer.push(val);
+        if self.buffer.len() == BLOCK_SIZE as usize {
+            self.flush();
+        }
+    }
+
+    pub fn flush(&mut self) {
+        if let Some((min_value, max_value)) = minmax(self.buffer.iter()) {
+            let mut bit_packer = BitPacker::new();
+            let num_bits_block = compute_num_bits(*max_value - min_value);
+            // todo performance: the padding handling could be done better, e.g. use a slice and
+            // return num_bytes written from bitpacker
+            self.compressed_blocks
+                .resize(self.compressed_blocks.len() - 8, 0); // remove padding for bitpacker
+            let offset = self.compressed_blocks.len() as u64;
+            // todo performance: for some bit_width we
+            // can encode multiple vals into the
+            // mini_buffer before checking to flush
+            // (to be done in BitPacker)
+            for val in self.buffer.iter() {
+                bit_packer
+                    .write(
+                        *val - min_value,
+                        num_bits_block,
+                        &mut self.compressed_blocks,
+                    )
+                    .expect("cannot write bitpacking to output"); // write to in memory can't fail
+            }
+            bit_packer.flush(&mut self.compressed_blocks).unwrap();
+            self.offset_and_bits
+                .push(BlockedBitpackerEntryMetaData::new(
+                    offset,
+                    num_bits_block,
+                    *min_value,
+                ));
+
+            self.buffer.clear();
+            self.compressed_blocks
+                .resize(self.compressed_blocks.len() + 8, 0); // add padding for bitpacker
+        }
+    }
+    #[inline]
+    pub fn get(&self, idx: usize) -> u64 {
+        let metadata_pos = idx / BLOCK_SIZE as usize;
+        let pos_in_block = idx % BLOCK_SIZE as usize;
+        if let Some(metadata) = self.offset_and_bits.get(metadata_pos) {
+            let unpacked = BitUnpacker::new(metadata.num_bits()).get(
+                pos_in_block as u64,
+                &self.compressed_blocks[metadata.offset() as usize..],
+            );
+            unpacked + metadata.base_value()
+        } else {
+            self.buffer[pos_in_block]
+        }
+    }
+
+    pub fn iter(&self) -> impl Iterator<Item = u64> + '_ {
+        // todo performance: we could decompress a whole block and cache it instead
+        let bitpacked_elems = self.offset_and_bits.len() * BLOCK_SIZE;
+        let iter = (0..bitpacked_elems)
+            .map(move |idx| self.get(idx))
+            .chain(self.buffer.iter().cloned());
+        iter
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    #[test]
+    fn blocked_bitpacker_empty() {
+        let blocked_bitpacker = BlockedBitpacker::new();
+        assert_eq!(blocked_bitpacker.iter().collect::<Vec<u64>>(), vec![]);
+    }
+    #[test]
+    fn blocked_bitpacker_one() {
+        let mut blocked_bitpacker = BlockedBitpacker::new();
+        blocked_bitpacker.add(50000);
+        assert_eq!(blocked_bitpacker.get(0), 50000);
+        assert_eq!(blocked_bitpacker.iter().collect::<Vec<u64>>(), vec![50000]);
+    }
+    #[test]
+    fn blocked_bitpacker_test() {
+        let mut blocked_bitpacker = BlockedBitpacker::new();
+        for val in 0..21500 {
+            blocked_bitpacker.add(val);
+        }
+        for val in 0..21500 {
+            assert_eq!(blocked_bitpacker.get(val as usize), val);
+        }
+        assert_eq!(blocked_bitpacker.iter().count(), 21500);
+        assert_eq!(blocked_bitpacker.iter().last().unwrap(), 21499);
+    }
+}

bitpacker/src/lib.rs

@@ -0,0 +1,80 @@
+mod bitpacker;
+mod blocked_bitpacker;
+
+pub use crate::bitpacker::{BitPacker, BitUnpacker};
+pub use crate::blocked_bitpacker::BlockedBitpacker;
+
+/// Computes the number of bits that will be used for bitpacking.
+///
+/// In general the target is the minimum number of bits
+/// required to express the amplitude given in argument.
+///
+/// e.g. If the amplitude is 10, we can store all ints on simply 4bits.
+///
+/// The logic is slightly more convoluted here as for optimization
+/// reasons, we want to ensure that a value spawns over at most 8 bytes
+/// of aligned bytes.
+///
+/// Spanning over 9 bytes is possible for instance, if we do
+/// bitpacking with an amplitude of 63 bits.
+/// In this case, the second int will start on bit
+/// 63 (which belongs to byte 7) and ends at byte 15;
+/// Hence 9 bytes (from byte 7 to byte 15 included).
+///
+/// To avoid this, we force the number of bits to 64bits
+/// when the result is greater than `64-8 = 56 bits`.
+///
+/// Note that this only affects rare use cases spawning over
+/// a very large range of values. Even in this case, it results
+/// in an extra cost of at most 12% compared to the optimal
+/// number of bits.
+pub fn compute_num_bits(n: u64) -> u8 {
+    let amplitude = (64u32 - n.leading_zeros()) as u8;
+    if amplitude <= 64 - 8 {
+        amplitude
+    } else {
+        64
+    }
+}
+
+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
+}
+
+#[test]
+fn test_compute_num_bits() {
+    assert_eq!(compute_num_bits(1), 1u8);
+    assert_eq!(compute_num_bits(0), 0u8);
+    assert_eq!(compute_num_bits(2), 2u8);
+    assert_eq!(compute_num_bits(3), 2u8);
+    assert_eq!(compute_num_bits(4), 3u8);
+    assert_eq!(compute_num_bits(255), 8u8);
+    assert_eq!(compute_num_bits(256), 9u8);
+    assert_eq!(compute_num_bits(5_000_000_000), 33u8);
+}
+
+#[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)));
+}

common/Cargo.toml

@@ -0,0 +1,17 @@
+[package]
+name = "tantivy-common"
+version = "0.2.0"
+authors = ["Paul Masurel <paul@quickwit.io>", "Pascal Seitz <pascal@quickwit.io>"]
+license = "MIT"
+edition = "2018"
+description = "common traits and utility functions used by multiple tantivy subcrates"
+
+# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
+
+[dependencies]
+byteorder = "1.4.3"
+ownedbytes = { version="0.2", path="../ownedbytes" }
+
+[dev-dependencies]
+proptest = "1.0.0"
+rand = "0.8.4"

common/src/bitset.rs

@@ -0,0 +1,745 @@
+use std::convert::TryInto;
+use std::io::Write;
+use std::{fmt, io, u64};
+
+use ownedbytes::OwnedBytes;
+
+#[derive(Clone, Copy, Eq, PartialEq)]
+pub struct TinySet(u64);
+
+impl fmt::Debug for TinySet {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        self.into_iter().collect::<Vec<u32>>().fmt(f)
+    }
+}
+
+pub struct TinySetIterator(TinySet);
+impl Iterator for TinySetIterator {
+    type Item = u32;
+
+    #[inline]
+    fn next(&mut self) -> Option<Self::Item> {
+        self.0.pop_lowest()
+    }
+}
+
+impl IntoIterator for TinySet {
+    type Item = u32;
+    type IntoIter = TinySetIterator;
+    fn into_iter(self) -> Self::IntoIter {
+        TinySetIterator(self)
+    }
+}
+
+impl TinySet {
+    pub fn serialize<T: Write>(&self, writer: &mut T) -> io::Result<()> {
+        writer.write_all(self.0.to_le_bytes().as_ref())
+    }
+
+    pub fn into_bytes(self) -> [u8; 8] {
+        self.0.to_le_bytes()
+    }
+
+    #[inline]
+    pub fn deserialize(data: [u8; 8]) -> Self {
+        let val: u64 = u64::from_le_bytes(data);
+        TinySet(val)
+    }
+
+    /// Returns an empty `TinySet`.
+    #[inline]
+    pub fn empty() -> TinySet {
+        TinySet(0u64)
+    }
+
+    /// Returns a full `TinySet`.
+    #[inline]
+    pub fn full() -> TinySet {
+        TinySet::empty().complement()
+    }
+
+    pub fn clear(&mut self) {
+        self.0 = 0u64;
+    }
+
+    /// Returns the complement of the set in `[0, 64[`.
+    ///
+    /// Careful on making this function public, as it will break the padding handling in the last
+    /// bucket.
+    #[inline]
+    fn complement(self) -> TinySet {
+        TinySet(!self.0)
+    }
+
+    /// Returns true iff the `TinySet` contains the element `el`.
+    #[inline]
+    pub fn contains(self, el: u32) -> bool {
+        !self.intersect(TinySet::singleton(el)).is_empty()
+    }
+
+    /// Returns the number of elements in the TinySet.
+    #[inline]
+    pub fn len(self) -> u32 {
+        self.0.count_ones()
+    }
+
+    /// Returns the intersection of `self` and `other`
+    #[inline]
+    #[must_use]
+    pub fn intersect(self, other: TinySet) -> TinySet {
+        TinySet(self.0 & other.0)
+    }
+
+    /// Creates a new `TinySet` containing only one element
+    /// within `[0; 64[`
+    #[inline]
+    pub fn singleton(el: u32) -> TinySet {
+        TinySet(1u64 << u64::from(el))
+    }
+
+    /// Insert a new element within [0..64)
+    #[inline]
+    #[must_use]
+    pub fn insert(self, el: u32) -> TinySet {
+        self.union(TinySet::singleton(el))
+    }
+
+    /// Removes an element within [0..64)
+    #[inline]
+    #[must_use]
+    pub fn remove(self, el: u32) -> TinySet {
+        self.intersect(TinySet::singleton(el).complement())
+    }
+
+    /// Insert a new element within [0..64)
+    ///
+    /// returns true if the set changed
+    #[inline]
+    pub fn insert_mut(&mut self, el: u32) -> bool {
+        let old = *self;
+        *self = old.insert(el);
+        old != *self
+    }
+
+    /// Remove a element within [0..64)
+    ///
+    /// returns true if the set changed
+    #[inline]
+    pub fn remove_mut(&mut self, el: u32) -> bool {
+        let old = *self;
+        *self = old.remove(el);
+        old != *self
+    }
+
+    /// Returns the union of two tinysets
+    #[inline]
+    #[must_use]
+    pub fn union(self, other: TinySet) -> TinySet {
+        TinySet(self.0 | other.0)
+    }
+
+    /// Returns true iff the `TinySet` is empty.
+    #[inline]
+    pub fn is_empty(self) -> bool {
+        self.0 == 0u64
+    }
+
+    /// Returns the lowest element in the `TinySet`
+    /// and removes it.
+    #[inline]
+    pub fn pop_lowest(&mut self) -> Option<u32> {
+        if self.is_empty() {
+            None
+        } else {
+            let lowest = self.0.trailing_zeros() as u32;
+            self.0 ^= TinySet::singleton(lowest).0;
+            Some(lowest)
+        }
+    }
+
+    /// Returns a `TinySet` than contains all values up
+    /// to limit excluded.
+    ///
+    /// The limit is assumed to be strictly lower than 64.
+    pub fn range_lower(upper_bound: u32) -> TinySet {
+        TinySet((1u64 << u64::from(upper_bound % 64u32)) - 1u64)
+    }
+
+    /// Returns a `TinySet` that contains all values greater
+    /// or equal to the given limit, included. (and up to 63)
+    ///
+    /// The limit is assumed to be strictly lower than 64.
+    pub fn range_greater_or_equal(from_included: u32) -> TinySet {
+        TinySet::range_lower(from_included).complement()
+    }
+}
+
+#[derive(Clone)]
+pub struct BitSet {
+    tinysets: Box<[TinySet]>,
+    len: u64,
+    max_value: u32,
+}
+
+fn num_buckets(max_val: u32) -> u32 {
+    (max_val + 63u32) / 64u32
+}
+
+impl BitSet {
+    /// serialize a `BitSet`.
+    pub fn serialize<T: Write>(&self, writer: &mut T) -> io::Result<()> {
+        writer.write_all(self.max_value.to_le_bytes().as_ref())?;
+        for tinyset in self.tinysets.iter().cloned() {
+            writer.write_all(&tinyset.into_bytes())?;
+        }
+        writer.flush()?;
+        Ok(())
+    }
+
+    /// Create a new `BitSet` that may contain elements
+    /// within `[0, max_val)`.
+    pub fn with_max_value(max_value: u32) -> BitSet {
+        let num_buckets = num_buckets(max_value);
+        let tinybitsets = vec![TinySet::empty(); num_buckets as usize].into_boxed_slice();
+        BitSet {
+            tinysets: tinybitsets,
+            len: 0,
+            max_value,
+        }
+    }
+
+    /// Create a new `BitSet` that may contain elements. Initially all values will be set.
+    /// within `[0, max_val)`.
+    pub fn with_max_value_and_full(max_value: u32) -> BitSet {
+        let num_buckets = num_buckets(max_value);
+        let mut tinybitsets = vec![TinySet::full(); num_buckets as usize].into_boxed_slice();
+
+        // Fix padding
+        let lower = max_value % 64u32;
+        if lower != 0 {
+            tinybitsets[tinybitsets.len() - 1] = TinySet::range_lower(lower);
+        }
+        BitSet {
+            tinysets: tinybitsets,
+            len: max_value as u64,
+            max_value,
+        }
+    }
+
+    /// Removes all elements from the `BitSet`.
+    pub fn clear(&mut self) {
+        for tinyset in self.tinysets.iter_mut() {
+            *tinyset = TinySet::empty();
+        }
+    }
+
+    /// Intersect with serialized bitset
+    pub fn intersect_update(&mut self, other: &ReadOnlyBitSet) {
+        self.intersect_update_with_iter(other.iter_tinysets());
+    }
+
+    /// Intersect with tinysets
+    fn intersect_update_with_iter(&mut self, other: impl Iterator<Item = TinySet>) {
+        self.len = 0;
+        for (left, right) in self.tinysets.iter_mut().zip(other) {
+            *left = left.intersect(right);
+            self.len += left.len() as u64;
+        }
+    }
+
+    /// Returns the number of elements in the `BitSet`.
+    #[inline]
+    pub fn len(&self) -> usize {
+        self.len as usize
+    }
+
+    /// Inserts an element in the `BitSet`
+    #[inline]
+    pub fn insert(&mut self, el: u32) {
+        // we do not check saturated els.
+        let higher = el / 64u32;
+        let lower = el % 64u32;
+        self.len += if self.tinysets[higher as usize].insert_mut(lower) {
+            1
+        } else {
+            0
+        };
+    }
+
+    /// Inserts an element in the `BitSet`
+    #[inline]
+    pub fn remove(&mut self, el: u32) {
+        // we do not check saturated els.
+        let higher = el / 64u32;
+        let lower = el % 64u32;
+        self.len -= if self.tinysets[higher as usize].remove_mut(lower) {
+            1
+        } else {
+            0
+        };
+    }
+
+    /// Returns true iff the elements is in the `BitSet`.
+    #[inline]
+    pub fn contains(&self, el: u32) -> bool {
+        self.tinyset(el / 64u32).contains(el % 64)
+    }
+
+    /// Returns the first non-empty `TinySet` associated to a bucket lower
+    /// or greater than bucket.
+    ///
+    /// Reminder: the tiny set with the bucket `bucket`, represents the
+    /// elements from `bucket * 64` to `(bucket+1) * 64`.
+    pub fn first_non_empty_bucket(&self, bucket: u32) -> Option<u32> {
+        self.tinysets[bucket as usize..]
+            .iter()
+            .cloned()
+            .position(|tinyset| !tinyset.is_empty())
+            .map(|delta_bucket| bucket + delta_bucket as u32)
+    }
+
+    #[inline]
+    pub fn max_value(&self) -> u32 {
+        self.max_value
+    }
+
+    /// Returns the tiny bitset representing the
+    /// the set restricted to the number range from
+    /// `bucket * 64` to `(bucket + 1) * 64`.
+    pub fn tinyset(&self, bucket: u32) -> TinySet {
+        self.tinysets[bucket as usize]
+    }
+}
+
+/// Serialized BitSet.
+#[derive(Clone)]
+pub struct ReadOnlyBitSet {
+    data: OwnedBytes,
+    max_value: u32,
+}
+
+pub fn intersect_bitsets(left: &ReadOnlyBitSet, other: &ReadOnlyBitSet) -> ReadOnlyBitSet {
+    assert_eq!(left.max_value(), other.max_value());
+    assert_eq!(left.data.len(), other.data.len());
+    let union_tinyset_it = left
+        .iter_tinysets()
+        .zip(other.iter_tinysets())
+        .map(|(left_tinyset, right_tinyset)| left_tinyset.intersect(right_tinyset));
+    let mut output_dataset: Vec<u8> = Vec::with_capacity(left.data.len());
+    for tinyset in union_tinyset_it {
+        output_dataset.extend_from_slice(&tinyset.into_bytes());
+    }
+    ReadOnlyBitSet {
+        data: OwnedBytes::new(output_dataset),
+        max_value: left.max_value(),
+    }
+}
+
+impl ReadOnlyBitSet {
+    pub fn open(data: OwnedBytes) -> Self {
+        let (max_value_data, data) = data.split(4);
+        assert_eq!(data.len() % 8, 0);
+        let max_value: u32 = u32::from_le_bytes(max_value_data.as_ref().try_into().unwrap());
+        ReadOnlyBitSet { data, max_value }
+    }
+
+    /// Number of elements in the bitset.
+    #[inline]
+    pub fn len(&self) -> usize {
+        self.iter_tinysets()
+            .map(|tinyset| tinyset.len() as usize)
+            .sum()
+    }
+
+    /// Iterate the tinyset on the fly from serialized data.
+    #[inline]
+    fn iter_tinysets(&self) -> impl Iterator<Item = TinySet> + '_ {
+        self.data.chunks_exact(8).map(move |chunk| {
+            let tinyset: TinySet = TinySet::deserialize(chunk.try_into().unwrap());
+            tinyset
+        })
+    }
+
+    /// Iterate over the positions of the elements.
+    #[inline]
+    pub fn iter(&self) -> impl Iterator<Item = u32> + '_ {
+        self.iter_tinysets()
+            .enumerate()
+            .flat_map(move |(chunk_num, tinyset)| {
+                let chunk_base_val = chunk_num as u32 * 64;
+                tinyset
+                    .into_iter()
+                    .map(move |val| val + chunk_base_val)
+                    .take_while(move |doc| *doc < self.max_value)
+            })
+    }
+
+    /// Returns true iff the elements is in the `BitSet`.
+    #[inline]
+    pub fn contains(&self, el: u32) -> bool {
+        let byte_offset = el / 8u32;
+        let b: u8 = self.data[byte_offset as usize];
+        let shift = (el % 8) as u8;
+        b & (1u8 << shift) != 0
+    }
+
+    /// Maximum value the bitset may contain.
+    /// (Note this is not the maximum value contained in the set.)
+    ///
+    /// A bitset has an intrinsic capacity.
+    /// It only stores elements within [0..max_value).
+    #[inline]
+    pub fn max_value(&self) -> u32 {
+        self.max_value
+    }
+
+    /// Number of bytes used in the bitset representation.
+    pub fn num_bytes(&self) -> usize {
+        self.data.len()
+    }
+}
+
+impl<'a> From<&'a BitSet> for ReadOnlyBitSet {
+    fn from(bitset: &'a BitSet) -> ReadOnlyBitSet {
+        let mut buffer = Vec::with_capacity(bitset.tinysets.len() * 8 + 4);
+        bitset
+            .serialize(&mut buffer)
+            .expect("serializing into a buffer should never fail");
+        ReadOnlyBitSet::open(OwnedBytes::new(buffer))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+
+    use std::collections::HashSet;
+
+    use ownedbytes::OwnedBytes;
+    use rand::distributions::Bernoulli;
+    use rand::rngs::StdRng;
+    use rand::{Rng, SeedableRng};
+
+    use super::{BitSet, ReadOnlyBitSet, TinySet};
+
+    #[test]
+    fn test_read_serialized_bitset_full_multi() {
+        for i in 0..1000 {
+            let bitset = BitSet::with_max_value_and_full(i);
+            let mut out = vec![];
+            bitset.serialize(&mut out).unwrap();
+
+            let bitset = ReadOnlyBitSet::open(OwnedBytes::new(out));
+            assert_eq!(bitset.len() as usize, i as usize);
+        }
+    }
+
+    #[test]
+    fn test_read_serialized_bitset_full_block() {
+        let bitset = BitSet::with_max_value_and_full(64);
+        let mut out = vec![];
+        bitset.serialize(&mut out).unwrap();
+
+        let bitset = ReadOnlyBitSet::open(OwnedBytes::new(out));
+        assert_eq!(bitset.len() as usize, 64);
+    }
+
+    #[test]
+    fn test_read_serialized_bitset_full() {
+        let mut bitset = BitSet::with_max_value_and_full(5);
+        bitset.remove(3);
+        let mut out = vec![];
+        bitset.serialize(&mut out).unwrap();
+
+        let bitset = ReadOnlyBitSet::open(OwnedBytes::new(out));
+        assert_eq!(bitset.len(), 4);
+    }
+
+    #[test]
+    fn test_bitset_intersect() {
+        let bitset_serialized = {
+            let mut bitset = BitSet::with_max_value_and_full(5);
+            bitset.remove(1);
+            bitset.remove(3);
+            let mut out = vec![];
+            bitset.serialize(&mut out).unwrap();
+
+            ReadOnlyBitSet::open(OwnedBytes::new(out))
+        };
+
+        let mut bitset = BitSet::with_max_value_and_full(5);
+        bitset.remove(1);
+        bitset.intersect_update(&bitset_serialized);
+
+        assert!(bitset.contains(0));
+        assert!(!bitset.contains(1));
+        assert!(bitset.contains(2));
+        assert!(!bitset.contains(3));
+        assert!(bitset.contains(4));
+
+        bitset.intersect_update_with_iter(vec![TinySet::singleton(0)].into_iter());
+
+        assert!(bitset.contains(0));
+        assert!(!bitset.contains(1));
+        assert!(!bitset.contains(2));
+        assert!(!bitset.contains(3));
+        assert!(!bitset.contains(4));
+        assert_eq!(bitset.len(), 1);
+
+        bitset.intersect_update_with_iter(vec![TinySet::singleton(1)].into_iter());
+        assert!(!bitset.contains(0));
+        assert!(!bitset.contains(1));
+        assert!(!bitset.contains(2));
+        assert!(!bitset.contains(3));
+        assert!(!bitset.contains(4));
+        assert_eq!(bitset.len(), 0);
+    }
+
+    #[test]
+    fn test_read_serialized_bitset_empty() {
+        let mut bitset = BitSet::with_max_value(5);
+        bitset.insert(3);
+        let mut out = vec![];
+        bitset.serialize(&mut out).unwrap();
+
+        let bitset = ReadOnlyBitSet::open(OwnedBytes::new(out));
+        assert_eq!(bitset.len(), 1);
+
+        {
+            let bitset = BitSet::with_max_value(5);
+            let mut out = vec![];
+            bitset.serialize(&mut out).unwrap();
+            let bitset = ReadOnlyBitSet::open(OwnedBytes::new(out));
+            assert_eq!(bitset.len(), 0);
+        }
+    }
+
+    #[test]
+    fn test_tiny_set_remove() {
+        {
+            let mut u = TinySet::empty().insert(63u32).insert(5).remove(63u32);
+            assert_eq!(u.pop_lowest(), Some(5u32));
+            assert!(u.pop_lowest().is_none());
+        }
+        {
+            let mut u = TinySet::empty()
+                .insert(63u32)
+                .insert(1)
+                .insert(5)
+                .remove(63u32);
+            assert_eq!(u.pop_lowest(), Some(1u32));
+            assert_eq!(u.pop_lowest(), Some(5u32));
+            assert!(u.pop_lowest().is_none());
+        }
+        {
+            let mut u = TinySet::empty().insert(1).remove(63u32);
+            assert_eq!(u.pop_lowest(), Some(1u32));
+            assert!(u.pop_lowest().is_none());
+        }
+        {
+            let mut u = TinySet::empty().insert(1).remove(1u32);
+            assert!(u.pop_lowest().is_none());
+        }
+    }
+    #[test]
+    fn test_tiny_set() {
+        assert!(TinySet::empty().is_empty());
+        {
+            let mut u = TinySet::empty().insert(1u32);
+            assert_eq!(u.pop_lowest(), Some(1u32));
+            assert!(u.pop_lowest().is_none())
+        }
+        {
+            let mut u = TinySet::empty().insert(1u32).insert(1u32);
+            assert_eq!(u.pop_lowest(), Some(1u32));
+            assert!(u.pop_lowest().is_none())
+        }
+        {
+            let mut u = TinySet::empty().insert(2u32);
+            assert_eq!(u.pop_lowest(), Some(2u32));
+            u.insert_mut(1u32);
+            assert_eq!(u.pop_lowest(), Some(1u32));
+            assert!(u.pop_lowest().is_none());
+        }
+        {
+            let mut u = TinySet::empty().insert(63u32);
+            assert_eq!(u.pop_lowest(), Some(63u32));
+            assert!(u.pop_lowest().is_none());
+        }
+        {
+            let mut u = TinySet::empty().insert(63u32).insert(5);
+            assert_eq!(u.pop_lowest(), Some(5u32));
+            assert_eq!(u.pop_lowest(), Some(63u32));
+            assert!(u.pop_lowest().is_none());
+        }
+        {
+            let original = TinySet::empty().insert(63u32).insert(5);
+            let after_serialize_deserialize = TinySet::deserialize(original.into_bytes());
+            assert_eq!(original, after_serialize_deserialize);
+        }
+    }
+
+    #[test]
+    fn test_bitset() {
+        let test_against_hashset = |els: &[u32], max_value: u32| {
+            let mut hashset: HashSet<u32> = HashSet::new();
+            let mut bitset = BitSet::with_max_value(max_value);
+            for &el in els {
+                assert!(el < max_value);
+                hashset.insert(el);
+                bitset.insert(el);
+            }
+            for el in 0..max_value {
+                assert_eq!(hashset.contains(&el), bitset.contains(el));
+            }
+            assert_eq!(bitset.max_value(), max_value);
+
+            // test deser
+            let mut data = vec![];
+            bitset.serialize(&mut data).unwrap();
+            let ro_bitset = ReadOnlyBitSet::open(OwnedBytes::new(data));
+            for el in 0..max_value {
+                assert_eq!(hashset.contains(&el), ro_bitset.contains(el));
+            }
+            assert_eq!(ro_bitset.max_value(), max_value);
+            assert_eq!(ro_bitset.len(), els.len());
+        };
+
+        test_against_hashset(&[], 0);
+        test_against_hashset(&[], 1);
+        test_against_hashset(&[0u32], 1);
+        test_against_hashset(&[0u32], 100);
+        test_against_hashset(&[1u32, 2u32], 4);
+        test_against_hashset(&[99u32], 100);
+        test_against_hashset(&[63u32], 64);
+        test_against_hashset(&[62u32, 63u32], 64);
+    }
+
+    #[test]
+    fn test_bitset_num_buckets() {
+        use super::num_buckets;
+        assert_eq!(num_buckets(0u32), 0);
+        assert_eq!(num_buckets(1u32), 1);
+        assert_eq!(num_buckets(64u32), 1);
+        assert_eq!(num_buckets(65u32), 2);
+        assert_eq!(num_buckets(128u32), 2);
+        assert_eq!(num_buckets(129u32), 3);
+    }
+
+    #[test]
+    fn test_tinyset_range() {
+        assert_eq!(
+            TinySet::range_lower(3).into_iter().collect::<Vec<u32>>(),
+            [0, 1, 2]
+        );
+        assert!(TinySet::range_lower(0).is_empty());
+        assert_eq!(
+            TinySet::range_lower(63).into_iter().collect::<Vec<u32>>(),
+            (0u32..63u32).collect::<Vec<_>>()
+        );
+        assert_eq!(
+            TinySet::range_lower(1).into_iter().collect::<Vec<u32>>(),
+            [0]
+        );
+        assert_eq!(
+            TinySet::range_lower(2).into_iter().collect::<Vec<u32>>(),
+            [0, 1]
+        );
+        assert_eq!(
+            TinySet::range_greater_or_equal(3)
+                .into_iter()
+                .collect::<Vec<u32>>(),
+            (3u32..64u32).collect::<Vec<_>>()
+        );
+    }
+
+    #[test]
+    fn test_bitset_len() {
+        let mut bitset = BitSet::with_max_value(1_000);
+        assert_eq!(bitset.len(), 0);
+        bitset.insert(3u32);
+        assert_eq!(bitset.len(), 1);
+        bitset.insert(103u32);
+        assert_eq!(bitset.len(), 2);
+        bitset.insert(3u32);
+        assert_eq!(bitset.len(), 2);
+        bitset.insert(103u32);
+        assert_eq!(bitset.len(), 2);
+        bitset.insert(104u32);
+        assert_eq!(bitset.len(), 3);
+        bitset.remove(105u32);
+        assert_eq!(bitset.len(), 3);
+        bitset.remove(104u32);
+        assert_eq!(bitset.len(), 2);
+        bitset.remove(3u32);
+        assert_eq!(bitset.len(), 1);
+        bitset.remove(103u32);
+        assert_eq!(bitset.len(), 0);
+    }
+
+    pub fn sample_with_seed(n: u32, ratio: f64, seed_val: u8) -> Vec<u32> {
+        StdRng::from_seed([seed_val; 32])
+            .sample_iter(&Bernoulli::new(ratio).unwrap())
+            .take(n as usize)
+            .enumerate()
+            .filter_map(|(val, keep)| if keep { Some(val as u32) } else { None })
+            .collect()
+    }
+
+    pub fn sample(n: u32, ratio: f64) -> Vec<u32> {
+        sample_with_seed(n, ratio, 4)
+    }
+
+    #[test]
+    fn test_bitset_clear() {
+        let mut bitset = BitSet::with_max_value(1_000);
+        let els = sample(1_000, 0.01f64);
+        for &el in &els {
+            bitset.insert(el);
+        }
+        assert!(els.iter().all(|el| bitset.contains(*el)));
+        bitset.clear();
+        for el in 0u32..1000u32 {
+            assert!(!bitset.contains(el));
+        }
+    }
+}
+
+#[cfg(all(test, feature = "unstable"))]
+mod bench {
+
+    use test;
+
+    use super::{BitSet, TinySet};
+
+    #[bench]
+    fn bench_tinyset_pop(b: &mut test::Bencher) {
+        b.iter(|| {
+            let mut tinyset = TinySet::singleton(test::black_box(31u32));
+            tinyset.pop_lowest();
+            tinyset.pop_lowest();
+            tinyset.pop_lowest();
+            tinyset.pop_lowest();
+            tinyset.pop_lowest();
+            tinyset.pop_lowest();
+        });
+    }
+
+    #[bench]
+    fn bench_tinyset_sum(b: &mut test::Bencher) {
+        let tiny_set = TinySet::empty().insert(10u32).insert(14u32).insert(21u32);
+        b.iter(|| {
+            assert_eq!(test::black_box(tiny_set).into_iter().sum::<u32>(), 45u32);
+        });
+    }
+
+    #[bench]
+    fn bench_tinyarr_sum(b: &mut test::Bencher) {
+        let v = [10u32, 14u32, 21u32];
+        b.iter(|| test::black_box(v).iter().cloned().sum::<u32>());
+    }
+
+    #[bench]
+    fn bench_bitset_initialize(b: &mut test::Bencher) {
+        b.iter(|| BitSet::with_max_value(1_000_000));
+    }
+}

common/src/lib.rs

@@ -1,70 +1,18 @@
-pub mod bitpacker;
-mod bitset;
-mod composite_file;
-mod counting_writer;
-mod serialize;
-mod vint;
+#![allow(clippy::len_without_is_empty)]
+
+use std::ops::Deref;
 
-pub use self::bitset::BitSet;
-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::{
-    read_u32_vint, read_u32_vint_no_advance, serialize_vint_u32, write_u32_vint, VInt,
-};
 pub use byteorder::LittleEndian as Endianness;
 
-/// 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;
+mod bitset;
+mod serialize;
+mod vint;
+mod writer;
 
-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.
-///
-/// In general the target is the minimum number of bits
-/// required to express the amplitude given in argument.
-///
-/// e.g. If the amplitude is 10, we can store all ints on simply 4bits.
-///
-/// The logic is slightly more convoluted here as for optimization
-/// reasons, we want to ensure that a value spawns over at most 8 bytes
-/// of aligns bytes.
-///
-/// Spanning over 9 bytes is possible for instance, if we do
-/// bitpacking with an amplitude of 63 bits.
-/// In this case, the second int will start on bit
-/// 63 (which belongs to byte 7) and ends at byte 15;
-/// Hence 9 bytes (from byte 7 to byte 15 included).
-///
-/// To avoid this, we force the number of bits to 64bits
-/// when the result is greater than `64-8 = 56 bits`.
-///
-/// Note that this only affects rare use cases spawning over
-/// a very large range of values. Even in this case, it results
-/// in an extra cost of at most 12% compared to the optimal
-/// number of bits.
-pub(crate) fn compute_num_bits(n: u64) -> u8 {
-    let amplitude = (64u32 - n.leading_zeros()) as u8;
-    if amplitude <= 64 - 8 {
-        amplitude
-    } else {
-        64
-    }
-}
+pub use bitset::*;
+pub use serialize::{BinarySerializable, DeserializeFrom, FixedSize};
+pub use vint::{read_u32_vint, read_u32_vint_no_advance, serialize_vint_u32, write_u32_vint, VInt};
+pub use writer::{AntiCallToken, CountingWriter, TerminatingWrite};
 
 /// Has length trait
 pub trait HasLen {
@@ -77,6 +25,12 @@ pub trait HasLen {
     }
 }
 
+impl<T: Deref<Target = [u8]>> HasLen for T {
+    fn len(&self) -> usize {
+        self.deref().len()
+    }
+}
+
 const HIGHEST_BIT: u64 = 1 << 63;
 
 /// Maps a `i64` to `u64`
@@ -99,13 +53,13 @@ const HIGHEST_BIT: u64 = 1 << 63;
 ///
 /// # See also
 /// The [reverse mapping is `u64_to_i64`](./fn.u64_to_i64.html).
-#[inline(always)]
+#[inline]
 pub fn i64_to_u64(val: i64) -> u64 {
     (val as u64) ^ HIGHEST_BIT
 }
 
 /// Reverse the mapping given by [`i64_to_u64`](./fn.i64_to_u64.html).
-#[inline(always)]
+#[inline]
 pub fn u64_to_i64(val: u64) -> i64 {
     (val ^ HIGHEST_BIT) as i64
 }
@@ -127,7 +81,7 @@ pub fn u64_to_i64(val: u64) -> i64 {
 ///
 /// # See also
 /// The [reverse mapping is `u64_to_f64`](./fn.u64_to_f64.html).
-#[inline(always)]
+#[inline]
 pub fn f64_to_u64(val: f64) -> u64 {
     let bits = val.to_bits();
     if val.is_sign_positive() {
@@ -138,7 +92,7 @@ pub fn f64_to_u64(val: f64) -> u64 {
 }
 
 /// Reverse the mapping given by [`i64_to_u64`](./fn.i64_to_u64.html).
-#[inline(always)]
+#[inline]
 pub fn u64_to_f64(val: u64) -> f64 {
     f64::from_bits(if val & HIGHEST_BIT != 0 {
         val ^ HIGHEST_BIT
@@ -148,14 +102,14 @@ pub fn u64_to_f64(val: u64) -> f64 {
 }
 
 #[cfg(test)]
-pub(crate) mod test {
+pub mod test {
 
-    pub use super::minmax;
-    pub use super::serialize::test::fixed_size_test;
-    use super::{compute_num_bits, f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64};
-    use proptest::prelude::*;
     use std::f64;
 
+    use proptest::prelude::*;
+
+    use super::{f64_to_u64, i64_to_u64, u64_to_f64, u64_to_i64, BinarySerializable, FixedSize};
+
     fn test_i64_converter_helper(val: i64) {
         assert_eq!(u64_to_i64(i64_to_u64(val)), val);
     }
@@ -164,6 +118,12 @@ pub(crate) mod test {
         assert_eq!(u64_to_f64(f64_to_u64(val)), val);
     }
 
+    pub fn fixed_size_test<O: BinarySerializable + FixedSize + Default>() {
+        let mut buffer = Vec::new();
+        O::default().serialize(&mut buffer).unwrap();
+        assert_eq!(buffer.len(), O::SIZE_IN_BYTES);
+    }
+
     proptest! {
         #[test]
         fn test_f64_converter_monotonicity_proptest((left, right) in (proptest::num::f64::NORMAL, proptest::num::f64::NORMAL)) {
@@ -198,49 +158,13 @@ pub(crate) mod test {
     #[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
+            .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);
-        assert_eq!(compute_num_bits(0), 0u8);
-        assert_eq!(compute_num_bits(2), 2u8);
-        assert_eq!(compute_num_bits(3), 2u8);
-        assert_eq!(compute_num_bits(4), 3u8);
-        assert_eq!(compute_num_bits(255), 8u8);
-        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)));
-    }
 }

common/src/serialize.rs

@@ -1,10 +1,9 @@
-use crate::common::Endianness;
-use crate::common::VInt;
+use std::io::{Read, Write};
+use std::{fmt, io};
+
 use byteorder::{ReadBytesExt, WriteBytesExt};
-use std::fmt;
-use std::io;
-use std::io::Read;
-use std::io::Write;
+
+use crate::{Endianness, VInt};
 
 /// Trait for a simple binary serialization.
 pub trait BinarySerializable: fmt::Debug + Sized {
@@ -14,6 +13,20 @@ pub trait BinarySerializable: fmt::Debug + Sized {
     fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self>;
 }
 
+pub trait DeserializeFrom<T: BinarySerializable> {
+    fn deserialize(&mut self) -> io::Result<T>;
+}
+
+/// Implement deserialize from &[u8] for all types which implement BinarySerializable.
+///
+/// TryFrom would actually be preferrable, but not possible because of the orphan
+/// rules (not completely sure if this could be resolved)
+impl<T: BinarySerializable> DeserializeFrom<T> for &[u8] {
+    fn deserialize(&mut self) -> io::Result<T> {
+        T::deserialize(self)
+    }
+}
+
 /// `FixedSize` marks a `BinarySerializable` as
 /// always serializing to the same size.
 pub trait FixedSize: BinarySerializable {
@@ -61,6 +74,11 @@ impl<Left: BinarySerializable, Right: BinarySerializable> BinarySerializable for
         Ok((Left::deserialize(reader)?, Right::deserialize(reader)?))
     }
 }
+impl<Left: BinarySerializable + FixedSize, Right: BinarySerializable + FixedSize> FixedSize
+    for (Left, Right)
+{
+    const SIZE_IN_BYTES: usize = Left::SIZE_IN_BYTES + Right::SIZE_IN_BYTES;
+}
 
 impl BinarySerializable for u32 {
     fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
@@ -141,6 +159,28 @@ impl FixedSize for u8 {
     const SIZE_IN_BYTES: usize = 1;
 }
 
+impl BinarySerializable for bool {
+    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
+        let val = if *self { 1 } else { 0 };
+        writer.write_u8(val)
+    }
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<bool> {
+        let val = reader.read_u8()?;
+        match val {
+            0 => Ok(false),
+            1 => Ok(true),
+            _ => Err(io::Error::new(
+                io::ErrorKind::InvalidData,
+                "invalid bool value on deserialization, data corrupted",
+            )),
+        }
+    }
+}
+
+impl FixedSize for bool {
+    const SIZE_IN_BYTES: usize = 1;
+}
+
 impl BinarySerializable for String {
     fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
         let data: &[u8] = self.as_bytes();
@@ -161,9 +201,8 @@ impl BinarySerializable for String {
 #[cfg(test)]
 pub mod test {
 
-    use super::*;
-    use crate::common::VInt;
-
+    use super::{VInt, *};
+    use crate::serialize::BinarySerializable;
     pub fn fixed_size_test<O: BinarySerializable + FixedSize + Default>() {
         let mut buffer = Vec::new();
         O::default().serialize(&mut buffer).unwrap();

common/src/vint.rs

@@ -1,8 +1,9 @@
-use super::BinarySerializable;
-use byteorder::{ByteOrder, LittleEndian};
 use std::io;
-use std::io::Read;
-use std::io::Write;
+use std::io::{Read, Write};
+
+use byteorder::{ByteOrder, LittleEndian};
+
+use super::BinarySerializable;
 
 ///   Wrapper over a `u64` that serializes as a variable int.
 #[derive(Clone, Copy, Debug, Eq, PartialEq)]
@@ -106,7 +107,7 @@ pub fn read_u32_vint_no_advance(data: &[u8]) -> (u32, usize) {
 pub fn write_u32_vint<W: io::Write>(val: u32, writer: &mut W) -> io::Result<()> {
     let mut buf = [0u8; 8];
     let data = serialize_vint_u32(val, &mut buf);
-    writer.write_all(&data)
+    writer.write_all(data)
 }
 
 impl VInt {
@@ -174,15 +175,13 @@ impl BinarySerializable for VInt {
 #[cfg(test)]
 mod tests {
 
-    use super::serialize_vint_u32;
-    use super::VInt;
-    use crate::common::BinarySerializable;
+    use super::{serialize_vint_u32, BinarySerializable, VInt};
 
     fn aux_test_vint(val: u64) {
         let mut v = [14u8; 10];
         let num_bytes = VInt(val).serialize_into(&mut v);
-        for i in num_bytes..10 {
-            assert_eq!(v[i], 14u8);
+        for el in &v[num_bytes..10] {
+            assert_eq!(el, &14u8);
         }
         assert!(num_bytes > 0);
         if num_bytes < 10 {

common/src/writer.rs

@@ -1,7 +1,4 @@
-use crate::directory::AntiCallToken;
-use crate::directory::TerminatingWrite;
-use std::io;
-use std::io::Write;
+use std::io::{self, BufWriter, Write};
 
 pub struct CountingWriter<W> {
     underlying: W,
@@ -16,47 +13,93 @@ impl<W: Write> CountingWriter<W> {
         }
     }
 
+    #[inline]
     pub fn written_bytes(&self) -> u64 {
         self.written_bytes
     }
 
     /// Returns the underlying write object.
     /// Note that this method does not trigger any flushing.
+    #[inline]
     pub fn finish(self) -> W {
         self.underlying
     }
 }
 
 impl<W: Write> Write for CountingWriter<W> {
+    #[inline]
     fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
         let written_size = self.underlying.write(buf)?;
         self.written_bytes += written_size as u64;
         Ok(written_size)
     }
 
+    #[inline]
     fn write_all(&mut self, buf: &[u8]) -> io::Result<()> {
         self.underlying.write_all(buf)?;
         self.written_bytes += buf.len() as u64;
         Ok(())
     }
 
+    #[inline]
     fn flush(&mut self) -> io::Result<()> {
         self.underlying.flush()
     }
 }
 
 impl<W: TerminatingWrite> TerminatingWrite for CountingWriter<W> {
+    #[inline]
     fn terminate_ref(&mut self, token: AntiCallToken) -> io::Result<()> {
         self.underlying.terminate_ref(token)
     }
 }
 
+/// Struct used to prevent from calling
+/// [`terminate_ref`](trait.TerminatingWrite.html#tymethod.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)
+    }
+}
+
+impl<'a> TerminatingWrite for &'a mut Vec<u8> {
+    fn terminate_ref(&mut self, _a: AntiCallToken) -> io::Result<()> {
+        self.flush()
+    }
+}
+
 #[cfg(test)]
 mod test {
 
-    use super::CountingWriter;
     use std::io::Write;
 
+    use super::CountingWriter;
+
     #[test]
     fn test_counting_writer() {
         let buffer: Vec<u8> = vec![];

doc/src/SUMMARY.md

@@ -7,6 +7,7 @@
 - [Segments](./basis.md)
 - [Defining your schema](./schema.md)
 - [Facetting](./facetting.md)
+- [Index Sorting](./index_sorting.md)
 - [Innerworkings](./innerworkings.md)
   - [Inverted index](./inverted_index.md)
 - [Best practise](./inverted_index.md)

doc/src/index_sorting.md

@@ -0,0 +1,61 @@
+
+- [Index Sorting](#index-sorting)
+    + [Why Sorting](#why-sorting)
+        * [Compression](#compression)
+        * [Top-N Optimization](#top-n-optimization)
+        * [Pruning](#pruning)
+        * [Other](#other)
+    + [Usage](#usage)
+
+# Index Sorting
+
+Tantivy allows you to sort the index according to a property.
+
+## Why Sorting
+
+Presorting an index has several advantages:
+
+###### Compression
+
+When data is sorted it is easier to compress the data. E.g. the numbers sequence [5, 2, 3, 1, 4] would be sorted to [1, 2, 3, 4, 5]. 
+If we apply delta encoding this list would be unsorted [5, -3, 1, -2, 3] vs. [1, 1, 1, 1, 1].
+Compression ratio is mainly affected on the fast field of the sorted property, every thing else is likely unaffected. 
+###### Top-N Optimization
+
+When data is presorted by a field and search queries request sorting by the same field, we can leverage the natural order of the documents. 
+E.g. if the data is sorted by timestamp and want the top n newest docs containing a term, we can simply leveraging the order of the docids.
+
+Note: Tantivy 0.16 does not do this optimization yet.
+
+###### Pruning
+
+Let's say we want all documents and want to apply the filter `>= 2010-08-11`. When the data is sorted, we could make a lookup in the fast field to find the docid range and use this as the filter.
+
+Note: Tantivy 0.16 does not do this optimization yet.
+
+###### Other?
+
+In principle there are many algorithms possible that exploit the monotonically increasing nature. (aggregations maybe?)
+
+## Usage
+The index sorting can be configured setting [`sort_by_field`](https://github.com/quickwit-oss/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/core/index_meta.rs#L238) on `IndexSettings` and passing it to a `IndexBuilder`. As of tantvy 0.16 only fast fields are allowed to be used.
+
+```
+let settings = IndexSettings {
+    sort_by_field: Some(IndexSortByField {
+        field: "intval".to_string(),
+        order: Order::Desc,
+    }),
+    ..Default::default()
+};
+let mut index_builder = Index::builder().schema(schema);
+index_builder = index_builder.settings(settings);
+let index = index_builder.create_in_ram().unwrap();
+```
+
+## Implementation details
+
+Sorting an index is applied in the serialization step. In general there are two serialization steps: [Finishing a single segment](https://github.com/quickwit-oss/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/indexer/segment_writer.rs#L338) and [merging multiple segments](https://github.com/quickwit-oss/tantivy/blob/000d76b11a139a84b16b9b95060a1c93e8b9851c/src/indexer/merger.rs#L1073).
+
+In both cases we generate a docid mapping reflecting the sort. This mapping is used when serializing the different components (doc store, fastfields, posting list, normfield, facets).
+

doc/src/json.md

@@ -0,0 +1,128 @@
+# Json
+
+As of tantivy 0.17, tantivy supports a json object type.
+This type can be used to allow for a schema-less search index.
+
+When indexing a json object, we "flatten" the JSON. This operation emits terms that represent a triplet `(json_path, value_type, value)`
+
+For instance,  if user is a json field, the following document:
+
+```json
+{
+    "user": {
+        "name": "Paul Masurel",
+        "address": {
+            "city": "Tokyo",
+            "country": "Japan"
+        },
+        "created_at": "2018-11-12T23:20:50.52Z"
+    }
+}
+```
+
+emits the following tokens:
+-  ("name", Text, "Paul")
+-  ("name", Text, "Masurel")
+-  ("address.city", Text, "Tokyo")
+-  ("address.country", Text, "Japan")
+-  ("created_at", Date, 15420648505)
+
+
+# Bytes-encoding and lexicographical sort.
+
+Like any other terms, these triplets are encoded into a binary format as follows.
+- `json_path`: the json path is a sequence of "segments". In the example above, `address.city`
+is just a debug representation of the json path `["address", "city"]`.
+Its representation is done by separating segments by a unicode char `\x01`, and ending the path by `\x00`.
+- `value type`: One byte represents the `Value` type.
+- `value`: The value representation is just the regular Value representation.
+
+This representation is designed to align the natural sort of Terms with the lexicographical sort
+of their binary representation (Tantivy's dictionary (whether fst or sstable) is sorted and does prefix encoding).
+
+In the example above, the terms will be sorted as
+-  ("address.city", Text, "Tokyo")
+-  ("address.country", Text, "Japan")
+-  ("name", Text, "Masurel")
+-  ("name", Text, "Paul")
+-  ("created_at", Date, 15420648505)
+
+As seen in "pitfalls", we may end up having to search for a value for a same path in several different fields. Putting the field code after the path makes it maximizes compression opportunities but also increases the chances for the two terms to end up in the actual same term dictionary block.
+
+
+# Pitfalls, limitation and corner cases.
+
+Json gives very little information about the type of the literals it stores.
+All numeric types end up mapped as a "Number" and there are no types for dates.
+
+At indexing, tantivy will try to interpret number and strings as different type with a
+priority order.
+
+Numbers will be interpreted as u64, i64 and f64 in that order.
+Strings will be interpreted as rfc3999 dates or simple strings.
+
+The first working type is picked and is the only term that is emitted for indexing.
+Note this interpretation happens on a per-document basis, and there is no effort to try to sniff
+a consistent field type at the scale of a segment.
+
+On the query parser side on the other hand, we may end up emitting more than one type.
+For instance, we do not even know if the type is a number or string based.
+
+So the query
+
+```
+my_path.my_segment:233
+```
+
+Will be interpreted as
+`(my_path.my_segment, String, 233) or (my_path.my_segment, u64, 233)`
+
+Likewise, we need to emit two tokens if the query contains an rfc3999 date.
+Indeed the date could have been actually a single token inside the text of a document at ingestion time. Generally speaking, we will always at least emit a string token in query parsing, and sometimes more.
+
+If one more json field is defined, things get even more complicated.
+
+
+## Default json field
+
+If the schema contains a text field called "text" and a json field that is set as a default field:
+`text:hello` could be reasonably interpreted as targetting the text field or as targetting the json field called `json_dynamic` with the json_path "text".
+
+If there is such an ambiguity, we decide to only search in the "text" field: `text:hello`.
+
+In other words, the parser will not search in default json fields if there is a schema hit.
+This is a product decision.
+
+The user can still target the JSON field by specifying its name explicitly:
+`json_dynamic.text:hello`.
+
+## Range queries are not supported.
+
+Json field do not support range queries.
+
+## Arrays do not work like nested object.
+
+If json object contains an array, a search query might return more documents
+than what might be expected.
+
+Let's take an example.
+
+```json
+{
+    "cart_id": 3234234 ,
+    "cart": [
+        {"product_type": "sneakers", "attributes": {"color": "white"} },
+        {"product_type": "t-shirt", "attributes": {"color": "red"}},
+    ]
+}
+```
+
+Despite the array structure, a document in tantivy is a bag of terms.
+The query:
+
+```
+cart.product_type:sneakers AND cart.attributes.color:red
+```
+
+Actually match the document above.
+

examples/aggregation.rs

@@ -0,0 +1,131 @@
+// # Aggregation example
+//
+// This example shows how you can use built-in aggregations.
+// We will use range buckets and compute the average in each bucket.
+//
+
+use serde_json::Value;
+use tantivy::aggregation::agg_req::{
+    Aggregation, Aggregations, BucketAggregation, BucketAggregationType, MetricAggregation,
+    RangeAggregation,
+};
+use tantivy::aggregation::agg_result::AggregationResults;
+use tantivy::aggregation::metric::AverageAggregation;
+use tantivy::aggregation::AggregationCollector;
+use tantivy::query::TermQuery;
+use tantivy::schema::{self, Cardinality, IndexRecordOption, Schema, TextFieldIndexing};
+use tantivy::{doc, Index, Term};
+
+fn main() -> tantivy::Result<()> {
+    let mut schema_builder = Schema::builder();
+    let text_fieldtype = schema::TextOptions::default()
+        .set_indexing_options(
+            TextFieldIndexing::default()
+                .set_tokenizer("default")
+                .set_index_option(IndexRecordOption::WithFreqs),
+        )
+        .set_stored();
+    let text_field = schema_builder.add_text_field("text", text_fieldtype);
+    let score_fieldtype =
+        crate::schema::NumericOptions::default().set_fast(Cardinality::SingleValue);
+    let highscore_field = schema_builder.add_f64_field("highscore", score_fieldtype.clone());
+    let price_field = schema_builder.add_f64_field("price", score_fieldtype.clone());
+
+    let schema = schema_builder.build();
+
+    // # Indexing documents
+    //
+    // Lets index a bunch of documents for this example.
+    let index = Index::create_in_ram(schema);
+
+    let mut index_writer = index.writer(50_000_000)?;
+    // writing the segment
+    index_writer.add_document(doc!(
+        text_field => "cool",
+        highscore_field => 1f64,
+        price_field => 0f64,
+    ))?;
+    index_writer.add_document(doc!(
+        text_field => "cool",
+        highscore_field => 3f64,
+        price_field => 1f64,
+    ))?;
+    index_writer.add_document(doc!(
+        text_field => "cool",
+        highscore_field => 5f64,
+        price_field => 1f64,
+    ))?;
+    index_writer.add_document(doc!(
+        text_field => "nohit",
+        highscore_field => 6f64,
+        price_field => 2f64,
+    ))?;
+    index_writer.add_document(doc!(
+        text_field => "cool",
+        highscore_field => 7f64,
+        price_field => 2f64,
+    ))?;
+    index_writer.commit()?;
+    index_writer.add_document(doc!(
+        text_field => "cool",
+        highscore_field => 11f64,
+        price_field => 10f64,
+    ))?;
+    index_writer.add_document(doc!(
+        text_field => "cool",
+        highscore_field => 14f64,
+        price_field => 15f64,
+    ))?;
+
+    index_writer.add_document(doc!(
+        text_field => "cool",
+        highscore_field => 15f64,
+        price_field => 20f64,
+    ))?;
+
+    index_writer.commit()?;
+
+    let reader = index.reader()?;
+    let text_field = reader.searcher().schema().get_field("text").unwrap();
+
+    let term_query = TermQuery::new(
+        Term::from_field_text(text_field, "cool"),
+        IndexRecordOption::Basic,
+    );
+
+    let sub_agg_req_1: Aggregations = vec![(
+        "average_price".to_string(),
+        Aggregation::Metric(MetricAggregation::Average(
+            AverageAggregation::from_field_name("price".to_string()),
+        )),
+    )]
+    .into_iter()
+    .collect();
+
+    let agg_req_1: Aggregations = vec![(
+        "score_ranges".to_string(),
+        Aggregation::Bucket(BucketAggregation {
+            bucket_agg: BucketAggregationType::Range(RangeAggregation {
+                field: "highscore".to_string(),
+                ranges: vec![
+                    (-1f64..9f64).into(),
+                    (9f64..14f64).into(),
+                    (14f64..20f64).into(),
+                ],
+            }),
+            sub_aggregation: sub_agg_req_1.clone(),
+        }),
+    )]
+    .into_iter()
+    .collect();
+
+    let collector = AggregationCollector::from_aggs(agg_req_1);
+
+    let searcher = reader.searcher();
+    let agg_res: AggregationResults = searcher.search(&term_query, &collector).unwrap();
+
+    let res: Value = serde_json::from_str(&serde_json::to_string(&agg_res)?)?;
+    println!("{}", serde_json::to_string_pretty(&res)?);
+
+    Ok(())
+}

examples/basic_search.rs

@@ -73,7 +73,7 @@ fn main() -> tantivy::Result<()> {
     // multithreaded.
     //
     // Here we give tantivy a budget of `50MB`.
-    // Using a bigger heap for the indexer may increase
+    // Using a bigger memory_arena for the indexer may increase
     // throughput, but 50 MB is already plenty.
     let mut index_writer = index.writer(50_000_000)?;
 
@@ -91,12 +91,12 @@ fn main() -> tantivy::Result<()> {
     old_man_doc.add_text(title, "The Old Man and the Sea");
     old_man_doc.add_text(
         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.",
+        "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.",
     );
 
     // ... and add it to the `IndexWriter`.
-    index_writer.add_document(old_man_doc);
+    index_writer.add_document(old_man_doc)?;
 
     // For convenience, tantivy also comes with a macro to
     // reduce the boilerplate above.
@@ -110,7 +110,7 @@ fn main() -> tantivy::Result<()> {
             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!(
@@ -120,7 +120,7 @@ fn main() -> tantivy::Result<()> {
              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
     // here. You can check out tantivy's tutorial to index

examples/custom_collector.rs

@@ -10,10 +10,9 @@
 // ---
 // Importing tantivy...
 use tantivy::collector::{Collector, SegmentCollector};
-use tantivy::fastfield::FastFieldReader;
+use tantivy::fastfield::{DynamicFastFieldReader, FastFieldReader};
 use tantivy::query::QueryParser;
-use tantivy::schema::Field;
-use tantivy::schema::{Schema, FAST, INDEXED, TEXT};
+use tantivy::schema::{Field, Schema, FAST, INDEXED, TEXT};
 use tantivy::{doc, Index, Score, SegmentReader};
 
 #[derive(Default)]
@@ -86,19 +85,17 @@ impl Collector for StatsCollector {
 
     fn merge_fruits(&self, segment_stats: Vec<Option<Stats>>) -> tantivy::Result<Option<Stats>> {
         let mut stats = Stats::default();
-        for segment_stats_opt in segment_stats {
-            if let Some(segment_stats) = segment_stats_opt {
-                stats.count += segment_stats.count;
-                stats.sum += segment_stats.sum;
-                stats.squared_sum += segment_stats.squared_sum;
-            }
+        for segment_stats in segment_stats.into_iter().flatten() {
+            stats.count += segment_stats.count;
+            stats.sum += segment_stats.sum;
+            stats.squared_sum += segment_stats.squared_sum;
         }
         Ok(stats.non_zero_count())
     }
 }
 
 struct StatsSegmentCollector {
-    fast_field_reader: FastFieldReader<u64>,
+    fast_field_reader: DynamicFastFieldReader<u64>,
     stats: Stats,
 }
 
@@ -139,7 +136,7 @@ fn main() -> tantivy::Result<()> {
     //
     // Lets index a bunch of fake documents for the sake of
     // this example.
-    let index = Index::create_in_ram(schema.clone());
+    let index = Index::create_in_ram(schema);
 
     let mut index_writer = index.writer(50_000_000)?;
     index_writer.add_document(doc!(
@@ -147,23 +144,23 @@ fn main() -> tantivy::Result<()> {
         product_description => "While it is ok for short distance travel, this broom \
         was designed quiditch. It will up your game.",
         price => 30_200u64
-    ));
+    ))?;
     index_writer.add_document(doc!(
         product_name => "Turbulobroom",
         product_description => "You might have heard of this broom before : it is the sponsor of the Wales team.\
             You'll enjoy its sharp turns, and rapid acceleration",
         price => 29_240u64
-    ));
+    ))?;
     index_writer.add_document(doc!(
         product_name => "Broomio",
         product_description => "Great value for the price. This broom is a market favorite",
         price => 21_240u64
-    ));
+    ))?;
     index_writer.add_document(doc!(
         product_name => "Whack a Mole",
         product_description => "Prime quality bat.",
         price => 5_200u64
-    ));
+    ))?;
     index_writer.commit()?;
 
     let reader = index.reader()?;

examples/custom_tokenizer.rs

@@ -62,13 +62,13 @@ fn main() -> tantivy::Result<()> {
     // multithreaded.
     //
     // Here we use a buffer of 50MB per thread. Using a bigger
-    // heap for the indexer can increase its throughput.
+    // memory arena 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."
-    ));
+    ))?;
     index_writer.add_document(doc!(
     title => "Of Mice and Men",
        body => r#"A few miles south of Soledad, the Salinas River drops in close to the hillside
@@ -79,14 +79,14 @@ fn main() -> tantivy::Result<()> {
                 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 => 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()?;
 
     let reader = index.reader()?;

examples/deleting_updating_documents.rs

@@ -56,8 +56,9 @@ fn main() -> tantivy::Result<()> {
     // If it is `text`, let's make sure to keep it `raw` and let's avoid
     // running any text processing on it.
     // This is done by associating this field to the tokenizer named `raw`.
-    // Rather than building our [`TextOptions`](//docs.rs/tantivy/~0/tantivy/schema/struct.TextOptions.html) manually,
-    // We use the `STRING` shortcut. `STRING` stands for indexed (without term frequency or positions)
+    // Rather than building our
+    // [`TextOptions`](//docs.rs/tantivy/~0/tantivy/schema/struct.TextOptions.html) manually, We
+    // use the `STRING` shortcut. `STRING` stands for indexed (without term frequency or positions)
     // and untokenized.
     //
     // Because we also want to be able to see this `id` in our returned documents,
@@ -76,21 +77,21 @@ fn main() -> tantivy::Result<()> {
     index_writer.add_document(doc!(
         isbn => "978-0099908401",
         title => "The old Man and the see"
-    ));
+    ))?;
     index_writer.add_document(doc!(
         isbn => "978-0140177398",
         title => "Of Mice and Men",
-    ));
+    ))?;
     index_writer.add_document(doc!(
        title => "Frankentein", //< Oops there is a typo here.
        isbn => "978-9176370711",
-    ));
+    ))?;
     index_writer.commit()?;
     let reader = index.reader()?;
 
     let frankenstein_isbn = Term::from_field_text(isbn, "978-9176370711");
 
-    // Oops our frankenstein doc seems mispelled
+    // Oops our frankenstein doc seems misspelled
     let frankenstein_doc_misspelled = extract_doc_given_isbn(&reader, &frankenstein_isbn)?.unwrap();
     assert_eq!(
         schema.to_json(&frankenstein_doc_misspelled),
@@ -122,7 +123,7 @@ fn main() -> tantivy::Result<()> {
     index_writer.add_document(doc!(
        title => "Frankenstein",
        isbn => "978-9176370711",
-    ));
+    ))?;
 
     // You are guaranteed that your clients will only observe your index in
     // the state it was in after a commit.

examples/faceted_search.rs

@@ -23,7 +23,7 @@ fn main() -> tantivy::Result<()> {
 
     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 classification = schema_builder.add_facet_field("classification", FacetOptions::default());
 
     let schema = schema_builder.build();
     let index = Index::create_in_ram(schema);
@@ -35,35 +35,35 @@ fn main() -> tantivy::Result<()> {
     index_writer.add_document(doc!(
         name => "Cat",
         classification => Facet::from("/Felidae/Felinae/Felis")
-    ));
+    ))?;
     index_writer.add_document(doc!(
         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()?;
 
     let reader = index.reader()?;
@@ -92,7 +92,7 @@ fn main() -> tantivy::Result<()> {
 
     // Check the reference doc for different ways to create a `Facet` object.
     {
-        let facet = Facet::from_text("/Felidae/Pantherinae");
+        let facet = Facet::from("/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);

examples/faceted_search_with_tweaked_score.rs

@@ -1,18 +1,18 @@
 use std::collections::HashSet;
+
 use tantivy::collector::TopDocs;
-use tantivy::doc;
 use tantivy::query::BooleanQuery;
 use tantivy::schema::*;
-use tantivy::{DocId, Index, Score, SegmentReader};
+use tantivy::{doc, 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 ingredient = schema_builder.add_facet_field("ingredient", FacetOptions::default());
 
     let schema = schema_builder.build();
-    let index = Index::create_in_ram(schema.clone());
+    let index = Index::create_in_ram(schema);
 
     let mut index_writer = index.writer(30_000_000)?;
 
@@ -20,14 +20,14 @@ fn main() -> tantivy::Result<()> {
         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"),
@@ -36,7 +36,7 @@ fn main() -> tantivy::Result<()> {
         ingredient => Facet::from("/ingredient/oil"),
         ingredient => Facet::from("/ingredient/tortilla-wrap"),
         ingredient => Facet::from("/ingredient/mushroom"),
-    ));
+    ))?;
     index_writer.commit()?;
 
     let reader = index.reader()?;
@@ -51,7 +51,7 @@ fn main() -> tantivy::Result<()> {
         let query = BooleanQuery::new_multiterms_query(
             facets
                 .iter()
-                .map(|key| Term::from_facet(ingredient, &key))
+                .map(|key| Term::from_facet(ingredient, key))
                 .collect(),
         );
         let top_docs_by_custom_score =
@@ -87,7 +87,7 @@ fn main() -> tantivy::Result<()> {
                     .unwrap()
                     .get_first(title)
                     .unwrap()
-                    .text()
+                    .as_text()
                     .unwrap()
                     .to_owned()
             })

examples/integer_range_search.rs

@@ -7,7 +7,7 @@ use tantivy::query::RangeQuery;
 use tantivy::schema::{Schema, INDEXED};
 use tantivy::{doc, Index, Result};
 
-fn run() -> Result<()> {
+fn main() -> Result<()> {
     // For the sake of simplicity, this schema will only have 1 field
     let mut schema_builder = Schema::builder();
 
@@ -19,7 +19,7 @@ fn run() -> Result<()> {
     {
         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.add_document(doc!(year_field => year))?;
         }
         index_writer.commit()?;
         // The index will be a range of years
@@ -33,7 +33,3 @@ fn run() -> Result<()> {
     assert_eq!(num_60s_books, 10);
     Ok(())
 }
-
-fn main() {
-    run().unwrap()
-}

examples/iterating_docs_and_positions.rs

@@ -1,4 +1,4 @@
-// # Iterating docs and positioms.
+// # Iterating docs and positions.
 //
 // At its core of tantivy, relies on a data structure
 // called an inverted index.
@@ -22,12 +22,12 @@ fn main() -> tantivy::Result<()> {
     let title = schema_builder.add_text_field("title", TEXT | STORED);
     let schema = schema_builder.build();
 
-    let index = Index::create_in_ram(schema.clone());
+    let index = Index::create_in_ram(schema);
 
     let mut index_writer = index.writer_with_num_threads(1, 50_000_000)?;
-    index_writer.add_document(doc!(title => "The Old Man and the Sea"));
-    index_writer.add_document(doc!(title => "Of Mice and Men"));
-    index_writer.add_document(doc!(title => "The modern Promotheus"));
+    index_writer.add_document(doc!(title => "The Old Man and the Sea"))?;
+    index_writer.add_document(doc!(title => "Of Mice and Men"))?;
+    index_writer.add_document(doc!(title => "The modern Promotheus"))?;
     index_writer.commit()?;
 
     let reader = index.reader()?;
@@ -52,11 +52,11 @@ fn main() -> tantivy::Result<()> {
         let term_the = Term::from_field_text(title, "the");
 
         // This segment posting object is like a cursor over the documents matching the term.
-        // The `IndexRecordOption` arguments tells tantivy we will be interested in both term frequencies
-        // and positions.
+        // The `IndexRecordOption` arguments tells tantivy we will be interested in both term
+        // frequencies and positions.
         //
-        // If you don't need all this information, you may get better performance by decompressing less
-        // information.
+        // If you don't need all this information, you may get better performance by decompressing
+        // less information.
         if let Some(mut segment_postings) =
             inverted_index.read_postings(&term_the, IndexRecordOption::WithFreqsAndPositions)?
         {
@@ -109,11 +109,11 @@ fn main() -> tantivy::Result<()> {
         let inverted_index = segment_reader.inverted_index(title)?;
 
         // This segment posting object is like a cursor over the documents matching the term.
-        // The `IndexRecordOption` arguments tells tantivy we will be interested in both term frequencies
-        // and positions.
+        // The `IndexRecordOption` arguments tells tantivy we will be interested in both term
+        // frequencies and positions.
         //
-        // If you don't need all this information, you may get better performance by decompressing less
-        // information.
+        // If you don't need all this information, you may get better performance by decompressing
+        // less information.
         if let Some(mut block_segment_postings) =
             inverted_index.read_block_postings(&term_the, IndexRecordOption::Basic)?
         {

examples/json_field.rs

@@ -0,0 +1,80 @@
+// # Json field example
+//
+// This example shows how the json field can be used
+// to make tantivy partially schemaless.
+
+use tantivy::collector::{Count, TopDocs};
+use tantivy::query::QueryParser;
+use tantivy::schema::{Schema, FAST, STORED, STRING, TEXT};
+use tantivy::Index;
+
+fn main() -> tantivy::Result<()> {
+    // # Defining the schema
+    //
+    // We need two fields:
+    // - a timestamp
+    // - a json object field
+    let mut schema_builder = Schema::builder();
+    schema_builder.add_date_field("timestamp", FAST | STORED);
+    let event_type = schema_builder.add_text_field("event_type", STRING | STORED);
+    let attributes = schema_builder.add_json_field("attributes", STORED | TEXT);
+    let schema = schema_builder.build();
+
+    // # Indexing documents
+    let index = Index::create_in_ram(schema.clone());
+
+    let mut index_writer = index.writer(50_000_000)?;
+    let doc = schema.parse_document(
+        r#"{
+        "timestamp": "2022-02-22T23:20:50.53Z",
+        "event_type": "click",
+        "attributes": {
+            "target": "submit-button",
+            "cart": {"product_id": 103},
+            "description": "the best vacuum cleaner ever"
+        }
+    }"#,
+    )?;
+    index_writer.add_document(doc)?;
+    let doc = schema.parse_document(
+        r#"{
+        "timestamp": "2022-02-22T23:20:51.53Z",
+        "event_type": "click",
+        "attributes": {
+            "target": "submit-button",
+            "cart": {"product_id": 133},
+            "description": "das keyboard"
+        }
+    }"#,
+    )?;
+    index_writer.add_document(doc)?;
+    index_writer.commit()?;
+
+    let reader = index.reader()?;
+    let searcher = reader.searcher();
+
+    let query_parser = QueryParser::for_index(&index, vec![event_type, attributes]);
+    {
+        let query = query_parser.parse_query("target:submit-button")?;
+        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        assert_eq!(count_docs.len(), 2);
+    }
+    {
+        let query = query_parser.parse_query("target:submit")?;
+        let count_docs = searcher.search(&*query, &TopDocs::with_limit(2))?;
+        assert_eq!(count_docs.len(), 2);
+    }
+    {
+        let query = query_parser.parse_query("cart.product_id:103")?;
+        let count_docs = searcher.search(&*query, &Count)?;
+        assert_eq!(count_docs, 1);
+    }
+    {
+        let query = query_parser
+            .parse_query("event_type:click AND cart.product_id:133")
+            .unwrap();
+        let hits = searcher.search(&*query, &TopDocs::with_limit(2)).unwrap();
+        assert_eq!(hits.len(), 1);
+    }
+    Ok(())
+}

examples/multiple_producer.rs

@@ -28,8 +28,9 @@
 use std::sync::{Arc, RwLock};
 use std::thread;
 use std::time::Duration;
+
 use tantivy::schema::{Schema, STORED, TEXT};
-use tantivy::{doc, Index, IndexWriter, Opstamp};
+use tantivy::{doc, Index, IndexWriter, Opstamp, TantivyError};
 
 fn main() -> tantivy::Result<()> {
     // # Defining the schema
@@ -59,10 +60,11 @@ fn main() -> tantivy::Result<()> {
                         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));
         }
+        Result::<(), TantivyError>::Ok(())
     });
 
     // # Second indexing thread.
@@ -78,19 +80,21 @@ fn main() -> tantivy::Result<()> {
                 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));
         }
+        Result::<(), TantivyError>::Ok(())
     });
 
     // # 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.
+            // 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()
+            index_writer_wlock.commit()?
         };
         println!("committed with opstamp {}", opstamp);
         thread::sleep(Duration::from_millis(500));

examples/pre_tokenized_text.rs

@@ -1,6 +1,6 @@
 // # Pre-tokenized text example
 //
-// This example shows how to use pre-tokenized text. Sometimes yout might
+// This example shows how to use pre-tokenized text. Sometimes you might
 // want to index and search through text which is already split into
 // tokens by some external tool.
 //
@@ -68,7 +68,7 @@ fn main() -> tantivy::Result<()> {
     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);
+    index_writer.add_document(old_man_doc)?;
 
     // Pretokenized text can also be fed as JSON
     let short_man_json = r#"{
@@ -82,9 +82,9 @@ fn main() -> tantivy::Result<()> {
         }]
     }"#;
 
-    let short_man_doc = schema.parse_document(&short_man_json)?;
+    let short_man_doc = schema.parse_document(short_man_json)?;
 
-    index_writer.add_document(short_man_doc);
+    index_writer.add_document(short_man_doc)?;
 
     // Let's commit changes
     index_writer.commit()?;
@@ -106,9 +106,7 @@ fn main() -> tantivy::Result<()> {
         IndexRecordOption::Basic,
     );
 
-    let (top_docs, count) = searcher
-        .search(&query, &(TopDocs::with_limit(2), Count))
-        .unwrap();
+    let (top_docs, count) = searcher.search(&query, &(TopDocs::with_limit(2), Count))?;
 
     assert_eq!(count, 2);
 
@@ -129,9 +127,7 @@ fn main() -> tantivy::Result<()> {
         IndexRecordOption::Basic,
     );
 
-    let (_top_docs, count) = searcher
-        .search(&query, &(TopDocs::with_limit(2), Count))
-        .unwrap();
+    let (_top_docs, count) = searcher.search(&query, &(TopDocs::with_limit(2), Count))?;
 
     assert_eq!(count, 0);
 

examples/snippet.rs

@@ -25,7 +25,7 @@ fn main() -> tantivy::Result<()> {
     let schema = schema_builder.build();
 
     // # Indexing documents
-    let index = Index::create_in_dir(&index_path, schema.clone())?;
+    let index = Index::create_in_dir(&index_path, schema)?;
 
     let mut index_writer = index.writer(50_000_000)?;
 
@@ -40,7 +40,7 @@ fn main() -> tantivy::Result<()> {
             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()?;
 
@@ -57,7 +57,10 @@ fn main() -> tantivy::Result<()> {
         let doc = searcher.doc(doc_address)?;
         let snippet = snippet_generator.snippet_from_doc(&doc);
         println!("Document score {}:", score);
-        println!("title: {}", doc.get_first(title).unwrap().text().unwrap());
+        println!(
+            "title: {}",
+            doc.get_first(title).unwrap().as_text().unwrap()
+        );
         println!("snippet: {}", snippet.to_html());
         println!("custom highlighting: {}", highlight(snippet));
     }
@@ -69,14 +72,14 @@ fn highlight(snippet: Snippet) -> String {
     let mut result = String::new();
     let mut start_from = 0;
 
-    for (start, end) in snippet.highlighted().iter().map(|h| h.bounds()) {
-        result.push_str(&snippet.fragments()[start_from..start]);
+    for fragment_range in snippet.highlighted() {
+        result.push_str(&snippet.fragment()[start_from..fragment_range.start]);
         result.push_str(" --> ");
-        result.push_str(&snippet.fragments()[start..end]);
+        result.push_str(&snippet.fragment()[fragment_range.clone()]);
         result.push_str(" <-- ");
-        start_from = end;
+        start_from = fragment_range.end;
     }
 
-    result.push_str(&snippet.fragments()[start_from..]);
+    result.push_str(&snippet.fragment()[start_from..]);
     result
 }

examples/stop_words.rs

@@ -68,7 +68,7 @@ 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",
@@ -80,7 +80,7 @@ fn main() -> tantivy::Result<()> {
             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",
@@ -88,7 +88,7 @@ fn main() -> tantivy::Result<()> {
              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()?;
 

examples/warmer.rs

@@ -0,0 +1,224 @@
+use std::cmp::Reverse;
+use std::collections::{HashMap, HashSet};
+use std::sync::{Arc, RwLock, Weak};
+
+use tantivy::collector::TopDocs;
+use tantivy::fastfield::FastFieldReader;
+use tantivy::query::QueryParser;
+use tantivy::schema::{Field, Schema, FAST, TEXT};
+use tantivy::{
+    doc, DocAddress, DocId, Index, IndexReader, Opstamp, Searcher, SearcherGeneration, SegmentId,
+    SegmentReader, Warmer,
+};
+
+// This example shows how warmers can be used to
+// load a values from an external sources using the Warmer API.
+//
+// In this example, we assume an e-commerce search engine.
+
+type ProductId = u64;
+
+/// Price
+type Price = u32;
+
+pub trait PriceFetcher: Send + Sync + 'static {
+    fn fetch_prices(&self, product_ids: &[ProductId]) -> Vec<Price>;
+}
+
+struct DynamicPriceColumn {
+    field: Field,
+    price_cache: RwLock<HashMap<(SegmentId, Option<Opstamp>), Arc<Vec<Price>>>>,
+    price_fetcher: Box<dyn PriceFetcher>,
+}
+
+impl DynamicPriceColumn {
+    pub fn with_product_id_field<T: PriceFetcher>(field: Field, price_fetcher: T) -> Self {
+        DynamicPriceColumn {
+            field,
+            price_cache: Default::default(),
+            price_fetcher: Box::new(price_fetcher),
+        }
+    }
+
+    pub fn price_for_segment(&self, segment_reader: &SegmentReader) -> Option<Arc<Vec<Price>>> {
+        let segment_key = (segment_reader.segment_id(), segment_reader.delete_opstamp());
+        self.price_cache.read().unwrap().get(&segment_key).cloned()
+    }
+}
+impl Warmer for DynamicPriceColumn {
+    fn warm(&self, searcher: &Searcher) -> tantivy::Result<()> {
+        for segment in searcher.segment_readers() {
+            let key = (segment.segment_id(), segment.delete_opstamp());
+            let product_id_reader = segment.fast_fields().u64(self.field)?;
+            let product_ids: Vec<ProductId> = segment
+                .doc_ids_alive()
+                .map(|doc| product_id_reader.get(doc))
+                .collect();
+            let mut prices_it = self.price_fetcher.fetch_prices(&product_ids).into_iter();
+            let mut price_vals: Vec<Price> = Vec::new();
+            for doc in 0..segment.max_doc() {
+                if segment.is_deleted(doc) {
+                    price_vals.push(0);
+                } else {
+                    price_vals.push(prices_it.next().unwrap())
+                }
+            }
+            self.price_cache
+                .write()
+                .unwrap()
+                .insert(key, Arc::new(price_vals));
+        }
+        Ok(())
+    }
+
+    fn garbage_collect(&self, live_generations: &[&SearcherGeneration]) {
+        let live_segment_id_and_delete_ops: HashSet<(SegmentId, Option<Opstamp>)> =
+            live_generations
+                .iter()
+                .flat_map(|gen| gen.segments())
+                .map(|(&segment_id, &opstamp)| (segment_id, opstamp))
+                .collect();
+        let mut price_cache_wrt = self.price_cache.write().unwrap();
+        // let price_cache = std::mem::take(&mut *price_cache_wrt);
+        // Drain would be nicer here.
+        *price_cache_wrt = std::mem::take(&mut *price_cache_wrt)
+            .into_iter()
+            .filter(|(seg_id_and_op, _)| !live_segment_id_and_delete_ops.contains(seg_id_and_op))
+            .collect();
+    }
+}
+
+/// For the sake of this example, the table is just an editable HashMap behind a RwLock.
+/// This map represents a map (ProductId -> Price)
+///
+/// In practise, it could be fetching things from an external service, like a SQL table.
+#[derive(Default, Clone)]
+pub struct ExternalPriceTable {
+    prices: Arc<RwLock<HashMap<ProductId, Price>>>,
+}
+
+impl ExternalPriceTable {
+    pub fn update_price(&self, product_id: ProductId, price: Price) {
+        let mut prices_wrt = self.prices.write().unwrap();
+        prices_wrt.insert(product_id, price);
+    }
+}
+
+impl PriceFetcher for ExternalPriceTable {
+    fn fetch_prices(&self, product_ids: &[ProductId]) -> Vec<Price> {
+        let prices_read = self.prices.read().unwrap();
+        product_ids
+            .iter()
+            .map(|product_id| prices_read.get(product_id).cloned().unwrap_or(0))
+            .collect()
+    }
+}
+
+fn main() -> tantivy::Result<()> {
+    // Declaring our schema.
+    let mut schema_builder = Schema::builder();
+    // The product id is assumed to be a primary id for our external price source.
+    let product_id = schema_builder.add_u64_field("product_id", FAST);
+    let text = schema_builder.add_text_field("text", TEXT);
+    let schema: Schema = schema_builder.build();
+
+    let price_table = ExternalPriceTable::default();
+    let price_dynamic_column = Arc::new(DynamicPriceColumn::with_product_id_field(
+        product_id,
+        price_table.clone(),
+    ));
+    price_table.update_price(OLIVE_OIL, 12);
+    price_table.update_price(GLOVES, 13);
+    price_table.update_price(SNEAKERS, 80);
+
+    const OLIVE_OIL: ProductId = 323423;
+    const GLOVES: ProductId = 3966623;
+    const SNEAKERS: ProductId = 23222;
+
+    let index = Index::create_in_ram(schema);
+    let mut writer = index.writer_with_num_threads(1, 10_000_000)?;
+    writer.add_document(doc!(product_id=>OLIVE_OIL, text=>"cooking olive oil from greece"))?;
+    writer.add_document(doc!(product_id=>GLOVES, text=>"kitchen gloves, perfect for cooking"))?;
+    writer.add_document(doc!(product_id=>SNEAKERS, text=>"uber sweet sneakers"))?;
+    writer.commit()?;
+
+    let warmers: Vec<Weak<dyn Warmer>> = vec![Arc::downgrade(
+        &(price_dynamic_column.clone() as Arc<dyn Warmer>),
+    )];
+    let reader: IndexReader = index
+        .reader_builder()
+        .warmers(warmers)
+        .num_searchers(1)
+        .try_into()?;
+    reader.reload()?;
+
+    let query_parser = QueryParser::for_index(&index, vec![text]);
+    let query = query_parser.parse_query("cooking")?;
+
+    let searcher = reader.searcher();
+    let score_by_price = move |segment_reader: &SegmentReader| {
+        let price = price_dynamic_column
+            .price_for_segment(segment_reader)
+            .unwrap();
+        move |doc_id: DocId| Reverse(price[doc_id as usize])
+    };
+
+    let most_expensive_first = TopDocs::with_limit(10).custom_score(score_by_price);
+
+    let hits = searcher.search(&query, &most_expensive_first)?;
+    assert_eq!(
+        &hits,
+        &[
+            (
+                Reverse(12u32),
+                DocAddress {
+                    segment_ord: 0,
+                    doc_id: 0u32
+                }
+            ),
+            (
+                Reverse(13u32),
+                DocAddress {
+                    segment_ord: 0,
+                    doc_id: 1u32
+                }
+            ),
+        ]
+    );
+
+    // Olive oil just got more expensive!
+    price_table.update_price(OLIVE_OIL, 15);
+
+    // The price update are directly reflected on `reload`.
+    //
+    // Be careful here though!...
+    // You may have spotted that we are still using the same `Searcher`.
+    //
+    // It is up to the `Warmer` implementer to decide how
+    // to control this behavior.
+
+    reader.reload()?;
+
+    let hits_with_new_prices = searcher.search(&query, &most_expensive_first)?;
+    assert_eq!(
+        &hits_with_new_prices,
+        &[
+            (
+                Reverse(13u32),
+                DocAddress {
+                    segment_ord: 0,
+                    doc_id: 1u32
+                }
+            ),
+            (
+                Reverse(15u32),
+                DocAddress {
+                    segment_ord: 0,
+                    doc_id: 0u32
+                }
+            ),
+        ]
+    );
+
+    Ok(())
+}

examples/working_with_json.rs

@@ -1,4 +1,3 @@
-use tantivy;
 use tantivy::schema::*;
 
 // # Document from json
@@ -22,7 +21,7 @@ fn main() -> tantivy::Result<()> {
     }"#;
 
     // We can parse our document
-    let _mice_and_men_doc = schema.parse_document(&mice_and_men_doc_json)?;
+    let _mice_and_men_doc = schema.parse_document(mice_and_men_doc_json)?;
 
     // Multi-valued field are allowed, they are
     // expressed in JSON by an array.
@@ -31,7 +30,7 @@ fn main() -> tantivy::Result<()> {
        "title": ["Frankenstein", "The Modern Prometheus"],
        "year": 1818
     }"#;
-    let _frankenstein_doc = schema.parse_document(&frankenstein_json)?;
+    let _frankenstein_doc = schema.parse_document(frankenstein_json)?;
 
     // Note that the schema is saved in your index directory.
     //

fastfield_codecs/Cargo.toml

@@ -0,0 +1,24 @@
+[package]
+name = "fastfield_codecs"
+version = "0.1.0"
+authors = ["Pascal Seitz <pascal@quickwit.io>"]
+license = "MIT"
+edition = "2018"
+description = "Fast field codecs used by tantivy"
+
+# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
+
+[dependencies]
+common = { version = "0.2", path = "../common/", package = "tantivy-common" }
+tantivy-bitpacker = { version="0.1.1", path = "../bitpacker/" }
+prettytable-rs = {version="0.8.0", optional= true}
+rand = {version="0.8.3", optional= true}
+
+[dev-dependencies]
+more-asserts = "0.2.1"
+rand = "0.8.3"
+
+[features]
+bin = ["prettytable-rs", "rand"]
+default = ["bin"]
+

fastfield_codecs/README.md

@@ -0,0 +1,68 @@
+
+
+# Fast Field Codecs
+
+This crate contains various fast field codecs, used to compress/decompress fast field data in tantivy.
+
+## Contributing
+
+Contributing is pretty straightforward. Since the bitpacking is the simplest compressor, you can check it for reference.
+
+A codec needs to implement 2 traits:
+
+- A reader implementing `FastFieldCodecReader` to read the codec.
+- A serializer implementing `FastFieldCodecSerializer` for compression estimation and codec name + id.
+
+### Tests
+
+Once the traits are implemented test and benchmark integration is pretty easy (see `test_with_codec_data_sets` and `bench.rs`).
+
+Make sure to add the codec to the main.rs, which tests the compression ratio and estimation against different data sets. You can run it with:
+```
+cargo run --features bin
+```
+
+### TODO
+- Add real world data sets in comparison
+- Add codec to cover sparse data sets
+
+
+### Codec Comparison
+```
++----------------------------------+-------------------+------------------------+
+|                                  | Compression Ratio | Compression Estimation |
++----------------------------------+-------------------+------------------------+
+| Autoincrement                    |                   |                        |
++----------------------------------+-------------------+------------------------+
+| LinearInterpol                   | 0.000039572664    | 0.000004396963         |
++----------------------------------+-------------------+------------------------+
+| MultiLinearInterpol              | 0.1477348         | 0.17275847             |
++----------------------------------+-------------------+------------------------+
+| Bitpacked                        | 0.28126493        | 0.28125                |
++----------------------------------+-------------------+------------------------+
+| Monotonically increasing concave |                   |                        |
++----------------------------------+-------------------+------------------------+
+| LinearInterpol                   | 0.25003937        | 0.26562938             |
++----------------------------------+-------------------+------------------------+
+| MultiLinearInterpol              | 0.190665          | 0.1883836              |
++----------------------------------+-------------------+------------------------+
+| Bitpacked                        | 0.31251436        | 0.3125                 |
++----------------------------------+-------------------+------------------------+
+| Monotonically increasing convex  |                   |                        |
++----------------------------------+-------------------+------------------------+
+| LinearInterpol                   | 0.25003937        | 0.28125438             |
++----------------------------------+-------------------+------------------------+
+| MultiLinearInterpol              | 0.18676           | 0.2040086              |
++----------------------------------+-------------------+------------------------+
+| Bitpacked                        | 0.31251436        | 0.3125                 |
++----------------------------------+-------------------+------------------------+
+| Almost monotonically increasing  |                   |                        |
++----------------------------------+-------------------+------------------------+
+| LinearInterpol                   | 0.14066513        | 0.1562544              |
++----------------------------------+-------------------+------------------------+
+| MultiLinearInterpol              | 0.16335973        | 0.17275847             |
++----------------------------------+-------------------+------------------------+
+| Bitpacked                        | 0.28126493        | 0.28125                |
++----------------------------------+-------------------+------------------------+
+
+```

fastfield_codecs/benches/bench.rs

@@ -0,0 +1,108 @@
+#![feature(test)]
+
+extern crate test;
+
+#[cfg(test)]
+mod tests {
+    use fastfield_codecs::bitpacked::{BitpackedFastFieldReader, BitpackedFastFieldSerializer};
+    use fastfield_codecs::linearinterpol::{
+        LinearInterpolFastFieldReader, LinearInterpolFastFieldSerializer,
+    };
+    use fastfield_codecs::multilinearinterpol::{
+        MultiLinearInterpolFastFieldReader, MultiLinearInterpolFastFieldSerializer,
+    };
+    use fastfield_codecs::*;
+
+    fn get_data() -> Vec<u64> {
+        let mut data: Vec<_> = (100..55000_u64)
+            .map(|num| num + rand::random::<u8>() as u64)
+            .collect();
+        data.push(99_000);
+        data.insert(1000, 2000);
+        data.insert(2000, 100);
+        data.insert(3000, 4100);
+        data.insert(4000, 100);
+        data.insert(5000, 800);
+        data
+    }
+
+    fn value_iter() -> impl Iterator<Item = u64> {
+        0..20_000
+    }
+    fn bench_get<S: FastFieldCodecSerializer, R: FastFieldCodecReader>(
+        b: &mut Bencher,
+        data: &[u64],
+    ) {
+        let mut bytes = vec![];
+        S::serialize(
+            &mut bytes,
+            &data,
+            stats_from_vec(data),
+            data.iter().cloned(),
+            data.iter().cloned(),
+        )
+        .unwrap();
+        let reader = R::open_from_bytes(&bytes).unwrap();
+        b.iter(|| {
+            for pos in value_iter() {
+                reader.get_u64(pos as u64, &bytes);
+            }
+        });
+    }
+    fn bench_create<S: FastFieldCodecSerializer>(b: &mut Bencher, data: &[u64]) {
+        let mut bytes = vec![];
+        b.iter(|| {
+            S::serialize(
+                &mut bytes,
+                &data,
+                stats_from_vec(data),
+                data.iter().cloned(),
+                data.iter().cloned(),
+            )
+            .unwrap();
+        });
+    }
+
+    use test::Bencher;
+    #[bench]
+    fn bench_fastfield_bitpack_create(b: &mut Bencher) {
+        let data: Vec<_> = get_data();
+        bench_create::<BitpackedFastFieldSerializer>(b, &data);
+    }
+    #[bench]
+    fn bench_fastfield_linearinterpol_create(b: &mut Bencher) {
+        let data: Vec<_> = get_data();
+        bench_create::<LinearInterpolFastFieldSerializer>(b, &data);
+    }
+    #[bench]
+    fn bench_fastfield_multilinearinterpol_create(b: &mut Bencher) {
+        let data: Vec<_> = get_data();
+        bench_create::<MultiLinearInterpolFastFieldSerializer>(b, &data);
+    }
+    #[bench]
+    fn bench_fastfield_bitpack_get(b: &mut Bencher) {
+        let data: Vec<_> = get_data();
+        bench_get::<BitpackedFastFieldSerializer, BitpackedFastFieldReader>(b, &data);
+    }
+    #[bench]
+    fn bench_fastfield_linearinterpol_get(b: &mut Bencher) {
+        let data: Vec<_> = get_data();
+        bench_get::<LinearInterpolFastFieldSerializer, LinearInterpolFastFieldReader>(b, &data);
+    }
+    #[bench]
+    fn bench_fastfield_multilinearinterpol_get(b: &mut Bencher) {
+        let data: Vec<_> = get_data();
+        bench_get::<MultiLinearInterpolFastFieldSerializer, MultiLinearInterpolFastFieldReader>(
+            b, &data,
+        );
+    }
+    pub fn stats_from_vec(data: &[u64]) -> FastFieldStats {
+        let min_value = data.iter().cloned().min().unwrap_or(0);
+        let max_value = data.iter().cloned().max().unwrap_or(0);
+        FastFieldStats {
+            min_value,
+            max_value,
+            num_vals: data.len() as u64,
+        }
+    }
+}

fastfield_codecs/src/bitpacked.rs

@@ -0,0 +1,172 @@
+use std::io::{self, Write};
+
+use common::BinarySerializable;
+use tantivy_bitpacker::{compute_num_bits, BitPacker, BitUnpacker};
+
+use crate::{FastFieldCodecReader, FastFieldCodecSerializer, FastFieldDataAccess, FastFieldStats};
+
+/// Depending on the field type, a different
+/// fast field is required.
+#[derive(Clone)]
+pub struct BitpackedFastFieldReader {
+    bit_unpacker: BitUnpacker,
+    pub min_value_u64: u64,
+    pub max_value_u64: u64,
+}
+
+impl<'data> FastFieldCodecReader for BitpackedFastFieldReader {
+    /// Opens a fast field given a file.
+    fn open_from_bytes(bytes: &[u8]) -> io::Result<Self> {
+        let (_data, mut footer) = bytes.split_at(bytes.len() - 16);
+        let min_value = u64::deserialize(&mut footer)?;
+        let amplitude = u64::deserialize(&mut footer)?;
+        let max_value = min_value + amplitude;
+        let num_bits = compute_num_bits(amplitude);
+        let bit_unpacker = BitUnpacker::new(num_bits);
+        Ok(BitpackedFastFieldReader {
+            min_value_u64: min_value,
+            max_value_u64: max_value,
+            bit_unpacker,
+        })
+    }
+    #[inline]
+    fn get_u64(&self, doc: u64, data: &[u8]) -> u64 {
+        self.min_value_u64 + self.bit_unpacker.get(doc, data)
+    }
+    #[inline]
+    fn min_value(&self) -> u64 {
+        self.min_value_u64
+    }
+    #[inline]
+    fn max_value(&self) -> u64 {
+        self.max_value_u64
+    }
+}
+pub struct BitpackedFastFieldSerializerLegacy<'a, W: 'a + Write> {
+    bit_packer: BitPacker,
+    write: &'a mut W,
+    min_value: u64,
+    amplitude: u64,
+    num_bits: u8,
+}
+
+impl<'a, W: Write> BitpackedFastFieldSerializerLegacy<'a, W> {
+    /// Creates a new fast field serializer.
+    ///
+    /// The serializer in fact encode the values by bitpacking
+    /// `(val - min_value)`.
+    ///
+    /// It requires a `min_value` and a `max_value` to compute
+    /// compute the minimum number of bits required to encode
+    /// values.
+    pub fn open(
+        write: &'a mut W,
+        min_value: u64,
+        max_value: u64,
+    ) -> io::Result<BitpackedFastFieldSerializerLegacy<'a, W>> {
+        assert!(min_value <= max_value);
+        let amplitude = max_value - min_value;
+        let num_bits = compute_num_bits(amplitude);
+        let bit_packer = BitPacker::new();
+        Ok(BitpackedFastFieldSerializerLegacy {
+            bit_packer,
+            write,
+            min_value,
+            amplitude,
+            num_bits,
+        })
+    }
+    /// Pushes a new value to the currently open u64 fast field.
+    #[inline]
+    pub fn add_val(&mut self, val: u64) -> io::Result<()> {
+        let val_to_write: u64 = val - self.min_value;
+        self.bit_packer
+            .write(val_to_write, self.num_bits, &mut self.write)?;
+        Ok(())
+    }
+    pub fn close_field(mut self) -> io::Result<()> {
+        self.bit_packer.close(&mut self.write)?;
+        self.min_value.serialize(&mut self.write)?;
+        self.amplitude.serialize(&mut self.write)?;
+        Ok(())
+    }
+}
+
+pub struct BitpackedFastFieldSerializer {}
+
+impl FastFieldCodecSerializer for BitpackedFastFieldSerializer {
+    const NAME: &'static str = "Bitpacked";
+    const ID: u8 = 1;
+    /// Serializes data with the BitpackedFastFieldSerializer.
+    ///
+    /// The serializer in fact encode the values by bitpacking
+    /// `(val - min_value)`.
+    ///
+    /// It requires a `min_value` and a `max_value` to compute
+    /// compute the minimum number of bits required to encode
+    /// values.
+    fn serialize(
+        write: &mut impl Write,
+        _fastfield_accessor: &impl FastFieldDataAccess,
+        stats: FastFieldStats,
+        data_iter: impl Iterator<Item = u64>,
+        _data_iter1: impl Iterator<Item = u64>,
+    ) -> io::Result<()> {
+        let mut serializer =
+            BitpackedFastFieldSerializerLegacy::open(write, stats.min_value, stats.max_value)?;
+
+        for val in data_iter {
+            serializer.add_val(val)?;
+        }
+        serializer.close_field()?;
+
+        Ok(())
+    }
+    fn is_applicable(
+        _fastfield_accessor: &impl FastFieldDataAccess,
+        _stats: FastFieldStats,
+    ) -> bool {
+        true
+    }
+    fn estimate(_fastfield_accessor: &impl FastFieldDataAccess, stats: FastFieldStats) -> f32 {
+        let amplitude = stats.max_value - stats.min_value;
+        let num_bits = compute_num_bits(amplitude);
+        let num_bits_uncompressed = 64;
+        num_bits as f32 / num_bits_uncompressed as f32
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::tests::get_codec_test_data_sets;
+
+    fn create_and_validate(data: &[u64], name: &str) {
+        crate::tests::create_and_validate::<BitpackedFastFieldSerializer, BitpackedFastFieldReader>(
+            data, name,
+        );
+    }
+
+    #[test]
+    fn test_with_codec_data_sets() {
+        let data_sets = get_codec_test_data_sets();
+        for (mut data, name) in data_sets {
+            create_and_validate(&data, name);
+            data.reverse();
+            create_and_validate(&data, name);
+        }
+    }
+
+    #[test]
+    fn bitpacked_fast_field_rand() {
+        for _ in 0..500 {
+            let mut data = (0..1 + rand::random::<u8>() as usize)
+                .map(|_| rand::random::<i64>() as u64 / 2)
+                .collect::<Vec<_>>();
+            create_and_validate(&data, "rand");
+
+            data.reverse();
+            create_and_validate(&data, "rand");
+        }
+    }
+}

fastfield_codecs/src/lib.rs

@@ -0,0 +1,225 @@
+#[cfg(test)]
+#[macro_use]
+extern crate more_asserts;
+
+use std::io;
+use std::io::Write;
+
+pub mod bitpacked;
+pub mod linearinterpol;
+pub mod multilinearinterpol;
+
+pub trait FastFieldCodecReader: Sized {
+    /// reads the metadata and returns the CodecReader
+    fn open_from_bytes(bytes: &[u8]) -> std::io::Result<Self>;
+
+    fn get_u64(&self, doc: u64, data: &[u8]) -> u64;
+
+    fn min_value(&self) -> u64;
+    fn max_value(&self) -> u64;
+}
+
+/// The FastFieldSerializerEstimate trait is required on all variants
+/// of fast field compressions, to decide which one to choose.
+pub trait FastFieldCodecSerializer {
+    /// A codex needs to provide a unique name and id, which is
+    /// used for debugging and de/serialization.
+    const NAME: &'static str;
+    const ID: u8;
+
+    /// Check if the Codec is able to compress the data
+    fn is_applicable(fastfield_accessor: &impl FastFieldDataAccess, stats: FastFieldStats) -> bool;
+
+    /// Returns an estimate of the compression ratio.
+    /// The baseline is uncompressed 64bit data.
+    ///
+    /// It could make sense to also return a value representing
+    /// computational complexity.
+    fn estimate(fastfield_accessor: &impl FastFieldDataAccess, stats: FastFieldStats) -> f32;
+
+    /// Serializes the data using the serializer into write.
+    /// There are multiple iterators, in case the codec needs to read the data multiple times.
+    /// The iterators should be preferred over using fastfield_accessor for performance reasons.
+    fn serialize(
+        write: &mut impl Write,
+        fastfield_accessor: &impl FastFieldDataAccess,
+        stats: FastFieldStats,
+        data_iter: impl Iterator<Item = u64>,
+        data_iter1: impl Iterator<Item = u64>,
+    ) -> io::Result<()>;
+}
+
+/// FastFieldDataAccess is the trait to access fast field data during serialization and estimation.
+pub trait FastFieldDataAccess {
+    /// Return the value associated to the given position.
+    ///
+    /// Whenever possible use the Iterator passed to the fastfield creation instead, for performance
+    /// reasons.
+    ///
+    /// # Panics
+    ///
+    /// May panic if `position` is greater than the index.
+    fn get_val(&self, position: u64) -> u64;
+}
+
+#[derive(Debug, Clone)]
+/// Statistics are used in codec detection and stored in the fast field footer.
+pub struct FastFieldStats {
+    pub min_value: u64,
+    pub max_value: u64,
+    pub num_vals: u64,
+}
+
+impl<'a> FastFieldDataAccess for &'a [u64] {
+    fn get_val(&self, position: u64) -> u64 {
+        self[position as usize]
+    }
+}
+
+impl FastFieldDataAccess for Vec<u64> {
+    fn get_val(&self, position: u64) -> u64 {
+        self[position as usize]
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use crate::bitpacked::{BitpackedFastFieldReader, BitpackedFastFieldSerializer};
+    use crate::linearinterpol::{LinearInterpolFastFieldReader, LinearInterpolFastFieldSerializer};
+    use crate::multilinearinterpol::{
+        MultiLinearInterpolFastFieldReader, MultiLinearInterpolFastFieldSerializer,
+    };
+
+    pub fn create_and_validate<S: FastFieldCodecSerializer, R: FastFieldCodecReader>(
+        data: &[u64],
+        name: &str,
+    ) -> (f32, f32) {
+        if !S::is_applicable(&data, crate::tests::stats_from_vec(data)) {
+            return (f32::MAX, 0.0);
+        }
+        let estimation = S::estimate(&data, crate::tests::stats_from_vec(data));
+        let mut out = vec![];
+        S::serialize(
+            &mut out,
+            &data,
+            crate::tests::stats_from_vec(data),
+            data.iter().cloned(),
+            data.iter().cloned(),
+        )
+        .unwrap();
+
+        let reader = R::open_from_bytes(&out).unwrap();
+        for (doc, orig_val) in data.iter().enumerate() {
+            let val = reader.get_u64(doc as u64, &out);
+            if val != *orig_val {
+                panic!(
+                    "val {:?} does not match orig_val {:?}, in data set {}, data {:?}",
+                    val, orig_val, name, data
+                );
+            }
+        }
+        let actual_compression = out.len() as f32 / (data.len() as f32 * 8.0);
+        (estimation, actual_compression)
+    }
+    pub fn get_codec_test_data_sets() -> Vec<(Vec<u64>, &'static str)> {
+        let mut data_and_names = vec![];
+
+        let data = (10..=20_u64).collect::<Vec<_>>();
+        data_and_names.push((data, "simple monotonically increasing"));
+
+        data_and_names.push((
+            vec![5, 6, 7, 8, 9, 10, 99, 100],
+            "offset in linear interpol",
+        ));
+        data_and_names.push((vec![5, 50, 3, 13, 1, 1000, 35], "rand small"));
+        data_and_names.push((vec![10], "single value"));
+
+        data_and_names
+    }
+
+    fn test_codec<S: FastFieldCodecSerializer, R: FastFieldCodecReader>() {
+        let codec_name = S::NAME;
+        for (data, data_set_name) in get_codec_test_data_sets() {
+            let (estimate, actual) =
+                crate::tests::create_and_validate::<S, R>(&data, data_set_name);
+            let result = if estimate == f32::MAX {
+                "Disabled".to_string()
+            } else {
+                format!("Estimate {:?} Actual {:?} ", estimate, actual)
+            };
+            println!(
+                "Codec {}, DataSet {}, {}",
+                codec_name, data_set_name, result
+            );
+        }
+    }
+    #[test]
+    fn test_codec_bitpacking() {
+        test_codec::<BitpackedFastFieldSerializer, BitpackedFastFieldReader>();
+    }
+    #[test]
+    fn test_codec_interpolation() {
+        test_codec::<LinearInterpolFastFieldSerializer, LinearInterpolFastFieldReader>();
+    }
+    #[test]
+    fn test_codec_multi_interpolation() {
+        test_codec::<MultiLinearInterpolFastFieldSerializer, MultiLinearInterpolFastFieldReader>();
+    }
+
+    use super::*;
+    pub fn stats_from_vec(data: &[u64]) -> FastFieldStats {
+        let min_value = data.iter().cloned().min().unwrap_or(0);
+        let max_value = data.iter().cloned().max().unwrap_or(0);
+        FastFieldStats {
+            min_value,
+            max_value,
+            num_vals: data.len() as u64,
+        }
+    }
+
+    #[test]
+    fn estimation_good_interpolation_case() {
+        let data = (10..=20000_u64).collect::<Vec<_>>();
+
+        let linear_interpol_estimation =
+            LinearInterpolFastFieldSerializer::estimate(&data, stats_from_vec(&data));
+        assert_le!(linear_interpol_estimation, 0.01);
+
+        let multi_linear_interpol_estimation =
+            MultiLinearInterpolFastFieldSerializer::estimate(&data, stats_from_vec(&data));
+        assert_le!(multi_linear_interpol_estimation, 0.2);
+        assert_le!(linear_interpol_estimation, multi_linear_interpol_estimation);
+
+        let bitpacked_estimation =
+            BitpackedFastFieldSerializer::estimate(&data, stats_from_vec(&data));
+        assert_le!(linear_interpol_estimation, bitpacked_estimation);
+    }
+    #[test]
+    fn estimation_test_bad_interpolation_case() {
+        let data = vec![200, 10, 10, 10, 10, 1000, 20];
+
+        let linear_interpol_estimation =
+            LinearInterpolFastFieldSerializer::estimate(&data, stats_from_vec(&data));
+        assert_le!(linear_interpol_estimation, 0.32);
+
+        let bitpacked_estimation =
+            BitpackedFastFieldSerializer::estimate(&data, stats_from_vec(&data));
+        assert_le!(bitpacked_estimation, linear_interpol_estimation);
+    }
+    #[test]
+    fn estimation_test_bad_interpolation_case_monotonically_increasing() {
+        let mut data = (200..=20000_u64).collect::<Vec<_>>();
+        data.push(1_000_000);
+
+        // in this case the linear interpolation can't in fact not be worse than bitpacking,
+        // but the estimator adds some threshold, which leads to estimated worse behavior
+        let linear_interpol_estimation =
+            LinearInterpolFastFieldSerializer::estimate(&data, stats_from_vec(&data));
+        assert_le!(linear_interpol_estimation, 0.35);
+
+        let bitpacked_estimation =
+            BitpackedFastFieldSerializer::estimate(&data, stats_from_vec(&data));
+        assert_le!(bitpacked_estimation, 0.32);
+        assert_le!(bitpacked_estimation, linear_interpol_estimation);
+    }
+}

fastfield_codecs/src/linearinterpol.rs

@@ -0,0 +1,300 @@
+use std::io::{self, Read, Write};
+use std::ops::Sub;
+
+use common::{BinarySerializable, FixedSize};
+use tantivy_bitpacker::{compute_num_bits, BitPacker, BitUnpacker};
+
+use crate::{FastFieldCodecReader, FastFieldCodecSerializer, FastFieldDataAccess, FastFieldStats};
+
+/// Depending on the field type, a different
+/// fast field is required.
+#[derive(Clone)]
+pub struct LinearInterpolFastFieldReader {
+    bit_unpacker: BitUnpacker,
+    pub footer: LinearInterpolFooter,
+    pub slope: f32,
+}
+
+#[derive(Clone, Debug)]
+pub struct LinearInterpolFooter {
+    pub relative_max_value: u64,
+    pub offset: u64,
+    pub first_val: u64,
+    pub last_val: u64,
+    pub num_vals: u64,
+    pub min_value: u64,
+    pub max_value: u64,
+}
+
+impl BinarySerializable for LinearInterpolFooter {
+    fn serialize<W: Write>(&self, write: &mut W) -> io::Result<()> {
+        self.relative_max_value.serialize(write)?;
+        self.offset.serialize(write)?;
+        self.first_val.serialize(write)?;
+        self.last_val.serialize(write)?;
+        self.num_vals.serialize(write)?;
+        self.min_value.serialize(write)?;
+        self.max_value.serialize(write)?;
+        Ok(())
+    }
+
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<LinearInterpolFooter> {
+        Ok(LinearInterpolFooter {
+            relative_max_value: u64::deserialize(reader)?,
+            offset: u64::deserialize(reader)?,
+            first_val: u64::deserialize(reader)?,
+            last_val: u64::deserialize(reader)?,
+            num_vals: u64::deserialize(reader)?,
+            min_value: u64::deserialize(reader)?,
+            max_value: u64::deserialize(reader)?,
+        })
+    }
+}
+
+impl FixedSize for LinearInterpolFooter {
+    const SIZE_IN_BYTES: usize = 56;
+}
+
+impl FastFieldCodecReader for LinearInterpolFastFieldReader {
+    /// Opens a fast field given a file.
+    fn open_from_bytes(bytes: &[u8]) -> io::Result<Self> {
+        let (_data, mut footer) = bytes.split_at(bytes.len() - LinearInterpolFooter::SIZE_IN_BYTES);
+        let footer = LinearInterpolFooter::deserialize(&mut footer)?;
+        let slope = get_slope(footer.first_val, footer.last_val, footer.num_vals);
+
+        let num_bits = compute_num_bits(footer.relative_max_value);
+        let bit_unpacker = BitUnpacker::new(num_bits);
+        Ok(LinearInterpolFastFieldReader {
+            bit_unpacker,
+            footer,
+            slope,
+        })
+    }
+    #[inline]
+    fn get_u64(&self, doc: u64, data: &[u8]) -> u64 {
+        let calculated_value = get_calculated_value(self.footer.first_val, doc, self.slope);
+        (calculated_value + self.bit_unpacker.get(doc, data)) - self.footer.offset
+    }
+
+    #[inline]
+    fn min_value(&self) -> u64 {
+        self.footer.min_value
+    }
+    #[inline]
+    fn max_value(&self) -> u64 {
+        self.footer.max_value
+    }
+}
+
+/// Fastfield serializer, which tries to guess values by linear interpolation
+/// and stores the difference bitpacked.
+pub struct LinearInterpolFastFieldSerializer {}
+
+#[inline]
+fn get_slope(first_val: u64, last_val: u64, num_vals: u64) -> f32 {
+    if num_vals <= 1 {
+        return 0.0;
+    }
+    //  We calculate the slope with f64 high precision and use the result in lower precision f32
+    //  This is done in order to handle estimations for very large values like i64::MAX
+    ((last_val as f64 - first_val as f64) / (num_vals as u64 - 1) as f64) as f32
+}
+
+#[inline]
+fn get_calculated_value(first_val: u64, pos: u64, slope: f32) -> u64 {
+    first_val + (pos as f32 * slope) as u64
+}
+
+impl FastFieldCodecSerializer for LinearInterpolFastFieldSerializer {
+    const NAME: &'static str = "LinearInterpol";
+    const ID: u8 = 2;
+    /// Creates a new fast field serializer.
+    fn serialize(
+        write: &mut impl Write,
+        fastfield_accessor: &impl FastFieldDataAccess,
+        stats: FastFieldStats,
+        data_iter: impl Iterator<Item = u64>,
+        data_iter1: impl Iterator<Item = u64>,
+    ) -> io::Result<()> {
+        assert!(stats.min_value <= stats.max_value);
+
+        let first_val = fastfield_accessor.get_val(0);
+        let last_val = fastfield_accessor.get_val(stats.num_vals as u64 - 1);
+        let slope = get_slope(first_val, last_val, stats.num_vals);
+        // calculate offset to ensure all values are positive
+        let mut offset = 0;
+        let mut rel_positive_max = 0;
+        for (pos, actual_value) in data_iter1.enumerate() {
+            let calculated_value = get_calculated_value(first_val, pos as u64, slope);
+            if calculated_value > actual_value {
+                // negative value we need to apply an offset
+                // we ignore negative values in the max value calculation, because negative values
+                // will be offset to 0
+                offset = offset.max(calculated_value - actual_value);
+            } else {
+                // positive value no offset reuqired
+                rel_positive_max = rel_positive_max.max(actual_value - calculated_value);
+            }
+        }
+
+        // rel_positive_max will be adjusted by offset
+        let relative_max_value = rel_positive_max + offset;
+
+        let num_bits = compute_num_bits(relative_max_value);
+        let mut bit_packer = BitPacker::new();
+        for (pos, val) in data_iter.enumerate() {
+            let calculated_value = get_calculated_value(first_val, pos as u64, slope);
+            let diff = (val + offset) - calculated_value;
+            bit_packer.write(diff, num_bits, write)?;
+        }
+        bit_packer.close(write)?;
+
+        let footer = LinearInterpolFooter {
+            relative_max_value,
+            offset,
+            first_val,
+            last_val,
+            num_vals: stats.num_vals,
+            min_value: stats.min_value,
+            max_value: stats.max_value,
+        };
+        footer.serialize(write)?;
+        Ok(())
+    }
+    fn is_applicable(
+        _fastfield_accessor: &impl FastFieldDataAccess,
+        stats: FastFieldStats,
+    ) -> bool {
+        if stats.num_vals < 3 {
+            return false; // disable compressor for this case
+        }
+        // On serialisation the offset is added to the actual value.
+        // We need to make sure this won't run into overflow calculation issues.
+        // For this we take the maximum theroretical offset and add this to the max value.
+        // If this doesn't overflow the algortihm should be fine
+        let theorethical_maximum_offset = stats.max_value - stats.min_value;
+        if stats
+            .max_value
+            .checked_add(theorethical_maximum_offset)
+            .is_none()
+        {
+            return false;
+        }
+        true
+    }
+    /// estimation for linear interpolation is hard because, you don't know
+    /// where the local maxima for the deviation of the calculated value are and
+    /// the offset to shift all values to >=0 is also unknown.
+    fn estimate(fastfield_accessor: &impl FastFieldDataAccess, stats: FastFieldStats) -> f32 {
+        let first_val = fastfield_accessor.get_val(0);
+        let last_val = fastfield_accessor.get_val(stats.num_vals as u64 - 1);
+        let slope = get_slope(first_val, last_val, stats.num_vals);
+
+        // let's sample at 0%, 5%, 10% .. 95%, 100%
+        let num_vals = stats.num_vals as f32 / 100.0;
+        let sample_positions = (0..20)
+            .map(|pos| (num_vals * pos as f32 * 5.0) as usize)
+            .collect::<Vec<_>>();
+
+        let max_distance = sample_positions
+            .iter()
+            .map(|pos| {
+                let calculated_value = get_calculated_value(first_val, *pos as u64, slope);
+                let actual_value = fastfield_accessor.get_val(*pos as u64);
+                distance(calculated_value, actual_value)
+            })
+            .max()
+            .unwrap_or(0);
+
+        // the theory would be that we don't have the actual max_distance, but we are close within
+        // 50% threshold.
+        // It is multiplied by 2 because in a log case scenario the line would be as much above as
+        // below. So the offset would = max_distance
+        //
+        let relative_max_value = (max_distance as f32 * 1.5) * 2.0;
+
+        let num_bits = compute_num_bits(relative_max_value as u64) as u64 * stats.num_vals as u64
+            + LinearInterpolFooter::SIZE_IN_BYTES as u64;
+        let num_bits_uncompressed = 64 * stats.num_vals;
+        num_bits as f32 / num_bits_uncompressed as f32
+    }
+}
+
+#[inline]
+fn distance<T: Sub<Output = T> + Ord>(x: T, y: T) -> T {
+    if x < y {
+        y - x
+    } else {
+        x - y
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::tests::get_codec_test_data_sets;
+
+    fn create_and_validate(data: &[u64], name: &str) -> (f32, f32) {
+        crate::tests::create_and_validate::<
+            LinearInterpolFastFieldSerializer,
+            LinearInterpolFastFieldReader,
+        >(data, name)
+    }
+
+    #[test]
+    fn test_compression() {
+        let data = (10..=6_000_u64).collect::<Vec<_>>();
+        let (estimate, actual_compression) =
+            create_and_validate(&data, "simple monotonically large");
+
+        assert!(actual_compression < 0.01);
+        assert!(estimate < 0.01);
+    }
+
+    #[test]
+    fn test_with_codec_data_sets() {
+        let data_sets = get_codec_test_data_sets();
+        for (mut data, name) in data_sets {
+            create_and_validate(&data, name);
+            data.reverse();
+            create_and_validate(&data, name);
+        }
+    }
+    #[test]
+    fn linear_interpol_fast_field_test_large_amplitude() {
+        let data = vec![
+            i64::MAX as u64 / 2,
+            i64::MAX as u64 / 3,
+            i64::MAX as u64 / 2,
+        ];
+
+        create_and_validate(&data, "large amplitude");
+    }
+    #[test]
+    fn linear_interpol_fast_concave_data() {
+        let data = vec![0, 1, 2, 5, 8, 10, 20, 50];
+        create_and_validate(&data, "concave data");
+    }
+    #[test]
+    fn linear_interpol_fast_convex_data() {
+        let data = vec![0, 40, 60, 70, 75, 77];
+        create_and_validate(&data, "convex data");
+    }
+    #[test]
+    fn linear_interpol_fast_field_test_simple() {
+        let data = (10..=20_u64).collect::<Vec<_>>();
+
+        create_and_validate(&data, "simple monotonically");
+    }
+
+    #[test]
+    fn linear_interpol_fast_field_rand() {
+        for _ in 0..5000 {
+            let mut data = (0..50).map(|_| rand::random::<u64>()).collect::<Vec<_>>();
+            create_and_validate(&data, "random");
+
+            data.reverse();
+            create_and_validate(&data, "random");
+        }
+    }
+}

fastfield_codecs/src/main.rs

@@ -0,0 +1,124 @@
+#[macro_use]
+extern crate prettytable;
+use fastfield_codecs::linearinterpol::LinearInterpolFastFieldSerializer;
+use fastfield_codecs::multilinearinterpol::MultiLinearInterpolFastFieldSerializer;
+use fastfield_codecs::{FastFieldCodecSerializer, FastFieldStats};
+use prettytable::{Cell, Row, Table};
+
+fn main() {
+    let mut table = Table::new();
+
+    // Add a row per time
+    table.add_row(row!["", "Compression Ratio", "Compression Estimation"]);
+
+    for (data, data_set_name) in get_codec_test_data_sets() {
+        let mut results = vec![];
+        let res = serialize_with_codec::<LinearInterpolFastFieldSerializer>(&data);
+        results.push(res);
+        let res = serialize_with_codec::<MultiLinearInterpolFastFieldSerializer>(&data);
+        results.push(res);
+        let res = serialize_with_codec::<fastfield_codecs::bitpacked::BitpackedFastFieldSerializer>(
+            &data,
+        );
+        results.push(res);
+
+        // let best_estimation_codec = results
+        //.iter()
+        //.min_by(|res1, res2| res1.partial_cmp(&res2).unwrap())
+        //.unwrap();
+        let best_compression_ratio_codec = results
+            .iter()
+            .min_by(|res1, res2| res1.partial_cmp(res2).unwrap())
+            .cloned()
+            .unwrap();
+
+        table.add_row(Row::new(vec![Cell::new(data_set_name).style_spec("Bbb")]));
+        for (is_applicable, est, comp, name) in results {
+            let (est_cell, ratio_cell) = if !is_applicable {
+                ("Codec Disabled".to_string(), "".to_string())
+            } else {
+                (est.to_string(), comp.to_string())
+            };
+            let style = if comp == best_compression_ratio_codec.1 {
+                "Fb"
+            } else {
+                ""
+            };
+
+            table.add_row(Row::new(vec![
+                Cell::new(name).style_spec("bFg"),
+                Cell::new(&ratio_cell).style_spec(style),
+                Cell::new(&est_cell).style_spec(""),
+            ]));
+        }
+    }
+
+    table.printstd();
+}
+
+pub fn get_codec_test_data_sets() -> Vec<(Vec<u64>, &'static str)> {
+    let mut data_and_names = vec![];
+
+    let data = (1000..=200_000_u64).collect::<Vec<_>>();
+    data_and_names.push((data, "Autoincrement"));
+
+    let mut current_cumulative = 0;
+    let data = (1..=200_000_u64)
+        .map(|num| {
+            let num = (num as f32 + num as f32).log10() as u64;
+            current_cumulative += num;
+            current_cumulative
+        })
+        .collect::<Vec<_>>();
+    // let data = (1..=200000_u64).map(|num| num + num).collect::<Vec<_>>();
+    data_and_names.push((data, "Monotonically increasing concave"));
+
+    let mut current_cumulative = 0;
+    let data = (1..=200_000_u64)
+        .map(|num| {
+            let num = (200_000.0 - num as f32).log10() as u64;
+            current_cumulative += num;
+            current_cumulative
+        })
+        .collect::<Vec<_>>();
+    data_and_names.push((data, "Monotonically increasing convex"));
+
+    let data = (1000..=200_000_u64)
+        .map(|num| num + rand::random::<u8>() as u64)
+        .collect::<Vec<_>>();
+    data_and_names.push((data, "Almost monotonically increasing"));
+
+    data_and_names
+}
+
+pub fn serialize_with_codec<S: FastFieldCodecSerializer>(
+    data: &[u64],
+) -> (bool, f32, f32, &'static str) {
+    let is_applicable = S::is_applicable(&data, stats_from_vec(data));
+    if !is_applicable {
+        return (false, 0.0, 0.0, S::NAME);
+    }
+    let estimation = S::estimate(&data, stats_from_vec(data));
+    let mut out = vec![];
+    S::serialize(
+        &mut out,
+        &data,
+        stats_from_vec(data),
+        data.iter().cloned(),
+        data.iter().cloned(),
+    )
+    .unwrap();
+
+    let actual_compression = out.len() as f32 / (data.len() * 8) as f32;
+    (true, estimation, actual_compression, S::NAME)
+}
+
+pub fn stats_from_vec(data: &[u64]) -> FastFieldStats {
+    let min_value = data.iter().cloned().min().unwrap_or(0);
+    let max_value = data.iter().cloned().max().unwrap_or(0);
+    FastFieldStats {
+        min_value,
+        max_value,
+        num_vals: data.len() as u64,
+    }
+}

fastfield_codecs/src/multilinearinterpol.rs

@@ -0,0 +1,427 @@
+//! MultiLinearInterpol compressor uses linear interpolation to guess a values and stores the
+//! offset, but in blocks of 512.
+//!
+//! With a CHUNK_SIZE of 512 and 29 byte metadata per block, we get a overhead for metadata of 232 /
+//! 512 = 0,45 bits per element. The additional space required per element in a block is the the
+//! maximum deviation of the linear interpolation estimation function.
+//!
+//! E.g. if the maximum deviation of an element is 12, all elements cost 4bits.
+//!
+//! Size per block:
+//! Num Elements * Maximum Deviation from Interpolation + 29 Byte Metadata
+
+use std::io::{self, Read, Write};
+use std::ops::Sub;
+
+use common::{BinarySerializable, CountingWriter, DeserializeFrom};
+use tantivy_bitpacker::{compute_num_bits, BitPacker, BitUnpacker};
+
+use crate::{FastFieldCodecReader, FastFieldCodecSerializer, FastFieldDataAccess, FastFieldStats};
+
+const CHUNK_SIZE: u64 = 512;
+
+/// Depending on the field type, a different
+/// fast field is required.
+#[derive(Clone)]
+pub struct MultiLinearInterpolFastFieldReader {
+    pub footer: MultiLinearInterpolFooter,
+}
+
+#[derive(Clone, Debug, Default)]
+struct Function {
+    // The offset in the data is required, because we have diffrent bit_widths per block
+    data_start_offset: u64,
+    // start_pos in the block will be CHUNK_SIZE * BLOCK_NUM
+    start_pos: u64,
+    // only used during serialization, 0 after deserialization
+    end_pos: u64,
+    // only used during serialization, 0 after deserialization
+    value_start_pos: u64,
+    // only used during serialization, 0 after deserialization
+    value_end_pos: u64,
+    slope: f32,
+    // The offset so that all values are positive when writing them
+    positive_val_offset: u64,
+    num_bits: u8,
+    bit_unpacker: BitUnpacker,
+}
+
+impl Function {
+    fn calc_slope(&mut self) {
+        let num_vals = self.end_pos - self.start_pos;
+        self.slope = get_slope(self.value_start_pos, self.value_end_pos, num_vals);
+    }
+    // split the interpolation into two function, change self and return the second split
+    fn split(&mut self, split_pos: u64, split_pos_value: u64) -> Function {
+        let mut new_function = Function {
+            start_pos: split_pos,
+            end_pos: self.end_pos,
+            value_start_pos: split_pos_value,
+            value_end_pos: self.value_end_pos,
+            ..Default::default()
+        };
+        new_function.calc_slope();
+        self.end_pos = split_pos;
+        self.value_end_pos = split_pos_value;
+        self.calc_slope();
+        new_function
+    }
+}
+
+impl BinarySerializable for Function {
+    fn serialize<W: Write>(&self, write: &mut W) -> io::Result<()> {
+        self.data_start_offset.serialize(write)?;
+        self.value_start_pos.serialize(write)?;
+        self.positive_val_offset.serialize(write)?;
+        self.slope.serialize(write)?;
+        self.num_bits.serialize(write)?;
+        Ok(())
+    }
+
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<Function> {
+        let data_start_offset = u64::deserialize(reader)?;
+        let value_start_pos = u64::deserialize(reader)?;
+        let offset = u64::deserialize(reader)?;
+        let slope = f32::deserialize(reader)?;
+        let num_bits = u8::deserialize(reader)?;
+        let interpolation = Function {
+            data_start_offset,
+            value_start_pos,
+            positive_val_offset: offset,
+            num_bits,
+            bit_unpacker: BitUnpacker::new(num_bits),
+            slope,
+            ..Default::default()
+        };
+
+        Ok(interpolation)
+    }
+}
+
+#[derive(Clone, Debug)]
+pub struct MultiLinearInterpolFooter {
+    pub num_vals: u64,
+    pub min_value: u64,
+    pub max_value: u64,
+    interpolations: Vec<Function>,
+}
+
+impl BinarySerializable for MultiLinearInterpolFooter {
+    fn serialize<W: Write>(&self, write: &mut W) -> io::Result<()> {
+        let mut out = vec![];
+        self.num_vals.serialize(&mut out)?;
+        self.min_value.serialize(&mut out)?;
+        self.max_value.serialize(&mut out)?;
+        self.interpolations.serialize(&mut out)?;
+        write.write_all(&out)?;
+        (out.len() as u32).serialize(write)?;
+        Ok(())
+    }
+
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<MultiLinearInterpolFooter> {
+        let mut footer = MultiLinearInterpolFooter {
+            num_vals: u64::deserialize(reader)?,
+            min_value: u64::deserialize(reader)?,
+            max_value: u64::deserialize(reader)?,
+            interpolations: Vec::<Function>::deserialize(reader)?,
+        };
+        for (num, interpol) in footer.interpolations.iter_mut().enumerate() {
+            interpol.start_pos = CHUNK_SIZE * num as u64;
+        }
+        Ok(footer)
+    }
+}
+
+#[inline]
+fn get_interpolation_position(doc: u64) -> usize {
+    let index = doc / CHUNK_SIZE;
+    index as usize
+}
+
+#[inline]
+fn get_interpolation_function(doc: u64, interpolations: &[Function]) -> &Function {
+    &interpolations[get_interpolation_position(doc)]
+}
+
+impl FastFieldCodecReader for MultiLinearInterpolFastFieldReader {
+    /// Opens a fast field given a file.
+    fn open_from_bytes(bytes: &[u8]) -> io::Result<Self> {
+        let footer_len: u32 = (&bytes[bytes.len() - 4..]).deserialize()?;
+
+        let (_data, mut footer) = bytes.split_at(bytes.len() - (4 + footer_len) as usize);
+        let footer = MultiLinearInterpolFooter::deserialize(&mut footer)?;
+
+        Ok(MultiLinearInterpolFastFieldReader { footer })
+    }
+
+    #[inline]
+    fn get_u64(&self, doc: u64, data: &[u8]) -> u64 {
+        let interpolation = get_interpolation_function(doc, &self.footer.interpolations);
+        let doc = doc - interpolation.start_pos;
+        let calculated_value =
+            get_calculated_value(interpolation.value_start_pos, doc, interpolation.slope);
+        let diff = interpolation
+            .bit_unpacker
+            .get(doc, &data[interpolation.data_start_offset as usize..]);
+        (calculated_value + diff) - interpolation.positive_val_offset
+    }
+
+    #[inline]
+    fn min_value(&self) -> u64 {
+        self.footer.min_value
+    }
+    #[inline]
+    fn max_value(&self) -> u64 {
+        self.footer.max_value
+    }
+}
+
+#[inline]
+fn get_slope(first_val: u64, last_val: u64, num_vals: u64) -> f32 {
+    ((last_val as f64 - first_val as f64) / (num_vals as u64 - 1) as f64) as f32
+}
+
+#[inline]
+fn get_calculated_value(first_val: u64, pos: u64, slope: f32) -> u64 {
+    (first_val as i64 + (pos as f32 * slope) as i64) as u64
+}
+
+/// Same as LinearInterpolFastFieldSerializer, but working on chunks of CHUNK_SIZE elements.
+pub struct MultiLinearInterpolFastFieldSerializer {}
+
+impl FastFieldCodecSerializer for MultiLinearInterpolFastFieldSerializer {
+    const NAME: &'static str = "MultiLinearInterpol";
+    const ID: u8 = 3;
+    /// Creates a new fast field serializer.
+    fn serialize(
+        write: &mut impl Write,
+        fastfield_accessor: &impl FastFieldDataAccess,
+        stats: FastFieldStats,
+        data_iter: impl Iterator<Item = u64>,
+        _data_iter1: impl Iterator<Item = u64>,
+    ) -> io::Result<()> {
+        assert!(stats.min_value <= stats.max_value);
+
+        let first_val = fastfield_accessor.get_val(0);
+        let last_val = fastfield_accessor.get_val(stats.num_vals as u64 - 1);
+
+        let mut first_function = Function {
+            end_pos: stats.num_vals,
+            value_start_pos: first_val,
+            value_end_pos: last_val,
+            ..Default::default()
+        };
+        first_function.calc_slope();
+        let mut interpolations = vec![first_function];
+
+        // Since we potentially apply multiple passes over the data, the data is cached.
+        // Multiple iteration can be expensive (merge with index sorting can add lot of overhead per
+        // iteration)
+        let data = data_iter.collect::<Vec<_>>();
+
+        //// let's split this into chunks of CHUNK_SIZE
+        for data_pos in (0..data.len() as u64).step_by(CHUNK_SIZE as usize).skip(1) {
+            let new_fun = {
+                let current_interpolation = interpolations.last_mut().unwrap();
+                current_interpolation.split(data_pos, data[data_pos as usize])
+            };
+            interpolations.push(new_fun);
+        }
+        // calculate offset and max (-> numbits) for each function
+        for interpolation in &mut interpolations {
+            let mut offset = 0;
+            let mut rel_positive_max = 0;
+            for (pos, actual_value) in data
+                [interpolation.start_pos as usize..interpolation.end_pos as usize]
+                .iter()
+                .cloned()
+                .enumerate()
+            {
+                let calculated_value = get_calculated_value(
+                    interpolation.value_start_pos,
+                    pos as u64,
+                    interpolation.slope,
+                );
+                if calculated_value > actual_value {
+                    // negative value we need to apply an offset
+                    // we ignore negative values in the max value calculation, because negative
+                    // values will be offset to 0
+                    offset = offset.max(calculated_value - actual_value);
+                } else {
+                    // positive value no offset reuqired
+                    rel_positive_max = rel_positive_max.max(actual_value - calculated_value);
+                }
+            }
+
+            interpolation.positive_val_offset = offset;
+            interpolation.num_bits = compute_num_bits(rel_positive_max + offset);
+        }
+        let mut bit_packer = BitPacker::new();
+
+        let write = &mut CountingWriter::wrap(write);
+        for interpolation in &mut interpolations {
+            interpolation.data_start_offset = write.written_bytes();
+            let num_bits = interpolation.num_bits;
+            for (pos, actual_value) in data
+                [interpolation.start_pos as usize..interpolation.end_pos as usize]
+                .iter()
+                .cloned()
+                .enumerate()
+            {
+                let calculated_value = get_calculated_value(
+                    interpolation.value_start_pos,
+                    pos as u64,
+                    interpolation.slope,
+                );
+                let diff = (actual_value + interpolation.positive_val_offset) - calculated_value;
+                bit_packer.write(diff, num_bits, write)?;
+            }
+            bit_packer.flush(write)?;
+        }
+        bit_packer.close(write)?;
+
+        let footer = MultiLinearInterpolFooter {
+            num_vals: stats.num_vals,
+            min_value: stats.min_value,
+            max_value: stats.max_value,
+            interpolations,
+        };
+        footer.serialize(write)?;
+        Ok(())
+    }
+
+    fn is_applicable(
+        _fastfield_accessor: &impl FastFieldDataAccess,
+        stats: FastFieldStats,
+    ) -> bool {
+        if stats.num_vals < 5_000 {
+            return false;
+        }
+        // On serialization the offset is added to the actual value.
+        // We need to make sure this won't run into overflow calculation issues.
+        // For this we take the maximum theroretical offset and add this to the max value.
+        // If this doesn't overflow the algortihm should be fine
+        let theorethical_maximum_offset = stats.max_value - stats.min_value;
+        if stats
+            .max_value
+            .checked_add(theorethical_maximum_offset)
+            .is_none()
+        {
+            return false;
+        }
+        true
+    }
+    /// estimation for linear interpolation is hard because, you don't know
+    /// where the local maxima are for the deviation of the calculated value and
+    /// the offset is also unknown.
+    fn estimate(fastfield_accessor: &impl FastFieldDataAccess, stats: FastFieldStats) -> f32 {
+        let first_val_in_first_block = fastfield_accessor.get_val(0);
+        let last_elem_in_first_chunk = CHUNK_SIZE.min(stats.num_vals);
+        let last_val_in_first_block =
+            fastfield_accessor.get_val(last_elem_in_first_chunk as u64 - 1);
+        let slope = get_slope(
+            first_val_in_first_block,
+            last_val_in_first_block,
+            stats.num_vals,
+        );
+
+        // let's sample at 0%, 5%, 10% .. 95%, 100%, but for the first block only
+        let sample_positions = (0..20)
+            .map(|pos| (last_elem_in_first_chunk as f32 / 100.0 * pos as f32 * 5.0) as usize)
+            .collect::<Vec<_>>();
+
+        let max_distance = sample_positions
+            .iter()
+            .map(|pos| {
+                let calculated_value =
+                    get_calculated_value(first_val_in_first_block, *pos as u64, slope);
+                let actual_value = fastfield_accessor.get_val(*pos as u64);
+                distance(calculated_value, actual_value)
+            })
+            .max()
+            .unwrap();
+
+        // Estimate one block and extrapolate the cost to all blocks.
+        // the theory would be that we don't have the actual max_distance, but we are close within
+        // 50% threshold.
+        // It is multiplied by 2 because in a log case scenario the line would be as much above as
+        // below. So the offset would = max_distance
+        //
+        let relative_max_value = (max_distance as f32 * 1.5) * 2.0;
+
+        let num_bits = compute_num_bits(relative_max_value as u64) as u64 * stats.num_vals as u64
+            // function metadata per block
+            + 29 * (stats.num_vals / CHUNK_SIZE);
+        let num_bits_uncompressed = 64 * stats.num_vals;
+        num_bits as f32 / num_bits_uncompressed as f32
+    }
+}
+
+fn distance<T: Sub<Output = T> + Ord>(x: T, y: T) -> T {
+    if x < y {
+        y - x
+    } else {
+        x - y
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::tests::get_codec_test_data_sets;
+
+    fn create_and_validate(data: &[u64], name: &str) -> (f32, f32) {
+        crate::tests::create_and_validate::<
+            MultiLinearInterpolFastFieldSerializer,
+            MultiLinearInterpolFastFieldReader,
+        >(data, name)
+    }
+
+    #[test]
+    fn test_compression() {
+        let data = (10..=6_000_u64).collect::<Vec<_>>();
+        let (estimate, actual_compression) =
+            create_and_validate(&data, "simple monotonically large");
+        assert!(actual_compression < 0.2);
+        assert!(estimate < 0.20);
+        assert!(estimate > 0.15);
+        assert!(actual_compression > 0.01);
+    }
+
+    #[test]
+    fn test_with_codec_data_sets() {
+        let data_sets = get_codec_test_data_sets();
+        for (mut data, name) in data_sets {
+            create_and_validate(&data, name);
+            data.reverse();
+            create_and_validate(&data, name);
+        }
+    }
+    #[test]
+    fn test_simple() {
+        let data = (10..=20_u64).collect::<Vec<_>>();
+        create_and_validate(&data, "simple monotonically");
+    }
+
+    #[test]
+    fn border_cases_1() {
+        let data = (0..1024).collect::<Vec<_>>();
+        create_and_validate(&data, "border case");
+    }
+    #[test]
+    fn border_case_2() {
+        let data = (0..1025).collect::<Vec<_>>();
+        create_and_validate(&data, "border case");
+    }
+    #[test]
+    fn rand() {
+        for _ in 0..10 {
+            let mut data = (5_000..20_000)
+                .map(|_| rand::random::<u32>() as u64)
+                .collect::<Vec<_>>();
+            let _ = create_and_validate(&data, "random");
+            data.reverse();
+            create_and_validate(&data, "random");
+        }
+    }
+}

ownedbytes/Cargo.toml

@@ -0,0 +1,11 @@
+[package]
+authors = ["Paul Masurel <paul@quickwit.io>", "Pascal Seitz <pascal@quickwit.io>"]
+name = "ownedbytes"
+version = "0.2.0"
+edition = "2018"
+description = "Expose data as static slice"
+license = "MIT"
+# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
+
+[dependencies]
+stable_deref_trait = "1.2.0"

ownedbytes/src/lib.rs

@@ -1,10 +1,9 @@
-use crate::directory::FileHandle;
-use stable_deref_trait::StableDeref;
 use std::convert::TryInto;
-use std::mem;
-use std::ops::Deref;
+use std::ops::{Deref, Range};
 use std::sync::Arc;
-use std::{fmt, io};
+use std::{fmt, io, mem};
+
+use stable_deref_trait::StableDeref;
 
 /// An OwnedBytes simply wraps an object that owns a slice of data and exposes
 /// this data as a static slice.
@@ -16,12 +15,6 @@ pub struct OwnedBytes {
     box_stable_deref: Arc<dyn Deref<Target = [u8]> + Sync + Send>,
 }
 
-impl FileHandle for OwnedBytes {
-    fn read_bytes(&self, from: usize, to: usize) -> io::Result<OwnedBytes> {
-        Ok(self.slice(from, to))
-    }
-}
-
 impl OwnedBytes {
     /// Creates an empty `OwnedBytes`.
     pub fn empty() -> OwnedBytes {
@@ -36,28 +29,30 @@ impl OwnedBytes {
         let bytes: &[u8] = box_stable_deref.as_ref();
         let data = unsafe { mem::transmute::<_, &'static [u8]>(bytes.deref()) };
         OwnedBytes {
-            box_stable_deref,
             data,
+            box_stable_deref,
         }
     }
 
     /// creates a fileslice that is just a view over a slice of the data.
-    pub fn slice(&self, from: usize, to: usize) -> Self {
+    #[must_use]
+    #[inline]
+    pub fn slice(&self, range: Range<usize>) -> Self {
         OwnedBytes {
-            data: &self.data[from..to],
+            data: &self.data[range],
             box_stable_deref: self.box_stable_deref.clone(),
         }
     }
 
     /// Returns the underlying slice of data.
     /// `Deref` and `AsRef` are also available.
-    #[inline(always)]
+    #[inline]
     pub fn as_slice(&self) -> &[u8] {
         self.data
     }
 
     /// Returns the len of the slice.
-    #[inline(always)]
+    #[inline]
     pub fn len(&self) -> usize {
         self.data.len()
     }
@@ -70,6 +65,8 @@ impl OwnedBytes {
     /// On the other hand, both `left` and `right` retain a handle over
     /// the entire slice of memory. In other words, the memory will only
     /// be released when both left and right are dropped.
+    #[inline]
+    #[must_use]
     pub fn split(self, split_len: usize) -> (OwnedBytes, OwnedBytes) {
         let right_box_stable_deref = self.box_stable_deref.clone();
         let left = OwnedBytes {
@@ -83,21 +80,33 @@ impl OwnedBytes {
         (left, right)
     }
 
+    /// Splits the right part of the `OwnedBytes` at the given offset.
+    ///
+    /// `self` is truncated to `split_len`, left with the remaining bytes.
+    pub fn split_off(&mut self, split_len: usize) -> OwnedBytes {
+        let right_box_stable_deref = self.box_stable_deref.clone();
+        let right_piece = OwnedBytes {
+            data: &self.data[split_len..],
+            box_stable_deref: right_box_stable_deref,
+        };
+        self.data = &self.data[..split_len];
+        right_piece
+    }
+
     /// Returns true iff this `OwnedBytes` is empty.
-    #[inline(always)]
+    #[inline]
     pub fn is_empty(&self) -> bool {
         self.as_slice().is_empty()
     }
 
     /// Drops the left most `advance_len` bytes.
-    ///
-    /// See also [.clip(clip_len: usize))](#method.clip).
-    #[inline(always)]
+    #[inline]
     pub fn advance(&mut self, advance_len: usize) {
         self.data = &self.data[advance_len..]
     }
 
     /// Reads an `u8` from the `OwnedBytes` and advance by one byte.
+    #[inline]
     pub fn read_u8(&mut self) -> u8 {
         assert!(!self.is_empty());
 
@@ -107,6 +116,7 @@ impl OwnedBytes {
     }
 
     /// Reads an `u64` encoded as little-endian from the `OwnedBytes` and advance by 8 bytes.
+    #[inline]
     pub fn read_u64(&mut self) -> u64 {
         assert!(self.len() > 7);
 
@@ -129,15 +139,45 @@ impl fmt::Debug for OwnedBytes {
     }
 }
 
+impl PartialEq for OwnedBytes {
+    fn eq(&self, other: &OwnedBytes) -> bool {
+        self.as_slice() == other.as_slice()
+    }
+}
+
+impl Eq for OwnedBytes {}
+
+impl PartialEq<[u8]> for OwnedBytes {
+    fn eq(&self, other: &[u8]) -> bool {
+        self.as_slice() == other
+    }
+}
+
+impl PartialEq<str> for OwnedBytes {
+    fn eq(&self, other: &str) -> bool {
+        self.as_slice() == other.as_bytes()
+    }
+}
+
+impl<'a, T: ?Sized> PartialEq<&'a T> for OwnedBytes
+where OwnedBytes: PartialEq<T>
+{
+    fn eq(&self, other: &&'a T) -> bool {
+        *self == **other
+    }
+}
+
 impl Deref for OwnedBytes {
     type Target = [u8];
 
+    #[inline]
     fn deref(&self) -> &Self::Target {
         self.as_slice()
     }
 }
 
 impl io::Read for OwnedBytes {
+    #[inline]
     fn read(&mut self, buf: &mut [u8]) -> io::Result<usize> {
         let read_len = {
             let data = self.as_slice();
@@ -154,6 +194,7 @@ impl io::Read for OwnedBytes {
         self.advance(read_len);
         Ok(read_len)
     }
+    #[inline]
     fn read_to_end(&mut self, buf: &mut Vec<u8>) -> io::Result<usize> {
         let read_len = {
             let data = self.as_slice();
@@ -163,6 +204,7 @@ impl io::Read for OwnedBytes {
         self.advance(read_len);
         Ok(read_len)
     }
+    #[inline]
     fn read_exact(&mut self, buf: &mut [u8]) -> io::Result<()> {
         let read_len = self.read(buf)?;
         if read_len != buf.len() {
@@ -176,6 +218,7 @@ impl io::Read for OwnedBytes {
 }
 
 impl AsRef<[u8]> for OwnedBytes {
+    #[inline]
     fn as_ref(&self) -> &[u8] {
         self.as_slice()
     }
@@ -287,4 +330,14 @@ mod tests {
             assert_eq!(right.as_slice(), b"");
         }
     }
+
+    #[test]
+    fn test_split_off() {
+        let mut data = OwnedBytes::new(b"abcdef".as_ref());
+        assert_eq!(data, "abcdef");
+        assert_eq!(data.split_off(2), "cdef");
+        assert_eq!(data, "ab");
+        assert_eq!(data.split_off(1), "b");
+        assert_eq!(data, "a");
+    }
 }

query-grammar/Cargo.toml

@@ -1,16 +1,17 @@
 [package]
 name = "tantivy-query-grammar"
-version = "0.14.0"
+version = "0.15.0"
 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"
+homepage = "https://github.com/quickwit-oss/tantivy"
+repository = "https://github.com/quickwit-oss/tantivy"
 readme = "README.md"
 keywords = ["search", "information", "retrieval"]
 edition = "2018"
 
 [dependencies]
 combine = {version="4", default-features=false, features=[] }
+once_cell = "1.7.2"
+regex ={ version = "1.5.4", default-features = false, features = ["std"] }

query-grammar/src/lib.rs

@@ -5,11 +5,11 @@ 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 use crate::user_input_ast::{UserInputAst, UserInputBound, UserInputLeaf, UserInputLiteral};
 
 pub struct Error;
 
-pub fn parse_query(query: &str) -> Result<UserInputAST, 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/query_grammar.rs

@@ -1,21 +1,47 @@
-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::char::{char, digit, space, spaces, string};
+use combine::parser::combinator::recognize;
+use combine::parser::range::{take_while, take_while1};
+use combine::parser::repeat::escaped;
 use combine::parser::Parser;
 use combine::{
     attempt, choice, eof, many, many1, one_of, optional, parser, satisfy, skip_many1, value,
 };
+use once_cell::sync::Lazy;
+use regex::Regex;
 
-fn field<'a>() -> impl Parser<&'a str, Output = String> {
-    (
-        (letter().or(char('_'))),
-        many(satisfy(|c: char| {
-            c.is_alphanumeric() || c == '_' || c == '-'
-        })),
-    )
-        .skip(char(':'))
-        .map(|(s1, s2): (char, String)| format!("{}{}", s1, s2))
+use super::user_input_ast::{UserInputAst, UserInputBound, UserInputLeaf, UserInputLiteral};
+use crate::Occur;
+
+// Note: '-' char is only forbidden at the beginning of a field name, would be clearer to add it to
+// special characters.
+const SPECIAL_CHARS: &[char] = &[
+    '+', '^', '`', ':', '{', '}', '"', '[', ']', '(', ')', '~', '!', '\\', '*', ' ',
+];
+const ESCAPED_SPECIAL_CHARS_PATTERN: &str = r#"\\(\+|\^|`|:|\{|\}|"|\[|\]|\(|\)|\~|!|\\|\*| )"#;
+
+/// Parses a field_name
+/// A field name must have at least one character and be followed by a colon.
+/// All characters are allowed including special characters `SPECIAL_CHARS`, but these
+/// need to be escaped with a backslack character '\'.
+fn field_name<'a>() -> impl Parser<&'a str, Output = String> {
+    static ESCAPED_SPECIAL_CHARS_RE: Lazy<Regex> =
+        Lazy::new(|| Regex::new(ESCAPED_SPECIAL_CHARS_PATTERN).unwrap());
+
+    recognize::<String, _, _>(escaped(
+        (
+            take_while1(|c| !SPECIAL_CHARS.contains(&c) && c != '-'),
+            take_while(|c| !SPECIAL_CHARS.contains(&c)),
+        ),
+        '\\',
+        satisfy(|c| SPECIAL_CHARS.contains(&c)),
+    ))
+    .skip(char(':'))
+    .map(|s| ESCAPED_SPECIAL_CHARS_RE.replace_all(&s, "$1").to_string())
+    .and_then(|s: String| match s.is_empty() {
+        true => Err(StringStreamError::UnexpectedParse),
+        _ => Ok(s),
+    })
 }
 
 fn word<'a>() -> impl Parser<&'a str, Output = String> {
@@ -35,6 +61,62 @@ fn word<'a>() -> impl Parser<&'a str, Output = String> {
         })
 }
 
+/// Parses a date time according to rfc3339
+/// 2015-08-02T18:54:42+02
+/// 2021-04-13T19:46:26.266051969+00:00
+///
+/// NOTE: also accepts 999999-99-99T99:99:99.266051969+99:99
+/// We delegate rejecting such invalid dates to the logical AST compuation code
+/// which invokes chrono::DateTime::parse_from_rfc3339 on the value to actually parse
+/// it (instead of merely extracting the datetime value as string as done here).
+fn date_time<'a>() -> impl Parser<&'a str, Output = String> {
+    let two_digits = || recognize::<String, _, _>((digit(), digit()));
+
+    // Parses a time zone
+    // -06:30
+    // Z
+    let time_zone = {
+        let utc = recognize::<String, _, _>(char('Z'));
+        let offset = recognize((
+            choice([char('-'), char('+')]),
+            two_digits(),
+            char(':'),
+            two_digits(),
+        ));
+
+        utc.or(offset)
+    };
+
+    // Parses a date
+    // 2010-01-30
+    let date = {
+        recognize::<String, _, _>((
+            many1::<String, _, _>(digit()),
+            char('-'),
+            two_digits(),
+            char('-'),
+            two_digits(),
+        ))
+    };
+
+    // Parses a time
+    // 12:30:02
+    // 19:46:26.266051969
+    let time = {
+        recognize::<String, _, _>((
+            two_digits(),
+            char(':'),
+            two_digits(),
+            char(':'),
+            two_digits(),
+            optional((char('.'), many1::<String, _, _>(digit()))),
+            time_zone,
+        ))
+    };
+
+    recognize((date, char('T'), time))
+}
+
 fn term_val<'a>() -> impl Parser<&'a str, Output = String> {
     let phrase = char('"').with(many1(satisfy(|c| c != '"'))).skip(char('"'));
     phrase.or(word())
@@ -42,7 +124,7 @@ fn term_val<'a>() -> impl Parser<&'a str, Output = String> {
 
 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(), term_val_with_field).map(|(field_name, phrase)| UserInputLiteral {
         field_name: Some(field_name),
         phrase,
     })
@@ -83,7 +165,8 @@ fn spaces1<'a>() -> impl Parser<&'a str, Output = ()> {
 /// [a TO *], [a TO c], [abc TO bcd}
 fn range<'a>() -> impl Parser<&'a str, Output = UserInputLeaf> {
     let range_term_val = || {
-        word()
+        attempt(date_time())
+            .or(word())
             .or(negative_number())
             .or(char('*').with(value("*".to_string())))
     };
@@ -138,7 +221,7 @@ fn range<'a>() -> impl Parser<&'a str, Output = UserInputLeaf> {
     );
 
     (
-        optional(field()).skip(spaces()),
+        optional(field_name()).skip(spaces()),
         // try elastic first, if it matches, the range is unbounded
         attempt(elastic_unbounded_range).or(lower_to_upper),
     )
@@ -152,21 +235,21 @@ fn range<'a>() -> impl Parser<&'a str, Output = UserInputLeaf> {
     })
 }
 
-fn negate(expr: UserInputAST) -> UserInputAST {
+fn negate(expr: UserInputAst) -> UserInputAst {
     expr.unary(Occur::MustNot)
 }
 
-fn leaf<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
+fn leaf<'a>() -> impl Parser<&'a str, Output = UserInputAst> {
     parser(|input| {
         char('(')
             .with(ast())
             .skip(char(')'))
-            .or(char('*').map(|_| UserInputAST::from(UserInputLeaf::All)))
+            .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))
+            .or(attempt(range().map(UserInputAst::from)))
+            .or(literal().map(UserInputAst::from))
             .parse_stream(input)
             .into_result()
     })
@@ -178,7 +261,7 @@ fn occur_symbol<'a>() -> impl Parser<&'a str, Output = Occur> {
         .or(char('+').map(|_| Occur::Must))
 }
 
-fn occur_leaf<'a>() -> impl Parser<&'a str, Output = (Option<Occur>, UserInputAST)> {
+fn occur_leaf<'a>() -> impl Parser<&'a str, Output = (Option<Occur>, UserInputAst)> {
     (optional(occur_symbol()), boosted_leaf())
 }
 
@@ -199,10 +282,10 @@ fn boost<'a>() -> impl Parser<&'a str, Output = f64> {
     (char('^'), positive_float_number()).map(|(_, boost)| boost)
 }
 
-fn boosted_leaf<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
+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::f64::EPSILON => {
-            UserInputAST::Boost(Box::new(leaf), boost)
+            UserInputAst::Boost(Box::new(leaf), boost)
         }
         _ => leaf,
     })
@@ -221,10 +304,10 @@ fn binary_operand<'a>() -> impl Parser<&'a str, Output = BinaryOperand> {
 }
 
 fn aggregate_binary_expressions(
-    left: UserInputAST,
-    others: Vec<(BinaryOperand, UserInputAST)>,
-) -> UserInputAST {
-    let mut dnf: Vec<Vec<UserInputAST>> = vec![vec![left]];
+    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 => {
@@ -238,33 +321,33 @@ fn aggregate_binary_expressions(
         }
     }
     if dnf.len() == 1 {
-        UserInputAST::and(dnf.into_iter().next().unwrap()) //< safe
+        UserInputAst::and(dnf.into_iter().next().unwrap()) //< safe
     } else {
-        let conjunctions = dnf.into_iter().map(UserInputAST::and).collect();
-        UserInputAST::or(conjunctions)
+        let conjunctions = dnf.into_iter().map(UserInputAst::and).collect();
+        UserInputAst::or(conjunctions)
     }
 }
 
-fn operand_leaf<'a>() -> impl Parser<&'a str, Output = (BinaryOperand, UserInputAST)> {
+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> {
+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)>| {
+        |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)]),
+                    Occur::MustNot => UserInputAst::Clause(vec![(Some(Occur::MustNot), ast)]),
                 }
             } else {
-                UserInputAST::Clause(subqueries.into_iter().collect())
+                UserInputAst::Clause(subqueries.into_iter().collect())
             }
         },
     );
@@ -272,10 +355,10 @@ pub fn ast<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
     spaces().with(expr).skip(spaces())
 }
 
-pub fn parse_to_ast<'a>() -> impl Parser<&'a str, Output = UserInputAST> {
+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))
+        .map(|opt_ast| opt_ast.unwrap_or_else(UserInputAst::empty_query))
 }
 
 #[cfg(test)]
@@ -283,9 +366,10 @@ mod test {
 
     type TestParseResult = Result<(), StringStreamError>;
 
-    use super::*;
     use combine::parser::Parser;
 
+    use super::*;
+
     pub fn nearly_equals(a: f64, b: f64) -> bool {
         (a - b).abs() < 0.0005 * (a + b).abs()
     }
@@ -324,6 +408,22 @@ mod test {
         error_parse("-1.");
     }
 
+    #[test]
+    fn test_date_time() {
+        let (val, remaining) = date_time()
+            .parse("2015-08-02T18:54:42+02:30")
+            .expect("cannot parse date");
+        assert_eq!(val, "2015-08-02T18:54:42+02:30");
+        assert_eq!(remaining, "");
+        assert!(date_time().parse("2015-08-02T18:54:42+02").is_err());
+
+        let (val, remaining) = date_time()
+            .parse("2021-04-13T19:46:26.266051969+00:00")
+            .expect("cannot parse fractional date");
+        assert_eq!(val, "2021-04-13T19:46:26.266051969+00:00");
+        assert_eq!(remaining, "");
+    }
+
     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);
@@ -391,21 +491,21 @@ mod test {
 
     #[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("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\":{\"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", "\"weight\":{\"*\" TO \"70\"]");
 
-        test_parse_query_to_ast_helper("weight: <= 70.5", "weight:{\"*\" TO \"70.5\"]");
+        test_parse_query_to_ast_helper("weight: <= 70.5", "\"weight\":{\"*\" TO \"70.5\"]");
     }
 
     #[test]
@@ -418,44 +518,100 @@ mod test {
     #[test]
     fn test_field_name() -> TestParseResult {
         assert_eq!(
-            super::field().parse("my-field-name:a")?,
-            ("my-field-name".to_string(), "a")
+            super::field_name().parse(".my.field.name:a"),
+            Ok((".my.field.name".to_string(), "a"))
         );
         assert_eq!(
-            super::field().parse("my_field_name:a")?,
-            ("my_field_name".to_string(), "a")
+            super::field_name().parse("my\\ field\\ name:a"),
+            Ok(("my field name".to_string(), "a"))
         );
-        assert!(super::field().parse(":a").is_err());
-        assert!(super::field().parse("-my_field:a").is_err());
+        assert!(super::field_name().parse("my field:a").is_err());
         assert_eq!(
-            super::field().parse("_my_field:a")?,
+            super::field_name().parse("\\(1\\+1\\):2"),
+            Ok(("(1+1)".to_string(), "2"))
+        );
+        assert_eq!(
+            super::field_name().parse("my_field_name:a"),
+            Ok(("my_field_name".to_string(), "a"))
+        );
+        assert!(super::field_name().parse("my_field_name").is_err());
+        assert!(super::field_name().parse(":a").is_err());
+        assert!(super::field_name().parse("-my_field:a").is_err());
+        assert_eq!(
+            super::field_name().parse("_my_field:a")?,
             ("_my_field".to_string(), "a")
         );
         Ok(())
     }
 
+    #[test]
+    fn test_field_name_re() {
+        let escaped_special_chars_re = Regex::new(ESCAPED_SPECIAL_CHARS_PATTERN).unwrap();
+        for special_char in SPECIAL_CHARS.iter() {
+            assert_eq!(
+                escaped_special_chars_re.replace_all(&format!("\\{}", special_char), "$1"),
+                special_char.to_string()
+            );
+        }
+    }
+
     #[test]
     fn test_range_parser() {
         // testing the range() parser separately
-        let res = range().parse("title: <hello").unwrap().0;
+        let res = range()
+            .parse("title: <hello")
+            .expect("Cannot parse felxible bound word")
+            .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;
+        let res2 = range()
+            .parse("title:{* TO hello}")
+            .expect("Cannot parse ununbounded to word")
+            .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;
+        let res3 = range()
+            .parse("weight: >=71.2")
+            .expect("Cannot parse flexible bound float")
+            .0;
+        let res4 = range()
+            .parse("weight:[71.2 TO *}")
+            .expect("Cannot parse float to unbounded")
+            .0;
         assert_eq!(res3, expected_weight);
         assert_eq!(res4, expected_weight);
+
+        let expected_dates = UserInputLeaf::Range {
+            field: Some("date_field".to_string()),
+            lower: UserInputBound::Exclusive("2015-08-02T18:54:42Z".to_string()),
+            upper: UserInputBound::Inclusive("2021-08-02T18:54:42+02:30".to_string()),
+        };
+        let res5 = range()
+            .parse("date_field:{2015-08-02T18:54:42Z TO 2021-08-02T18:54:42+02:30]")
+            .expect("Cannot parse date range")
+            .0;
+        assert_eq!(res5, expected_dates);
+
+        let expected_flexible_dates = UserInputLeaf::Range {
+            field: Some("date_field".to_string()),
+            lower: UserInputBound::Unbounded,
+            upper: UserInputBound::Inclusive("2021-08-02T18:54:42.12345+02:30".to_string()),
+        };
+
+        let res6 = range()
+            .parse("date_field: <=2021-08-02T18:54:42.12345+02:30")
+            .expect("Cannot parse date range")
+            .0;
+        assert_eq!(res6, expected_flexible_dates);
     }
 
     #[test]
@@ -492,12 +648,14 @@ mod test {
 
     #[test]
     fn test_single_term_with_field() {
-        test_parse_query_to_ast_helper("abc:toto", "abc:\"toto\"");
+        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_parse_query_to_ast_helper("abc:1.1", "\"abc\":\"1.1\"");
+        test_parse_query_to_ast_helper("a.b.c:1.1", "\"a.b.c\":\"1.1\"");
+        test_parse_query_to_ast_helper("a\\ b\\ c:1.1", "\"a b c\":\"1.1\"");
     }
 
     #[test]
@@ -513,22 +671,27 @@ mod test {
     #[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_parse_query_to_ast_helper("+abc:toto", "\"abc\":\"toto\"");
+        test_parse_query_to_ast_helper("+a\\+b\\+c:toto", "\"a+b+c\":\"toto\"");
+        test_parse_query_to_ast_helper("(+abc:toto -titi)", "(+\"abc\":\"toto\" -\"titi\")");
+        test_parse_query_to_ast_helper("-abc:toto", "(-\"abc\":\"toto\")");
+        test_is_parse_err("--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_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(
+            "1.2.foo.bar:[1.1 TO *}",
+            "\"1.2.foo.bar\":[\"1.1\" TO \"*\"}",
+        );
         test_is_parse_err("abc +    ");
     }
 }

query-grammar/src/user_input_ast.rs

@@ -24,7 +24,7 @@ impl Debug for UserInputLeaf {
                 ref upper,
             } => {
                 if let Some(ref field) = field {
-                    write!(formatter, "{}:", field)?;
+                    write!(formatter, "\"{}\":", field)?;
                 }
                 lower.display_lower(formatter)?;
                 write!(formatter, " TO ")?;
@@ -45,7 +45,7 @@ pub struct UserInputLiteral {
 impl fmt::Debug for UserInputLiteral {
     fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
         match self.field_name {
-            Some(ref field_name) => write!(formatter, "{}:\"{}\"", field_name, self.phrase),
+            Some(ref field_name) => write!(formatter, "\"{}\":\"{}\"", field_name, self.phrase),
             None => write!(formatter, "\"{}\"", self.phrase),
         }
     }
@@ -59,7 +59,7 @@ pub enum UserInputBound {
 }
 
 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),
@@ -67,7 +67,7 @@ impl UserInputBound {
         }
     }
 
-    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),
@@ -79,46 +79,47 @@ impl UserInputBound {
         match *self {
             UserInputBound::Inclusive(ref contents) => contents,
             UserInputBound::Exclusive(ref contents) => contents,
-            UserInputBound::Unbounded => &"*",
+            UserInputBound::Unbounded => "*",
         }
     }
 }
 
-pub enum UserInputAST {
-    Clause(Vec<(Option<Occur>, UserInputAST)>),
+pub enum UserInputAst {
+    Clause(Vec<(Option<Occur>, UserInputAst)>),
     Leaf(Box<UserInputLeaf>),
-    Boost(Box<UserInputAST>, f64),
+    Boost(Box<UserInputAst>, f64),
 }
 
-impl UserInputAST {
-    pub fn unary(self, occur: Occur) -> UserInputAST {
-        UserInputAST::Clause(vec![(Some(occur), self)])
+impl UserInputAst {
+    #[must_use]
+    pub fn unary(self, occur: Occur) -> UserInputAst {
+        UserInputAst::Clause(vec![(Some(occur), self)])
     }
 
-    fn compose(occur: Occur, asts: Vec<UserInputAST>) -> UserInputAST {
+    fn compose(occur: Occur, asts: Vec<UserInputAst>) -> UserInputAst {
         assert_ne!(occur, Occur::MustNot);
         assert!(!asts.is_empty());
         if asts.len() == 1 {
             asts.into_iter().next().unwrap() //< safe
         } else {
-            UserInputAST::Clause(
+            UserInputAst::Clause(
                 asts.into_iter()
-                    .map(|ast: UserInputAST| (Some(occur), ast))
+                    .map(|ast: UserInputAst| (Some(occur), ast))
                     .collect::<Vec<_>>(),
             )
         }
     }
 
-    pub fn empty_query() -> UserInputAST {
-        UserInputAST::Clause(Vec::default())
+    pub fn empty_query() -> UserInputAst {
+        UserInputAst::Clause(Vec::default())
     }
 
-    pub fn and(asts: Vec<UserInputAST>) -> UserInputAST {
-        UserInputAST::compose(Occur::Must, asts)
+    pub fn and(asts: Vec<UserInputAst>) -> UserInputAst {
+        UserInputAst::compose(Occur::Must, asts)
     }
 
-    pub fn or(asts: Vec<UserInputAST>) -> UserInputAST {
-        UserInputAST::compose(Occur::Should, asts)
+    pub fn or(asts: Vec<UserInputAst>) -> UserInputAst {
+        UserInputAst::compose(Occur::Should, asts)
     }
 }
 
@@ -128,15 +129,15 @@ impl From<UserInputLiteral> for UserInputLeaf {
     }
 }
 
-impl From<UserInputLeaf> for UserInputAST {
-    fn from(leaf: UserInputLeaf) -> UserInputAST {
-        UserInputAST::Leaf(Box::new(leaf))
+impl From<UserInputLeaf> for UserInputAst {
+    fn from(leaf: UserInputLeaf) -> UserInputAst {
+        UserInputAst::Leaf(Box::new(leaf))
     }
 }
 
 fn print_occur_ast(
     occur_opt: Option<Occur>,
-    ast: &UserInputAST,
+    ast: &UserInputAst,
     formatter: &mut fmt::Formatter,
 ) -> fmt::Result {
     if let Some(occur) = occur_opt {
@@ -147,10 +148,10 @@ fn print_occur_ast(
     Ok(())
 }
 
-impl fmt::Debug for UserInputAST {
+impl fmt::Debug for UserInputAst {
     fn fmt(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
         match *self {
-            UserInputAST::Clause(ref subqueries) => {
+            UserInputAst::Clause(ref subqueries) => {
                 if subqueries.is_empty() {
                     write!(formatter, "<emptyclause>")?;
                 } else {
@@ -164,8 +165,8 @@ impl fmt::Debug for UserInputAST {
                 }
                 Ok(())
             }
-            UserInputAST::Leaf(ref subquery) => write!(formatter, "{:?}", subquery),
-            UserInputAST::Boost(ref leaf, boost) => write!(formatter, "({:?})^{}", leaf, boost),
+            UserInputAst::Leaf(ref subquery) => write!(formatter, "{:?}", subquery),
+            UserInputAst::Boost(ref leaf, boost) => write!(formatter, "({:?})^{}", leaf, boost),
         }
     }
 }

rustfmt.toml

@@ -1 +1,7 @@
-use_try_shorthand = true
+comment_width = 120
+format_strings = true
+group_imports = "StdExternalCrate"
+imports_granularity = "Module"
+normalize_comments = true
+where_single_line = true
+wrap_comments = true

src/aggregation/README.md

@@ -0,0 +1,36 @@
+# Contributing
+
+When adding new bucket aggregation make sure to extend the "test_aggregation_flushing" test for at least 2 levels.
+
+
+
+# Code Organization
+
+Tantivy's aggregations have been designed to mimic the 
+[aggregations of elasticsearch](https://www.elastic.co/guide/en/elasticsearch/reference/current/search-aggregations.html).
+
+The code is organized in submodules:
+
+## bucket
+Contains all bucket aggregations, like range aggregation. These bucket aggregations group documents into buckets and can contain sub-aggegations.
+
+## metric
+Contains all metric aggregations, like average aggregation. Metric aggregations do not have sub aggregations.
+
+#### agg_req
+agg_req contains the users aggregation request. Deserialization from json is compatible with elasticsearch aggregation requests.
+
+#### agg_req_with_accessor
+agg_req_with_accessor contains the users aggregation request enriched with fast field accessors etc, which are
+used during collection.
+
+#### segment_agg_result
+segment_agg_result contains the aggregation result tree, which is used for collection of a segment.
+The tree from agg_req_with_accessor is passed during collection.
+
+#### intermediate_agg_result
+intermediate_agg_result contains the aggregation tree for merging with other trees.
+
+#### agg_result
+agg_result contains the final aggregation tree.
+

src/aggregation/agg_req.rs

@@ -0,0 +1,270 @@
+//! Contains the aggregation request tree. Used to build an
+//! [AggregationCollector](super::AggregationCollector).
+//!
+//! [Aggregations] is the top level entry point to create a request, which is a `HashMap<String,
+//! Aggregation>`.
+//!
+//! Requests are compatible with the json format of elasticsearch.
+//!
+//! # Example
+//!
+//! ```
+//! use tantivy::aggregation::bucket::RangeAggregation;
+//! use tantivy::aggregation::agg_req::BucketAggregationType;
+//! use tantivy::aggregation::agg_req::{Aggregation, Aggregations};
+//! use tantivy::aggregation::agg_req::BucketAggregation;
+//! let agg_req1: Aggregations = vec![
+//!     (
+//!         "range".to_string(),
+//!         Aggregation::Bucket(BucketAggregation {
+//!             bucket_agg: BucketAggregationType::Range(RangeAggregation{
+//!                 field: "score".to_string(),
+//!                 ranges: vec![(3f64..7f64).into(), (7f64..20f64).into()],
+//!             }),
+//!             sub_aggregation: Default::default(),
+//!         }),
+//!     ),
+//! ]
+//! .into_iter()
+//! .collect();
+//!
+//! let elasticsearch_compatible_json_req = r#"
+//! {
+//!   "range": {
+//!     "range": {
+//!       "field": "score",
+//!       "ranges": [
+//!         { "from": 3.0, "to": 7.0 },
+//!         { "from": 7.0, "to": 20.0 }
+//!       ]
+//!     }
+//!   }
+//! }"#;
+//! let agg_req2: Aggregations = serde_json::from_str(elasticsearch_compatible_json_req).unwrap();
+//! assert_eq!(agg_req1, agg_req2);
+//! ```
+
+use std::collections::{HashMap, HashSet};
+
+use serde::{Deserialize, Serialize};
+
+pub use super::bucket::RangeAggregation;
+use super::metric::{AverageAggregation, StatsAggregation};
+
+/// The top-level aggregation request structure, which contains [Aggregation] and their user defined
+/// names. It is also used in [buckets](BucketAggregation) to define sub-aggregations.
+///
+/// The key is the user defined name of the aggregation.
+pub type Aggregations = HashMap<String, Aggregation>;
+
+/// Extract all fast field names used in the tree.
+pub fn get_fast_field_names(aggs: &Aggregations) -> HashSet<String> {
+    let mut fast_field_names = Default::default();
+    for el in aggs.values() {
+        el.get_fast_field_names(&mut fast_field_names)
+    }
+    fast_field_names
+}
+
+/// Aggregation request of [BucketAggregation] or [MetricAggregation].
+///
+/// An aggregation is either a bucket or a metric.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+#[serde(untagged)]
+pub enum Aggregation {
+    /// Bucket aggregation, see [BucketAggregation] for details.
+    Bucket(BucketAggregation),
+    /// Metric aggregation, see [MetricAggregation] for details.
+    Metric(MetricAggregation),
+}
+
+impl Aggregation {
+    fn get_fast_field_names(&self, fast_field_names: &mut HashSet<String>) {
+        match self {
+            Aggregation::Bucket(bucket) => bucket.get_fast_field_names(fast_field_names),
+            Aggregation::Metric(metric) => metric.get_fast_field_names(fast_field_names),
+        }
+    }
+}
+
+/// BucketAggregations create buckets of documents. Each bucket is associated with a rule which
+/// determines whether or not a document in the falls into it. In other words, the buckets
+/// effectively define document sets. Buckets are not necessarily disjunct, therefore a document can
+/// fall into multiple buckets. In addition to the buckets themselves, the bucket aggregations also
+/// compute and return the number of documents for each bucket. Bucket aggregations, as opposed to
+/// metric aggregations, can hold sub-aggregations. These sub-aggregations will be aggregated for
+/// the buckets created by their "parent" bucket aggregation. There are different bucket
+/// aggregators, each with a different "bucketing" strategy. Some define a single bucket, some
+/// define fixed number of multiple buckets, and others dynamically create the buckets during the
+/// aggregation process.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct BucketAggregation {
+    /// Bucket aggregation strategy to group documents.
+    #[serde(flatten)]
+    pub bucket_agg: BucketAggregationType,
+    /// The sub_aggregations in the buckets. Each bucket will aggregate on the document set in the
+    /// bucket.
+    #[serde(rename = "aggs")]
+    #[serde(default)]
+    #[serde(skip_serializing_if = "Aggregations::is_empty")]
+    pub sub_aggregation: Aggregations,
+}
+
+impl BucketAggregation {
+    fn get_fast_field_names(&self, fast_field_names: &mut HashSet<String>) {
+        self.bucket_agg.get_fast_field_names(fast_field_names);
+        fast_field_names.extend(get_fast_field_names(&self.sub_aggregation));
+    }
+}
+
+/// The bucket aggregation types.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub enum BucketAggregationType {
+    /// Put data into buckets of user-defined ranges.
+    #[serde(rename = "range")]
+    Range(RangeAggregation),
+}
+
+impl BucketAggregationType {
+    fn get_fast_field_names(&self, fast_field_names: &mut HashSet<String>) {
+        match self {
+            BucketAggregationType::Range(range) => fast_field_names.insert(range.field.to_string()),
+        };
+    }
+}
+
+/// The aggregations in this family compute metrics based on values extracted
+/// from the documents that are being aggregated. Values are extracted from the fast field of
+/// the document.
+
+/// Some aggregations output a single numeric metric (e.g. Average) and are called
+/// single-value numeric metrics aggregation, others generate multiple metrics (e.g. Stats) and are
+/// called multi-value numeric metrics aggregation.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub enum MetricAggregation {
+    /// Calculates the average.
+    #[serde(rename = "avg")]
+    Average(AverageAggregation),
+    /// Calculates stats sum, average, min, max, standard_deviation on a field.
+    #[serde(rename = "stats")]
+    Stats(StatsAggregation),
+}
+
+impl MetricAggregation {
+    fn get_fast_field_names(&self, fast_field_names: &mut HashSet<String>) {
+        match self {
+            MetricAggregation::Average(avg) => fast_field_names.insert(avg.field.to_string()),
+            MetricAggregation::Stats(stats) => fast_field_names.insert(stats.field.to_string()),
+        };
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn serialize_to_json_test() {
+        let agg_req1: Aggregations = vec![(
+            "range".to_string(),
+            Aggregation::Bucket(BucketAggregation {
+                bucket_agg: BucketAggregationType::Range(RangeAggregation {
+                    field: "score".to_string(),
+                    ranges: vec![
+                        (f64::MIN..3f64).into(),
+                        (3f64..7f64).into(),
+                        (7f64..20f64).into(),
+                        (20f64..f64::MAX).into(),
+                    ],
+                }),
+                sub_aggregation: Default::default(),
+            }),
+        )]
+        .into_iter()
+        .collect();
+
+        let elasticsearch_compatible_json_req = r#"{
+  "range": {
+    "range": {
+      "field": "score",
+      "ranges": [
+        {
+          "to": 3.0
+        },
+        {
+          "from": 3.0,
+          "to": 7.0
+        },
+        {
+          "from": 7.0,
+          "to": 20.0
+        },
+        {
+          "from": 20.0
+        }
+      ]
+    }
+  }
+}"#;
+        let agg_req2: String = serde_json::to_string_pretty(&agg_req1).unwrap();
+        assert_eq!(agg_req2, elasticsearch_compatible_json_req);
+    }
+
+    #[test]
+    fn test_get_fast_field_names() {
+        let agg_req2: Aggregations = vec![
+            (
+                "range".to_string(),
+                Aggregation::Bucket(BucketAggregation {
+                    bucket_agg: BucketAggregationType::Range(RangeAggregation {
+                        field: "score2".to_string(),
+                        ranges: vec![
+                            (f64::MIN..3f64).into(),
+                            (3f64..7f64).into(),
+                            (7f64..20f64).into(),
+                            (20f64..f64::MAX).into(),
+                        ],
+                    }),
+                    sub_aggregation: Default::default(),
+                }),
+            ),
+            (
+                "metric".to_string(),
+                Aggregation::Metric(MetricAggregation::Average(
+                    AverageAggregation::from_field_name("field123".to_string()),
+                )),
+            ),
+        ]
+        .into_iter()
+        .collect();
+
+        let agg_req1: Aggregations = vec![(
+            "range".to_string(),
+            Aggregation::Bucket(BucketAggregation {
+                bucket_agg: BucketAggregationType::Range(RangeAggregation {
+                    field: "score".to_string(),
+                    ranges: vec![
+                        (f64::MIN..3f64).into(),
+                        (3f64..7f64).into(),
+                        (7f64..20f64).into(),
+                        (20f64..f64::MAX).into(),
+                    ],
+                }),
+                sub_aggregation: agg_req2,
+            }),
+        )]
+        .into_iter()
+        .collect();
+
+        assert_eq!(
+            get_fast_field_names(&agg_req1),
+            vec![
+                "score".to_string(),
+                "score2".to_string(),
+                "field123".to_string()
+            ]
+            .into_iter()
+            .collect()
+        )
+    }
+}

src/aggregation/agg_req_with_accessor.rs

@@ -0,0 +1,146 @@
+//! This will enhance the request tree with access to the fastfield and metadata.
+
+use super::agg_req::{Aggregation, Aggregations, BucketAggregationType, MetricAggregation};
+use super::bucket::RangeAggregation;
+use super::metric::{AverageAggregation, StatsAggregation};
+use super::VecWithNames;
+use crate::fastfield::{type_and_cardinality, DynamicFastFieldReader, FastType};
+use crate::schema::{Cardinality, Type};
+use crate::{SegmentReader, TantivyError};
+
+#[derive(Clone, Default)]
+pub(crate) struct AggregationsWithAccessor {
+    pub metrics: VecWithNames<MetricAggregationWithAccessor>,
+    pub buckets: VecWithNames<BucketAggregationWithAccessor>,
+}
+
+impl AggregationsWithAccessor {
+    fn from_data(
+        metrics: VecWithNames<MetricAggregationWithAccessor>,
+        buckets: VecWithNames<BucketAggregationWithAccessor>,
+    ) -> Self {
+        Self { metrics, buckets }
+    }
+
+    pub fn is_empty(&self) -> bool {
+        self.metrics.is_empty() && self.buckets.is_empty()
+    }
+}
+
+#[derive(Clone)]
+pub struct BucketAggregationWithAccessor {
+    /// In general there can be buckets without fast field access, e.g. buckets that are created
+    /// based on search terms. So eventually this needs to be Option or moved.
+    pub(crate) accessor: DynamicFastFieldReader<u64>,
+    pub(crate) field_type: Type,
+    pub(crate) bucket_agg: BucketAggregationType,
+    pub(crate) sub_aggregation: AggregationsWithAccessor,
+}
+
+impl BucketAggregationWithAccessor {
+    fn try_from_bucket(
+        bucket: &BucketAggregationType,
+        sub_aggregation: &Aggregations,
+        reader: &SegmentReader,
+    ) -> crate::Result<BucketAggregationWithAccessor> {
+        let (accessor, field_type) = match &bucket {
+            BucketAggregationType::Range(RangeAggregation {
+                field: field_name,
+                ranges: _,
+            }) => get_ff_reader_and_validate(reader, field_name)?,
+        };
+        let sub_aggregation = sub_aggregation.clone();
+        Ok(BucketAggregationWithAccessor {
+            accessor,
+            field_type,
+            sub_aggregation: get_aggs_with_accessor_and_validate(&sub_aggregation, reader)?,
+            bucket_agg: bucket.clone(),
+        })
+    }
+}
+
+/// Contains the metric request and the fast field accessor.
+#[derive(Clone)]
+pub struct MetricAggregationWithAccessor {
+    pub metric: MetricAggregation,
+    pub field_type: Type,
+    pub accessor: DynamicFastFieldReader<u64>,
+}
+
+impl MetricAggregationWithAccessor {
+    fn try_from_metric(
+        metric: &MetricAggregation,
+        reader: &SegmentReader,
+    ) -> crate::Result<MetricAggregationWithAccessor> {
+        match &metric {
+            MetricAggregation::Average(AverageAggregation { field: field_name })
+            | MetricAggregation::Stats(StatsAggregation { field: field_name }) => {
+                let (accessor, field_type) = get_ff_reader_and_validate(reader, field_name)?;
+
+                Ok(MetricAggregationWithAccessor {
+                    accessor,
+                    field_type,
+                    metric: metric.clone(),
+                })
+            }
+        }
+    }
+}
+
+pub(crate) fn get_aggs_with_accessor_and_validate(
+    aggs: &Aggregations,
+    reader: &SegmentReader,
+) -> crate::Result<AggregationsWithAccessor> {
+    let mut metrics = vec![];
+    let mut buckets = vec![];
+    for (key, agg) in aggs.iter() {
+        match agg {
+            Aggregation::Bucket(bucket) => buckets.push((
+                key.to_string(),
+                BucketAggregationWithAccessor::try_from_bucket(
+                    &bucket.bucket_agg,
+                    &bucket.sub_aggregation,
+                    reader,
+                )?,
+            )),
+            Aggregation::Metric(metric) => metrics.push((
+                key.to_string(),
+                MetricAggregationWithAccessor::try_from_metric(metric, reader)?,
+            )),
+        }
+    }
+    Ok(AggregationsWithAccessor::from_data(
+        VecWithNames::from_entries(metrics),
+        VecWithNames::from_entries(buckets),
+    ))
+}
+
+fn get_ff_reader_and_validate(
+    reader: &SegmentReader,
+    field_name: &str,
+) -> crate::Result<(DynamicFastFieldReader<u64>, Type)> {
+    let field = reader
+        .schema()
+        .get_field(field_name)
+        .ok_or_else(|| TantivyError::FieldNotFound(field_name.to_string()))?;
+    let field_type = reader.schema().get_field_entry(field).field_type();
+
+    if let Some((ff_type, cardinality)) = type_and_cardinality(field_type) {
+        if cardinality == Cardinality::MultiValues || ff_type == FastType::Date {
+            return Err(TantivyError::InvalidArgument(format!(
+                "Invalid field type in aggregation {:?}, only Cardinality::SingleValue supported",
+                field_type.value_type()
+            )));
+        }
+    } else {
+        return Err(TantivyError::InvalidArgument(format!(
+            "Only single value fast fields of type f64, u64, i64 are supported, but got {:?} ",
+            field_type.value_type()
+        )));
+    };
+
+    let ff_fields = reader.fast_fields();
+    ff_fields
+        .u64_lenient(field)
+        .map(|field| (field, field_type.value_type()))
+}

src/aggregation/agg_result.rs

@@ -0,0 +1,170 @@
+//! Contains the final aggregation tree.
+//! This tree can be converted via the `into()` method from `IntermediateAggregationResults`.
+//! This conversion computes the final result. For example: The intermediate result contains
+//! intermediate average results, which is the sum and the number of values. The actual average is
+//! calculated on the step from intermediate to final aggregation result tree.
+
+use std::cmp::Ordering;
+use std::collections::HashMap;
+
+use itertools::Itertools;
+use serde::{Deserialize, Serialize};
+
+use super::intermediate_agg_result::{
+    IntermediateAggregationResult, IntermediateAggregationResults, IntermediateBucketResult,
+    IntermediateMetricResult, IntermediateRangeBucketEntry,
+};
+use super::metric::{SingleMetricResult, Stats};
+use super::Key;
+
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+/// The final aggegation result.
+pub struct AggregationResults(pub HashMap<String, AggregationResult>);
+
+impl From<IntermediateAggregationResults> for AggregationResults {
+    fn from(tree: IntermediateAggregationResults) -> Self {
+        Self(
+            tree.0
+                .into_iter()
+                .map(|(key, agg)| (key, agg.into()))
+                .collect(),
+        )
+    }
+}
+
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+#[serde(untagged)]
+/// An aggregation is either a bucket or a metric.
+pub enum AggregationResult {
+    /// Bucket result variant.
+    BucketResult(BucketResult),
+    /// Metric result variant.
+    MetricResult(MetricResult),
+}
+impl From<IntermediateAggregationResult> for AggregationResult {
+    fn from(tree: IntermediateAggregationResult) -> Self {
+        match tree {
+            IntermediateAggregationResult::Bucket(bucket) => {
+                AggregationResult::BucketResult(bucket.into())
+            }
+            IntermediateAggregationResult::Metric(metric) => {
+                AggregationResult::MetricResult(metric.into())
+            }
+        }
+    }
+}
+
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+#[serde(untagged)]
+/// MetricResult
+pub enum MetricResult {
+    /// Average metric result.
+    Average(SingleMetricResult),
+    /// Stats metric result.
+    Stats(Stats),
+}
+
+impl From<IntermediateMetricResult> for MetricResult {
+    fn from(metric: IntermediateMetricResult) -> Self {
+        match metric {
+            IntermediateMetricResult::Average(avg_data) => {
+                MetricResult::Average(avg_data.finalize().into())
+            }
+            IntermediateMetricResult::Stats(intermediate_stats) => {
+                MetricResult::Stats(intermediate_stats.finalize())
+            }
+        }
+    }
+}
+
+/// BucketEntry holds bucket aggregation result types.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+#[serde(untagged)]
+pub enum BucketResult {
+    /// This is the default entry for a bucket, which contains a key, count, and optionally
+    /// sub_aggregations.
+    Range {
+        /// The range buckets sorted by range.
+        buckets: Vec<RangeBucketEntry>,
+    },
+}
+
+impl From<IntermediateBucketResult> for BucketResult {
+    fn from(result: IntermediateBucketResult) -> Self {
+        match result {
+            IntermediateBucketResult::Range(range_map) => {
+                let mut buckets: Vec<RangeBucketEntry> = range_map
+                    .into_iter()
+                    .map(|(_, bucket)| bucket.into())
+                    .collect_vec();
+
+                buckets.sort_by(|a, b| {
+                    a.from
+                        .unwrap_or(f64::MIN)
+                        .partial_cmp(&b.from.unwrap_or(f64::MIN))
+                        .unwrap_or(Ordering::Equal)
+                });
+                BucketResult::Range { buckets }
+            }
+        }
+    }
+}
+
+/// This is the range entry for a bucket, which contains a key, count, and optionally
+/// sub_aggregations.
+///
+/// # JSON Format
+/// ```json
+/// {
+///   ...
+///     "my_ranges": {
+///       "buckets": [
+///         {
+///           "key": "*-10",
+///           "to": 10,
+///           "doc_count": 5
+///         },
+///         {
+///           "key": "10-20",
+///           "from": 10,
+///           "to": 20,
+///           "doc_count": 2
+///         },
+///         {
+///           "key": "20-*",
+///           "from": 20,
+///           "doc_count": 3
+///         }
+///       ]
+///    }
+///    ...
+/// }
+///  ```
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct RangeBucketEntry {
+    /// The identifier of the bucket.
+    pub key: Key,
+    /// Number of documents in the bucket.
+    pub doc_count: u64,
+    #[serde(flatten)]
+    /// sub-aggregations in this bucket.
+    pub sub_aggregation: AggregationResults,
+    /// The from range of the bucket. Equals f64::MIN when None.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub from: Option<f64>,
+    /// The to range of the bucket. Equals f64::MAX when None.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub to: Option<f64>,
+}
+
+impl From<IntermediateRangeBucketEntry> for RangeBucketEntry {
+    fn from(entry: IntermediateRangeBucketEntry) -> Self {
+        RangeBucketEntry {
+            key: entry.key,
+            doc_count: entry.doc_count,
+            sub_aggregation: entry.sub_aggregation.into(),
+            to: entry.to,
+            from: entry.from,
+        }
+    }
+}

src/aggregation/bucket/mod.rs

@@ -0,0 +1,13 @@
+//! Module for all bucket aggregations.
+//!
+//! BucketAggregations create buckets of documents
+//! [BucketAggregation](super::agg_req::BucketAggregation).
+//!
+//! Results of final buckets are [BucketResult](super::agg_result::BucketResult).
+//! Results of intermediate buckets are
+//! [IntermediateBucketResult](super::intermediate_agg_result::IntermediateBucketResult)
+
+mod range;
+
+pub(crate) use range::SegmentRangeCollector;
+pub use range::*;

src/aggregation/bucket/range.rs

@@ -0,0 +1,563 @@
+use std::ops::Range;
+
+use serde::{Deserialize, Serialize};
+
+use crate::aggregation::agg_req_with_accessor::{
+    AggregationsWithAccessor, BucketAggregationWithAccessor,
+};
+use crate::aggregation::intermediate_agg_result::IntermediateBucketResult;
+use crate::aggregation::segment_agg_result::{
+    SegmentAggregationResultsCollector, SegmentRangeBucketEntry,
+};
+use crate::aggregation::{f64_from_fastfield_u64, f64_to_fastfield_u64, Key};
+use crate::fastfield::FastFieldReader;
+use crate::schema::Type;
+use crate::{DocId, TantivyError};
+
+/// Provide user-defined buckets to aggregate on.
+/// Two special buckets will automatically be created to cover the whole range of values.
+/// The provided buckets have to be continous.
+/// During the aggregation, the values extracted from the fast_field `field` will be checked
+/// against each bucket range. Note that this aggregation includes the from value and excludes the
+/// to value for each range.
+///
+/// Result type is [BucketResult](crate::aggregation::agg_result::BucketResult) with
+/// [RangeBucketEntry](crate::aggregation::agg_result::RangeBucketEntry) on the
+/// AggregationCollector.
+///
+/// Result type is
+/// [crate::aggregation::intermediate_agg_result::IntermediateBucketResult] with
+/// [crate::aggregation::intermediate_agg_result::IntermediateRangeBucketEntry] on the
+/// DistributedAggregationCollector.
+///
+/// # Request JSON Format
+/// ```json
+/// {
+///     "range": {
+///         "field": "score",
+///         "ranges": [
+///             { "to": 3.0 },
+///             { "from": 3.0, "to": 7.0 },
+///             { "from": 7.0, "to": 20.0 }
+///             { "from": 20.0 }
+///         ]
+///     }
+///  }
+///  ```
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct RangeAggregation {
+    /// The field to aggregate on.
+    pub field: String,
+    /// Note that this aggregation includes the from value and excludes the to value for each
+    /// range. Extra buckets will be created until the first to, and last from, if necessary.
+    pub ranges: Vec<RangeAggregationRange>,
+}
+
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+/// The range for one range bucket.
+pub struct RangeAggregationRange {
+    /// The from range value, which is inclusive in the range.
+    /// None equals to an open ended interval.
+    #[serde(skip_serializing_if = "Option::is_none", default)]
+    pub from: Option<f64>,
+    /// The to range value, which is not inclusive in the range.
+    /// None equals to an open ended interval.
+    #[serde(skip_serializing_if = "Option::is_none", default)]
+    pub to: Option<f64>,
+}
+
+impl From<Range<f64>> for RangeAggregationRange {
+    fn from(range: Range<f64>) -> Self {
+        let from = if range.start == f64::MIN {
+            None
+        } else {
+            Some(range.start)
+        };
+        let to = if range.end == f64::MAX {
+            None
+        } else {
+            Some(range.end)
+        };
+        RangeAggregationRange { from, to }
+    }
+}
+
+#[derive(Clone, Debug, PartialEq)]
+pub(crate) struct SegmentRangeAndBucketEntry {
+    range: Range<u64>,
+    bucket: SegmentRangeBucketEntry,
+}
+
+/// The collector puts values from the fast field into the correct buckets and does a conversion to
+/// the correct datatype.
+#[derive(Clone, Debug, PartialEq)]
+pub struct SegmentRangeCollector {
+    /// The buckets containing the aggregation data.
+    buckets: Vec<SegmentRangeAndBucketEntry>,
+    field_type: Type,
+}
+
+impl SegmentRangeCollector {
+    pub fn into_intermediate_bucket_result(self) -> IntermediateBucketResult {
+        let field_type = self.field_type;
+
+        let buckets = self
+            .buckets
+            .into_iter()
+            .map(move |range_bucket| {
+                (
+                    range_to_string(&range_bucket.range, &field_type),
+                    range_bucket.bucket.into(),
+                )
+            })
+            .collect();
+
+        IntermediateBucketResult::Range(buckets)
+    }
+
+    pub(crate) fn from_req_and_validate(
+        req: &RangeAggregation,
+        sub_aggregation: &AggregationsWithAccessor,
+        field_type: Type,
+    ) -> crate::Result<Self> {
+        // The range input on the request is f64.
+        // We need to convert to u64 ranges, because we read the values as u64.
+        // The mapping from the conversion is monotonic so ordering is preserved.
+        let buckets = extend_validate_ranges(&req.ranges, &field_type)?
+            .iter()
+            .map(|range| {
+                let to = if range.end == u64::MAX {
+                    None
+                } else {
+                    Some(f64_from_fastfield_u64(range.end, &field_type))
+                };
+                let from = if range.start == u64::MIN {
+                    None
+                } else {
+                    Some(f64_from_fastfield_u64(range.start, &field_type))
+                };
+                let sub_aggregation = if sub_aggregation.is_empty() {
+                    None
+                } else {
+                    Some(SegmentAggregationResultsCollector::from_req_and_validate(
+                        sub_aggregation,
+                    )?)
+                };
+                Ok(SegmentRangeAndBucketEntry {
+                    range: range.clone(),
+                    bucket: SegmentRangeBucketEntry {
+                        key: range_to_key(range, &field_type),
+                        doc_count: 0,
+                        sub_aggregation,
+                        from,
+                        to,
+                    },
+                })
+            })
+            .collect::<crate::Result<_>>()?;
+
+        Ok(SegmentRangeCollector {
+            buckets,
+            field_type,
+        })
+    }
+
+    #[inline]
+    pub(crate) fn collect_block(
+        &mut self,
+        doc: &[DocId],
+        bucket_with_accessor: &BucketAggregationWithAccessor,
+        force_flush: bool,
+    ) {
+        let mut iter = doc.chunks_exact(4);
+        for docs in iter.by_ref() {
+            let val1 = bucket_with_accessor.accessor.get(docs[0]);
+            let val2 = bucket_with_accessor.accessor.get(docs[1]);
+            let val3 = bucket_with_accessor.accessor.get(docs[2]);
+            let val4 = bucket_with_accessor.accessor.get(docs[3]);
+            let bucket_pos1 = self.get_bucket_pos(val1);
+            let bucket_pos2 = self.get_bucket_pos(val2);
+            let bucket_pos3 = self.get_bucket_pos(val3);
+            let bucket_pos4 = self.get_bucket_pos(val4);
+
+            self.increment_bucket(bucket_pos1, docs[0], &bucket_with_accessor.sub_aggregation);
+            self.increment_bucket(bucket_pos2, docs[1], &bucket_with_accessor.sub_aggregation);
+            self.increment_bucket(bucket_pos3, docs[2], &bucket_with_accessor.sub_aggregation);
+            self.increment_bucket(bucket_pos4, docs[3], &bucket_with_accessor.sub_aggregation);
+        }
+        for doc in iter.remainder() {
+            let val = bucket_with_accessor.accessor.get(*doc);
+            let bucket_pos = self.get_bucket_pos(val);
+            self.increment_bucket(bucket_pos, *doc, &bucket_with_accessor.sub_aggregation);
+        }
+        if force_flush {
+            for bucket in &mut self.buckets {
+                if let Some(sub_aggregation) = &mut bucket.bucket.sub_aggregation {
+                    sub_aggregation
+                        .flush_staged_docs(&bucket_with_accessor.sub_aggregation, force_flush);
+                }
+            }
+        }
+    }
+
+    #[inline]
+    fn increment_bucket(
+        &mut self,
+        bucket_pos: usize,
+        doc: DocId,
+        bucket_with_accessor: &AggregationsWithAccessor,
+    ) {
+        let bucket = &mut self.buckets[bucket_pos];
+
+        bucket.bucket.doc_count += 1;
+        if let Some(sub_aggregation) = &mut bucket.bucket.sub_aggregation {
+            sub_aggregation.collect(doc, bucket_with_accessor);
+        }
+    }
+
+    #[inline]
+    fn get_bucket_pos(&self, val: u64) -> usize {
+        let pos = self
+            .buckets
+            .binary_search_by_key(&val, |probe| probe.range.start)
+            .unwrap_or_else(|pos| pos - 1);
+        debug_assert!(self.buckets[pos].range.contains(&val));
+        pos
+    }
+}
+
+/// Converts the user provided f64 range value to fast field value space.
+///
+/// Internally fast field values are always stored as u64.
+/// If the fast field has u64 [1,2,5], these values are stored as is in the fast field.
+/// A fast field with f64 [1.0, 2.0, 5.0] is converted to u64 space, using a
+/// monotonic mapping function, so the order is preserved.
+///
+/// Consequently, a f64 user range 1.0..3.0 needs to be converted to fast field value space using
+/// the same monotonic mapping function, so that the provided ranges contain the u64 values in the
+/// fast field.
+/// The alternative would be that every value read would be converted to the f64 range, but that is
+/// more computational expensive when many documents are hit.
+fn to_u64_range(range: &RangeAggregationRange, field_type: &Type) -> crate::Result<Range<u64>> {
+    let start = if let Some(from) = range.from {
+        f64_to_fastfield_u64(from, field_type)
+            .ok_or_else(|| TantivyError::InvalidArgument("invalid field type".to_string()))?
+    } else {
+        u64::MIN
+    };
+
+    let end = if let Some(to) = range.to {
+        f64_to_fastfield_u64(to, field_type)
+            .ok_or_else(|| TantivyError::InvalidArgument("invalid field type".to_string()))?
+    } else {
+        u64::MAX
+    };
+
+    Ok(start..end)
+}
+
+/// Extends the provided buckets to contain the whole value range, by inserting buckets at the
+/// beginning and end.
+fn extend_validate_ranges(
+    buckets: &[RangeAggregationRange],
+    field_type: &Type,
+) -> crate::Result<Vec<Range<u64>>> {
+    let mut converted_buckets = buckets
+        .iter()
+        .map(|range| to_u64_range(range, field_type))
+        .collect::<crate::Result<Vec<_>>>()?;
+
+    converted_buckets.sort_by_key(|bucket| bucket.start);
+    if converted_buckets[0].start != u64::MIN {
+        converted_buckets.insert(0, u64::MIN..converted_buckets[0].start);
+    }
+
+    if converted_buckets[converted_buckets.len() - 1].end != u64::MAX {
+        converted_buckets.push(converted_buckets[converted_buckets.len() - 1].end..u64::MAX);
+    }
+
+    // fill up holes in the ranges
+    let find_hole = |converted_buckets: &[Range<u64>]| {
+        for (pos, ranges) in converted_buckets.windows(2).enumerate() {
+            if ranges[0].end > ranges[1].start {
+                return Err(TantivyError::InvalidArgument(format!(
+                    "Overlapping ranges not supported range {:?}, range+1 {:?}",
+                    ranges[0], ranges[1]
+                )));
+            }
+            if ranges[0].end != ranges[1].start {
+                return Ok(Some(pos));
+            }
+        }
+        Ok(None)
+    };
+
+    while let Some(hole_pos) = find_hole(&converted_buckets)? {
+        let new_range = converted_buckets[hole_pos].end..converted_buckets[hole_pos + 1].start;
+        converted_buckets.insert(hole_pos + 1, new_range);
+    }
+
+    Ok(converted_buckets)
+}
+
+pub(crate) fn range_to_string(range: &Range<u64>, field_type: &Type) -> String {
+    // is_start is there for malformed requests, e.g. ig the user passes the range u64::MIN..0.0,
+    // it should be rendererd as "*-0" and not "*-*"
+    let to_str = |val: u64, is_start: bool| {
+        if (is_start && val == u64::MIN) || (!is_start && val == u64::MAX) {
+            "*".to_string()
+        } else {
+            f64_from_fastfield_u64(val, field_type).to_string()
+        }
+    };
+
+    format!("{}-{}", to_str(range.start, true), to_str(range.end, false))
+}
+
+pub(crate) fn range_to_key(range: &Range<u64>, field_type: &Type) -> Key {
+    Key::Str(range_to_string(range, field_type))
+}
+
+#[cfg(test)]
+mod tests {
+
+    use serde_json::Value;
+
+    use super::*;
+    use crate::aggregation::agg_req::{
+        Aggregation, Aggregations, BucketAggregation, BucketAggregationType,
+    };
+    use crate::aggregation::tests::get_test_index_with_num_docs;
+    use crate::aggregation::AggregationCollector;
+    use crate::fastfield::FastValue;
+    use crate::query::AllQuery;
+
+    pub fn get_collector_from_ranges(
+        ranges: Vec<RangeAggregationRange>,
+        field_type: Type,
+    ) -> SegmentRangeCollector {
+        let req = RangeAggregation {
+            field: "dummy".to_string(),
+            ranges,
+        };
+
+        SegmentRangeCollector::from_req_and_validate(&req, &Default::default(), field_type).unwrap()
+    }
+
+    #[test]
+    fn range_fraction_test() -> crate::Result<()> {
+        let index = get_test_index_with_num_docs(false, 100)?;
+
+        let agg_req: Aggregations = vec![(
+            "range".to_string(),
+            Aggregation::Bucket(BucketAggregation {
+                bucket_agg: BucketAggregationType::Range(RangeAggregation {
+                    field: "fraction_f64".to_string(),
+                    ranges: vec![(0f64..0.1f64).into(), (0.1f64..0.2f64).into()],
+                }),
+                sub_aggregation: Default::default(),
+            }),
+        )]
+        .into_iter()
+        .collect();
+
+        let collector = AggregationCollector::from_aggs(agg_req);
+
+        let reader = index.reader()?;
+        let searcher = reader.searcher();
+        let agg_res = searcher.search(&AllQuery, &collector).unwrap();
+
+        let res: Value = serde_json::from_str(&serde_json::to_string(&agg_res)?)?;
+
+        assert_eq!(res["range"]["buckets"][0]["key"], "*-0");
+        assert_eq!(res["range"]["buckets"][0]["doc_count"], 0);
+        assert_eq!(res["range"]["buckets"][1]["key"], "0-0.1");
+        assert_eq!(res["range"]["buckets"][1]["doc_count"], 10);
+        assert_eq!(res["range"]["buckets"][2]["key"], "0.1-0.2");
+        assert_eq!(res["range"]["buckets"][2]["doc_count"], 10);
+        assert_eq!(res["range"]["buckets"][3]["key"], "0.2-*");
+        assert_eq!(res["range"]["buckets"][3]["doc_count"], 80);
+
+        Ok(())
+    }
+
+    #[test]
+    fn bucket_test_extend_range_hole() {
+        let buckets = vec![(10f64..20f64).into(), (30f64..40f64).into()];
+        let collector = get_collector_from_ranges(buckets, Type::F64);
+
+        let buckets = collector.buckets;
+        assert_eq!(buckets[0].range.start, u64::MIN);
+        assert_eq!(buckets[0].range.end, 10f64.to_u64());
+        assert_eq!(buckets[1].range.start, 10f64.to_u64());
+        assert_eq!(buckets[1].range.end, 20f64.to_u64());
+        // Added bucket to fill hole
+        assert_eq!(buckets[2].range.start, 20f64.to_u64());
+        assert_eq!(buckets[2].range.end, 30f64.to_u64());
+        assert_eq!(buckets[3].range.start, 30f64.to_u64());
+        assert_eq!(buckets[3].range.end, 40f64.to_u64());
+    }
+
+    #[test]
+    fn bucket_test_range_conversion_special_case() {
+        // the monotonic conversion between f64 and u64, does not map f64::MIN.to_u64() ==
+        // u64::MIN, but the into trait converts f64::MIN/MAX to None
+        let buckets = vec![
+            (f64::MIN..10f64).into(),
+            (10f64..20f64).into(),
+            (20f64..f64::MAX).into(),
+        ];
+        let collector = get_collector_from_ranges(buckets, Type::F64);
+
+        let buckets = collector.buckets;
+        assert_eq!(buckets[0].range.start, u64::MIN);
+        assert_eq!(buckets[0].range.end, 10f64.to_u64());
+        assert_eq!(buckets[1].range.start, 10f64.to_u64());
+        assert_eq!(buckets[1].range.end, 20f64.to_u64());
+        assert_eq!(buckets[2].range.start, 20f64.to_u64());
+        assert_eq!(buckets[2].range.end, u64::MAX);
+        assert_eq!(buckets.len(), 3);
+    }
+
+    #[test]
+    fn bucket_range_test_negative_vals() {
+        let buckets = vec![(-10f64..-1f64).into()];
+        let collector = get_collector_from_ranges(buckets, Type::F64);
+
+        let buckets = collector.buckets;
+        assert_eq!(&buckets[0].bucket.key.to_string(), "*--10");
+        assert_eq!(&buckets[buckets.len() - 1].bucket.key.to_string(), "-1-*");
+    }
+    #[test]
+    fn bucket_range_test_positive_vals() {
+        let buckets = vec![(0f64..10f64).into()];
+        let collector = get_collector_from_ranges(buckets, Type::F64);
+
+        let buckets = collector.buckets;
+        assert_eq!(&buckets[0].bucket.key.to_string(), "*-0");
+        assert_eq!(&buckets[buckets.len() - 1].bucket.key.to_string(), "10-*");
+    }
+
+    #[test]
+    fn range_binary_search_test_u64() {
+        let check_ranges = |ranges: Vec<RangeAggregationRange>| {
+            let collector = get_collector_from_ranges(ranges, Type::U64);
+            let search = |val: u64| collector.get_bucket_pos(val);
+
+            assert_eq!(search(u64::MIN), 0);
+            assert_eq!(search(9), 0);
+            assert_eq!(search(10), 1);
+            assert_eq!(search(11), 1);
+            assert_eq!(search(99), 1);
+            assert_eq!(search(100), 2);
+            assert_eq!(search(u64::MAX - 1), 2); // Since the end range is never included, the max
+                                                 // value
+        };
+
+        let ranges = vec![(10.0..100.0).into()];
+        check_ranges(ranges);
+
+        let ranges = vec![
+            RangeAggregationRange {
+                to: Some(10.0),
+                from: None,
+            },
+            (10.0..100.0).into(),
+        ];
+        check_ranges(ranges);
+
+        let ranges = vec![
+            RangeAggregationRange {
+                to: Some(10.0),
+                from: None,
+            },
+            (10.0..100.0).into(),
+            RangeAggregationRange {
+                to: None,
+                from: Some(100.0),
+            },
+        ];
+        check_ranges(ranges);
+    }
+
+    #[test]
+    fn range_binary_search_test_f64() {
+        let ranges = vec![
+            //(f64::MIN..10.0).into(),
+            (10.0..100.0).into(),
+            //(100.0..f64::MAX).into(),
+        ];
+
+        let collector = get_collector_from_ranges(ranges, Type::F64);
+        let search = |val: u64| collector.get_bucket_pos(val);
+
+        assert_eq!(search(u64::MIN), 0);
+        assert_eq!(search(9f64.to_u64()), 0);
+        assert_eq!(search(10f64.to_u64()), 1);
+        assert_eq!(search(11f64.to_u64()), 1);
+        assert_eq!(search(99f64.to_u64()), 1);
+        assert_eq!(search(100f64.to_u64()), 2);
+        assert_eq!(search(u64::MAX - 1), 2); // Since the end range is never included,
+                                             // the max value
+    }
+}
+
+#[cfg(all(test, feature = "unstable"))]
+mod bench {
+
+    use itertools::Itertools;
+    use rand::seq::SliceRandom;
+    use rand::thread_rng;
+
+    use super::*;
+    use crate::aggregation::bucket::range::tests::get_collector_from_ranges;
+
+    const TOTAL_DOCS: u64 = 1_000_000u64;
+    const NUM_DOCS: u64 = 50_000u64;
+
+    fn get_collector_with_buckets(num_buckets: u64, num_docs: u64) -> SegmentRangeCollector {
+        let bucket_size = num_docs / num_buckets;
+        let mut buckets: Vec<RangeAggregationRange> = vec![];
+        for i in 0..num_buckets {
+            let bucket_start = (i * bucket_size) as f64;
+            buckets.push((bucket_start..bucket_start + bucket_size as f64).into())
+        }
+
+        get_collector_from_ranges(buckets, Type::U64)
+    }
+
+    fn get_rand_docs(total_docs: u64, num_docs_returned: u64) -> Vec<u64> {
+        let mut rng = thread_rng();
+
+        let all_docs = (0..total_docs - 1).collect_vec();
+        let mut vals = all_docs
+            .as_slice()
+            .choose_multiple(&mut rng, num_docs_returned as usize)
+            .cloned()
+            .collect_vec();
+        vals.sort();
+        vals
+    }
+
+    fn bench_range_binary_search(b: &mut test::Bencher, num_buckets: u64) {
+        let collector = get_collector_with_buckets(num_buckets, TOTAL_DOCS);
+        let vals = get_rand_docs(TOTAL_DOCS, NUM_DOCS);
+        b.iter(|| {
+            let mut bucket_pos = 0;
+            for val in &vals {
+                bucket_pos = collector.get_bucket_pos(*val);
+            }
+            bucket_pos
+        })
+    }
+
+    #[bench]
+    fn bench_range_100_buckets(b: &mut test::Bencher) {
+        bench_range_binary_search(b, 100)
+    }
+
+    #[bench]
+    fn bench_range_10_buckets(b: &mut test::Bencher) {
+        bench_range_binary_search(b, 10)
+    }
+}

src/aggregation/collector.rs

@@ -0,0 +1,149 @@
+use super::agg_req::Aggregations;
+use super::agg_req_with_accessor::AggregationsWithAccessor;
+use super::agg_result::AggregationResults;
+use super::intermediate_agg_result::IntermediateAggregationResults;
+use super::segment_agg_result::SegmentAggregationResultsCollector;
+use crate::aggregation::agg_req_with_accessor::get_aggs_with_accessor_and_validate;
+use crate::collector::{Collector, SegmentCollector};
+use crate::{SegmentReader, TantivyError};
+
+/// Collector for aggregations.
+///
+/// The collector collects all aggregations by the underlying aggregation request.
+pub struct AggregationCollector {
+    agg: Aggregations,
+}
+
+impl AggregationCollector {
+    /// Create collector from aggregation request.
+    pub fn from_aggs(agg: Aggregations) -> Self {
+        Self { agg }
+    }
+}
+
+/// Collector for distributed aggregations.
+///
+/// The collector collects all aggregations by the underlying aggregation request.
+///
+/// # Purpose
+/// AggregationCollector returns `IntermediateAggregationResults` and not the final
+/// `AggregationResults`, so that results from differenct indices can be merged and then converted
+/// into the final `AggregationResults` via the `into()` method.
+pub struct DistributedAggregationCollector {
+    agg: Aggregations,
+}
+
+impl DistributedAggregationCollector {
+    /// Create collector from aggregation request.
+    pub fn from_aggs(agg: Aggregations) -> Self {
+        Self { agg }
+    }
+}
+
+impl Collector for DistributedAggregationCollector {
+    type Fruit = IntermediateAggregationResults;
+
+    type Child = AggregationSegmentCollector;
+
+    fn for_segment(
+        &self,
+        _segment_local_id: crate::SegmentOrdinal,
+        reader: &crate::SegmentReader,
+    ) -> crate::Result<Self::Child> {
+        AggregationSegmentCollector::from_agg_req_and_reader(&self.agg, reader)
+    }
+
+    fn requires_scoring(&self) -> bool {
+        false
+    }
+
+    fn merge_fruits(
+        &self,
+        segment_fruits: Vec<<Self::Child as SegmentCollector>::Fruit>,
+    ) -> crate::Result<Self::Fruit> {
+        merge_fruits(segment_fruits)
+    }
+}
+
+impl Collector for AggregationCollector {
+    type Fruit = AggregationResults;
+
+    type Child = AggregationSegmentCollector;
+
+    fn for_segment(
+        &self,
+        _segment_local_id: crate::SegmentOrdinal,
+        reader: &crate::SegmentReader,
+    ) -> crate::Result<Self::Child> {
+        let aggs_with_accessor = get_aggs_with_accessor_and_validate(&self.agg, reader)?;
+        let result =
+            SegmentAggregationResultsCollector::from_req_and_validate(&aggs_with_accessor)?;
+        Ok(AggregationSegmentCollector {
+            aggs: aggs_with_accessor,
+            result,
+        })
+    }
+
+    fn requires_scoring(&self) -> bool {
+        false
+    }
+
+    fn merge_fruits(
+        &self,
+        segment_fruits: Vec<<Self::Child as SegmentCollector>::Fruit>,
+    ) -> crate::Result<Self::Fruit> {
+        merge_fruits(segment_fruits).map(|res| res.into())
+    }
+}
+
+fn merge_fruits(
+    mut segment_fruits: Vec<IntermediateAggregationResults>,
+) -> crate::Result<IntermediateAggregationResults> {
+    if let Some(mut fruit) = segment_fruits.pop() {
+        for next_fruit in segment_fruits {
+            fruit.merge_fruits(&next_fruit);
+        }
+        Ok(fruit)
+    } else {
+        Err(TantivyError::InvalidArgument(
+            "no fruits provided in merge_fruits".to_string(),
+        ))
+    }
+}
+
+/// AggregationSegmentCollector does the aggregation collection on a segment.
+pub struct AggregationSegmentCollector {
+    aggs: AggregationsWithAccessor,
+    result: SegmentAggregationResultsCollector,
+}
+
+impl AggregationSegmentCollector {
+    /// Creates an AggregationSegmentCollector from an [Aggregations] request and a segment reader.
+    /// Also includes validation, e.g. checking field types and existence.
+    pub fn from_agg_req_and_reader(
+        agg: &Aggregations,
+        reader: &SegmentReader,
+    ) -> crate::Result<Self> {
+        let aggs_with_accessor = get_aggs_with_accessor_and_validate(agg, reader)?;
+        let result =
+            SegmentAggregationResultsCollector::from_req_and_validate(&aggs_with_accessor)?;
+        Ok(AggregationSegmentCollector {
+            aggs: aggs_with_accessor,
+            result,
+        })
+    }
+}
+
+impl SegmentCollector for AggregationSegmentCollector {
+    type Fruit = IntermediateAggregationResults;
+
+    #[inline]
+    fn collect(&mut self, doc: crate::DocId, _score: crate::Score) {
+        self.result.collect(doc, &self.aggs);
+    }
+
+    fn harvest(mut self) -> Self::Fruit {
+        self.result.flush_staged_docs(&self.aggs, true);
+        self.result.into()
+    }
+}

src/aggregation/intermediate_agg_result.rs

@@ -0,0 +1,304 @@
+//! Contains the intermediate aggregation tree, that can be merged.
+//! Intermediate aggregation results can be used to merge results between segments or between
+//! indices.
+
+use std::collections::HashMap;
+
+use serde::{Deserialize, Serialize};
+
+use super::metric::{IntermediateAverage, IntermediateStats};
+use super::segment_agg_result::{
+    SegmentAggregationResultsCollector, SegmentBucketResultCollector, SegmentMetricResultCollector,
+    SegmentRangeBucketEntry,
+};
+use super::{Key, SerializedKey, VecWithNames};
+
+/// Contains the intermediate aggregation result, which is optimized to be merged with other
+/// intermediate results.
+#[derive(Default, Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct IntermediateAggregationResults(pub(crate) VecWithNames<IntermediateAggregationResult>);
+
+impl From<SegmentAggregationResultsCollector> for IntermediateAggregationResults {
+    fn from(tree: SegmentAggregationResultsCollector) -> Self {
+        let mut data = vec![];
+        for (key, bucket) in tree.buckets.into_iter() {
+            data.push((key, IntermediateAggregationResult::Bucket(bucket.into())));
+        }
+        for (key, metric) in tree.metrics.into_iter() {
+            data.push((key, IntermediateAggregationResult::Metric(metric.into())));
+        }
+        Self(VecWithNames::from_entries(data))
+    }
+}
+
+impl IntermediateAggregationResults {
+    /// Merge an other intermediate aggregation result into this result.
+    ///
+    /// The order of the values need to be the same on both results. This is ensured when the same
+    /// (key values) are present on the underlying VecWithNames struct.
+    pub fn merge_fruits(&mut self, other: &IntermediateAggregationResults) {
+        for (tree_left, tree_right) in self.0.values_mut().zip(other.0.values()) {
+            tree_left.merge_fruits(tree_right);
+        }
+    }
+}
+
+/// An aggregation is either a bucket or a metric.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub enum IntermediateAggregationResult {
+    /// Bucket variant
+    Bucket(IntermediateBucketResult),
+    /// Metric variant
+    Metric(IntermediateMetricResult),
+}
+
+impl IntermediateAggregationResult {
+    fn merge_fruits(&mut self, other: &IntermediateAggregationResult) {
+        match (self, other) {
+            (
+                IntermediateAggregationResult::Bucket(res_left),
+                IntermediateAggregationResult::Bucket(res_right),
+            ) => {
+                res_left.merge_fruits(res_right);
+            }
+            (
+                IntermediateAggregationResult::Metric(res_left),
+                IntermediateAggregationResult::Metric(res_right),
+            ) => {
+                res_left.merge_fruits(res_right);
+            }
+            _ => {
+                panic!("incompatible types in aggregation tree on merge fruits");
+            }
+        }
+    }
+}
+
+/// Holds the intermediate data for metric results
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub enum IntermediateMetricResult {
+    /// Average containing intermediate average data result
+    Average(IntermediateAverage),
+    /// AverageData variant
+    Stats(IntermediateStats),
+}
+
+impl From<SegmentMetricResultCollector> for IntermediateMetricResult {
+    fn from(tree: SegmentMetricResultCollector) -> Self {
+        match tree {
+            SegmentMetricResultCollector::Average(collector) => {
+                IntermediateMetricResult::Average(IntermediateAverage::from_collector(collector))
+            }
+            SegmentMetricResultCollector::Stats(collector) => {
+                IntermediateMetricResult::Stats(collector.stats)
+            }
+        }
+    }
+}
+
+impl IntermediateMetricResult {
+    fn merge_fruits(&mut self, other: &IntermediateMetricResult) {
+        match (self, other) {
+            (
+                IntermediateMetricResult::Average(avg_data_left),
+                IntermediateMetricResult::Average(avg_data_right),
+            ) => {
+                avg_data_left.merge_fruits(avg_data_right);
+            }
+            (
+                IntermediateMetricResult::Stats(stats_left),
+                IntermediateMetricResult::Stats(stats_right),
+            ) => {
+                stats_left.merge_fruits(stats_right);
+            }
+            _ => {
+                panic!("incompatible fruit types in tree {:?}", other);
+            }
+        }
+    }
+}
+
+/// The intermediate bucket results. Internally they can be easily merged via the keys of the
+/// buckets.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub enum IntermediateBucketResult {
+    /// This is the range entry for a bucket, which contains a key, count, from, to, and optionally
+    /// sub_aggregations.
+    Range(HashMap<SerializedKey, IntermediateRangeBucketEntry>),
+}
+
+impl From<SegmentBucketResultCollector> for IntermediateBucketResult {
+    fn from(collector: SegmentBucketResultCollector) -> Self {
+        match collector {
+            SegmentBucketResultCollector::Range(range) => range.into_intermediate_bucket_result(),
+        }
+    }
+}
+
+impl IntermediateBucketResult {
+    fn merge_fruits(&mut self, other: &IntermediateBucketResult) {
+        match (self, other) {
+            (
+                IntermediateBucketResult::Range(entries_left),
+                IntermediateBucketResult::Range(entries_right),
+            ) => {
+                for (name, entry_left) in entries_left.iter_mut() {
+                    if let Some(entry_right) = entries_right.get(name) {
+                        entry_left.merge_fruits(entry_right);
+                    }
+                }
+
+                for (key, res) in entries_right.iter() {
+                    if !entries_left.contains_key(key) {
+                        entries_left.insert(key.clone(), res.clone());
+                    }
+                }
+            }
+        }
+    }
+}
+
+/// This is the range entry for a bucket, which contains a key, count, and optionally
+/// sub_aggregations.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct IntermediateRangeBucketEntry {
+    /// The unique the bucket is identified.
+    pub key: Key,
+    /// The number of documents in the bucket.
+    pub doc_count: u64,
+    pub(crate) values: Option<Vec<u64>>,
+    /// The sub_aggregation in this bucket.
+    pub sub_aggregation: IntermediateAggregationResults,
+    /// The from range of the bucket. Equals f64::MIN when None.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub from: Option<f64>,
+    /// The to range of the bucket. Equals f64::MAX when None.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub to: Option<f64>,
+}
+
+impl From<SegmentRangeBucketEntry> for IntermediateRangeBucketEntry {
+    fn from(entry: SegmentRangeBucketEntry) -> Self {
+        let sub_aggregation = if let Some(sub_aggregation) = entry.sub_aggregation {
+            sub_aggregation.into()
+        } else {
+            Default::default()
+        };
+        // let sub_aggregation = entry.sub_aggregation.into();
+
+        IntermediateRangeBucketEntry {
+            key: entry.key,
+            doc_count: entry.doc_count,
+            values: None,
+            sub_aggregation,
+            to: entry.to,
+            from: entry.from,
+        }
+    }
+}
+
+impl IntermediateRangeBucketEntry {
+    fn merge_fruits(&mut self, other: &IntermediateRangeBucketEntry) {
+        self.doc_count += other.doc_count;
+        self.sub_aggregation.merge_fruits(&other.sub_aggregation);
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use pretty_assertions::assert_eq;
+
+    use super::*;
+
+    fn get_sub_test_tree(data: &[(String, u64)]) -> IntermediateAggregationResults {
+        let mut map = HashMap::new();
+        let mut buckets = HashMap::new();
+        for (key, doc_count) in data {
+            buckets.insert(
+                key.to_string(),
+                IntermediateRangeBucketEntry {
+                    key: Key::Str(key.to_string()),
+                    doc_count: *doc_count,
+                    values: None,
+                    sub_aggregation: Default::default(),
+                    from: None,
+                    to: None,
+                },
+            );
+        }
+        map.insert(
+            "my_agg_level2".to_string(),
+            IntermediateAggregationResult::Bucket(IntermediateBucketResult::Range(buckets)),
+        );
+        IntermediateAggregationResults(VecWithNames::from_entries(map.into_iter().collect()))
+    }
+
+    fn get_test_tree(data: &[(String, u64, String, u64)]) -> IntermediateAggregationResults {
+        let mut map = HashMap::new();
+        let mut buckets = HashMap::new();
+        for (key, doc_count, sub_aggregation_key, sub_aggregation_count) in data {
+            buckets.insert(
+                key.to_string(),
+                IntermediateRangeBucketEntry {
+                    key: Key::Str(key.to_string()),
+                    doc_count: *doc_count,
+                    values: None,
+                    from: None,
+                    to: None,
+                    sub_aggregation: get_sub_test_tree(&[(
+                        sub_aggregation_key.to_string(),
+                        *sub_aggregation_count,
+                    )]),
+                },
+            );
+        }
+        map.insert(
+            "my_agg_level1".to_string(),
+            IntermediateAggregationResult::Bucket(IntermediateBucketResult::Range(buckets)),
+        );
+        IntermediateAggregationResults(VecWithNames::from_entries(map.into_iter().collect()))
+    }
+
+    #[test]
+    fn test_merge_fruits_tree_1() {
+        let mut tree_left = get_test_tree(&[
+            ("red".to_string(), 50, "1900".to_string(), 25),
+            ("blue".to_string(), 30, "1900".to_string(), 30),
+        ]);
+        let tree_right = get_test_tree(&[
+            ("red".to_string(), 60, "1900".to_string(), 30),
+            ("blue".to_string(), 25, "1900".to_string(), 50),
+        ]);
+
+        tree_left.merge_fruits(&tree_right);
+
+        let tree_expected = get_test_tree(&[
+            ("red".to_string(), 110, "1900".to_string(), 55),
+            ("blue".to_string(), 55, "1900".to_string(), 80),
+        ]);
+
+        assert_eq!(tree_left, tree_expected);
+    }
+
+    #[test]
+    fn test_merge_fruits_tree_2() {
+        let mut tree_left = get_test_tree(&[
+            ("red".to_string(), 50, "1900".to_string(), 25),
+            ("blue".to_string(), 30, "1900".to_string(), 30),
+        ]);
+        let tree_right = get_test_tree(&[
+            ("red".to_string(), 60, "1900".to_string(), 30),
+            ("green".to_string(), 25, "1900".to_string(), 50),
+        ]);
+
+        tree_left.merge_fruits(&tree_right);
+
+        let tree_expected = get_test_tree(&[
+            ("red".to_string(), 110, "1900".to_string(), 55),
+            ("blue".to_string(), 30, "1900".to_string(), 30),
+            ("green".to_string(), 25, "1900".to_string(), 50),
+        ]);
+
+        assert_eq!(tree_left, tree_expected);
+    }
+}

src/aggregation/metric/average.rs

@@ -0,0 +1,114 @@
+use std::fmt::Debug;
+
+use serde::{Deserialize, Serialize};
+
+use crate::aggregation::f64_from_fastfield_u64;
+use crate::fastfield::{DynamicFastFieldReader, FastFieldReader};
+use crate::schema::Type;
+use crate::DocId;
+
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+/// A single-value metric aggregation that computes the average of numeric values that are
+/// extracted from the aggregated documents.
+/// Supported field types are u64, i64, and f64.
+/// See [super::SingleMetricResult] for return value.
+///
+/// # JSON Format
+/// ```json
+/// {
+///     "avg": {
+///         "field": "score",
+///     }
+///  }
+///  ```
+pub struct AverageAggregation {
+    /// The field name to compute the stats on.
+    pub field: String,
+}
+impl AverageAggregation {
+    /// Create new AverageAggregation from a field.
+    pub fn from_field_name(field_name: String) -> Self {
+        AverageAggregation { field: field_name }
+    }
+    /// Return the field name.
+    pub fn field_name(&self) -> &str {
+        &self.field
+    }
+}
+
+#[derive(Clone, PartialEq)]
+pub(crate) struct SegmentAverageCollector {
+    pub data: IntermediateAverage,
+    field_type: Type,
+}
+
+impl Debug for SegmentAverageCollector {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("AverageCollector")
+            .field("data", &self.data)
+            .finish()
+    }
+}
+
+impl SegmentAverageCollector {
+    pub fn from_req(field_type: Type) -> Self {
+        Self {
+            field_type,
+            data: Default::default(),
+        }
+    }
+    pub(crate) fn collect_block(&mut self, doc: &[DocId], field: &DynamicFastFieldReader<u64>) {
+        let mut iter = doc.chunks_exact(4);
+        for docs in iter.by_ref() {
+            let val1 = field.get(docs[0]);
+            let val2 = field.get(docs[1]);
+            let val3 = field.get(docs[2]);
+            let val4 = field.get(docs[3]);
+            let val1 = f64_from_fastfield_u64(val1, &self.field_type);
+            let val2 = f64_from_fastfield_u64(val2, &self.field_type);
+            let val3 = f64_from_fastfield_u64(val3, &self.field_type);
+            let val4 = f64_from_fastfield_u64(val4, &self.field_type);
+            self.data.collect(val1);
+            self.data.collect(val2);
+            self.data.collect(val3);
+            self.data.collect(val4);
+        }
+        for doc in iter.remainder() {
+            let val = field.get(*doc);
+            let val = f64_from_fastfield_u64(val, &self.field_type);
+            self.data.collect(val);
+        }
+    }
+}
+
+/// Contains mergeable version of average data.
+#[derive(Default, Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct IntermediateAverage {
+    pub(crate) sum: f64,
+    pub(crate) doc_count: u64,
+}
+
+impl IntermediateAverage {
+    pub(crate) fn from_collector(collector: SegmentAverageCollector) -> Self {
+        collector.data
+    }
+
+    /// Merge average data into this instance.
+    pub fn merge_fruits(&mut self, other: &IntermediateAverage) {
+        self.sum += other.sum;
+        self.doc_count += other.doc_count;
+    }
+    /// compute final result
+    pub fn finalize(&self) -> Option<f64> {
+        if self.doc_count == 0 {
+            None
+        } else {
+            Some(self.sum / self.doc_count as f64)
+        }
+    }
+    #[inline]
+    fn collect(&mut self, val: f64) {
+        self.doc_count += 1;
+        self.sum += val;
+    }
+}

src/aggregation/metric/mod.rs

@@ -0,0 +1,30 @@
+//! Module for all metric aggregations.
+//!
+//! The aggregations in this family compute metrics, see [super::agg_req::MetricAggregation] for
+//! details.
+mod average;
+mod stats;
+pub use average::*;
+use serde::{Deserialize, Serialize};
+pub use stats::*;
+
+/// Single-metric aggregations use this common result structure.
+///
+/// Main reason to wrap it in value is to match elasticsearch output structure.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct SingleMetricResult {
+    /// The value of the single value metric.
+    pub value: Option<f64>,
+}
+
+impl From<f64> for SingleMetricResult {
+    fn from(value: f64) -> Self {
+        Self { value: Some(value) }
+    }
+}
+
+impl From<Option<f64>> for SingleMetricResult {
+    fn from(value: Option<f64>) -> Self {
+        Self { value }
+    }
+}

src/aggregation/metric/stats.rs

@@ -0,0 +1,314 @@
+use serde::{Deserialize, Serialize};
+
+use crate::aggregation::f64_from_fastfield_u64;
+use crate::fastfield::{DynamicFastFieldReader, FastFieldReader};
+use crate::schema::Type;
+use crate::DocId;
+
+/// A multi-value metric aggregation that computes stats of numeric values that are
+/// extracted from the aggregated documents.
+/// Supported field types are u64, i64, and f64.
+/// See [Stats] for returned statistics.
+///
+/// # JSON Format
+/// ```json
+/// {
+///     "stats": {
+///         "field": "score",
+///     }
+///  }
+///  ```
+
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct StatsAggregation {
+    /// The field name to compute the stats on.
+    pub field: String,
+}
+
+impl StatsAggregation {
+    /// Create new StatsAggregation from a field.
+    pub fn from_field_name(field_name: String) -> Self {
+        StatsAggregation { field: field_name }
+    }
+    /// Return the field name.
+    pub fn field_name(&self) -> &str {
+        &self.field
+    }
+}
+
+/// Stats contains a collection of statistics.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct Stats {
+    /// The number of documents.
+    pub count: usize,
+    /// The sum of the fast field values.
+    pub sum: f64,
+    /// The standard deviation of the fast field values. None for count == 0.
+    pub standard_deviation: Option<f64>,
+    /// The min value of the fast field values.
+    pub min: Option<f64>,
+    /// The max value of the fast field values.
+    pub max: Option<f64>,
+    /// The average of the values. None for count == 0.
+    pub avg: Option<f64>,
+}
+
+/// IntermediateStats contains the mergeable version for stats.
+#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
+pub struct IntermediateStats {
+    count: usize,
+    sum: f64,
+    squared_sum: f64,
+    min: f64,
+    max: f64,
+}
+
+impl IntermediateStats {
+    fn new() -> Self {
+        Self {
+            count: 0,
+            sum: 0.0,
+            squared_sum: 0.0,
+            min: f64::MAX,
+            max: f64::MIN,
+        }
+    }
+
+    pub(crate) fn avg(&self) -> Option<f64> {
+        if self.count == 0 {
+            None
+        } else {
+            Some(self.sum / (self.count as f64))
+        }
+    }
+
+    fn square_mean(&self) -> f64 {
+        self.squared_sum / (self.count as f64)
+    }
+
+    pub(crate) fn standard_deviation(&self) -> Option<f64> {
+        self.avg()
+            .map(|average| (self.square_mean() - average * average).sqrt())
+    }
+
+    /// Merge data from other stats into this instance.
+    pub fn merge_fruits(&mut self, other: &IntermediateStats) {
+        self.count += other.count;
+        self.sum += other.sum;
+        self.squared_sum += other.squared_sum;
+        self.min = self.min.min(other.min);
+        self.max = self.max.max(other.max);
+    }
+
+    /// compute final resultimprove_docs
+    pub fn finalize(&self) -> Stats {
+        let min = if self.count == 0 {
+            None
+        } else {
+            Some(self.min)
+        };
+        let max = if self.count == 0 {
+            None
+        } else {
+            Some(self.max)
+        };
+        Stats {
+            count: self.count,
+            sum: self.sum,
+            standard_deviation: self.standard_deviation(),
+            min,
+            max,
+            avg: self.avg(),
+        }
+    }
+
+    #[inline]
+    fn collect(&mut self, value: f64) {
+        self.count += 1;
+        self.sum += value;
+        self.squared_sum += value * value;
+        self.min = self.min.min(value);
+        self.max = self.max.max(value);
+    }
+}
+
+#[derive(Clone, Debug, PartialEq)]
+pub(crate) struct SegmentStatsCollector {
+    pub(crate) stats: IntermediateStats,
+    field_type: Type,
+}
+
+impl SegmentStatsCollector {
+    pub fn from_req(field_type: Type) -> Self {
+        Self {
+            field_type,
+            stats: IntermediateStats::new(),
+        }
+    }
+    pub(crate) fn collect_block(&mut self, doc: &[DocId], field: &DynamicFastFieldReader<u64>) {
+        let mut iter = doc.chunks_exact(4);
+        for docs in iter.by_ref() {
+            let val1 = field.get(docs[0]);
+            let val2 = field.get(docs[1]);
+            let val3 = field.get(docs[2]);
+            let val4 = field.get(docs[3]);
+            let val1 = f64_from_fastfield_u64(val1, &self.field_type);
+            let val2 = f64_from_fastfield_u64(val2, &self.field_type);
+            let val3 = f64_from_fastfield_u64(val3, &self.field_type);
+            let val4 = f64_from_fastfield_u64(val4, &self.field_type);
+            self.stats.collect(val1);
+            self.stats.collect(val2);
+            self.stats.collect(val3);
+            self.stats.collect(val4);
+        }
+        for doc in iter.remainder() {
+            let val = field.get(*doc);
+            let val = f64_from_fastfield_u64(val, &self.field_type);
+            self.stats.collect(val);
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+
+    use std::iter;
+
+    use serde_json::Value;
+
+    use crate::aggregation::agg_req::{
+        Aggregation, Aggregations, BucketAggregation, BucketAggregationType, MetricAggregation,
+        RangeAggregation,
+    };
+    use crate::aggregation::agg_result::AggregationResults;
+    use crate::aggregation::metric::StatsAggregation;
+    use crate::aggregation::tests::get_test_index_2_segments;
+    use crate::aggregation::AggregationCollector;
+    use crate::query::TermQuery;
+    use crate::schema::IndexRecordOption;
+    use crate::Term;
+
+    #[test]
+    fn test_aggregation_stats() -> crate::Result<()> {
+        let index = get_test_index_2_segments(false)?;
+
+        let reader = index.reader()?;
+        let text_field = reader.searcher().schema().get_field("text").unwrap();
+
+        let term_query = TermQuery::new(
+            Term::from_field_text(text_field, "cool"),
+            IndexRecordOption::Basic,
+        );
+
+        let agg_req_1: Aggregations = vec![
+            (
+                "stats_i64".to_string(),
+                Aggregation::Metric(MetricAggregation::Stats(StatsAggregation::from_field_name(
+                    "score_i64".to_string(),
+                ))),
+            ),
+            (
+                "stats_f64".to_string(),
+                Aggregation::Metric(MetricAggregation::Stats(StatsAggregation::from_field_name(
+                    "score_f64".to_string(),
+                ))),
+            ),
+            (
+                "stats".to_string(),
+                Aggregation::Metric(MetricAggregation::Stats(StatsAggregation::from_field_name(
+                    "score".to_string(),
+                ))),
+            ),
+            (
+                "range".to_string(),
+                Aggregation::Bucket(BucketAggregation {
+                    bucket_agg: BucketAggregationType::Range(RangeAggregation {
+                        field: "score".to_string(),
+                        ranges: vec![
+                            (3f64..7f64).into(),
+                            (7f64..19f64).into(),
+                            (19f64..20f64).into(),
+                        ],
+                    }),
+                    sub_aggregation: iter::once((
+                        "stats".to_string(),
+                        Aggregation::Metric(MetricAggregation::Stats(
+                            StatsAggregation::from_field_name("score".to_string()),
+                        )),
+                    ))
+                    .collect(),
+                }),
+            ),
+        ]
+        .into_iter()
+        .collect();
+
+        let collector = AggregationCollector::from_aggs(agg_req_1);
+
+        let searcher = reader.searcher();
+        let agg_res: AggregationResults = searcher.search(&term_query, &collector).unwrap();
+
+        let res: Value = serde_json::from_str(&serde_json::to_string(&agg_res)?)?;
+        assert_eq!(
+            res["stats"],
+            json!({
+                "avg": 12.142857142857142,
+                "count": 7,
+                "max": 44.0,
+                "min": 1.0,
+                "standard_deviation": 13.65313748796613,
+                "sum": 85.0
+            })
+        );
+
+        assert_eq!(
+            res["stats_i64"],
+            json!({
+                "avg": 12.142857142857142,
+                "count": 7,
+                "max": 44.0,
+                "min": 1.0,
+                "standard_deviation": 13.65313748796613,
+                "sum": 85.0
+            })
+        );
+
+        assert_eq!(
+            res["stats_f64"],
+            json!({
+                "avg":  12.214285714285714,
+                "count": 7,
+                "max": 44.5,
+                "min": 1.0,
+                "standard_deviation": 13.819905785437443,
+                "sum": 85.5
+            })
+        );
+
+        assert_eq!(
+            res["range"]["buckets"][2]["stats"],
+            json!({
+                "avg": 10.666666666666666,
+                "count": 3,
+                "max": 14.0,
+                "min": 7.0,
+                "standard_deviation": 2.867441755680877,
+                "sum": 32.0
+            })
+        );
+
+        assert_eq!(
+            res["range"]["buckets"][3]["stats"],
+            json!({
+                "avg": serde_json::Value::Null,
+                "count": 0,
+                "max": serde_json::Value::Null,
+                "min": serde_json::Value::Null,
+                "standard_deviation": serde_json::Value::Null,
+                "sum": 0.0,
+            })
+        );
+
+        Ok(())
+    }
+}

src/aggregation/segment_agg_result.rs

@@ -0,0 +1,204 @@
+//! Contains aggregation trees which is used during collection in a segment.
+//! This tree contains datastructrues optimized for fast collection.
+//! The tree can be converted to an intermediate tree, which contains datastructrues optimized for
+//! merging.
+
+use std::fmt::Debug;
+
+use super::agg_req::MetricAggregation;
+use super::agg_req_with_accessor::{
+    AggregationsWithAccessor, BucketAggregationWithAccessor, MetricAggregationWithAccessor,
+};
+use super::bucket::SegmentRangeCollector;
+use super::metric::{
+    AverageAggregation, SegmentAverageCollector, SegmentStatsCollector, StatsAggregation,
+};
+use super::{Key, VecWithNames};
+use crate::aggregation::agg_req::BucketAggregationType;
+use crate::DocId;
+
+pub(crate) const DOC_BLOCK_SIZE: usize = 256;
+pub(crate) type DocBlock = [DocId; DOC_BLOCK_SIZE];
+
+#[derive(Clone, PartialEq)]
+pub(crate) struct SegmentAggregationResultsCollector {
+    pub(crate) metrics: VecWithNames<SegmentMetricResultCollector>,
+    pub(crate) buckets: VecWithNames<SegmentBucketResultCollector>,
+    staged_docs: DocBlock,
+    num_staged_docs: usize,
+}
+
+impl Debug for SegmentAggregationResultsCollector {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("SegmentAggregationResultsCollector")
+            .field("metrics", &self.metrics)
+            .field("buckets", &self.buckets)
+            .field("staged_docs", &&self.staged_docs[..self.num_staged_docs])
+            .field("num_staged_docs", &self.num_staged_docs)
+            .finish()
+    }
+}
+
+impl SegmentAggregationResultsCollector {
+    pub(crate) fn from_req_and_validate(req: &AggregationsWithAccessor) -> crate::Result<Self> {
+        let buckets = req
+            .buckets
+            .entries()
+            .map(|(key, req)| {
+                Ok((
+                    key.to_string(),
+                    SegmentBucketResultCollector::from_req_and_validate(req)?,
+                ))
+            })
+            .collect::<crate::Result<_>>()?;
+        let metrics = req
+            .metrics
+            .entries()
+            .map(|(key, req)| {
+                Ok((
+                    key.to_string(),
+                    SegmentMetricResultCollector::from_req_and_validate(req)?,
+                ))
+            })
+            .collect::<crate::Result<_>>()?;
+        Ok(SegmentAggregationResultsCollector {
+            metrics: VecWithNames::from_entries(metrics),
+            buckets: VecWithNames::from_entries(buckets),
+            staged_docs: [0; DOC_BLOCK_SIZE],
+            num_staged_docs: 0,
+        })
+    }
+
+    #[inline]
+    pub(crate) fn collect(
+        &mut self,
+        doc: crate::DocId,
+        agg_with_accessor: &AggregationsWithAccessor,
+    ) {
+        self.staged_docs[self.num_staged_docs] = doc;
+        self.num_staged_docs += 1;
+        if self.num_staged_docs == self.staged_docs.len() {
+            self.flush_staged_docs(agg_with_accessor, false);
+        }
+    }
+
+    #[inline(never)]
+    pub(crate) fn flush_staged_docs(
+        &mut self,
+        agg_with_accessor: &AggregationsWithAccessor,
+        force_flush: bool,
+    ) {
+        for (agg_with_accessor, collector) in agg_with_accessor
+            .metrics
+            .values()
+            .zip(self.metrics.values_mut())
+        {
+            collector.collect_block(&self.staged_docs[..self.num_staged_docs], agg_with_accessor);
+        }
+        for (agg_with_accessor, collector) in agg_with_accessor
+            .buckets
+            .values()
+            .zip(self.buckets.values_mut())
+        {
+            collector.collect_block(
+                &self.staged_docs[..self.num_staged_docs],
+                agg_with_accessor,
+                force_flush,
+            );
+        }
+
+        self.num_staged_docs = 0;
+    }
+}
+
+#[derive(Clone, Debug, PartialEq)]
+pub(crate) enum SegmentMetricResultCollector {
+    Average(SegmentAverageCollector),
+    Stats(SegmentStatsCollector),
+}
+
+impl SegmentMetricResultCollector {
+    pub fn from_req_and_validate(req: &MetricAggregationWithAccessor) -> crate::Result<Self> {
+        match &req.metric {
+            MetricAggregation::Average(AverageAggregation { field: _ }) => {
+                Ok(SegmentMetricResultCollector::Average(
+                    SegmentAverageCollector::from_req(req.field_type),
+                ))
+            }
+            MetricAggregation::Stats(StatsAggregation { field: _ }) => {
+                Ok(SegmentMetricResultCollector::Stats(
+                    SegmentStatsCollector::from_req(req.field_type),
+                ))
+            }
+        }
+    }
+    pub(crate) fn collect_block(&mut self, doc: &[DocId], metric: &MetricAggregationWithAccessor) {
+        match self {
+            SegmentMetricResultCollector::Average(avg_collector) => {
+                avg_collector.collect_block(doc, &metric.accessor);
+            }
+            SegmentMetricResultCollector::Stats(stats_collector) => {
+                stats_collector.collect_block(doc, &metric.accessor);
+            }
+        }
+    }
+}
+
+/// SegmentBucketAggregationResultCollectors will have specialized buckets for collection inside
+/// segments.
+/// The typical structure of Map<Key, Bucket> is not suitable during collection for performance
+/// reasons.
+#[derive(Clone, Debug, PartialEq)]
+pub(crate) enum SegmentBucketResultCollector {
+    Range(SegmentRangeCollector),
+}
+
+impl SegmentBucketResultCollector {
+    pub fn from_req_and_validate(req: &BucketAggregationWithAccessor) -> crate::Result<Self> {
+        match &req.bucket_agg {
+            BucketAggregationType::Range(range_req) => {
+                Ok(Self::Range(SegmentRangeCollector::from_req_and_validate(
+                    range_req,
+                    &req.sub_aggregation,
+                    req.field_type,
+                )?))
+            }
+        }
+    }
+
+    #[inline]
+    pub(crate) fn collect_block(
+        &mut self,
+        doc: &[DocId],
+        bucket_with_accessor: &BucketAggregationWithAccessor,
+        force_flush: bool,
+    ) {
+        match self {
+            SegmentBucketResultCollector::Range(range) => {
+                range.collect_block(doc, bucket_with_accessor, force_flush);
+            }
+        }
+    }
+}
+
+#[derive(Clone, PartialEq)]
+pub(crate) struct SegmentRangeBucketEntry {
+    pub key: Key,
+    pub doc_count: u64,
+    pub sub_aggregation: Option<SegmentAggregationResultsCollector>,
+    /// The from range of the bucket. Equals f64::MIN when None.
+    pub from: Option<f64>,
+    /// The to range of the bucket. Equals f64::MAX when None.
+    pub to: Option<f64>,
+}
+
+impl Debug for SegmentRangeBucketEntry {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("SegmentRangeBucketEntry")
+            .field("key", &self.key)
+            .field("doc_count", &self.doc_count)
+            .field("from", &self.from)
+            .field("to", &self.to)
+            .finish()
+    }
+}

src/collector/count_collector.rs

@@ -1,9 +1,6 @@
 use super::Collector;
 use crate::collector::SegmentCollector;
-use crate::DocId;
-use crate::Score;
-use crate::SegmentLocalId;
-use crate::SegmentReader;
+use crate::{DocId, Score, SegmentOrdinal, SegmentReader};
 
 /// `CountCollector` collector only counts how many
 /// documents match the query.
@@ -20,10 +17,10 @@ use crate::SegmentReader;
 /// let index = Index::create_in_ram(schema);
 ///
 /// 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"));
+/// index_writer.add_document(doc!(title => "The Name of the Wind")).unwrap();
+/// index_writer.add_document(doc!(title => "The Diary of Muadib")).unwrap();
+/// index_writer.add_document(doc!(title => "A Dairy Cow")).unwrap();
+/// index_writer.add_document(doc!(title => "The Diary of a Young Girl")).unwrap();
 /// assert!(index_writer.commit().is_ok());
 ///
 /// let reader = index.reader().unwrap();
@@ -45,7 +42,7 @@ impl Collector for Count {
 
     fn for_segment(
         &self,
-        _: SegmentLocalId,
+        _: SegmentOrdinal,
         _: &SegmentReader,
     ) -> crate::Result<SegmentCountCollector> {
         Ok(SegmentCountCollector::default())
@@ -80,8 +77,7 @@ impl SegmentCollector for SegmentCountCollector {
 #[cfg(test)]
 mod tests {
     use super::{Count, SegmentCountCollector};
-    use crate::collector::Collector;
-    use crate::collector::SegmentCollector;
+    use crate::collector::{Collector, SegmentCollector};
 
     #[test]
     fn test_count_collect_does_not_requires_scoring() {

src/collector/custom_score_top_collector.rs

@@ -8,8 +8,7 @@ pub(crate) struct CustomScoreTopCollector<TCustomScorer, TScore = Score> {
 }
 
 impl<TCustomScorer, TScore> CustomScoreTopCollector<TCustomScorer, TScore>
-where
-    TScore: Clone + PartialOrd,
+where TScore: Clone + PartialOrd
 {
     pub(crate) fn new(
         custom_scorer: TCustomScorer,
@@ -58,9 +57,7 @@ where
         segment_local_id: u32,
         segment_reader: &SegmentReader,
     ) -> crate::Result<Self::Child> {
-        let segment_collector = self
-            .collector
-            .for_segment(segment_local_id, segment_reader)?;
+        let segment_collector = self.collector.for_segment(segment_local_id, segment_reader);
         let segment_scorer = self.custom_scorer.segment_scorer(segment_reader)?;
         Ok(CustomScoreTopSegmentCollector {
             segment_collector,
@@ -116,8 +113,7 @@ where
 }
 
 impl<F, TScore> CustomSegmentScorer<TScore> for F
-where
-    F: 'static + FnMut(DocId) -> TScore,
+where F: 'static + FnMut(DocId) -> TScore
 {
     fn score(&mut self, doc: DocId) -> TScore {
         (self)(doc)

src/collector/docset_collector.rs

@@ -1,8 +1,7 @@
 use std::collections::HashSet;
 
-use crate::{DocAddress, DocId, Score};
-
 use super::{Collector, SegmentCollector};
+use crate::{DocAddress, DocId, Score};
 
 /// Collectors that returns the set of DocAddress that matches the query.
 ///
@@ -15,7 +14,7 @@ impl Collector for DocSetCollector {
 
     fn for_segment(
         &self,
-        segment_local_id: crate::SegmentLocalId,
+        segment_local_id: crate::SegmentOrdinal,
         _segment: &crate::SegmentReader,
     ) -> crate::Result<Self::Child> {
         Ok(DocSetChildCollector {
@@ -36,7 +35,7 @@ impl Collector for DocSetCollector {
         let mut result = HashSet::with_capacity(len);
         for (segment_local_id, docs) in segment_fruits {
             for doc in docs {
-                result.insert(DocAddress(segment_local_id, doc));
+                result.insert(DocAddress::new(segment_local_id, doc));
             }
         }
         Ok(result)

src/collector/facet_collector.rs

@@ -1,21 +1,14 @@
-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 std::cmp::Ordering;
-use std::collections::btree_map;
-use std::collections::BTreeMap;
-use std::collections::BTreeSet;
-use std::collections::BinaryHeap;
-use std::collections::Bound;
+use std::collections::{btree_map, BTreeMap, BTreeSet, BinaryHeap};
 use std::iter::Peekable;
+use std::ops::Bound;
 use std::{u64, usize};
 
+use crate::collector::{Collector, SegmentCollector};
+use crate::fastfield::FacetReader;
+use crate::schema::{Facet, Field};
+use crate::{DocId, Score, SegmentOrdinal, SegmentReader};
+
 struct Hit<'a> {
     count: u64,
     facet: &'a Facet,
@@ -37,7 +30,10 @@ impl<'a> PartialOrd<Hit<'a>> for Hit<'a> {
 
 impl<'a> Ord for Hit<'a> {
     fn cmp(&self, other: &Self) -> Ordering {
-        other.count.cmp(&self.count)
+        other
+            .count
+            .cmp(&self.count)
+            .then(self.facet.cmp(other.facet))
     }
 }
 
@@ -80,7 +76,7 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
 /// ```rust
 /// use tantivy::collector::FacetCollector;
 /// use tantivy::query::AllQuery;
-/// use tantivy::schema::{Facet, Schema, TEXT};
+/// use tantivy::schema::{Facet, Schema, FacetOptions, TEXT};
 /// use tantivy::{doc, Index};
 ///
 /// fn example() -> tantivy::Result<()> {
@@ -89,7 +85,7 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
 ///     // Facet have their own specific type.
 ///     // It is not a bad practise to put all of your
 ///     // facet information in the same field.
-///     let facet = schema_builder.add_facet_field("facet");
+///     let facet = schema_builder.add_facet_field("facet", FacetOptions::default());
 ///     let title = schema_builder.add_text_field("title", TEXT);
 ///     let schema = schema_builder.build();
 ///     let index = Index::create_in_ram(schema);
@@ -100,23 +96,23 @@ fn facet_depth(facet_bytes: &[u8]) -> usize {
 ///             title => "The Name of the Wind",
 ///             facet => Facet::from("/lang/en"),
 ///             facet => Facet::from("/category/fiction/fantasy")
-///         ));
+///         ))?;
 ///         index_writer.add_document(doc!(
 ///             title => "Dune",
 ///             facet => Facet::from("/lang/en"),
 ///             facet => Facet::from("/category/fiction/sci-fi")
-///         ));
+///         ))?;
 ///         index_writer.add_document(doc!(
 ///             title => "La Vénus d'Ille",
 ///             facet => Facet::from("/lang/fr"),
 ///             facet => Facet::from("/category/fiction/fantasy"),
 ///             facet => Facet::from("/category/fiction/horror")
-///         ));
+///         ))?;
 ///         index_writer.add_document(doc!(
 ///             title => "The Diary of a Young Girl",
 ///             facet => Facet::from("/lang/en"),
 ///             facet => Facet::from("/category/biography")
-///         ));
+///         ))?;
 ///         index_writer.commit()?;
 ///     }
 ///     let reader = index.reader()?;
@@ -237,9 +233,7 @@ impl FacetCollector {
     /// If you need the correct number of unique documents for two such facets,
     /// just add them in separate `FacetCollector`.
     pub fn add_facet<T>(&mut self, facet_from: T)
-    where
-        Facet: From<T>,
-    {
+    where Facet: From<T> {
         let facet = Facet::from(facet_from);
         for old_facet in &self.facets {
             assert!(
@@ -262,7 +256,7 @@ impl Collector for FacetCollector {
 
     fn for_segment(
         &self,
-        _: SegmentLocalId,
+        _: SegmentOrdinal,
         reader: &SegmentReader,
     ) -> crate::Result<FacetSegmentCollector> {
         let facet_reader = reader.facet_reader(self.field)?;
@@ -294,10 +288,8 @@ impl Collector for FacetCollector {
                                 if depth == collapse_depth + 1 {
                                     collapsed_id = collapse_facet_ords.len();
                                     collapse_facet_ords.push(facet_streamer.term_ord());
-                                    collapse_mapping.push(collapsed_id);
-                                } else {
-                                    collapse_mapping.push(collapsed_id);
                                 }
+                                collapse_mapping.push(collapsed_id);
                             }
                             break;
                         }
@@ -399,11 +391,9 @@ impl<'a> Iterator for FacetChildIterator<'a> {
 
 impl FacetCounts {
     /// Returns an iterator over all of the facet count pairs inside this result.
-    /// See the documentation for `FacetCollector` for a usage example.
+    /// See the documentation for [FacetCollector] for a usage example.
     pub fn get<T>(&self, facet_from: T) -> FacetChildIterator<'_>
-    where
-        Facet: From<T>,
-    {
+    where Facet: From<T> {
         let facet = Facet::from(facet_from);
         let left_bound = Bound::Excluded(facet.clone());
         let right_bound = if facet.is_root() {
@@ -420,11 +410,9 @@ impl FacetCounts {
     }
 
     /// Returns a vector of top `k` facets with their counts, sorted highest-to-lowest by counts.
-    /// See the documentation for `FacetCollector` for a usage example.
+    /// See the documentation for [FacetCollector] for a usage example.
     pub fn top_k<T>(&self, facet: T, k: usize) -> Vec<(&Facet, u64)>
-    where
-        Facet: From<T>,
-    {
+    where Facet: From<T> {
         let mut heap = BinaryHeap::with_capacity(k);
         let mut it = self.get(facet);
 
@@ -457,25 +445,27 @@ impl FacetCounts {
 
 #[cfg(test)]
 mod tests {
+    use std::iter;
+
+    use rand::distributions::Uniform;
+    use rand::prelude::SliceRandom;
+    use rand::{thread_rng, Rng};
+
     use super::{FacetCollector, FacetCounts};
     use crate::collector::Count;
     use crate::core::Index;
     use crate::query::{AllQuery, QueryParser, TermQuery};
-    use crate::schema::{Document, Facet, Field, IndexRecordOption, Schema};
+    use crate::schema::{Document, Facet, FacetOptions, Field, IndexRecordOption, Schema};
     use crate::Term;
-    use rand::distributions::Uniform;
-    use rand::prelude::SliceRandom;
-    use rand::{thread_rng, Rng};
-    use std::iter;
 
     #[test]
-    fn test_facet_collector_drilldown() {
+    fn test_facet_collector_drilldown() -> crate::Result<()> {
         let mut schema_builder = Schema::builder();
-        let facet_field = schema_builder.add_facet_field("facet");
+        let facet_field = schema_builder.add_facet_field("facet", FacetOptions::default());
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
 
-        let mut index_writer = index.writer_for_tests().unwrap();
+        let mut index_writer = index.writer_for_tests()?;
         let num_facets: usize = 3 * 4 * 5;
         let facets: Vec<Facet> = (0..num_facets)
             .map(|mut n| {
@@ -490,14 +480,14 @@ mod tests {
         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.add_document(doc)?;
         }
-        index_writer.commit().unwrap();
-        let reader = index.reader().unwrap();
+        index_writer.commit()?;
+        let reader = index.reader()?;
         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();
+        let counts = searcher.search(&AllQuery, &facet_collector)?;
 
         {
             let facets: Vec<(String, u64)> = counts
@@ -517,11 +507,13 @@ mod tests {
                 .collect::<Vec<_>>()
             );
         }
+        Ok(())
     }
 
     #[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::from_field_id(0));
         facet_collector.add_facet(Facet::from("/country"));
@@ -529,55 +521,56 @@ mod tests {
     }
 
     #[test]
-    fn test_doc_unsorted_multifacet() {
+    fn test_doc_unsorted_multifacet() -> crate::Result<()> {
         let mut schema_builder = Schema::builder();
-        let facet_field = schema_builder.add_facet_field("facets");
+        let facet_field = schema_builder.add_facet_field("facets", FacetOptions::default());
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
-        let mut index_writer = index.writer_for_tests().unwrap();
+        let mut index_writer = index.writer_for_tests()?;
         index_writer.add_document(doc!(
-            facet_field => Facet::from_text(&"/subjects/A/a"),
-            facet_field => Facet::from_text(&"/subjects/B/a"),
-            facet_field => Facet::from_text(&"/subjects/A/b"),
-            facet_field => Facet::from_text(&"/subjects/B/b"),
-        ));
-        index_writer.commit().unwrap();
-        let reader = index.reader().unwrap();
+            facet_field => Facet::from_text(&"/subjects/A/a").unwrap(),
+            facet_field => Facet::from_text(&"/subjects/B/a").unwrap(),
+            facet_field => Facet::from_text(&"/subjects/A/b").unwrap(),
+            facet_field => Facet::from_text(&"/subjects/B/b").unwrap(),
+        ))?;
+        index_writer.commit()?;
+        let reader = index.reader()?;
         let searcher = reader.searcher();
         assert_eq!(searcher.num_docs(), 1);
         let mut facet_collector = FacetCollector::for_field(facet_field);
         facet_collector.add_facet("/subjects");
-        let counts = searcher.search(&AllQuery, &facet_collector).unwrap();
+        let counts = searcher.search(&AllQuery, &facet_collector)?;
         let facets: Vec<(&Facet, u64)> = counts.get("/subjects").collect();
         assert_eq!(facets[0].1, 1);
+        Ok(())
     }
 
     #[test]
     fn test_doc_search_by_facet() -> crate::Result<()> {
         let mut schema_builder = Schema::builder();
-        let facet_field = schema_builder.add_facet_field("facet");
+        let facet_field = schema_builder.add_facet_field("facet", FacetOptions::default());
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
         let mut index_writer = index.writer_for_tests()?;
         index_writer.add_document(doc!(
-            facet_field => Facet::from_text(&"/A/A"),
-        ));
+            facet_field => Facet::from_text(&"/A/A").unwrap(),
+        ))?;
         index_writer.add_document(doc!(
-            facet_field => Facet::from_text(&"/A/B"),
-        ));
+            facet_field => Facet::from_text(&"/A/B").unwrap(),
+        ))?;
         index_writer.add_document(doc!(
-            facet_field => Facet::from_text(&"/A/C/A"),
-        ));
+            facet_field => Facet::from_text(&"/A/C/A").unwrap(),
+        ))?;
         index_writer.add_document(doc!(
-            facet_field => Facet::from_text(&"/D/C/A"),
-        ));
+            facet_field => Facet::from_text(&"/D/C/A").unwrap(),
+        ))?;
         index_writer.commit()?;
         let reader = index.reader()?;
         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));
+            let term = Term::from_facet(facet_field, &Facet::from_text(facet_str).unwrap());
             searcher
                 .search(&TermQuery::new(term, IndexRecordOption::Basic), &Count)
                 .unwrap()
@@ -612,7 +605,7 @@ mod tests {
     #[test]
     fn test_facet_collector_topk() {
         let mut schema_builder = Schema::builder();
-        let facet_field = schema_builder.add_facet_field("facet");
+        let facet_field = schema_builder.add_facet_field("facet", FacetOptions::default());
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
 
@@ -636,7 +629,7 @@ mod tests {
 
         let mut index_writer = index.writer_for_tests().unwrap();
         for doc in docs {
-            index_writer.add_document(doc);
+            index_writer.add_document(doc).unwrap();
         }
         index_writer.commit().unwrap();
         let searcher = index.reader().unwrap().searcher();
@@ -657,23 +650,59 @@ mod tests {
             );
         }
     }
+
+    #[test]
+    fn test_facet_collector_topk_tie_break() -> crate::Result<()> {
+        let mut schema_builder = Schema::builder();
+        let facet_field = schema_builder.add_facet_field("facet", FacetOptions::default());
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+
+        let docs: Vec<Document> = vec![("b", 2), ("a", 2), ("c", 4)]
+            .into_iter()
+            .flat_map(|(c, count)| {
+                let facet = Facet::from(&format!("/facet/{}", c));
+                let doc = doc!(facet_field => facet);
+                iter::repeat(doc).take(count)
+            })
+            .collect();
+
+        let mut index_writer = index.writer_for_tests()?;
+        for doc in docs {
+            index_writer.add_document(doc)?;
+        }
+        index_writer.commit()?;
+
+        let searcher = index.reader()?.searcher();
+        let mut facet_collector = FacetCollector::for_field(facet_field);
+        facet_collector.add_facet("/facet");
+        let counts: FacetCounts = searcher.search(&AllQuery, &facet_collector)?;
+
+        let facets: Vec<(&Facet, u64)> = counts.top_k("/facet", 2);
+        assert_eq!(
+            facets,
+            vec![(&Facet::from("/facet/c"), 4), (&Facet::from("/facet/a"), 2)]
+        );
+        Ok(())
+    }
 }
 
 #[cfg(all(test, feature = "unstable"))]
 mod bench {
 
-    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 crate::collector::FacetCollector;
+    use crate::query::AllQuery;
+    use crate::schema::{Facet, Schema, INDEXED};
+    use crate::Index;
+
     #[bench]
     fn bench_facet_collector(b: &mut Bencher) {
         let mut schema_builder = Schema::builder();
-        let facet_field = schema_builder.add_facet_field("facet");
+        let facet_field = schema_builder.add_facet_field("facet", INDEXED);
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
 
@@ -689,7 +718,7 @@ mod bench {
 
         let mut index_writer = index.writer_for_tests().unwrap();
         for doc in docs {
-            index_writer.add_document(doc);
+            index_writer.add_document(doc).unwrap();
         }
         index_writer.commit().unwrap();
         let reader = index.reader().unwrap();

src/collector/filter_collector_wrapper.rs

@@ -12,12 +12,13 @@
 use std::marker::PhantomData;
 
 use crate::collector::{Collector, SegmentCollector};
-use crate::fastfield::{FastFieldReader, FastValue};
+use crate::fastfield::{DynamicFastFieldReader, FastFieldReader, FastValue};
 use crate::schema::Field;
 use crate::{Score, SegmentReader, TantivyError};
 
-/// The `FilterCollector` collector filters docs using a u64 fast field value and a predicate.
-/// Only the documents for which the predicate returned "true" will be passed on to the next collector.
+/// The `FilterCollector` filters docs using a fast field value and a predicate.
+/// Only the documents for which the predicate returned "true" will be passed on to the next
+/// collector.
 ///
 /// ```rust
 /// use tantivy::collector::{TopDocs, FilterCollector};
@@ -25,42 +26,44 @@ use crate::{Score, SegmentReader, TantivyError};
 /// use tantivy::schema::{Schema, TEXT, INDEXED, FAST};
 /// use tantivy::{doc, DocAddress, Index};
 ///
+/// # fn main() -> tantivy::Result<()> {
 /// let mut schema_builder = Schema::builder();
 /// let title = schema_builder.add_text_field("title", TEXT);
 /// let price = schema_builder.add_u64_field("price", INDEXED | FAST);
 /// let schema = schema_builder.build();
 /// let index = Index::create_in_ram(schema);
 ///
-/// let mut index_writer = index.writer_with_num_threads(1, 10_000_000).unwrap();
-/// index_writer.add_document(doc!(title => "The Name of the Wind", price => 30_200u64));
-/// index_writer.add_document(doc!(title => "The Diary of Muadib", price => 29_240u64));
-/// index_writer.add_document(doc!(title => "A Dairy Cow", price => 21_240u64));
-/// index_writer.add_document(doc!(title => "The Diary of a Young Girl", price => 20_120u64));
-/// assert!(index_writer.commit().is_ok());
+/// let mut index_writer = index.writer_with_num_threads(1, 10_000_000)?;
+/// index_writer.add_document(doc!(title => "The Name of the Wind", price => 30_200u64))?;
+/// index_writer.add_document(doc!(title => "The Diary of Muadib", price => 29_240u64))?;
+/// index_writer.add_document(doc!(title => "A Dairy Cow", price => 21_240u64))?;
+/// index_writer.add_document(doc!(title => "The Diary of a Young Girl", price => 20_120u64))?;
+/// index_writer.commit()?;
 ///
-/// let reader = index.reader().unwrap();
+/// let reader = index.reader()?;
 /// let searcher = reader.searcher();
 ///
 /// let query_parser = QueryParser::for_index(&index, vec![title]);
-/// let query = query_parser.parse_query("diary").unwrap();
+/// let query = query_parser.parse_query("diary")?;
 /// let no_filter_collector = FilterCollector::new(price, &|value: u64| value > 20_120u64, TopDocs::with_limit(2));
-/// let top_docs = searcher.search(&query, &no_filter_collector).unwrap();
+/// let top_docs = searcher.search(&query, &no_filter_collector)?;
 ///
 /// assert_eq!(top_docs.len(), 1);
-/// assert_eq!(top_docs[0].1, DocAddress(0, 1));
+/// assert_eq!(top_docs[0].1, DocAddress::new(0, 1));
 ///
 /// let filter_all_collector: FilterCollector<_, _, u64> = FilterCollector::new(price, &|value| value < 5u64, TopDocs::with_limit(2));
-/// let filtered_top_docs = searcher.search(&query, &filter_all_collector).unwrap();
+/// let filtered_top_docs = searcher.search(&query, &filter_all_collector)?;
 ///
 /// assert_eq!(filtered_top_docs.len(), 0);
+/// # Ok(())
+/// # }
 /// ```
 pub struct FilterCollector<TCollector, TPredicate, TPredicateValue: FastValue>
-where
-    TPredicate: 'static,
+where TPredicate: 'static + Clone
 {
     field: Field,
     collector: TCollector,
-    predicate: &'static TPredicate,
+    predicate: TPredicate,
     t_predicate_value: PhantomData<TPredicateValue>,
 }
 
@@ -68,12 +71,12 @@ impl<TCollector, TPredicate, TPredicateValue: FastValue>
     FilterCollector<TCollector, TPredicate, TPredicateValue>
 where
     TCollector: Collector + Send + Sync,
-    TPredicate: Fn(TPredicateValue) -> bool + Send + Sync,
+    TPredicate: Fn(TPredicateValue) -> bool + Send + Sync + Clone,
 {
     /// Create a new FilterCollector.
     pub fn new(
         field: Field,
-        predicate: &'static TPredicate,
+        predicate: TPredicate,
         collector: TCollector,
     ) -> FilterCollector<TCollector, TPredicate, TPredicateValue> {
         FilterCollector {
@@ -89,8 +92,8 @@ impl<TCollector, TPredicate, TPredicateValue: FastValue> Collector
     for FilterCollector<TCollector, TPredicate, TPredicateValue>
 where
     TCollector: Collector + Send + Sync,
-    TPredicate: 'static + Fn(TPredicateValue) -> bool + Send + Sync,
-    TPredicateValue: 'static + FastValue,
+    TPredicate: 'static + Fn(TPredicateValue) -> bool + Send + Sync + Clone,
+    TPredicateValue: FastValue,
 {
     // That's the type of our result.
     // Our standard deviation will be a float.
@@ -133,7 +136,7 @@ where
         Ok(FilterSegmentCollector {
             fast_field_reader,
             segment_collector,
-            predicate: self.predicate,
+            predicate: self.predicate.clone(),
             t_predicate_value: PhantomData,
         })
     }
@@ -153,11 +156,11 @@ where
 pub struct FilterSegmentCollector<TSegmentCollector, TPredicate, TPredicateValue>
 where
     TPredicate: 'static,
-    TPredicateValue: 'static + FastValue,
+    TPredicateValue: FastValue,
 {
-    fast_field_reader: FastFieldReader<TPredicateValue>,
+    fast_field_reader: DynamicFastFieldReader<TPredicateValue>,
     segment_collector: TSegmentCollector,
-    predicate: &'static TPredicate,
+    predicate: TPredicate,
     t_predicate_value: PhantomData<TPredicateValue>,
 }
 
@@ -166,7 +169,7 @@ impl<TSegmentCollector, TPredicate, TPredicateValue> SegmentCollector
 where
     TSegmentCollector: SegmentCollector,
     TPredicate: 'static + Fn(TPredicateValue) -> bool + Send + Sync,
-    TPredicateValue: 'static + FastValue,
+    TPredicateValue: FastValue,
 {
     type Fruit = TSegmentCollector::Fruit;
 

src/collector/histogram_collector.rs

@@ -0,0 +1,293 @@
+use fastdivide::DividerU64;
+
+use crate::collector::{Collector, SegmentCollector};
+use crate::fastfield::{DynamicFastFieldReader, FastFieldReader, FastValue};
+use crate::schema::{Field, Type};
+use crate::{DocId, Score};
+
+/// Histogram builds an histogram of the values of a fastfield for the
+/// collected DocSet.
+///
+/// At construction, it is given parameters that define a partition of an interval
+/// [min_val, max_val) into N buckets with the same width.
+/// The ith bucket is then defined by `[min_val + i * bucket_width, min_val + (i+1) * bucket_width)`
+///
+/// An histogram is then defined as a `Vec<u64>` of length `num_buckets`, that contains a count of
+/// documents for each value bucket.
+///
+/// See also [`HistogramCollector::new()`].
+///
+/// # Warning
+///
+/// f64 fields are not supported.
+#[derive(Clone)]
+pub struct HistogramCollector {
+    min_value: u64,
+    num_buckets: usize,
+    divider: DividerU64,
+    field: Field,
+}
+
+impl HistogramCollector {
+    /// Builds a new HistogramCollector.
+    ///
+    /// The scale/range of the histogram is not dynamic. It is required to
+    /// define it by supplying following parameter:
+    ///  - `min_value`: the minimum value that can be recorded in the histogram.
+    ///  - `bucket_width`: the length of the interval that is associated to each buckets.
+    ///  - `num_buckets`: The overall number of buckets.
+    ///
+    /// Together, this parameters define a partition of `[min_value, min_value + num_buckets *
+    /// bucket_width)` into `num_buckets` intervals of width bucket that we call `bucket`.
+    ///
+    /// # Disclaimer
+    /// This function panics if the field given is of type f64.
+    pub fn new<TFastValue: FastValue>(
+        field: Field,
+        min_value: TFastValue,
+        bucket_width: u64,
+        num_buckets: usize,
+    ) -> HistogramCollector {
+        let fast_type = TFastValue::to_type();
+        assert!(fast_type == Type::U64 || fast_type == Type::I64 || fast_type == Type::Date);
+        HistogramCollector {
+            min_value: min_value.to_u64(),
+            num_buckets,
+            field,
+            divider: DividerU64::divide_by(bucket_width),
+        }
+    }
+}
+
+struct HistogramComputer {
+    counts: Vec<u64>,
+    min_value: u64,
+    divider: DividerU64,
+}
+
+impl HistogramComputer {
+    #[inline]
+    pub(crate) fn add_value(&mut self, value: u64) {
+        if value < self.min_value {
+            return;
+        }
+        let delta = value - self.min_value;
+        let delta_u64 = delta.to_u64();
+        let bucket_id: usize = self.divider.divide(delta_u64) as usize;
+        if bucket_id < self.counts.len() {
+            self.counts[bucket_id] += 1;
+        }
+    }
+
+    fn harvest(self) -> Vec<u64> {
+        self.counts
+    }
+}
+pub struct SegmentHistogramCollector {
+    histogram_computer: HistogramComputer,
+    ff_reader: DynamicFastFieldReader<u64>,
+}
+
+impl SegmentCollector for SegmentHistogramCollector {
+    type Fruit = Vec<u64>;
+
+    fn collect(&mut self, doc: DocId, _score: Score) {
+        let value = self.ff_reader.get(doc);
+        self.histogram_computer.add_value(value);
+    }
+
+    fn harvest(self) -> Self::Fruit {
+        self.histogram_computer.harvest()
+    }
+}
+
+impl Collector for HistogramCollector {
+    type Fruit = Vec<u64>;
+    type Child = SegmentHistogramCollector;
+
+    fn for_segment(
+        &self,
+        _segment_local_id: crate::SegmentOrdinal,
+        segment: &crate::SegmentReader,
+    ) -> crate::Result<Self::Child> {
+        let ff_reader = segment.fast_fields().u64_lenient(self.field)?;
+        Ok(SegmentHistogramCollector {
+            histogram_computer: HistogramComputer {
+                counts: vec![0; self.num_buckets],
+                min_value: self.min_value,
+                divider: self.divider,
+            },
+            ff_reader,
+        })
+    }
+
+    fn requires_scoring(&self) -> bool {
+        false
+    }
+
+    fn merge_fruits(&self, child_histograms: Vec<Vec<u64>>) -> crate::Result<Vec<u64>> {
+        Ok(add_vecs(child_histograms, self.num_buckets))
+    }
+}
+
+pub fn add_arrays_into(acc: &mut [u64], add: &[u64]) {
+    assert_eq!(acc.len(), add.len());
+    for (dest_bucket, bucket_count) in acc.iter_mut().zip(add) {
+        *dest_bucket += bucket_count;
+    }
+}
+
+fn add_vecs(mut vals_list: Vec<Vec<u64>>, len: usize) -> Vec<u64> {
+    let mut acc = vals_list.pop().unwrap_or_else(|| vec![0u64; len]);
+    assert_eq!(acc.len(), len);
+    for vals in vals_list {
+        add_arrays_into(&mut acc, &vals);
+    }
+    acc
+}
+
+#[cfg(test)]
+mod tests {
+    use fastdivide::DividerU64;
+    use query::AllQuery;
+
+    use super::{add_vecs, HistogramCollector, HistogramComputer};
+    use crate::chrono::{TimeZone, Utc};
+    use crate::schema::{Schema, FAST};
+    use crate::{doc, query, Index};
+
+    #[test]
+    fn test_add_histograms_simple() {
+        assert_eq!(
+            add_vecs(vec![vec![1, 0, 3], vec![11, 2, 3], vec![0, 0, 1]], 3),
+            vec![12, 2, 7]
+        )
+    }
+
+    #[test]
+    fn test_add_histograms_empty() {
+        assert_eq!(add_vecs(vec![], 3), vec![0, 0, 0])
+    }
+
+    #[test]
+    fn test_histogram_builder_simple() {
+        // [1..3)
+        // [3..5)
+        // ..
+        // [9..11)
+        let mut histogram_computer = HistogramComputer {
+            counts: vec![0; 5],
+            min_value: 1,
+            divider: DividerU64::divide_by(2),
+        };
+        histogram_computer.add_value(1);
+        histogram_computer.add_value(7);
+        assert_eq!(histogram_computer.harvest(), vec![1, 0, 0, 1, 0]);
+    }
+
+    #[test]
+    fn test_histogram_too_low_is_ignored() {
+        let mut histogram_computer = HistogramComputer {
+            counts: vec![0; 5],
+            min_value: 2,
+            divider: DividerU64::divide_by(2),
+        };
+        histogram_computer.add_value(0);
+        assert_eq!(histogram_computer.harvest(), vec![0, 0, 0, 0, 0]);
+    }
+
+    #[test]
+    fn test_histogram_too_high_is_ignored() {
+        let mut histogram_computer = HistogramComputer {
+            counts: vec![0u64; 5],
+            min_value: 0,
+            divider: DividerU64::divide_by(2),
+        };
+        histogram_computer.add_value(10);
+        assert_eq!(histogram_computer.harvest(), vec![0, 0, 0, 0, 0]);
+    }
+    #[test]
+    fn test_no_segments() -> crate::Result<()> {
+        let mut schema_builder = Schema::builder();
+        let val_field = schema_builder.add_u64_field("val_field", FAST);
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        let reader = index.reader()?;
+        let searcher = reader.searcher();
+        let all_query = AllQuery;
+        let histogram_collector = HistogramCollector::new(val_field, 0u64, 2, 5);
+        let histogram = searcher.search(&all_query, &histogram_collector)?;
+        assert_eq!(histogram, vec![0; 5]);
+        Ok(())
+    }
+
+    #[test]
+    fn test_histogram_i64() -> crate::Result<()> {
+        let mut schema_builder = Schema::builder();
+        let val_field = schema_builder.add_i64_field("val_field", FAST);
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        let mut writer = index.writer_with_num_threads(1, 4_000_000)?;
+        writer.add_document(doc!(val_field=>12i64))?;
+        writer.add_document(doc!(val_field=>-30i64))?;
+        writer.add_document(doc!(val_field=>-12i64))?;
+        writer.add_document(doc!(val_field=>-10i64))?;
+        writer.commit()?;
+        let reader = index.reader()?;
+        let searcher = reader.searcher();
+        let all_query = AllQuery;
+        let histogram_collector = HistogramCollector::new(val_field, -20i64, 10u64, 4);
+        let histogram = searcher.search(&all_query, &histogram_collector)?;
+        assert_eq!(histogram, vec![1, 1, 0, 1]);
+        Ok(())
+    }
+
+    #[test]
+    fn test_histogram_merge() -> crate::Result<()> {
+        let mut schema_builder = Schema::builder();
+        let val_field = schema_builder.add_i64_field("val_field", FAST);
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        let mut writer = index.writer_with_num_threads(1, 4_000_000)?;
+        writer.add_document(doc!(val_field=>12i64))?;
+        writer.commit()?;
+        writer.add_document(doc!(val_field=>-30i64))?;
+        writer.commit()?;
+        writer.add_document(doc!(val_field=>-12i64))?;
+        writer.commit()?;
+        writer.add_document(doc!(val_field=>-10i64))?;
+        writer.commit()?;
+        let reader = index.reader()?;
+        let searcher = reader.searcher();
+        let all_query = AllQuery;
+        let histogram_collector = HistogramCollector::new(val_field, -20i64, 10u64, 4);
+        let histogram = searcher.search(&all_query, &histogram_collector)?;
+        assert_eq!(histogram, vec![1, 1, 0, 1]);
+        Ok(())
+    }
+
+    #[test]
+    fn test_histogram_dates() -> crate::Result<()> {
+        let mut schema_builder = Schema::builder();
+        let date_field = schema_builder.add_date_field("date_field", FAST);
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema);
+        let mut writer = index.writer_with_num_threads(1, 4_000_000)?;
+        writer.add_document(doc!(date_field=>Utc.ymd(1982, 9, 17).and_hms(0, 0,0)))?;
+        writer.add_document(doc!(date_field=>Utc.ymd(1986, 3, 9).and_hms(0, 0, 0)))?;
+        writer.add_document(doc!(date_field=>Utc.ymd(1983, 9, 27).and_hms(0, 0, 0)))?;
+        writer.commit()?;
+        let reader = index.reader()?;
+        let searcher = reader.searcher();
+        let all_query = AllQuery;
+        let week_histogram_collector = HistogramCollector::new(
+            date_field,
+            Utc.ymd(1980, 1, 1).and_hms(0, 0, 0),
+            3600 * 24 * 365, // it is just for a unit test... sorry leap years.
+            10,
+        );
+        let week_histogram = searcher.search(&all_query, &week_histogram_collector)?;
+        assert_eq!(week_histogram, vec![0, 0, 1, 1, 0, 0, 1, 0, 0, 0]);
+        Ok(())
+    }
+}

src/collector/mod.rs

@@ -1,98 +1,96 @@
-/*!
+//! # Collectors
+//!
+//! Collectors define the information you want to extract from the documents matching the queries.
+//! In tantivy jargon, we call this information your search "fruit".
+//!
+//! Your fruit could for instance be :
+//! - [the count of matching documents](./struct.Count.html)
+//! - [the top 10 documents, by relevancy or by a fast field](./struct.TopDocs.html)
+//! - [facet counts](./struct.FacetCollector.html)
+//!
+//! At one point in your code, you will trigger the actual search operation by calling
+//! [the `search(...)` method of your `Searcher` object](../struct.Searcher.html#method.search).
+//! This call will look like this.
+//!
+//! ```verbatim
+//! let fruit = searcher.search(&query, &collector)?;
+//! ```
+//!
+//! Here the type of fruit is actually determined as an associated type of the collector
+//! (`Collector::Fruit`).
+//!
+//!
+//! # Combining several collectors
+//!
+//! A rich search experience often requires to run several collectors on your search query.
+//! For instance,
+//! - selecting the top-K products matching your query
+//! - counting the matching documents
+//! - computing several facets
+//! - computing statistics about the matching product prices
+//!
+//! A simple and efficient way to do that is to pass your collectors as one tuple.
+//! The resulting `Fruit` will then be a typed tuple with each collector's original fruits
+//! in their respective position.
+//!
+//! ```rust
+//! # use tantivy::schema::*;
+//! # use tantivy::*;
+//! # use tantivy::query::*;
+//! use tantivy::collector::{Count, TopDocs};
+//! #
+//! # fn main() -> tantivy::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.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)>) =
+//! searcher.search(&query, &(Count, TopDocs::with_limit(2)))?;
+//! #     Ok(())
+//! # }
+//! ```
+//!
+//! 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`](./struct.MultiCollector.html).
+//!
+//! # Combining several collectors dynamically
+//!
+//! Combining collectors into a tuple is a zero-cost abstraction: everything
+//! happens as if you had manually implemented a single collector
+//! combining all of our features.
+//!
+//! Unfortunately it requires you to know at compile time your collector types.
+//! If on the other hand, the collectors depend on some query parameter,
+//! you can rely on `MultiCollector`'s.
+//!
+//!
+//! # Implementing your own collectors.
+//!
+//! See the `custom_collector` example.
 
-# Collectors
-
-Collectors define the information you want to extract from the documents matching the queries.
-In tantivy jargon, we call this information your search "fruit".
-
-Your fruit could for instance be :
-- [the count of matching documents](./struct.Count.html)
-- [the top 10 documents, by relevancy or by a fast field](./struct.TopDocs.html)
-- [facet counts](./struct.FacetCollector.html)
-
-At one point in your code, you will trigger the actual search operation by calling
-[the `search(...)` method of your `Searcher` object](../struct.Searcher.html#method.search).
-This call will look like this.
-
-```verbatim
-let fruit = searcher.search(&query, &collector)?;
-```
-
-Here the type of fruit is actually determined as an associated type of the collector (`Collector::Fruit`).
-
-
-# Combining several collectors
-
-A rich search experience often requires to run several collectors on your search query.
-For instance,
-- selecting the top-K products matching your query
-- counting the matching documents
-- computing several facets
-- computing statistics about the matching product prices
-
-A simple and efficient way to do that is to pass your collectors as one tuple.
-The resulting `Fruit` will then be a typed tuple with each collector's original fruits
-in their respective position.
-
-```rust
-# use tantivy::schema::*;
-# use tantivy::*;
-# use tantivy::query::*;
-use tantivy::collector::{Count, TopDocs};
-#
-# fn main() -> tantivy::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.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)>) =
-    searcher.search(&query, &(Count, TopDocs::with_limit(2)))?;
-#     Ok(())
-# }
-```
-
-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`](./struct.MultiCollector.html).
-
-# Combining several collectors dynamically
-
-Combining collectors into a tuple is a zero-cost abstraction: everything
-happens as if you had manually implemented a single collector
-combining all of our features.
-
-Unfortunately it requires you to know at compile time your collector types.
-If on the other hand, the collectors depend on some query parameter,
-you can rely on `MultiCollector`'s.
-
-
-# Implementing your own collectors.
-
-See the `custom_collector` example.
-
-*/
-
-use crate::DocId;
-use crate::Score;
-use crate::SegmentLocalId;
-use crate::SegmentReader;
 use downcast_rs::impl_downcast;
 
+use crate::{DocId, Score, SegmentOrdinal, SegmentReader};
+
 mod count_collector;
 pub use self::count_collector::Count;
 
+mod histogram_collector;
+pub use histogram_collector::HistogramCollector;
+
 mod multi_collector;
 pub use self::multi_collector::MultiCollector;
 
@@ -108,8 +106,7 @@ mod tweak_score_top_collector;
 pub use self::tweak_score_top_collector::{ScoreSegmentTweaker, ScoreTweaker};
 
 mod facet_collector;
-pub use self::facet_collector::FacetCollector;
-pub use self::facet_collector::FacetCounts;
+pub use self::facet_collector::{FacetCollector, FacetCounts};
 use crate::query::Weight;
 
 mod docset_collector;
@@ -152,7 +149,7 @@ pub trait Collector: Sync + Send {
     /// on this segment.
     fn for_segment(
         &self,
-        segment_local_id: SegmentLocalId,
+        segment_local_id: SegmentOrdinal,
         segment: &SegmentReader,
     ) -> crate::Result<Self::Child>;
 
@@ -175,9 +172,9 @@ pub trait Collector: Sync + Send {
     ) -> 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() {
+        if let Some(alive_bitset) = reader.alive_bitset() {
             weight.for_each(reader, &mut |doc, score| {
-                if delete_bitset.is_alive(doc) {
+                if alive_bitset.is_alive(doc) {
                     segment_collector.collect(doc, score);
                 }
             })?;
@@ -190,6 +187,61 @@ pub trait Collector: Sync + Send {
     }
 }
 
+impl<TSegmentCollector: SegmentCollector> SegmentCollector for Option<TSegmentCollector> {
+    type Fruit = Option<TSegmentCollector::Fruit>;
+
+    fn collect(&mut self, doc: DocId, score: Score) {
+        if let Some(segment_collector) = self {
+            segment_collector.collect(doc, score);
+        }
+    }
+
+    fn harvest(self) -> Self::Fruit {
+        self.map(|segment_collector| segment_collector.harvest())
+    }
+}
+
+impl<TCollector: Collector> Collector for Option<TCollector> {
+    type Fruit = Option<TCollector::Fruit>;
+
+    type Child = Option<<TCollector as Collector>::Child>;
+
+    fn for_segment(
+        &self,
+        segment_local_id: SegmentOrdinal,
+        segment: &SegmentReader,
+    ) -> crate::Result<Self::Child> {
+        Ok(if let Some(inner) = self {
+            let inner_segment_collector = inner.for_segment(segment_local_id, segment)?;
+            Some(inner_segment_collector)
+        } else {
+            None
+        })
+    }
+
+    fn requires_scoring(&self) -> bool {
+        self.as_ref()
+            .map(|inner| inner.requires_scoring())
+            .unwrap_or(false)
+    }
+
+    fn merge_fruits(
+        &self,
+        segment_fruits: Vec<<Self::Child as SegmentCollector>::Fruit>,
+    ) -> crate::Result<Self::Fruit> {
+        if let Some(inner) = self.as_ref() {
+            let inner_segment_fruits: Vec<_> = segment_fruits
+                .into_iter()
+                .flat_map(|fruit_opt| fruit_opt.into_iter())
+                .collect();
+            let fruit = inner.merge_fruits(inner_segment_fruits)?;
+            Ok(Some(fruit))
+        } else {
+            Ok(None)
+        }
+    }
+}
+
 /// The `SegmentCollector` is the trait in charge of defining the
 /// collect operation at the scale of the segment.
 ///

src/collector/multi_collector.rs

@@ -1,14 +1,10 @@
-use super::Collector;
-use super::SegmentCollector;
-use crate::collector::Fruit;
-use crate::DocId;
-use crate::Score;
-use crate::SegmentLocalId;
-use crate::SegmentReader;
-use crate::TantivyError;
 use std::marker::PhantomData;
 use std::ops::Deref;
 
+use super::{Collector, SegmentCollector};
+use crate::collector::Fruit;
+use crate::{DocId, Score, SegmentOrdinal, SegmentReader, TantivyError};
+
 pub struct MultiFruit {
     sub_fruits: Vec<Option<Box<dyn Fruit>>>,
 }
@@ -104,7 +100,8 @@ impl<TFruit: Fruit> FruitHandle<TFruit> {
 ///
 /// 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).
+/// [Combining several collectors section of the collector
+/// documentation](./index.html#combining-several-collectors).
 ///
 /// ```rust
 /// use tantivy::collector::{Count, TopDocs, MultiCollector};
@@ -112,19 +109,19 @@ impl<TFruit: Fruit> FruitHandle<TFruit> {
 /// use tantivy::schema::{Schema, TEXT};
 /// use tantivy::{doc, Index};
 ///
+/// # fn main() -> tantivy::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()?;
 ///
-/// 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 reader = index.reader().unwrap();
+/// let reader = index.reader()?;
 /// let searcher = reader.searcher();
 ///
 /// let mut collectors = MultiCollector::new();
@@ -139,6 +136,8 @@ impl<TFruit: Fruit> FruitHandle<TFruit> {
 ///
 /// assert_eq!(count, 2);
 /// assert_eq!(top_docs.len(), 2);
+/// # Ok(())
+/// # }
 /// ```
 #[allow(clippy::type_complexity)]
 #[derive(Default)]
@@ -175,7 +174,7 @@ impl<'a> Collector for MultiCollector<'a> {
 
     fn for_segment(
         &self,
-        segment_local_id: SegmentLocalId,
+        segment_local_id: SegmentOrdinal,
         segment: &SegmentReader,
     ) -> crate::Result<MultiCollectorChild> {
         let children = self
@@ -246,39 +245,38 @@ mod tests {
     use super::*;
     use crate::collector::{Count, TopDocs};
     use crate::query::TermQuery;
-    use crate::schema::IndexRecordOption;
-    use crate::schema::{Schema, TEXT};
-    use crate::Index;
-    use crate::Term;
+    use crate::schema::{IndexRecordOption, Schema, TEXT};
+    use crate::{Index, Term};
 
     #[test]
-    fn test_multi_collector() {
+    fn test_multi_collector() -> crate::Result<()> {
         let mut schema_builder = Schema::builder();
         let text = schema_builder.add_text_field("text", TEXT);
         let schema = schema_builder.build();
 
         let index = Index::create_in_ram(schema);
         {
-            let mut index_writer = index.writer_for_tests().unwrap();
-            index_writer.add_document(doc!(text=>"abc"));
-            index_writer.add_document(doc!(text=>"abc abc abc"));
-            index_writer.add_document(doc!(text=>"abc abc"));
-            index_writer.commit().unwrap();
-            index_writer.add_document(doc!(text=>""));
-            index_writer.add_document(doc!(text=>"abc abc abc abc"));
-            index_writer.add_document(doc!(text=>"abc"));
-            index_writer.commit().unwrap();
+            let mut index_writer = index.writer_for_tests()?;
+            index_writer.add_document(doc!(text=>"abc"))?;
+            index_writer.add_document(doc!(text=>"abc abc abc"))?;
+            index_writer.add_document(doc!(text=>"abc abc"))?;
+            index_writer.commit()?;
+            index_writer.add_document(doc!(text=>""))?;
+            index_writer.add_document(doc!(text=>"abc abc abc abc"))?;
+            index_writer.add_document(doc!(text=>"abc"))?;
+            index_writer.commit()?;
         }
-        let searcher = index.reader().unwrap().searcher();
+        let searcher = index.reader()?.searcher();
         let term = Term::from_field_text(text, "abc");
         let query = TermQuery::new(term, IndexRecordOption::Basic);
 
         let mut collectors = MultiCollector::new();
         let topdocs_handler = collectors.add_collector(TopDocs::with_limit(2));
         let count_handler = collectors.add_collector(Count);
-        let mut multifruits = searcher.search(&query, &mut collectors).unwrap();
+        let mut multifruits = searcher.search(&query, &collectors).unwrap();
 
         assert_eq!(count_handler.extract(&mut multifruits), 5);
         assert_eq!(topdocs_handler.extract(&mut multifruits).len(), 2);
+        Ok(())
     }
 }

src/collector/tests.rs

@@ -1,20 +1,13 @@
-use super::*;
-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;
-
-use crate::collector::{FilterCollector, TopDocs};
-use crate::query::QueryParser;
-use crate::schema::{Schema, FAST, TEXT};
-use crate::DateTime;
-use crate::{doc, Index};
 use std::str::FromStr;
 
+use super::*;
+use crate::collector::{Count, FilterCollector, TopDocs};
+use crate::core::SegmentReader;
+use crate::fastfield::{BytesFastFieldReader, DynamicFastFieldReader, FastFieldReader};
+use crate::query::{AllQuery, QueryParser};
+use crate::schema::{Field, Schema, FAST, TEXT};
+use crate::{doc, DateTime, DocAddress, DocId, Document, Index, Score, Searcher, SegmentOrdinal};
+
 pub const TEST_COLLECTOR_WITH_SCORE: TestCollector = TestCollector {
     compute_score: true,
 };
@@ -24,7 +17,7 @@ pub const TEST_COLLECTOR_WITHOUT_SCORE: TestCollector = TestCollector {
 };
 
 #[test]
-pub fn test_filter_collector() {
+pub fn test_filter_collector() -> crate::Result<()> {
     let mut schema_builder = Schema::builder();
     let title = schema_builder.add_text_field("title", TEXT);
     let price = schema_builder.add_u64_field("price", FAST);
@@ -32,28 +25,28 @@ pub fn test_filter_collector() {
     let schema = schema_builder.build();
     let index = Index::create_in_ram(schema);
 
-    let mut index_writer = index.writer_with_num_threads(1, 10_000_000).unwrap();
-    index_writer.add_document(doc!(title => "The Name of the Wind", price => 30_200u64, date => DateTime::from_str("1898-04-09T00:00:00+00:00").unwrap()));
-    index_writer.add_document(doc!(title => "The Diary of Muadib", price => 29_240u64, date => DateTime::from_str("2020-04-09T00:00:00+00:00").unwrap()));
-    index_writer.add_document(doc!(title => "The Diary of Anne Frank", price => 18_240u64, date => DateTime::from_str("2019-04-20T00:00:00+00:00").unwrap()));
-    index_writer.add_document(doc!(title => "A Dairy Cow", price => 21_240u64, date => DateTime::from_str("2019-04-09T00:00:00+00:00").unwrap()));
-    index_writer.add_document(doc!(title => "The Diary of a Young Girl", price => 20_120u64, date => DateTime::from_str("2018-04-09T00:00:00+00:00").unwrap()));
-    assert!(index_writer.commit().is_ok());
+    let mut index_writer = index.writer_with_num_threads(1, 10_000_000)?;
+    index_writer.add_document(doc!(title => "The Name of the Wind", price => 30_200u64, date => DateTime::from_str("1898-04-09T00:00:00+00:00").unwrap()))?;
+    index_writer.add_document(doc!(title => "The Diary of Muadib", price => 29_240u64, date => DateTime::from_str("2020-04-09T00:00:00+00:00").unwrap()))?;
+    index_writer.add_document(doc!(title => "The Diary of Anne Frank", price => 18_240u64, date => DateTime::from_str("2019-04-20T00:00:00+00:00").unwrap()))?;
+    index_writer.add_document(doc!(title => "A Dairy Cow", price => 21_240u64, date => DateTime::from_str("2019-04-09T00:00:00+00:00").unwrap()))?;
+    index_writer.add_document(doc!(title => "The Diary of a Young Girl", price => 20_120u64, date => DateTime::from_str("2018-04-09T00:00:00+00:00").unwrap()))?;
+    index_writer.commit()?;
 
-    let reader = index.reader().unwrap();
+    let reader = index.reader()?;
     let searcher = reader.searcher();
 
     let query_parser = QueryParser::for_index(&index, vec![title]);
-    let query = query_parser.parse_query("diary").unwrap();
+    let query = query_parser.parse_query("diary")?;
     let filter_some_collector = FilterCollector::new(
         price,
         &|value: u64| value > 20_120u64,
         TopDocs::with_limit(2),
     );
-    let top_docs = searcher.search(&query, &filter_some_collector).unwrap();
+    let top_docs = searcher.search(&query, &filter_some_collector)?;
 
     assert_eq!(top_docs.len(), 1);
-    assert_eq!(top_docs[0].1, DocAddress(0, 1));
+    assert_eq!(top_docs[0].1, DocAddress::new(0, 1));
 
     let filter_all_collector: FilterCollector<_, _, u64> =
         FilterCollector::new(price, &|value| value < 5u64, TopDocs::with_limit(2));
@@ -66,9 +59,10 @@ pub fn test_filter_collector() {
     }
 
     let filter_dates_collector = FilterCollector::new(date, &date_filter, TopDocs::with_limit(5));
-    let filtered_date_docs = searcher.search(&query, &filter_dates_collector).unwrap();
+    let filtered_date_docs = searcher.search(&query, &filter_dates_collector)?;
 
     assert_eq!(filtered_date_docs.len(), 2);
+    Ok(())
 }
 
 /// Stores all of the doc ids.
@@ -82,7 +76,7 @@ pub struct TestCollector {
 }
 
 pub struct TestSegmentCollector {
-    segment_id: SegmentLocalId,
+    segment_id: SegmentOrdinal,
     fruit: TestFruit,
 }
 
@@ -108,7 +102,7 @@ impl Collector for TestCollector {
 
     fn for_segment(
         &self,
-        segment_id: SegmentLocalId,
+        segment_id: SegmentOrdinal,
         _reader: &SegmentReader,
     ) -> crate::Result<TestSegmentCollector> {
         Ok(TestSegmentCollector {
@@ -126,7 +120,7 @@ impl Collector for TestCollector {
             if fruit.docs().is_empty() {
                 0
             } else {
-                fruit.docs()[0].segment_ord()
+                fruit.docs()[0].segment_ord
             }
         });
         let mut docs = vec![];
@@ -143,7 +137,7 @@ impl SegmentCollector for TestSegmentCollector {
     type Fruit = TestFruit;
 
     fn collect(&mut self, doc: DocId, score: Score) {
-        self.fruit.docs.push(DocAddress(self.segment_id, doc));
+        self.fruit.docs.push(DocAddress::new(self.segment_id, doc));
         self.fruit.scores.push(score);
     }
 
@@ -162,7 +156,7 @@ pub struct FastFieldTestCollector {
 
 pub struct FastFieldSegmentCollector {
     vals: Vec<u64>,
-    reader: FastFieldReader<u64>,
+    reader: DynamicFastFieldReader<u64>,
 }
 
 impl FastFieldTestCollector {
@@ -177,7 +171,7 @@ impl Collector for FastFieldTestCollector {
 
     fn for_segment(
         &self,
-        _: SegmentLocalId,
+        _: SegmentOrdinal,
         segment_reader: &SegmentReader,
     ) -> crate::Result<FastFieldSegmentCollector> {
         let reader = segment_reader
@@ -268,3 +262,30 @@ impl SegmentCollector for BytesFastFieldSegmentCollector {
         self.vals
     }
 }
+
+fn make_test_searcher() -> crate::Result<crate::LeasedItem<Searcher>> {
+    let schema = Schema::builder().build();
+    let index = Index::create_in_ram(schema);
+    let mut index_writer = index.writer_for_tests()?;
+    index_writer.add_document(Document::default())?;
+    index_writer.add_document(Document::default())?;
+    index_writer.commit()?;
+    Ok(index.reader()?.searcher())
+}
+
+#[test]
+fn test_option_collector_some() -> crate::Result<()> {
+    let searcher = make_test_searcher()?;
+    let counts = searcher.search(&AllQuery, &Some(Count))?;
+    assert_eq!(counts, Some(2));
+    Ok(())
+}
+
+#[test]
+fn test_option_collector_none() -> crate::Result<()> {
+    let searcher = make_test_searcher()?;
+    let none_collector: Option<Count> = None;
+    let counts = searcher.search(&AllQuery, &none_collector)?;
+    assert_eq!(counts, None);
+    Ok(())
+}

src/collector/top_collector.rs

@@ -1,11 +1,9 @@
-use crate::DocAddress;
-use crate::DocId;
-use crate::SegmentLocalId;
-use crate::SegmentReader;
 use std::cmp::Ordering;
 use std::collections::BinaryHeap;
 use std::marker::PhantomData;
 
+use crate::{DocAddress, DocId, SegmentOrdinal, 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
@@ -62,17 +60,14 @@ pub(crate) struct TopCollector<T> {
 }
 
 impl<T> TopCollector<T>
-where
-    T: PartialOrd + Clone,
+where T: PartialOrd + Clone
 {
     /// Creates a top collector, with a number of documents equal to "limit".
     ///
     /// # Panics
     /// The method panics if limit is 0
     pub fn with_limit(limit: usize) -> TopCollector<T> {
-        if limit < 1 {
-            panic!("Limit must be strictly greater than 0.");
-        }
+        assert!(limit >= 1, "Limit must be strictly greater than 0.");
         Self {
             limit,
             offset: 0,
@@ -118,13 +113,10 @@ where
 
     pub(crate) fn for_segment<F: PartialOrd>(
         &self,
-        segment_id: SegmentLocalId,
+        segment_id: SegmentOrdinal,
         _: &SegmentReader,
-    ) -> crate::Result<TopSegmentCollector<F>> {
-        Ok(TopSegmentCollector::new(
-            segment_id,
-            self.limit + self.offset,
-        ))
+    ) -> TopSegmentCollector<F> {
+        TopSegmentCollector::new(segment_id, self.limit + self.offset)
     }
 
     /// Create a new TopCollector with the same limit and offset.
@@ -150,37 +142,39 @@ where
 pub(crate) struct TopSegmentCollector<T> {
     limit: usize,
     heap: BinaryHeap<ComparableDoc<T, DocId>>,
-    segment_id: u32,
+    segment_ord: u32,
 }
 
 impl<T: PartialOrd> TopSegmentCollector<T> {
-    fn new(segment_id: SegmentLocalId, limit: usize) -> TopSegmentCollector<T> {
+    fn new(segment_ord: SegmentOrdinal, limit: usize) -> TopSegmentCollector<T> {
         TopSegmentCollector {
             limit,
             heap: BinaryHeap::with_capacity(limit),
-            segment_id,
+            segment_ord,
         }
     }
 }
 
 impl<T: PartialOrd + Clone> TopSegmentCollector<T> {
     pub fn harvest(self) -> Vec<(T, DocAddress)> {
-        let segment_id = self.segment_id;
+        let segment_ord = self.segment_ord;
         self.heap
             .into_sorted_vec()
             .into_iter()
             .map(|comparable_doc| {
                 (
                     comparable_doc.feature,
-                    DocAddress(segment_id, comparable_doc.doc),
+                    DocAddress {
+                        segment_ord,
+                        doc_id: comparable_doc.doc,
+                    },
                 )
             })
             .collect()
     }
 
-    /// Return true iff at least K documents have gone through
-    /// the collector.
-    #[inline(always)]
+    /// Return true if more documents have been collected than the limit.
+    #[inline]
     pub(crate) fn at_capacity(&self) -> bool {
         self.heap.len() >= self.limit
     }
@@ -189,7 +183,7 @@ impl<T: PartialOrd + Clone> TopSegmentCollector<T> {
     ///
     /// It collects documents until it has reached the max capacity. Once it reaches capacity, it
     /// will compare the lowest scoring item with the given one and keep whichever is greater.
-    #[inline(always)]
+    #[inline]
     pub fn collect(&mut self, doc: DocId, feature: T) {
         if self.at_capacity() {
             // It's ok to unwrap as long as a limit of 0 is forbidden.
@@ -223,9 +217,9 @@ mod tests {
         assert_eq!(
             top_collector.harvest(),
             vec![
-                (0.8, DocAddress(0, 1)),
-                (0.3, DocAddress(0, 5)),
-                (0.2, DocAddress(0, 3))
+                (0.8, DocAddress::new(0, 1)),
+                (0.3, DocAddress::new(0, 5)),
+                (0.2, DocAddress::new(0, 3))
             ]
         );
     }
@@ -241,10 +235,10 @@ mod tests {
         assert_eq!(
             top_collector.harvest(),
             vec![
-                (0.9, DocAddress(0, 7)),
-                (0.8, DocAddress(0, 1)),
-                (0.3, DocAddress(0, 5)),
-                (0.2, DocAddress(0, 3))
+                (0.9, DocAddress::new(0, 7)),
+                (0.8, DocAddress::new(0, 1)),
+                (0.3, DocAddress::new(0, 5)),
+                (0.2, DocAddress::new(0, 3))
             ]
         );
     }
@@ -255,7 +249,7 @@ mod tests {
         // 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 score = 3.3f32;
 
         let mut top_collector_limit_2 = TopSegmentCollector::new(0, 2);
         for id in &doc_ids_collection {
@@ -279,17 +273,17 @@ mod tests {
 
         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)),
+                (0.9, DocAddress::new(0, 1)),
+                (0.8, DocAddress::new(0, 2)),
+                (0.7, DocAddress::new(0, 3)),
+                (0.6, DocAddress::new(0, 4)),
+                (0.5, DocAddress::new(0, 5)),
             ]])
             .unwrap();
 
         assert_eq!(
             results,
-            vec![(0.8, DocAddress(0, 2)), (0.7, DocAddress(0, 3)),]
+            vec![(0.8, DocAddress::new(0, 2)), (0.7, DocAddress::new(0, 3)),]
         );
     }
 
@@ -298,10 +292,13 @@ mod tests {
         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))]])
+            .merge_fruits(vec![vec![
+                (0.9, DocAddress::new(0, 1)),
+                (0.8, DocAddress::new(0, 2)),
+            ]])
             .unwrap();
 
-        assert_eq!(results, vec![(0.8, DocAddress(0, 2)),]);
+        assert_eq!(results, vec![(0.8, DocAddress::new(0, 2)),]);
     }
 
     #[test]
@@ -309,7 +306,10 @@ mod tests {
         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))]])
+            .merge_fruits(vec![vec![
+                (0.9, DocAddress::new(0, 1)),
+                (0.8, DocAddress::new(0, 2)),
+            ]])
             .unwrap();
 
         assert_eq!(results, vec![]);
@@ -318,9 +318,10 @@ mod tests {
 
 #[cfg(all(test, feature = "unstable"))]
 mod bench {
-    use super::TopSegmentCollector;
     use test::Bencher;
 
+    use super::TopSegmentCollector;
+
     #[bench]
     fn bench_top_segment_collector_collect_not_at_capacity(b: &mut Bencher) {
         let mut top_collector = TopSegmentCollector::new(0, 400);

src/collector/top_score_collector.rs

@@ -1,21 +1,18 @@
+use std::collections::BinaryHeap;
+use std::fmt;
+use std::marker::PhantomData;
+
 use super::Collector;
-use crate::collector::top_collector::{ComparableDoc, TopCollector};
+use crate::collector::custom_score_top_collector::CustomScoreTopCollector;
+use crate::collector::top_collector::{ComparableDoc, TopCollector, TopSegmentCollector};
 use crate::collector::tweak_score_top_collector::TweakedScoreTopCollector;
 use crate::collector::{
     CustomScorer, CustomSegmentScorer, ScoreSegmentTweaker, ScoreTweaker, SegmentCollector,
 };
-use crate::fastfield::FastFieldReader;
+use crate::fastfield::{DynamicFastFieldReader, FastFieldReader, FastValue};
 use crate::query::Weight;
 use crate::schema::Field;
-use crate::DocAddress;
-use crate::DocId;
-use crate::Score;
-use crate::SegmentLocalId;
-use crate::SegmentReader;
-use crate::{collector::custom_score_top_collector::CustomScoreTopCollector, fastfield::FastValue};
-use crate::{collector::top_collector::TopSegmentCollector, TantivyError};
-use std::fmt;
-use std::{collections::BinaryHeap, marker::PhantomData};
+use crate::{DocAddress, DocId, Score, SegmentOrdinal, SegmentReader, TantivyError};
 
 struct FastFieldConvertCollector<
     TCollector: Collector<Fruit = Vec<(u64, DocAddress)>>,
@@ -29,7 +26,7 @@ struct FastFieldConvertCollector<
 impl<TCollector, TFastValue> Collector for FastFieldConvertCollector<TCollector, TFastValue>
 where
     TCollector: Collector<Fruit = Vec<(u64, DocAddress)>>,
-    TFastValue: FastValue + 'static,
+    TFastValue: FastValue,
 {
     type Fruit = Vec<(TFastValue, DocAddress)>;
 
@@ -37,7 +34,7 @@ where
 
     fn for_segment(
         &self,
-        segment_local_id: crate::SegmentLocalId,
+        segment_local_id: crate::SegmentOrdinal,
         segment: &SegmentReader,
     ) -> crate::Result<Self::Child> {
         let schema = segment.schema();
@@ -94,27 +91,30 @@ where
 /// use tantivy::schema::{Schema, TEXT};
 /// use tantivy::{doc, DocAddress, Index};
 ///
+/// # fn main() -> tantivy::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, 10_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 index_writer = index.writer_with_num_threads(1, 10_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()?;
 ///
-/// let reader = index.reader().unwrap();
+/// let reader = index.reader()?;
 /// 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)).unwrap();
+/// let query = query_parser.parse_query("diary")?;
+/// let top_docs = searcher.search(&query, &TopDocs::with_limit(2))?;
 ///
-/// assert_eq!(top_docs[0].1, DocAddress(0, 1));
-/// assert_eq!(top_docs[1].1, DocAddress(0, 3));
+/// assert_eq!(top_docs[0].1, DocAddress::new(0, 1));
+/// assert_eq!(top_docs[1].1, DocAddress::new(0, 3));
+/// # Ok(())
+/// # }
 /// ```
 pub struct TopDocs(TopCollector<Score>);
 
@@ -129,7 +129,7 @@ impl fmt::Debug for TopDocs {
 }
 
 struct ScorerByFastFieldReader {
-    ff_reader: FastFieldReader<u64>,
+    ff_reader: DynamicFastFieldReader<u64>,
 }
 
 impl CustomSegmentScorer<u64> for ScorerByFastFieldReader {
@@ -151,7 +151,7 @@ impl CustomScorer<u64> for ScorerByField {
         // mapping is monotonic, so it is sufficient to compute our top-K docs.
         //
         // The conversion will then happen only on the top-K docs.
-        let ff_reader: FastFieldReader<u64> = segment_reader
+        let ff_reader = segment_reader
             .fast_fields()
             .typed_fast_field_reader(self.field)?;
         Ok(ScorerByFastFieldReader { ff_reader })
@@ -180,41 +180,46 @@ impl TopDocs {
     /// use tantivy::schema::{Schema, TEXT};
     /// use tantivy::{doc, DocAddress, Index};
     ///
+    /// # fn main() -> tantivy::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, 10_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 mut index_writer = index.writer_with_num_threads(1, 10_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.add_document(doc!(title => "The Diary of Lena Mukhina"))?;
+    /// index_writer.commit()?;
     ///
-    /// let reader = index.reader().unwrap();
+    /// let reader = index.reader()?;
     /// 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();
+    /// let query = query_parser.parse_query("diary")?;
+    /// let top_docs = searcher.search(&query, &TopDocs::with_limit(2).and_offset(1))?;
     ///
     /// assert_eq!(top_docs.len(), 2);
-    /// assert_eq!(top_docs[0].1, DocAddress(0, 4));
-    /// assert_eq!(top_docs[1].1, DocAddress(0, 3));
+    /// assert_eq!(top_docs[0].1, DocAddress::new(0, 4));
+    /// assert_eq!(top_docs[1].1, DocAddress::new(0, 3));
+    /// Ok(())
+    /// # }
     /// ```
+    #[must_use]
     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.
     ///
-    /// If the field is not a fast or does not exist, this method returns successfully (it is not aware of any schema).
-    /// An error will be returned at the moment of search.
+    /// If the field is not a fast or does not exist, this method returns successfully (it is not
+    /// aware of any schema). An error will be returned at the moment of search.
     ///
-    /// If the field is a FAST field but not a u64 field, search will return successfully but it will return
-    /// returns a monotonic u64-representation (ie. the order is still correct) of the requested field type.
+    /// If the field is a FAST field but not a u64 field, search will return successfully but it
+    /// will return returns a monotonic u64-representation (ie. the order is still correct) of
+    /// the requested field type.
     ///
     /// # Example
     ///
@@ -234,17 +239,17 @@ impl TopDocs {
     /// #
     /// #   let index = Index::create_in_ram(schema);
     /// #   let mut index_writer = index.writer_with_num_threads(1, 10_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());
+    /// #   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()?;
     /// #   let reader = index.reader()?;
     /// #   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))]);
+    /// #            vec![(97u64, DocAddress::new(0u32, 1)),
+    /// #                 (80u64, DocAddress::new(0u32, 3))]);
     /// #   Ok(())
     /// # }
     /// /// Searches the document matching the given query, and
@@ -289,14 +294,15 @@ impl TopDocs {
 
     /// Set top-K to rank documents by a given fast field.
     ///
-    /// If the field is not a fast field, or its field type does not match the generic type, this method does not panic,
-    /// but an explicit error will be returned at the moment of collection.
+    /// If the field is not a fast field, or its field type does not match the generic type, this
+    /// method does not panic, but an explicit error will be returned at the moment of
+    /// collection.
     ///
     /// Note that this method is a generic. The requested fast field type will be often
     /// inferred in your code by the rust compiler.
     ///
-    /// Implementation-wise, for performance reason, tantivy will manipulate the u64 representation of your fast
-    /// field until the last moment.
+    /// Implementation-wise, for performance reason, tantivy will manipulate the u64 representation
+    /// of your fast field until the last moment.
     ///
     /// # Example
     ///
@@ -316,15 +322,15 @@ impl TopDocs {
     /// #
     /// #   let index = Index::create_in_ram(schema);
     /// #   let mut index_writer = index.writer_with_num_threads(1, 10_000_000)?;
-    /// #   index_writer.add_document(doc!(title => "MadCow Inc.", rating => 92_000_000i64));
-    /// #   index_writer.add_document(doc!(title => "Zozo Cow KKK", rating => 119_000_000i64));
-    /// #   index_writer.add_document(doc!(title => "Declining Cow", rating => -63_000_000i64));
+    /// #   index_writer.add_document(doc!(title => "MadCow Inc.", rating => 92_000_000i64))?;
+    /// #   index_writer.add_document(doc!(title => "Zozo Cow KKK", rating => 119_000_000i64))?;
+    /// #   index_writer.add_document(doc!(title => "Declining Cow", rating => -63_000_000i64))?;
     /// #   assert!(index_writer.commit().is_ok());
     /// #   let reader = index.reader()?;
     /// #   let top_docs = docs_sorted_by_revenue(&reader.searcher(), &AllQuery, rating)?;
     /// #   assert_eq!(top_docs,
-    /// #            vec![(119_000_000i64, DocAddress(0, 1)),
-    /// #                 (92_000_000i64, DocAddress(0, 0))]);
+    /// #            vec![(119_000_000i64, DocAddress::new(0, 1)),
+    /// #                 (92_000_000i64, DocAddress::new(0, 0))]);
     /// #   Ok(())
     /// # }
     /// /// Searches the document matching the given query, and
@@ -361,7 +367,7 @@ impl TopDocs {
         fast_field: Field,
     ) -> impl Collector<Fruit = Vec<(TFastValue, DocAddress)>>
     where
-        TFastValue: FastValue + 'static,
+        TFastValue: FastValue,
     {
         let u64_collector = self.order_by_u64_field(fast_field);
         FastFieldConvertCollector {
@@ -401,6 +407,7 @@ impl TopDocs {
     /// # use tantivy::query::QueryParser;
     /// use tantivy::SegmentReader;
     /// use tantivy::collector::TopDocs;
+    /// use tantivy::fastfield::FastFieldReader;
     /// use tantivy::schema::Field;
     ///
     /// fn create_schema() -> Schema {
@@ -416,9 +423,9 @@ impl TopDocs {
     ///   let mut index_writer = index.writer_with_num_threads(1, 10_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.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)
     /// }
@@ -508,6 +515,7 @@ impl TopDocs {
     /// use tantivy::SegmentReader;
     /// use tantivy::collector::TopDocs;
     /// use tantivy::schema::Field;
+    /// use tantivy::fastfield::FastFieldReader;
     ///
     /// # fn create_schema() -> Schema {
     /// #    let mut schema_builder = Schema::builder();
@@ -525,9 +533,9 @@ impl TopDocs {
     /// #
     /// 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.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";
@@ -600,10 +608,10 @@ impl Collector for TopDocs {
 
     fn for_segment(
         &self,
-        segment_local_id: SegmentLocalId,
+        segment_local_id: SegmentOrdinal,
         reader: &SegmentReader,
     ) -> crate::Result<Self::Child> {
-        let collector = self.0.for_segment(segment_local_id, reader)?;
+        let collector = self.0.for_segment(segment_local_id, reader);
         Ok(TopScoreSegmentCollector(collector))
     }
 
@@ -627,10 +635,10 @@ impl Collector for TopDocs {
         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() {
+        if let Some(alive_bitset) = reader.alive_bitset() {
             let mut threshold = Score::MIN;
             weight.for_each_pruning(threshold, reader, &mut |doc, score| {
-                if delete_bitset.is_deleted(doc) {
+                if alive_bitset.is_deleted(doc) {
                     return threshold;
                 }
                 let heap_item = ComparableDoc {
@@ -671,7 +679,15 @@ impl Collector for TopDocs {
         let fruit = heap
             .into_sorted_vec()
             .into_iter()
-            .map(|cid| (cid.feature, DocAddress(segment_ord, cid.doc)))
+            .map(|cid| {
+                (
+                    cid.feature,
+                    DocAddress {
+                        segment_ord,
+                        doc_id: cid.doc,
+                    },
+                )
+            })
             .collect();
         Ok(fruit)
     }
@@ -698,25 +714,20 @@ mod tests {
     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};
+    use crate::{DocAddress, DocId, Index, IndexWriter, Score, SegmentReader};
 
-    fn make_index() -> Index {
+    fn make_index() -> crate::Result<Index> {
         let mut schema_builder = Schema::builder();
         let text_field = schema_builder.add_text_field("text", TEXT);
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
-        {
-            // writing the segment
-            let mut index_writer = index.writer_with_num_threads(1, 10_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
+        // writing the segment
+        let mut index_writer = index.writer_with_num_threads(1, 10_000_000)?;
+        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"))?;
+        index_writer.commit()?;
+        Ok(index)
     }
 
     fn assert_results_equals(results: &[(Score, DocAddress)], expected: &[(Score, DocAddress)]) {
@@ -727,30 +738,29 @@ mod tests {
     }
 
     #[test]
-    fn test_top_collector_not_at_capacity_without_offset() {
-        let index = make_index();
+    fn test_top_collector_not_at_capacity_without_offset() -> crate::Result<()> {
+        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 text_query = query_parser.parse_query("droopy tax")?;
         let score_docs: Vec<(Score, DocAddress)> = index
-            .reader()
-            .unwrap()
+            .reader()?
             .searcher()
-            .search(&text_query, &TopDocs::with_limit(4))
-            .unwrap();
+            .search(&text_query, &TopDocs::with_limit(4))?;
         assert_results_equals(
             &score_docs,
             &[
-                (0.81221175, DocAddress(0u32, 1)),
-                (0.5376842, DocAddress(0u32, 2)),
-                (0.48527452, DocAddress(0, 0)),
+                (0.81221175, DocAddress::new(0u32, 1)),
+                (0.5376842, DocAddress::new(0u32, 2)),
+                (0.48527452, DocAddress::new(0, 0)),
             ],
         );
+        Ok(())
     }
 
     #[test]
     fn test_top_collector_not_at_capacity_with_offset() {
-        let index = make_index();
+        let index = make_index().unwrap();
         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();
@@ -760,12 +770,12 @@ mod tests {
             .searcher()
             .search(&text_query, &TopDocs::with_limit(4).and_offset(2))
             .unwrap();
-        assert_results_equals(&score_docs[..], &[(0.48527452, DocAddress(0, 0))]);
+        assert_results_equals(&score_docs[..], &[(0.48527452, DocAddress::new(0, 0))]);
     }
 
     #[test]
     fn test_top_collector_at_capacity() {
-        let index = make_index();
+        let index = make_index().unwrap();
         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();
@@ -778,15 +788,15 @@ mod tests {
         assert_results_equals(
             &score_docs,
             &[
-                (0.81221175, DocAddress(0u32, 1)),
-                (0.5376842, DocAddress(0u32, 2)),
+                (0.81221175, DocAddress::new(0u32, 1)),
+                (0.5376842, DocAddress::new(0u32, 2)),
             ],
         );
     }
 
     #[test]
     fn test_top_collector_at_capacity_with_offset() {
-        let index = make_index();
+        let index = make_index().unwrap();
         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();
@@ -799,15 +809,15 @@ mod tests {
         assert_results_equals(
             &score_docs[..],
             &[
-                (0.5376842, DocAddress(0u32, 2)),
-                (0.48527452, DocAddress(0, 0)),
+                (0.5376842, DocAddress::new(0u32, 2)),
+                (0.48527452, DocAddress::new(0, 0)),
             ],
         );
     }
 
     #[test]
     fn test_top_collector_stable_sorting() {
-        let index = make_index();
+        let index = make_index().unwrap();
 
         // using AllQuery to get a constant score
         let searcher = index.reader().unwrap().searcher();
@@ -838,37 +848,44 @@ mod tests {
     const SIZE: &str = "size";
 
     #[test]
-    fn test_top_field_collector_not_at_capacity() {
+    fn test_top_field_collector_not_at_capacity() -> crate::Result<()> {
         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,
-            ));
+            index_writer
+                .add_document(doc!(
+                    title => "bottle of beer",
+                    size => 12u64,
+                ))
+                .unwrap();
+            index_writer
+                .add_document(doc!(
+                    title => "growler of beer",
+                    size => 64u64,
+                ))
+                .unwrap();
+            index_writer
+                .add_document(doc!(
+                    title => "pint of beer",
+                    size => 16u64,
+                ))
+                .unwrap();
         });
-        let searcher = index.reader().unwrap().searcher();
+        let searcher = index.reader()?.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();
+        let top_docs: Vec<(u64, DocAddress)> = searcher.search(&query, &top_collector)?;
         assert_eq!(
             &top_docs[..],
             &[
-                (64, DocAddress(0, 1)),
-                (16, DocAddress(0, 2)),
-                (12, DocAddress(0, 0))
+                (64, DocAddress::new(0, 1)),
+                (16, DocAddress::new(0, 2)),
+                (12, DocAddress::new(0, 0))
             ]
         );
+        Ok(())
     }
 
     #[test]
@@ -884,12 +901,12 @@ mod tests {
         index_writer.add_document(doc!(
             name => "Paul Robeson",
             birthday => pr_birthday
-        ));
+        ))?;
         let mr_birthday = crate::DateTime::from_str("1947-11-08T00:00:00+00:00")?;
         index_writer.add_document(doc!(
             name => "Minnie Riperton",
             birthday => mr_birthday
-        ));
+        ))?;
         index_writer.commit()?;
         let searcher = index.reader()?.searcher();
         let top_collector = TopDocs::with_limit(3).order_by_fast_field(birthday);
@@ -898,8 +915,8 @@ mod tests {
         assert_eq!(
             &top_docs[..],
             &[
-                (mr_birthday, DocAddress(0, 1)),
-                (pr_birthday, DocAddress(0, 0)),
+                (mr_birthday, DocAddress::new(0, 1)),
+                (pr_birthday, DocAddress::new(0, 0)),
             ]
         );
         Ok(())
@@ -916,18 +933,21 @@ mod tests {
         index_writer.add_document(doc!(
                 city => "georgetown",
                 altitude =>  -1i64,
-        ));
+        ))?;
         index_writer.add_document(doc!(
             city => "tokyo",
             altitude =>  40i64,
-        ));
+        ))?;
         index_writer.commit()?;
         let searcher = index.reader()?.searcher();
         let top_collector = TopDocs::with_limit(3).order_by_fast_field(altitude);
         let top_docs: Vec<(i64, DocAddress)> = searcher.search(&AllQuery, &top_collector)?;
         assert_eq!(
             &top_docs[..],
-            &[(40i64, DocAddress(0, 1)), (-1i64, DocAddress(0, 0)),]
+            &[
+                (40i64, DocAddress::new(0, 1)),
+                (-1i64, DocAddress::new(0, 0)),
+            ]
         );
         Ok(())
     }
@@ -943,18 +963,21 @@ mod tests {
         index_writer.add_document(doc!(
                 city => "georgetown",
                 altitude =>  -1.0f64,
-        ));
+        ))?;
         index_writer.add_document(doc!(
             city => "tokyo",
             altitude =>  40f64,
-        ));
+        ))?;
         index_writer.commit()?;
         let searcher = index.reader()?.searcher();
         let top_collector = TopDocs::with_limit(3).order_by_fast_field(altitude);
         let top_docs: Vec<(f64, DocAddress)> = searcher.search(&AllQuery, &top_collector)?;
         assert_eq!(
             &top_docs[..],
-            &[(40f64, DocAddress(0, 1)), (-1.0f64, DocAddress(0, 0)),]
+            &[
+                (40f64, DocAddress::new(0, 1)),
+                (-1.0f64, DocAddress::new(0, 0)),
+            ]
         );
         Ok(())
     }
@@ -967,10 +990,12 @@ mod tests {
         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,
-            ));
+            index_writer
+                .add_document(doc!(
+                    title => "bottle of beer",
+                    size => 12u64,
+                ))
+                .unwrap();
         });
         let searcher = index.reader().unwrap().searcher();
         let top_collector = TopDocs::with_limit(4).order_by_u64_field(Field::from_field_id(2));
@@ -987,7 +1012,7 @@ mod tests {
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
         let mut index_writer = index.writer_for_tests()?;
-        index_writer.add_document(doc!(size=>1u64));
+        index_writer.add_document(doc!(size=>1u64))?;
         index_writer.commit()?;
         let searcher = index.reader()?.searcher();
         let segment = searcher.segment_reader(0);
@@ -1004,7 +1029,7 @@ mod tests {
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
         let mut index_writer = index.writer_for_tests()?;
-        index_writer.add_document(doc!(size=>1u64));
+        index_writer.add_document(doc!(size=>1u64))?;
         index_writer.commit()?;
         let searcher = index.reader()?.searcher();
         let segment = searcher.segment_reader(0);
@@ -1017,30 +1042,26 @@ mod tests {
     }
 
     #[test]
-    fn test_tweak_score_top_collector_with_offset() {
-        let index = make_index();
+    fn test_tweak_score_top_collector_with_offset() -> crate::Result<()> {
+        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 text_query = query_parser.parse_query("droopy tax")?;
         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();
-
+        let score_docs: Vec<(u32, DocAddress)> =
+            index.reader()?.searcher().search(&text_query, &collector)?;
         assert_eq!(
             score_docs,
-            vec![(1, DocAddress(0, 1)), (0, DocAddress(0, 0)),]
+            vec![(1, DocAddress::new(0, 1)), (0, DocAddress::new(0, 0)),]
         );
+        Ok(())
     }
 
     #[test]
     fn test_custom_score_top_collector_with_offset() {
-        let index = make_index();
+        let index = make_index().unwrap();
         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();
@@ -1056,7 +1077,7 @@ mod tests {
 
         assert_eq!(
             score_docs,
-            vec![(1, DocAddress(0, 1)), (0, DocAddress(0, 0)),]
+            vec![(1, DocAddress::new(0, 1)), (0, DocAddress::new(0, 0)),]
         );
     }
 
@@ -1064,7 +1085,7 @@ mod tests {
         query: &str,
         query_field: Field,
         schema: Schema,
-        mut doc_adder: impl FnMut(&mut IndexWriter) -> (),
+        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, 10_000_000).unwrap();

src/collector/tweak_score_top_collector.rs

@@ -1,7 +1,6 @@
 use crate::collector::top_collector::{TopCollector, TopSegmentCollector};
 use crate::collector::{Collector, SegmentCollector};
-use crate::DocAddress;
-use crate::{DocId, Result, Score, SegmentReader};
+use crate::{DocAddress, DocId, Result, Score, SegmentReader};
 
 pub(crate) struct TweakedScoreTopCollector<TScoreTweaker, TScore = Score> {
     score_tweaker: TScoreTweaker,
@@ -9,8 +8,7 @@ pub(crate) struct TweakedScoreTopCollector<TScoreTweaker, TScore = Score> {
 }
 
 impl<TScoreTweaker, TScore> TweakedScoreTopCollector<TScoreTweaker, TScore>
-where
-    TScore: Clone + PartialOrd,
+where TScore: Clone + PartialOrd
 {
     pub fn new(
         score_tweaker: TScoreTweaker,
@@ -62,9 +60,7 @@ where
         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)?;
+        let segment_collector = self.collector.for_segment(segment_local_id, segment_reader);
         Ok(TopTweakedScoreSegmentCollector {
             segment_collector,
             segment_scorer,
@@ -120,8 +116,7 @@ where
 }
 
 impl<F, TScore> ScoreSegmentTweaker<TScore> for F
-where
-    F: 'static + FnMut(DocId, Score) -> TScore,
+where F: 'static + FnMut(DocId, Score) -> TScore
 {
     fn score(&mut self, doc: DocId, score: Score) -> TScore {
         (self)(doc, score)

src/common/bitset.rs

@@ -1,396 +0,0 @@
-use std::fmt;
-use std::u64;
-
-#[derive(Clone, Copy, Eq, PartialEq)]
-pub(crate) struct TinySet(u64);
-
-impl fmt::Debug for TinySet {
-    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
-        self.into_iter().collect::<Vec<u32>>().fmt(f)
-    }
-}
-
-pub struct TinySetIterator(TinySet);
-impl Iterator for TinySetIterator {
-    type Item = u32;
-
-    fn next(&mut self) -> Option<Self::Item> {
-        self.0.pop_lowest()
-    }
-}
-
-impl IntoIterator for TinySet {
-    type Item = u32;
-    type IntoIter = TinySetIterator;
-    fn into_iter(self) -> Self::IntoIter {
-        TinySetIterator(self)
-    }
-}
-
-impl TinySet {
-    /// Returns an empty `TinySet`.
-    pub fn empty() -> 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)
-    }
-
-    /// Returns true iff the `TinySet` contains the element `el`.
-    pub fn contains(self, el: u32) -> bool {
-        !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)
-    }
-
-    /// Creates a new `TinySet` containing only one element
-    /// within `[0; 64[`
-    #[inline(always)]
-    pub fn singleton(el: u32) -> TinySet {
-        TinySet(1u64 << u64::from(el))
-    }
-
-    /// Insert a new element within [0..64[
-    #[inline(always)]
-    pub fn insert(self, el: u32) -> TinySet {
-        self.union(TinySet::singleton(el))
-    }
-
-    /// Insert a new element within [0..64[
-    #[inline(always)]
-    pub fn insert_mut(&mut self, el: u32) -> bool {
-        let old = *self;
-        *self = old.insert(el);
-        old != *self
-    }
-
-    /// Returns the union of two tinysets
-    #[inline(always)]
-    pub fn union(self, other: TinySet) -> TinySet {
-        TinySet(self.0 | other.0)
-    }
-
-    /// Returns true iff the `TinySet` is empty.
-    #[inline(always)]
-    pub fn is_empty(self) -> bool {
-        self.0 == 0u64
-    }
-
-    /// Returns the lowest element in the `TinySet`
-    /// and removes it.
-    #[inline(always)]
-    pub fn pop_lowest(&mut self) -> Option<u32> {
-        if self.is_empty() {
-            None
-        } else {
-            let lowest = self.0.trailing_zeros() as u32;
-            self.0 ^= TinySet::singleton(lowest).0;
-            Some(lowest)
-        }
-    }
-
-    /// Returns a `TinySet` than contains all values up
-    /// to limit excluded.
-    ///
-    /// The limit is assumed to be strictly lower than 64.
-    pub fn range_lower(upper_bound: u32) -> TinySet {
-        TinySet((1u64 << u64::from(upper_bound % 64u32)) - 1u64)
-    }
-
-    /// Returns a `TinySet` that contains all values greater
-    /// or equal to the given limit, included. (and up to 63)
-    ///
-    /// The limit is assumed to be strictly lower than 64.
-    pub fn range_greater_or_equal(from_included: u32) -> TinySet {
-        TinySet::range_lower(from_included).complement()
-    }
-}
-
-#[derive(Clone)]
-pub struct BitSet {
-    tinysets: Box<[TinySet]>,
-    len: usize,
-    max_value: u32,
-}
-
-fn num_buckets(max_val: u32) -> u32 {
-    (max_val + 63u32) / 64u32
-}
-
-impl BitSet {
-    /// Create a new `BitSet` that may contain elements
-    /// within `[0, max_val[`.
-    pub fn with_max_value(max_value: u32) -> BitSet {
-        let num_buckets = num_buckets(max_value);
-        let tinybisets = vec![TinySet::empty(); num_buckets as usize].into_boxed_slice();
-        BitSet {
-            tinysets: tinybisets,
-            len: 0,
-            max_value,
-        }
-    }
-
-    /// Removes all elements from the `BitSet`.
-    pub fn clear(&mut self) {
-        for tinyset in self.tinysets.iter_mut() {
-            *tinyset = TinySet::empty();
-        }
-    }
-
-    /// Returns the number of elements in the `BitSet`.
-    pub fn len(&self) -> usize {
-        self.len
-    }
-
-    /// Inserts an element in the `BitSet`
-    pub fn insert(&mut self, el: u32) {
-        // we do not check saturated els.
-        let higher = el / 64u32;
-        let lower = el % 64u32;
-        self.len += if self.tinysets[higher as usize].insert_mut(lower) {
-            1
-        } else {
-            0
-        };
-    }
-
-    /// Returns true iff the elements is in the `BitSet`.
-    pub fn contains(&self, el: u32) -> bool {
-        self.tinyset(el / 64u32).contains(el % 64)
-    }
-
-    /// Returns the first non-empty `TinySet` associated to a bucket lower
-    /// or greater than bucket.
-    ///
-    /// Reminder: the tiny set with the bucket `bucket`, represents the
-    /// elements from `bucket * 64` to `(bucket+1) * 64`.
-    pub(crate) fn first_non_empty_bucket(&self, bucket: u32) -> Option<u32> {
-        self.tinysets[bucket as usize..]
-            .iter()
-            .cloned()
-            .position(|tinyset| !tinyset.is_empty())
-            .map(|delta_bucket| bucket + delta_bucket as u32)
-    }
-
-    pub fn max_value(&self) -> u32 {
-        self.max_value
-    }
-
-    /// Returns the tiny bitset representing the
-    /// the set restricted to the number range from
-    /// `bucket * 64` to `(bucket + 1) * 64`.
-    pub(crate) fn tinyset(&self, bucket: u32) -> TinySet {
-        self.tinysets[bucket as usize]
-    }
-}
-
-#[cfg(test)]
-mod tests {
-
-    use super::BitSet;
-    use super::TinySet;
-    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;
-
-    #[test]
-    fn test_tiny_set() {
-        assert!(TinySet::empty().is_empty());
-        {
-            let mut u = TinySet::empty().insert(1u32);
-            assert_eq!(u.pop_lowest(), Some(1u32));
-            assert!(u.pop_lowest().is_none())
-        }
-        {
-            let mut u = TinySet::empty().insert(1u32).insert(1u32);
-            assert_eq!(u.pop_lowest(), Some(1u32));
-            assert!(u.pop_lowest().is_none())
-        }
-        {
-            let mut u = TinySet::empty().insert(2u32);
-            assert_eq!(u.pop_lowest(), Some(2u32));
-            u.insert_mut(1u32);
-            assert_eq!(u.pop_lowest(), Some(1u32));
-            assert!(u.pop_lowest().is_none());
-        }
-        {
-            let mut u = TinySet::empty().insert(63u32);
-            assert_eq!(u.pop_lowest(), Some(63u32));
-            assert!(u.pop_lowest().is_none());
-        }
-    }
-
-    #[test]
-    fn test_bitset() {
-        let test_against_hashset = |els: &[u32], max_value: u32| {
-            let mut hashset: HashSet<u32> = HashSet::new();
-            let mut bitset = BitSet::with_max_value(max_value);
-            for &el in els {
-                assert!(el < max_value);
-                hashset.insert(el);
-                bitset.insert(el);
-            }
-            for el in 0..max_value {
-                assert_eq!(hashset.contains(&el), bitset.contains(el));
-            }
-            assert_eq!(bitset.max_value(), max_value);
-        };
-
-        test_against_hashset(&[], 0);
-        test_against_hashset(&[], 1);
-        test_against_hashset(&[0u32], 1);
-        test_against_hashset(&[0u32], 100);
-        test_against_hashset(&[1u32, 2u32], 4);
-        test_against_hashset(&[99u32], 100);
-        test_against_hashset(&[63u32], 64);
-        test_against_hashset(&[62u32, 63u32], 64);
-    }
-
-    #[test]
-    fn test_bitset_large() {
-        let arr = generate_nonunique_unsorted(100_000, 5_000);
-        let mut btreeset: BTreeSet<u32> = BTreeSet::new();
-        let mut bitset = BitSet::with_max_value(100_000);
-        for el in arr {
-            btreeset.insert(el);
-            bitset.insert(el);
-        }
-        for i in 0..100_000 {
-            assert_eq!(btreeset.contains(&i), bitset.contains(i));
-        }
-        assert_eq!(btreeset.len(), bitset.len());
-        let mut bitset_docset = BitSetDocSet::from(bitset);
-        let mut remaining = true;
-        for el in btreeset.into_iter() {
-            assert!(remaining);
-            assert_eq!(bitset_docset.doc(), el);
-            remaining = bitset_docset.advance() != TERMINATED;
-        }
-        assert!(!remaining);
-    }
-
-    #[test]
-    fn test_bitset_num_buckets() {
-        use super::num_buckets;
-        assert_eq!(num_buckets(0u32), 0);
-        assert_eq!(num_buckets(1u32), 1);
-        assert_eq!(num_buckets(64u32), 1);
-        assert_eq!(num_buckets(65u32), 2);
-        assert_eq!(num_buckets(128u32), 2);
-        assert_eq!(num_buckets(129u32), 3);
-    }
-
-    #[test]
-    fn test_tinyset_range() {
-        assert_eq!(
-            TinySet::range_lower(3).into_iter().collect::<Vec<u32>>(),
-            [0, 1, 2]
-        );
-        assert!(TinySet::range_lower(0).is_empty());
-        assert_eq!(
-            TinySet::range_lower(63).into_iter().collect::<Vec<u32>>(),
-            (0u32..63u32).collect::<Vec<_>>()
-        );
-        assert_eq!(
-            TinySet::range_lower(1).into_iter().collect::<Vec<u32>>(),
-            [0]
-        );
-        assert_eq!(
-            TinySet::range_lower(2).into_iter().collect::<Vec<u32>>(),
-            [0, 1]
-        );
-        assert_eq!(
-            TinySet::range_greater_or_equal(3)
-                .into_iter()
-                .collect::<Vec<u32>>(),
-            (3u32..64u32).collect::<Vec<_>>()
-        );
-    }
-
-    #[test]
-    fn test_bitset_len() {
-        let mut bitset = BitSet::with_max_value(1_000);
-        assert_eq!(bitset.len(), 0);
-        bitset.insert(3u32);
-        assert_eq!(bitset.len(), 1);
-        bitset.insert(103u32);
-        assert_eq!(bitset.len(), 2);
-        bitset.insert(3u32);
-        assert_eq!(bitset.len(), 2);
-        bitset.insert(103u32);
-        assert_eq!(bitset.len(), 2);
-        bitset.insert(104u32);
-        assert_eq!(bitset.len(), 3);
-    }
-
-    #[test]
-    fn test_bitset_clear() {
-        let mut bitset = BitSet::with_max_value(1_000);
-        let els = tests::sample(1_000, 0.01f64);
-        for &el in &els {
-            bitset.insert(el);
-        }
-        assert!(els.iter().all(|el| bitset.contains(*el)));
-        bitset.clear();
-        for el in 0u32..1000u32 {
-            assert!(!bitset.contains(el));
-        }
-    }
-}
-
-#[cfg(all(test, feature = "unstable"))]
-mod bench {
-
-    use super::BitSet;
-    use super::TinySet;
-    use test;
-
-    #[bench]
-    fn bench_tinyset_pop(b: &mut test::Bencher) {
-        b.iter(|| {
-            let mut tinyset = TinySet::singleton(test::black_box(31u32));
-            tinyset.pop_lowest();
-            tinyset.pop_lowest();
-            tinyset.pop_lowest();
-            tinyset.pop_lowest();
-            tinyset.pop_lowest();
-            tinyset.pop_lowest();
-        });
-    }
-
-    #[bench]
-    fn bench_tinyset_sum(b: &mut test::Bencher) {
-        let tiny_set = TinySet::empty().insert(10u32).insert(14u32).insert(21u32);
-        b.iter(|| {
-            assert_eq!(test::black_box(tiny_set).into_iter().sum::<u32>(), 45u32);
-        });
-    }
-
-    #[bench]
-    fn bench_tinyarr_sum(b: &mut test::Bencher) {
-        let v = [10u32, 14u32, 21u32];
-        b.iter(|| test::black_box(v).iter().cloned().sum::<u32>());
-    }
-
-    #[bench]
-    fn bench_bitset_initialize(b: &mut test::Bencher) {
-        b.iter(|| BitSet::with_max_value(1_000_000));
-    }
-}

src/core/executor.rs

@@ -57,7 +57,11 @@ impl Executor {
                                 let (idx, arg) = arg_with_idx;
                                 let fruit = f(arg);
                                 if let Err(err) = fruit_sender.send((idx, fruit)) {
-                                    error!("Failed to send search task. It probably means all search threads have panicked. {:?}", err);
+                                    error!(
+                                        "Failed to send search task. It probably means all search \
+                                         threads have panicked. {:?}",
+                                        err
+                                    );
                                 }
                             });
                         }

src/core/index.rs

@@ -1,35 +1,27 @@
-use super::segment::Segment;
-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::error::OpenReadError;
-use crate::directory::ManagedDirectory;
-#[cfg(feature = "mmap")]
-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::collections::HashSet;
 use std::fmt;
-
 #[cfg(feature = "mmap")]
 use std::path::Path;
 use std::path::PathBuf;
 use std::sync::Arc;
 
+use super::segment::Segment;
+use super::IndexSettings;
+use crate::core::{
+    Executor, IndexMeta, SegmentId, SegmentMeta, SegmentMetaInventory, META_FILEPATH,
+};
+use crate::directory::error::OpenReadError;
+#[cfg(feature = "mmap")]
+use crate::directory::MmapDirectory;
+use crate::directory::{Directory, ManagedDirectory, RamDirectory, INDEX_WRITER_LOCK};
+use crate::error::{DataCorruption, TantivyError};
+use crate::indexer::index_writer::{MAX_NUM_THREAD, MEMORY_ARENA_NUM_BYTES_MIN};
+use crate::indexer::segment_updater::save_new_metas;
+use crate::reader::{IndexReader, IndexReaderBuilder};
+use crate::schema::{Field, FieldType, Schema};
+use crate::tokenizer::{TextAnalyzer, TokenizerManager};
+use crate::IndexWriter;
+
 fn load_metas(
     directory: &dyn Directory,
     inventory: &SegmentMetaInventory,
@@ -42,7 +34,7 @@ fn load_metas(
             "Meta file does not contain valid utf8 file.".to_string(),
         )
     })?;
-    IndexMeta::deserialize(&meta_string, &inventory)
+    IndexMeta::deserialize(&meta_string, inventory)
         .map_err(|e| {
             DataCorruption::new(
                 META_FILEPATH.to_path_buf(),
@@ -55,21 +47,160 @@ fn load_metas(
         .map_err(From::from)
 }
 
+/// IndexBuilder can be used to create an index.
+///
+/// Use in conjunction with `SchemaBuilder`. Global index settings
+/// can be configured with `IndexSettings`
+///
+/// # Examples
+///
+/// ```
+/// use tantivy::schema::*;
+/// use tantivy::{Index, IndexSettings, IndexSortByField, Order};
+///
+/// let mut schema_builder = Schema::builder();
+/// let id_field = schema_builder.add_text_field("id", STRING);
+/// let title_field = schema_builder.add_text_field("title", TEXT);
+/// let body_field = schema_builder.add_text_field("body", TEXT);
+/// let number_field = schema_builder.add_u64_field(
+///     "number",
+///     NumericOptions::default().set_fast(Cardinality::SingleValue),
+/// );
+///
+/// let schema = schema_builder.build();
+/// let settings = IndexSettings{sort_by_field: Some(IndexSortByField{field:"number".to_string(), order:Order::Asc}), ..Default::default()};
+/// let index = Index::builder().schema(schema).settings(settings).create_in_ram();
+/// ```
+pub struct IndexBuilder {
+    schema: Option<Schema>,
+    index_settings: IndexSettings,
+}
+impl Default for IndexBuilder {
+    fn default() -> Self {
+        IndexBuilder::new()
+    }
+}
+impl IndexBuilder {
+    /// Creates a new `IndexBuilder`
+    pub fn new() -> Self {
+        Self {
+            schema: None,
+            index_settings: IndexSettings::default(),
+        }
+    }
+
+    /// Set the settings
+    #[must_use]
+    pub fn settings(mut self, settings: IndexSettings) -> Self {
+        self.index_settings = settings;
+        self
+    }
+
+    /// Set the schema
+    #[must_use]
+    pub fn schema(mut self, schema: Schema) -> Self {
+        self.schema = Some(schema);
+        self
+    }
+
+    /// Creates a new index using the `RAMDirectory`.
+    ///
+    /// The index will be allocated in anonymous memory.
+    /// This should only be used for unit tests.
+    pub fn create_in_ram(self) -> Result<Index, TantivyError> {
+        let ram_directory = RamDirectory::create();
+        Ok(self
+            .create(ram_directory)
+            .expect("Creating a RAMDirectory should never fail"))
+    }
+
+    /// Creates a new index in a given filepath.
+    /// The index will use the `MMapDirectory`.
+    ///
+    /// If a previous index was in this directory, it returns an `IndexAlreadyExists` error.
+    #[cfg(feature = "mmap")]
+    pub fn create_in_dir<P: AsRef<Path>>(self, directory_path: P) -> crate::Result<Index> {
+        let mmap_directory: Box<dyn Directory> = Box::new(MmapDirectory::open(directory_path)?);
+        if Index::exists(&*mmap_directory)? {
+            return Err(TantivyError::IndexAlreadyExists);
+        }
+        self.create(mmap_directory)
+    }
+
+    /// Creates a new index in a temp directory.
+    ///
+    /// The index will use the `MMapDirectory` in a newly created directory.
+    /// The temp directory will be destroyed automatically when the `Index` object
+    /// is destroyed.
+    ///
+    /// 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(self) -> crate::Result<Index> {
+        let mmap_directory: Box<dyn Directory> = Box::new(MmapDirectory::create_from_tempdir()?);
+        self.create(mmap_directory)
+    }
+
+    fn get_expect_schema(&self) -> crate::Result<Schema> {
+        self.schema
+            .as_ref()
+            .cloned()
+            .ok_or(TantivyError::IndexBuilderMissingArgument("schema"))
+    }
+
+    /// Opens or creates a new index in the provided directory
+    pub fn open_or_create<T: Into<Box<dyn Directory>>>(self, dir: T) -> crate::Result<Index> {
+        let dir = dir.into();
+        if !Index::exists(&*dir)? {
+            return self.create(dir);
+        }
+        let index = Index::open(dir)?;
+        if index.schema() == self.get_expect_schema()? {
+            Ok(index)
+        } else {
+            Err(TantivyError::SchemaError(
+                "An index exists but the schema does not match.".to_string(),
+            ))
+        }
+    }
+    /// Creates a new index given an implementation of the trait `Directory`.
+    ///
+    /// If a directory previously existed, it will be erased.
+    fn create<T: Into<Box<dyn Directory>>>(self, dir: T) -> crate::Result<Index> {
+        let dir = dir.into();
+        let directory = ManagedDirectory::wrap(dir)?;
+        save_new_metas(
+            self.get_expect_schema()?,
+            self.index_settings.clone(),
+            &directory,
+        )?;
+        let mut metas = IndexMeta::with_schema(self.get_expect_schema()?);
+        metas.index_settings = self.index_settings;
+        let index = Index::open_from_metas(directory, &metas, SegmentMetaInventory::default());
+        Ok(index)
+    }
+}
+
 /// Search Index
 #[derive(Clone)]
 pub struct Index {
     directory: ManagedDirectory,
     schema: Schema,
+    settings: IndexSettings,
     executor: Arc<Executor>,
     tokenizers: TokenizerManager,
     inventory: SegmentMetaInventory,
 }
 
 impl Index {
+    /// Creates a new builder.
+    pub fn builder() -> IndexBuilder {
+        IndexBuilder::new()
+    }
     /// Examines the directory to see if it contains an index.
     ///
     /// Effectively, it only checks for the presence of the `meta.json` file.
-    pub fn exists<Dir: Directory>(dir: &Dir) -> Result<bool, OpenReadError> {
+    pub fn exists(dir: &dyn Directory) -> Result<bool, OpenReadError> {
         dir.exists(&META_FILEPATH)
     }
 
@@ -86,7 +217,7 @@ 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) -> crate::Result<()> {
-        self.executor = Arc::new(Executor::multi_thread(num_threads, "thrd-tantivy-search-")?);
+        self.executor = Arc::new(Executor::multi_thread(num_threads, "tantivy-search-")?);
         Ok(())
     }
 
@@ -97,44 +228,36 @@ impl Index {
         self.set_multithread_executor(default_num_threads)
     }
 
-    /// Creates a new index using the `RAMDirectory`.
+    /// Creates a new index using the `RamDirectory`.
     ///
     /// The index will be allocated in anonymous memory.
-    /// This should only be used for unit tests.
+    /// This is useful for indexing small set of documents
+    /// for instances like unit test or temporary in memory index.
     pub fn create_in_ram(schema: Schema) -> Index {
-        let ram_directory = RAMDirectory::create();
-        Index::create(ram_directory, schema).expect("Creating a RAMDirectory should never fail")
+        IndexBuilder::new().schema(schema).create_in_ram().unwrap()
     }
 
     /// Creates a new index in a given filepath.
     /// The index will use the `MMapDirectory`.
     ///
-    /// If a previous index was in this directory, then its meta file will be destroyed.
+    /// If a previous index was in this directory, then it returns  an `IndexAlreadyExists` error.
     #[cfg(feature = "mmap")]
     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)
+        IndexBuilder::new()
+            .schema(schema)
+            .create_in_dir(directory_path)
     }
 
     /// Opens or creates a new index in the provided directory
-    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 {
-            Err(TantivyError::SchemaError(
-                "An index exists but the schema does not match.".to_string(),
-            ))
-        }
+    pub fn open_or_create<T: Into<Box<dyn Directory>>>(
+        dir: T,
+        schema: Schema,
+    ) -> crate::Result<Index> {
+        let dir = dir.into();
+        IndexBuilder::new().schema(schema).open_or_create(dir)
     }
 
     /// Creates a new index in a temp directory.
@@ -144,44 +267,41 @@ impl Index {
     /// is destroyed.
     ///
     /// The temp directory is only used for testing the `MmapDirectory`.
-    /// For other unit tests, prefer the `RAMDirectory`, see: `create_in_ram`.
+    /// For other unit tests, prefer the `RamDirectory`, see: `create_in_ram`.
     #[cfg(feature = "mmap")]
     pub fn create_from_tempdir(schema: Schema) -> crate::Result<Index> {
-        let mmap_directory = MmapDirectory::create_from_tempdir()?;
-        Index::create(mmap_directory, schema)
+        IndexBuilder::new().schema(schema).create_from_tempdir()
     }
 
     /// Creates a new index given an implementation of the trait `Directory`.
     ///
     /// If a directory previously existed, it will be erased.
-    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(directory: ManagedDirectory, schema: Schema) -> crate::Result<Index> {
-        save_new_metas(schema.clone(), &directory)?;
-        let metas = IndexMeta::with_schema(schema);
-        Index::create_from_metas(directory, &metas, SegmentMetaInventory::default())
+    pub fn create<T: Into<Box<dyn Directory>>>(
+        dir: T,
+        schema: Schema,
+        settings: IndexSettings,
+    ) -> crate::Result<Index> {
+        let dir: Box<dyn Directory> = dir.into();
+        let mut builder = IndexBuilder::new().schema(schema);
+        builder = builder.settings(settings);
+        builder.create(dir)
     }
 
     /// Creates a new index given a directory and an `IndexMeta`.
-    fn create_from_metas(
+    fn open_from_metas(
         directory: ManagedDirectory,
         metas: &IndexMeta,
         inventory: SegmentMetaInventory,
-    ) -> crate::Result<Index> {
+    ) -> Index {
         let schema = metas.schema.clone();
-        Ok(Index {
+        Index {
+            settings: metas.index_settings.clone(),
             directory,
             schema,
             tokenizers: TokenizerManager::default(),
             executor: Arc::new(Executor::single_thread()),
             inventory,
-        })
+        }
     }
 
     /// Accessor for the tokenizer manager.
@@ -238,7 +358,7 @@ impl 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> {
+    pub(crate) fn list_all_segment_metas(&self) -> Vec<SegmentMeta> {
         self.inventory.all()
     }
 
@@ -252,11 +372,13 @@ impl Index {
     }
 
     /// Open the index using the provided directory
-    pub fn open<D: Directory>(directory: D) -> crate::Result<Index> {
+    pub fn open<T: Into<Box<dyn Directory>>>(directory: T) -> crate::Result<Index> {
+        let directory = directory.into();
         let directory = ManagedDirectory::wrap(directory)?;
         let inventory = SegmentMetaInventory::default();
         let metas = load_metas(&directory, &inventory)?;
-        Index::create_from_metas(directory, &metas, inventory)
+        let index = Index::open_from_metas(directory, &metas, inventory);
+        Ok(index)
     }
 
     /// Reads the index meta file from the directory.
@@ -275,19 +397,18 @@ impl Index {
     /// - `num_threads` defines the number of indexing workers that
     /// should work at the same time.
     ///
-    /// - `overall_heap_size_in_bytes` sets the amount of memory
+    /// - `overall_memory_arena_in_bytes` sets the amount of memory
     /// allocated for all indexing thread.
-    /// Each thread will receive a budget of  `overall_heap_size_in_bytes / num_threads`.
+    /// Each thread will receive a budget of  `overall_memory_arena_in_bytes / num_threads`.
     ///
     /// # Errors
-    /// If the lockfile already exists, returns `Error::DirectoryLockBusy` or an `Error::IOError`.
-    ///
-    /// # Panics
-    /// If the heap size per thread is too small, panics.
+    /// If the lockfile already exists, returns `Error::DirectoryLockBusy` or an `Error::IoError`.
+    /// If the memory arena per thread is too small or too big, returns
+    /// `TantivyError::InvalidArgument`
     pub fn writer_with_num_threads(
         &self,
         num_threads: usize,
-        overall_heap_size_in_bytes: usize,
+        overall_memory_arena_in_bytes: usize,
     ) -> crate::Result<IndexWriter> {
         let directory_lock = self
             .directory
@@ -296,26 +417,25 @@ impl Index {
                 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."
+                        "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;
+        let memory_arena_in_bytes_per_thread = overall_memory_arena_in_bytes / num_threads;
         IndexWriter::new(
             self,
             num_threads,
-            heap_size_in_bytes_per_thread,
+            memory_arena_in_bytes_per_thread,
             directory_lock,
         )
     }
 
     /// Helper to create an index writer for tests.
     ///
-    /// That index writer only simply has a single thread and a heap of 5 MB.
+    /// That index writer only simply has a single thread and a memory arena of 10 MB.
     /// Using a single thread gives us a deterministic allocation of DocId.
     #[cfg(test)]
     pub fn writer_for_tests(&self) -> crate::Result<IndexWriter> {
@@ -324,21 +444,32 @@ impl Index {
 
     /// Creates a multithreaded writer
     ///
-    /// Tantivy will automatically define the number of threads to use.
-    /// `overall_heap_size_in_bytes` is the total target memory usage that will be split
+    /// Tantivy will automatically define the number of threads to use, but
+    /// no more than 8 threads.
+    /// `overall_memory_arena_in_bytes` is the total target memory usage that will be split
     /// between a given number of threads.
     ///
     /// # Errors
     /// 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) -> 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 {
-            num_threads = (overall_heap_size_in_bytes / HEAP_SIZE_MIN).max(1);
+    /// If the memory arena per thread is too small or too big, returns
+    /// `TantivyError::InvalidArgument`
+    pub fn writer(&self, memory_arena_num_bytes: usize) -> crate::Result<IndexWriter> {
+        let mut num_threads = std::cmp::min(num_cpus::get(), MAX_NUM_THREAD);
+        let memory_arena_num_bytes_per_thread = memory_arena_num_bytes / num_threads;
+        if memory_arena_num_bytes_per_thread < MEMORY_ARENA_NUM_BYTES_MIN {
+            num_threads = (memory_arena_num_bytes / MEMORY_ARENA_NUM_BYTES_MIN).max(1);
         }
-        self.writer_with_num_threads(num_threads, overall_heap_size_in_bytes)
+        self.writer_with_num_threads(num_threads, memory_arena_num_bytes)
+    }
+
+    /// Accessor to the index settings
+    pub fn settings(&self) -> &IndexSettings {
+        &self.settings
+    }
+
+    /// Accessor to the index settings
+    pub fn settings_mut(&mut self) -> &mut IndexSettings {
+        &mut self.settings
     }
 
     /// Accessor to the index schema
@@ -397,7 +528,22 @@ impl Index {
 
     /// Returns the set of corrupted files
     pub fn validate_checksum(&self) -> crate::Result<HashSet<PathBuf>> {
-        self.directory.list_damaged().map_err(Into::into)
+        let managed_files = self.directory.list_managed_files();
+        let active_segments_files: HashSet<PathBuf> = self
+            .searchable_segment_metas()?
+            .iter()
+            .flat_map(|segment_meta| segment_meta.list_files())
+            .collect();
+        let active_existing_files: HashSet<&PathBuf> =
+            active_segments_files.intersection(&managed_files).collect();
+
+        let mut damaged_files = HashSet::new();
+        for path in active_existing_files {
+            if !self.directory.validate_checksum(path)? {
+                damaged_files.insert((*path).clone());
+            }
+        }
+        Ok(damaged_files)
     }
 }
 
@@ -409,12 +555,9 @@ impl fmt::Debug for Index {
 
 #[cfg(test)]
 mod tests {
-    use crate::directory::{RAMDirectory, WatchCallback};
-    use crate::schema::Field;
-    use crate::schema::{Schema, INDEXED, TEXT};
-    use crate::IndexReader;
-    use crate::ReloadPolicy;
-    use crate::{Directory, Index};
+    use crate::directory::{RamDirectory, WatchCallback};
+    use crate::schema::{Field, Schema, INDEXED, TEXT};
+    use crate::{Directory, Index, IndexReader, IndexSettings, ReloadPolicy};
 
     #[test]
     fn test_indexer_for_field() {
@@ -432,15 +575,20 @@ mod tests {
 
     #[test]
     fn test_index_exists() {
-        let directory = RAMDirectory::create();
-        assert!(!Index::exists(&directory).unwrap());
-        assert!(Index::create(directory.clone(), throw_away_schema()).is_ok());
-        assert!(Index::exists(&directory).unwrap());
+        let directory: Box<dyn Directory> = Box::new(RamDirectory::create());
+        assert!(!Index::exists(directory.as_ref()).unwrap());
+        assert!(Index::create(
+            directory.clone(),
+            throw_away_schema(),
+            IndexSettings::default()
+        )
+        .is_ok());
+        assert!(Index::exists(directory.as_ref()).unwrap());
     }
 
     #[test]
     fn open_or_create_should_create() {
-        let directory = RAMDirectory::create();
+        let directory = RamDirectory::create();
         assert!(!Index::exists(&directory).unwrap());
         assert!(Index::open_or_create(directory.clone(), throw_away_schema()).is_ok());
         assert!(Index::exists(&directory).unwrap());
@@ -448,24 +596,44 @@ mod tests {
 
     #[test]
     fn open_or_create_should_open() {
-        let directory = RAMDirectory::create();
-        assert!(Index::create(directory.clone(), throw_away_schema()).is_ok());
-        assert!(Index::exists(&directory).unwrap());
+        let directory: Box<dyn Directory> = Box::new(RamDirectory::create());
+        assert!(Index::create(
+            directory.clone(),
+            throw_away_schema(),
+            IndexSettings::default()
+        )
+        .is_ok());
+        assert!(Index::exists(directory.as_ref()).unwrap());
         assert!(Index::open_or_create(directory, throw_away_schema()).is_ok());
     }
 
     #[test]
     fn create_should_wipeoff_existing() {
-        let directory = RAMDirectory::create();
-        assert!(Index::create(directory.clone(), throw_away_schema()).is_ok());
-        assert!(Index::exists(&directory).unwrap());
-        assert!(Index::create(directory.clone(), Schema::builder().build()).is_ok());
+        let directory: Box<dyn Directory> = Box::new(RamDirectory::create());
+        assert!(Index::create(
+            directory.clone(),
+            throw_away_schema(),
+            IndexSettings::default()
+        )
+        .is_ok());
+        assert!(Index::exists(directory.as_ref()).unwrap());
+        assert!(Index::create(
+            directory,
+            Schema::builder().build(),
+            IndexSettings::default()
+        )
+        .is_ok());
     }
 
     #[test]
     fn open_or_create_exists_but_schema_does_not_match() {
-        let directory = RAMDirectory::create();
-        assert!(Index::create(directory.clone(), throw_away_schema()).is_ok());
+        let directory = RamDirectory::create();
+        assert!(Index::create(
+            directory.clone(),
+            throw_away_schema(),
+            IndexSettings::default()
+        )
+        .is_ok());
         assert!(Index::exists(&directory).unwrap());
         assert!(Index::open_or_create(directory.clone(), throw_away_schema()).is_ok());
         let err = Index::open_or_create(directory, Schema::builder().build());
@@ -482,7 +650,7 @@ mod tests {
     }
 
     #[test]
-    fn test_index_on_commit_reload_policy() {
+    fn test_index_on_commit_reload_policy() -> crate::Result<()> {
         let schema = throw_away_schema();
         let field = schema.get_field("num_likes").unwrap();
         let index = Index::create_in_ram(schema);
@@ -492,19 +660,21 @@ mod tests {
             .try_into()
             .unwrap();
         assert_eq!(reader.searcher().num_docs(), 0);
-        test_index_on_commit_reload_policy_aux(field, &index, &reader);
+        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;
 
+        use super::*;
+        use crate::Directory;
+
         #[test]
-        fn test_index_on_commit_reload_policy_mmap() {
+        fn test_index_on_commit_reload_policy_mmap() -> crate::Result<()> {
             let schema = throw_away_schema();
             let field = schema.get_field("num_likes").unwrap();
             let tempdir = TempDir::new().unwrap();
@@ -516,7 +686,7 @@ mod tests {
                 .try_into()
                 .unwrap();
             assert_eq!(reader.searcher().num_docs(), 0);
-            test_index_on_commit_reload_policy_aux(field, &index, &reader);
+            test_index_on_commit_reload_policy_aux(field, &index, &reader)
         }
 
         #[test]
@@ -531,7 +701,7 @@ mod tests {
                 .reload_policy(ReloadPolicy::Manual)
                 .try_into()?;
             assert_eq!(reader.searcher().num_docs(), 0);
-            writer.add_document(doc!(field=>1u64));
+            writer.add_document(doc!(field=>1u64))?;
             let (sender, receiver) = crossbeam::channel::unbounded();
             let _handle = index.directory_mut().watch(WatchCallback::new(move || {
                 let _ = sender.send(());
@@ -545,7 +715,7 @@ mod tests {
         }
 
         #[test]
-        fn test_index_on_commit_reload_policy_different_directories() {
+        fn test_index_on_commit_reload_policy_different_directories() -> crate::Result<()> {
             let schema = throw_away_schema();
             let field = schema.get_field("num_likes").unwrap();
             let tempdir = TempDir::new().unwrap();
@@ -558,10 +728,14 @@ mod tests {
                 .try_into()
                 .unwrap();
             assert_eq!(reader.searcher().num_docs(), 0);
-            test_index_on_commit_reload_policy_aux(field, &write_index, &reader);
+            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) {
+    fn test_index_on_commit_reload_policy_aux(
+        field: Field,
+        index: &Index,
+        reader: &IndexReader,
+    ) -> crate::Result<()> {
         let mut reader_index = reader.index();
         let (sender, receiver) = crossbeam::channel::unbounded();
         let _watch_handle = reader_index
@@ -569,9 +743,9 @@ mod tests {
             .watch(WatchCallback::new(move || {
                 let _ = sender.send(());
             }));
-        let mut writer = index.writer_for_tests().unwrap();
+        let mut writer = index.writer_for_tests()?;
         assert_eq!(reader.searcher().num_docs(), 0);
-        writer.add_document(doc!(field=>1u64));
+        writer.add_document(doc!(field=>1u64))?;
         writer.commit().unwrap();
         // We need a loop here because it is possible for notify to send more than
         // one modify event. It was observed on CI on MacOS.
@@ -581,7 +755,7 @@ mod tests {
                 break;
             }
         }
-        writer.add_document(doc!(field=>2u64));
+        writer.add_document(doc!(field=>2u64))?;
         writer.commit().unwrap();
         // ... Same as above
         loop {
@@ -590,37 +764,37 @@ mod tests {
                 break;
             }
         }
+        Ok(())
     }
 
     // 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();
+    fn garbage_collect_works_as_intended() -> crate::Result<()> {
+        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 index = Index::create(directory.clone(), schema, IndexSettings::default())?;
 
         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));
+            writer.add_document(doc!(field => i))?;
         }
         let (sender, receiver) = crossbeam::channel::unbounded();
         let _handle = directory.watch(WatchCallback::new(move || {
             let _ = sender.send(());
         }));
-        writer.commit().unwrap();
+        writer.commit()?;
         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();
+            .try_into()?;
 
         assert_eq!(reader.searcher().num_docs(), 8_000);
-        writer.wait_merging_threads().unwrap();
+        writer.wait_merging_threads()?;
         let mem_right_after_merge_finished = directory.total_mem_usage();
 
         reader.reload().unwrap();
@@ -632,5 +806,6 @@ mod tests {
             mem_right_after_merge_finished,
             mem_right_after_commit
         );
+        Ok(())
     }
 }

src/core/index_meta.rs

@@ -1,12 +1,16 @@
-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;
+use std::sync::atomic::AtomicBool;
+use std::sync::Arc;
+
+use serde::{Deserialize, Serialize};
+
+use super::SegmentComponent;
+use crate::core::SegmentId;
+use crate::schema::Schema;
+use crate::store::Compressor;
+use crate::{Inventory, Opstamp, TrackedObject};
 
 #[derive(Clone, Debug, Serialize, Deserialize)]
 struct DeleteMeta {
@@ -33,6 +37,7 @@ impl SegmentMetaInventory {
         let inner = InnerSegmentMeta {
             segment_id,
             max_doc,
+            include_temp_doc_store: Arc::new(AtomicBool::new(true)),
             deletes: None,
         };
         SegmentMeta::from(self.inventory.track(inner))
@@ -80,6 +85,15 @@ impl SegmentMeta {
         self.tracked.segment_id
     }
 
+    /// Removes the Component::TempStore from the alive list and
+    /// therefore marks the temp docstore file to be deleted by
+    /// the garbage collection.
+    pub fn untrack_temp_docstore(&self) {
+        self.tracked
+            .include_temp_doc_store
+            .store(false, std::sync::atomic::Ordering::Relaxed);
+    }
+
     /// Returns the number of deleted documents.
     pub fn num_deleted_docs(&self) -> u32 {
         self.tracked
@@ -91,14 +105,26 @@ impl SegmentMeta {
 
     /// Returns the list of files that
     /// are required for the segment meta.
+    /// Note: Some of the returned files may not exist depending on the state of the segment.
     ///
     /// 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>>()
+        if self
+            .tracked
+            .include_temp_doc_store
+            .load(std::sync::atomic::Ordering::Relaxed)
+        {
+            SegmentComponent::iterator()
+                .map(|component| self.relative_path(*component))
+                .collect::<HashSet<PathBuf>>()
+        } else {
+            SegmentComponent::iterator()
+                .filter(|comp| *comp != &SegmentComponent::TempStore)
+                .map(|component| self.relative_path(*component))
+                .collect::<HashSet<PathBuf>>()
+        }
     }
 
     /// Returns the relative path of a component of our segment.
@@ -108,14 +134,14 @@ impl SegmentMeta {
     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)),
+            SegmentComponent::Postings => ".idx".to_string(),
+            SegmentComponent::Positions => ".pos".to_string(),
+            SegmentComponent::Terms => ".term".to_string(),
+            SegmentComponent::Store => ".store".to_string(),
+            SegmentComponent::TempStore => ".store.temp".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)
     }
@@ -160,12 +186,18 @@ impl SegmentMeta {
             segment_id: inner_meta.segment_id,
             max_doc,
             deletes: None,
+            include_temp_doc_store: Arc::new(AtomicBool::new(true)),
         });
         SegmentMeta { tracked }
     }
 
     #[doc(hidden)]
+    #[must_use]
     pub fn with_delete_meta(self, num_deleted_docs: u32, opstamp: Opstamp) -> SegmentMeta {
+        assert!(
+            num_deleted_docs <= self.max_doc(),
+            "There cannot be more deleted docs than there are docs."
+        );
         let delete_meta = DeleteMeta {
             num_deleted_docs,
             opstamp,
@@ -173,6 +205,7 @@ impl SegmentMeta {
         let tracked = self.tracked.map(move |inner_meta| InnerSegmentMeta {
             segment_id: inner_meta.segment_id,
             max_doc: inner_meta.max_doc,
+            include_temp_doc_store: Arc::new(AtomicBool::new(true)),
             deletes: Some(delete_meta),
         });
         SegmentMeta { tracked }
@@ -184,6 +217,14 @@ struct InnerSegmentMeta {
     segment_id: SegmentId,
     max_doc: u32,
     deletes: Option<DeleteMeta>,
+    /// If you want to avoid the SegmentComponent::TempStore file to be covered by
+    /// garbage collection and deleted, set this to true. This is used during merge.
+    #[serde(skip)]
+    #[serde(default = "default_temp_store")]
+    pub(crate) include_temp_doc_store: Arc<AtomicBool>,
+}
+fn default_temp_store() -> Arc<AtomicBool> {
+    Arc::new(AtomicBool::new(false))
 }
 
 impl InnerSegmentMeta {
@@ -194,6 +235,51 @@ impl InnerSegmentMeta {
     }
 }
 
+/// Search Index Settings.
+///
+/// Contains settings which are applied on the whole
+/// index, like presort documents.
+#[derive(Clone, Default, Serialize, Deserialize, Eq, PartialEq)]
+pub struct IndexSettings {
+    /// Sorts the documents by information
+    /// provided in `IndexSortByField`
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub sort_by_field: Option<IndexSortByField>,
+    /// The `Compressor` used to compress the doc store.
+    #[serde(default)]
+    pub docstore_compression: Compressor,
+}
+/// Settings to presort the documents in an index
+///
+/// Presorting documents can greatly performance
+/// in some scenarios, by applying top n
+/// optimizations.
+#[derive(Clone, Serialize, Deserialize, Eq, PartialEq)]
+pub struct IndexSortByField {
+    /// The field to sort the documents by
+    pub field: String,
+    /// The order to sort the documents by
+    pub order: Order,
+}
+/// The order to sort by
+#[derive(Clone, Serialize, Deserialize, Eq, PartialEq)]
+pub enum Order {
+    /// Ascending Order
+    Asc,
+    /// Descending Order
+    Desc,
+}
+impl Order {
+    /// return if the Order is ascending
+    pub fn is_asc(&self) -> bool {
+        self == &Order::Asc
+    }
+    /// return if the Order is descending
+    pub fn is_desc(&self) -> bool {
+        self == &Order::Desc
+    }
+}
+
 /// Meta information about the `Index`.
 ///
 /// This object is serialized on disk in the `meta.json` file.
@@ -201,9 +287,11 @@ impl InnerSegmentMeta {
 /// * the searchable segments,
 /// * the index `docstamp`
 /// * the schema
-///
 #[derive(Clone, Serialize)]
 pub struct IndexMeta {
+    /// `IndexSettings` to configure index options.
+    #[serde(default)]
+    pub index_settings: IndexSettings,
     /// List of `SegmentMeta` informations associated to each finalized segment of the index.
     pub segments: Vec<SegmentMeta>,
     /// Index `Schema`
@@ -222,6 +310,8 @@ pub struct IndexMeta {
 #[derive(Deserialize)]
 struct UntrackedIndexMeta {
     pub segments: Vec<InnerSegmentMeta>,
+    #[serde(default)]
+    pub index_settings: IndexSettings,
     pub schema: Schema,
     pub opstamp: Opstamp,
     #[serde(skip_serializing_if = "Option::is_none")]
@@ -231,6 +321,7 @@ struct UntrackedIndexMeta {
 impl UntrackedIndexMeta {
     pub fn track(self, inventory: &SegmentMetaInventory) -> IndexMeta {
         IndexMeta {
+            index_settings: self.index_settings,
             segments: self
                 .segments
                 .into_iter()
@@ -251,6 +342,7 @@ impl IndexMeta {
     /// Opstamp will the value `0u64`.
     pub fn with_schema(schema: Schema) -> IndexMeta {
         IndexMeta {
+            index_settings: IndexSettings::default(),
             segments: vec![],
             schema,
             opstamp: 0u64,
@@ -283,7 +375,7 @@ mod tests {
 
     use super::IndexMeta;
     use crate::schema::{Schema, TEXT};
-    use serde_json;
+    use crate::{IndexSettings, IndexSortByField, Order};
 
     #[test]
     fn test_serialize_metas() {
@@ -293,6 +385,13 @@ mod tests {
             schema_builder.build()
         };
         let index_metas = IndexMeta {
+            index_settings: IndexSettings {
+                sort_by_field: Some(IndexSortByField {
+                    field: "text".to_string(),
+                    order: Order::Asc,
+                }),
+                ..Default::default()
+            },
             segments: Vec::new(),
             schema,
             opstamp: 0u64,
@@ -301,7 +400,7 @@ mod tests {
         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}"#
+            r#"{"index_settings":{"sort_by_field":{"field":"text","order":"Asc"},"docstore_compression":"lz4"},"segments":[],"schema":[{"name":"text","type":"text","options":{"indexing":{"record":"position","fieldnorms":true,"tokenizer":"default"},"stored":false}}],"opstamp":0}"#
         );
     }
 }

src/core/inverted_index_reader.rs

@@ -1,12 +1,11 @@
 use std::io;
 
-use crate::common::BinarySerializable;
+use common::BinarySerializable;
+
 use crate::directory::FileSlice;
 use crate::positions::PositionReader;
-use crate::postings::TermInfo;
-use crate::postings::{BlockSegmentPostings, SegmentPostings};
-use crate::schema::IndexRecordOption;
-use crate::schema::Term;
+use crate::postings::{BlockSegmentPostings, SegmentPostings, TermInfo};
+use crate::schema::{IndexRecordOption, Term};
 use crate::termdict::TermDictionary;
 
 /// The inverted index reader is in charge of accessing
@@ -26,7 +25,6 @@ pub struct InvertedIndexReader {
     termdict: TermDictionary,
     postings_file_slice: FileSlice,
     positions_file_slice: FileSlice,
-    positions_idx_file_slice: FileSlice,
     record_option: IndexRecordOption,
     total_num_tokens: u64,
 }
@@ -37,7 +35,6 @@ impl InvertedIndexReader {
         termdict: TermDictionary,
         postings_file_slice: FileSlice,
         positions_file_slice: FileSlice,
-        positions_idx_file_slice: FileSlice,
         record_option: IndexRecordOption,
     ) -> io::Result<InvertedIndexReader> {
         let (total_num_tokens_slice, postings_body) = postings_file_slice.split(8);
@@ -46,7 +43,6 @@ impl InvertedIndexReader {
             termdict,
             postings_file_slice: postings_body,
             positions_file_slice,
-            positions_idx_file_slice,
             record_option,
             total_num_tokens,
         })
@@ -59,7 +55,6 @@ impl InvertedIndexReader {
             termdict: TermDictionary::empty(),
             postings_file_slice: FileSlice::empty(),
             positions_file_slice: FileSlice::empty(),
-            positions_idx_file_slice: FileSlice::empty(),
             record_option,
             total_num_tokens: 0u64,
         }
@@ -90,10 +85,11 @@ impl InvertedIndexReader {
         term_info: &TermInfo,
         block_postings: &mut BlockSegmentPostings,
     ) -> io::Result<()> {
-        let start_offset = term_info.postings_start_offset as usize;
-        let stop_offset = term_info.postings_stop_offset as usize;
-        let postings_slice = self.postings_file_slice.slice(start_offset, stop_offset);
-        block_postings.reset(term_info.doc_freq, postings_slice.read_bytes()?);
+        let postings_slice = self
+            .postings_file_slice
+            .slice(term_info.postings_range.clone());
+        let postings_bytes = postings_slice.read_bytes()?;
+        block_postings.reset(term_info.doc_freq, postings_bytes)?;
         Ok(())
     }
 
@@ -120,10 +116,9 @@ impl InvertedIndexReader {
         term_info: &TermInfo,
         requested_option: IndexRecordOption,
     ) -> io::Result<BlockSegmentPostings> {
-        let postings_data = self.postings_file_slice.slice(
-            term_info.postings_start_offset as usize,
-            term_info.postings_stop_offset as usize,
-        );
+        let postings_data = self
+            .postings_file_slice
+            .slice(term_info.postings_range.clone());
         BlockSegmentPostings::open(
             term_info.doc_freq,
             postings_data,
@@ -142,12 +137,12 @@ impl InvertedIndexReader {
         option: IndexRecordOption,
     ) -> io::Result<SegmentPostings> {
         let block_postings = self.read_block_postings_from_terminfo(term_info, option)?;
-        let position_stream = {
+        let position_reader = {
             if option.has_positions() {
-                let position_reader = self.positions_file_slice.clone();
-                let skip_reader = self.positions_idx_file_slice.clone();
-                let position_reader =
-                    PositionReader::new(position_reader, skip_reader, term_info.positions_idx)?;
+                let positions_data = self
+                    .positions_file_slice
+                    .read_bytes_slice(term_info.positions_range.clone())?;
+                let position_reader = PositionReader::open(positions_data)?;
                 Some(position_reader)
             } else {
                 None
@@ -155,7 +150,7 @@ impl InvertedIndexReader {
         };
         Ok(SegmentPostings::from_block_postings(
             block_postings,
-            position_stream,
+            position_reader,
         ))
     }
 
@@ -203,3 +198,36 @@ impl InvertedIndexReader {
             .unwrap_or(0u32))
     }
 }
+
+#[cfg(feature = "quickwit")]
+impl InvertedIndexReader {
+    pub(crate) async fn get_term_info_async(
+        &self,
+        term: &Term,
+    ) -> crate::AsyncIoResult<Option<TermInfo>> {
+        self.termdict.get_async(term.value_bytes()).await
+    }
+
+    /// Returns a block postings given a `Term`.
+    /// This method is for an advanced usage only.
+    ///
+    /// Most user should prefer using `read_postings` instead.
+    pub async fn warm_postings(
+        &self,
+        term: &Term,
+        with_positions: bool,
+    ) -> crate::AsyncIoResult<()> {
+        let term_info_opt = self.get_term_info_async(term).await?;
+        if let Some(term_info) = term_info_opt {
+            self.postings_file_slice
+                .read_bytes_slice_async(term_info.postings_range.clone())
+                .await?;
+            if with_positions {
+                self.positions_file_slice
+                    .read_bytes_slice_async(term_info.positions_range.clone())
+                    .await?;
+            }
+        }
+        Ok(())
+    }
+}

src/core/mod.rs

@@ -8,20 +8,22 @@ mod segment_component;
 mod segment_id;
 mod segment_reader;
 
+use std::path::Path;
+
+use once_cell::sync::Lazy;
+
 pub use self::executor::Executor;
-pub use self::index::Index;
-pub use self::index_meta::{IndexMeta, SegmentMeta, SegmentMetaInventory};
+pub use self::index::{Index, IndexBuilder};
+pub use self::index_meta::{
+    IndexMeta, IndexSettings, IndexSortByField, Order, SegmentMeta, SegmentMetaInventory,
+};
 pub use self::inverted_index_reader::InvertedIndexReader;
-pub use self::searcher::Searcher;
+pub use self::searcher::{Searcher, SearcherGeneration};
 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_reader::SegmentReader;
 
-use once_cell::sync::Lazy;
-use std::path::Path;
-
 /// 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"));

src/core/searcher.rs

@@ -1,28 +1,73 @@
-use crate::collector::Collector;
-use crate::core::Executor;
+use std::collections::BTreeMap;
+use std::{fmt, io};
 
-use crate::core::SegmentReader;
+use crate::collector::Collector;
+use crate::core::{Executor, SegmentReader};
 use crate::query::Query;
-use crate::schema::Document;
-use crate::schema::Schema;
-use crate::schema::Term;
+use crate::schema::{Document, Schema, Term};
 use crate::space_usage::SearcherSpaceUsage;
 use crate::store::StoreReader;
-use crate::DocAddress;
-use crate::Index;
+use crate::{DocAddress, Index, Opstamp, SegmentId, TrackedObject};
 
-use std::{fmt, io};
+/// Identifies the searcher generation accessed by a [Searcher].
+///
+/// While this might seem redundant, a [SearcherGeneration] contains
+/// both a `generation_id` AND a list of `(SegmentId, DeleteOpstamp)`.
+///
+/// This is on purpose. This object is used by the `Warmer` API.
+/// Having both information makes it possible to identify which
+/// artifact should be refreshed or garbage collected.
+///
+/// Depending on the use case, `Warmer`'s implementers can decide to
+/// produce artifacts per:
+/// - `generation_id` (e.g. some searcher level aggregates)
+/// - `(segment_id, delete_opstamp)` (e.g. segment level aggregates)
+/// - `segment_id` (e.g. for immutable document level information)
+/// - `(generation_id, segment_id)` (e.g. for consistent dynamic column)
+/// - ...
+#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub struct SearcherGeneration {
+    segments: BTreeMap<SegmentId, Option<Opstamp>>,
+    generation_id: u64,
+}
+
+impl SearcherGeneration {
+    pub(crate) fn from_segment_readers(
+        segment_readers: &[SegmentReader],
+        generation_id: u64,
+    ) -> Self {
+        let mut segment_id_to_del_opstamp = BTreeMap::new();
+        for segment_reader in segment_readers {
+            segment_id_to_del_opstamp
+                .insert(segment_reader.segment_id(), segment_reader.delete_opstamp());
+        }
+        Self {
+            segments: segment_id_to_del_opstamp,
+            generation_id,
+        }
+    }
+
+    /// Returns the searcher generation id.
+    pub fn generation_id(&self) -> u64 {
+        self.generation_id
+    }
+
+    /// Return a `(SegmentId -> DeleteOpstamp)` mapping.
+    pub fn segments(&self) -> &BTreeMap<SegmentId, Option<Opstamp>> {
+        &self.segments
+    }
+}
 
 /// Holds a list of `SegmentReader`s ready for search.
 ///
 /// It guarantees that the `Segment` will not be removed before
 /// the destruction of the `Searcher`.
-///
 pub struct Searcher {
     schema: Schema,
     index: Index,
     segment_readers: Vec<SegmentReader>,
     store_readers: Vec<StoreReader>,
+    generation: TrackedObject<SearcherGeneration>,
 }
 
 impl Searcher {
@@ -31,6 +76,7 @@ impl Searcher {
         schema: Schema,
         index: Index,
         segment_readers: Vec<SegmentReader>,
+        generation: TrackedObject<SearcherGeneration>,
     ) -> io::Result<Searcher> {
         let store_readers: Vec<StoreReader> = segment_readers
             .iter()
@@ -41,6 +87,7 @@ impl Searcher {
             index,
             segment_readers,
             store_readers,
+            generation,
         })
     }
 
@@ -49,14 +96,25 @@ impl Searcher {
         &self.index
     }
 
+    /// [SearcherGeneration] which identifies the version of the snapshot held by this `Searcher`.
+    pub fn generation(&self) -> &SearcherGeneration {
+        self.generation.as_ref()
+    }
+
     /// Fetches a document from tantivy's store given a `DocAddress`.
     ///
     /// The searcher uses the segment ordinal to route the
     /// the request to the right `Segment`.
     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)
+        let store_reader = &self.store_readers[doc_address.segment_ord as usize];
+        store_reader.get(doc_address.doc_id)
+    }
+
+    /// Fetches a document in an asynchronous manner.
+    #[cfg(feature = "quickwit")]
+    pub async fn doc_async(&self, doc_address: DocAddress) -> crate::Result<Document> {
+        let store_reader = &self.store_readers[doc_address.segment_ord as usize];
+        store_reader.get_async(doc_address.doc_id).await
     }
 
     /// Access the schema associated to the index of this searcher.
@@ -89,7 +147,7 @@ impl Searcher {
         &self.segment_readers
     }
 
-    /// Returns the segment_reader associated with the given segment_ordinal
+    /// Returns the segment_reader associated with the given segment_ord
     pub fn segment_reader(&self, segment_ord: u32) -> &SegmentReader {
         &self.segment_readers[segment_ord as usize]
     }

src/core/segment.rs

@@ -1,16 +1,13 @@
-use super::SegmentComponent;
-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::{FileSlice, WritePtr};
-use crate::indexer::segment_serializer::SegmentSerializer;
-use crate::schema::Schema;
-use crate::Opstamp;
 use std::fmt;
 use std::path::PathBuf;
 
+use super::SegmentComponent;
+use crate::core::{Index, SegmentId, SegmentMeta};
+use crate::directory::error::{OpenReadError, OpenWriteError};
+use crate::directory::{Directory, FileSlice, WritePtr};
+use crate::schema::Schema;
+use crate::Opstamp;
+
 /// A segment is a piece of the index.
 #[derive(Clone)]
 pub struct Segment {
@@ -57,6 +54,7 @@ impl Segment {
     }
 
     #[doc(hidden)]
+    #[must_use]
     pub fn with_delete_meta(self, num_deleted_docs: u32, opstamp: Opstamp) -> Segment {
         Segment {
             index: self.index,
@@ -90,12 +88,3 @@ impl Segment {
         Ok(write)
     }
 }
-
-pub trait SerializableSegment {
-    /// Writes a view of a segment by pushing information
-    /// to the `SegmentSerializer`.
-    ///
-    /// # Returns
-    /// The number of documents in the segment.
-    fn write(&self, serializer: SegmentSerializer) -> crate::Result<u32>;
-}

src/core/segment_component.rs

@@ -4,42 +4,42 @@ use std::slice;
 /// Each component is stored in its own file,
 /// using the pattern `segment_uuid`.`component_extension`,
 /// except the delete component that takes an `segment_uuid`.`delete_opstamp`.`component_extension`
-#[derive(Copy, Clone)]
+#[derive(Copy, Clone, Eq, PartialEq)]
 pub enum SegmentComponent {
     /// Postings (or inverted list). Sorted lists of document ids, associated to terms
-    POSTINGS,
+    Postings,
     /// Positions of terms in each document.
-    POSITIONS,
-    /// Index to seek within the position file
-    POSITIONSSKIP,
+    Positions,
     /// Column-oriented random-access storage of fields.
-    FASTFIELDS,
+    FastFields,
     /// Stores the sum  of the length (in terms) of each field for each document.
     /// Field norms are stored as a special u64 fast field.
-    FIELDNORMS,
+    FieldNorms,
     /// Dictionary associating `Term`s to `TermInfo`s which is
     /// simply an address into the `postings` file and the `positions` file.
-    TERMS,
+    Terms,
     /// Row-oriented, compressed storage of the documents.
     /// Accessing a document from the store is relatively slow, as it
     /// requires to decompress the entire block it belongs to.
-    STORE,
+    Store,
+    /// Temporary storage of the documents, before streamed to `Store`.
+    TempStore,
     /// Bitset describing which document of the segment is deleted.
-    DELETE,
+    Delete,
 }
 
 impl SegmentComponent {
     /// Iterates through the components.
     pub fn iterator() -> slice::Iter<'static, SegmentComponent> {
         static SEGMENT_COMPONENTS: [SegmentComponent; 8] = [
-            SegmentComponent::POSTINGS,
-            SegmentComponent::POSITIONS,
-            SegmentComponent::POSITIONSSKIP,
-            SegmentComponent::FASTFIELDS,
-            SegmentComponent::FIELDNORMS,
-            SegmentComponent::TERMS,
-            SegmentComponent::STORE,
-            SegmentComponent::DELETE,
+            SegmentComponent::Postings,
+            SegmentComponent::Positions,
+            SegmentComponent::FastFields,
+            SegmentComponent::FieldNorms,
+            SegmentComponent::Terms,
+            SegmentComponent::Store,
+            SegmentComponent::TempStore,
+            SegmentComponent::Delete,
         ];
         SEGMENT_COMPONENTS.iter()
     }

src/core/segment_id.rs

@@ -1,14 +1,14 @@
 use std::cmp::{Ord, Ordering};
+use std::error::Error;
 use std::fmt;
-use uuid::Uuid;
+use std::str::FromStr;
+#[cfg(test)]
+use std::sync::atomic;
 
 #[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;
+use uuid::Uuid;
 
 /// Uuid identifying a segment.
 ///
@@ -22,7 +22,7 @@ use std::sync::atomic;
 pub struct SegmentId(Uuid);
 
 #[cfg(test)]
-static AUTO_INC_COUNTER: Lazy<atomic::AtomicUsize> = Lazy::new(|| atomic::AtomicUsize::default());
+static AUTO_INC_COUNTER: Lazy<atomic::AtomicUsize> = Lazy::new(atomic::AtomicUsize::default);
 
 #[cfg(test)]
 const ZERO_ARRAY: [u8; 8] = [0u8; 8];
@@ -108,6 +108,12 @@ impl fmt::Debug for SegmentId {
     }
 }
 
+impl fmt::Display for SegmentId {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "Seg({:?})", self.short_uuid_string())
+    }
+}
+
 impl PartialOrd for SegmentId {
     fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
         Some(self.cmp(other))

src/core/segment_reader.rs

@@ -1,26 +1,19 @@
-use crate::common::HasLen;
-use crate::core::InvertedIndexReader;
-use crate::core::Segment;
-use crate::core::SegmentComponent;
-use crate::core::SegmentId;
-use crate::directory::FileSlice;
-use crate::fastfield::DeleteBitSet;
-use crate::fastfield::FacetReader;
-use crate::fastfield::FastFieldReaders;
+use std::collections::HashMap;
+use std::sync::{Arc, RwLock};
+use std::{fmt, io};
+
+use fail::fail_point;
+
+use crate::core::{InvertedIndexReader, Segment, SegmentComponent, SegmentId};
+use crate::directory::{CompositeFile, FileSlice};
+use crate::error::DataCorruption;
+use crate::fastfield::{intersect_alive_bitsets, AliveBitSet, FacetReader, FastFieldReaders};
 use crate::fieldnorm::{FieldNormReader, FieldNormReaders};
-use crate::schema::FieldType;
-use crate::schema::Schema;
-use crate::schema::{Field, IndexRecordOption};
+use crate::schema::{Field, FieldType, IndexRecordOption, Schema};
 use crate::space_usage::SegmentSpaceUsage;
 use crate::store::StoreReader;
 use crate::termdict::TermDictionary;
-use crate::DocId;
-use crate::{common::CompositeFile, error::DataCorruption};
-use fail::fail_point;
-use std::fmt;
-use std::sync::Arc;
-use std::sync::RwLock;
-use std::{collections::HashMap, io};
+use crate::{DocId, Opstamp};
 
 /// Entry point to access all of the datastructures of the `Segment`
 ///
@@ -32,43 +25,36 @@ use std::{collections::HashMap, io};
 ///
 /// The segment reader has a very low memory footprint,
 /// as close to all of the memory data is mmapped.
-///
-///
-/// TODO fix not decoding docfreq
 #[derive(Clone)]
 pub struct SegmentReader {
     inv_idx_reader_cache: Arc<RwLock<HashMap<Field, Arc<InvertedIndexReader>>>>,
 
     segment_id: SegmentId,
+    delete_opstamp: Option<Opstamp>,
+
     max_doc: DocId,
     num_docs: DocId,
 
     termdict_composite: CompositeFile,
     postings_composite: CompositeFile,
     positions_composite: CompositeFile,
-    positions_idx_composite: CompositeFile,
     fast_fields_readers: Arc<FastFieldReaders>,
     fieldnorm_readers: FieldNormReaders,
 
     store_file: FileSlice,
-    delete_bitset_opt: Option<DeleteBitSet>,
+    alive_bitset_opt: Option<AliveBitSet>,
     schema: Schema,
 }
 
 impl SegmentReader {
     /// Returns the highest document id ever attributed in
     /// this segment + 1.
-    /// Today, `tantivy` does not handle deletes, so it happens
-    /// to also be the number of documents in the index.
     pub fn max_doc(&self) -> DocId {
         self.max_doc
     }
 
-    /// Returns the number of documents.
+    /// Returns the number of alive documents.
     /// Deleted documents are not counted.
-    ///
-    /// Today, `tantivy` does not handle deletes so max doc and
-    /// num_docs are the same.
     pub fn num_docs(&self) -> DocId {
         self.num_docs
     }
@@ -81,14 +67,12 @@ impl SegmentReader {
     /// Return the number of documents that have been
     /// deleted in the segment.
     pub fn num_deleted_docs(&self) -> DocId {
-        self.delete_bitset()
-            .map(|delete_set| delete_set.len() as DocId)
-            .unwrap_or(0u32)
+        self.max_doc - self.num_docs
     }
 
-    /// Returns true iff some of the documents of the segment have been deleted.
+    /// Returns true if some of the documents of the segment have been deleted.
     pub fn has_deletes(&self) -> bool {
-        self.delete_bitset().is_some()
+        self.num_deleted_docs() > 0
     }
 
     /// Accessor to a segment's fast field reader given a field.
@@ -108,19 +92,22 @@ impl SegmentReader {
     /// Accessor to the `FacetReader` associated to a given `Field`.
     pub fn facet_reader(&self, field: Field) -> crate::Result<FacetReader> {
         let field_entry = self.schema.get_field_entry(field);
-        if field_entry.field_type() != &FieldType::HierarchicalFacet {
-            return Err(crate::TantivyError::InvalidArgument(format!(
+
+        match field_entry.field_type() {
+            FieldType::Facet(_) => {
+                let term_ords_reader = self.fast_fields().u64s(field)?;
+                let termdict = self
+                    .termdict_composite
+                    .open_read(field)
+                    .map(TermDictionary::open)
+                    .unwrap_or_else(|| Ok(TermDictionary::empty()))?;
+                Ok(FacetReader::new(term_ords_reader, termdict))
+            }
+            _ => Err(crate::TantivyError::InvalidArgument(format!(
                 "Field {:?} is not a facet field.",
                 field_entry.name()
-            )));
+            ))),
         }
-        let term_ords_reader = self.fast_fields().u64s(field)?;
-        let termdict = self
-            .termdict_composite
-            .open_read(field)
-            .map(TermDictionary::open)
-            .unwrap_or_else(|| Ok(TermDictionary::empty()))?;
-        Ok(FacetReader::new(term_ords_reader, termdict))
     }
 
     /// Accessor to the segment's `Field norms`'s reader.
@@ -134,13 +121,17 @@ impl SegmentReader {
         self.fieldnorm_readers.get_field(field)?.ok_or_else(|| {
             let field_name = self.schema.get_field_name(field);
             let err_msg = format!(
-                "Field norm not found for field {:?}. Was it marked as indexed during indexing?",
-                field_name
+                "Field norm not found for field {field_name:?}. Was the field set to record norm \
+                 during indexing?"
             );
             crate::TantivyError::SchemaError(err_msg)
         })
     }
 
+    pub(crate) fn fieldnorms_readers(&self) -> &FieldNormReaders {
+        &self.fieldnorm_readers
+    }
+
     /// Accessor to the segment's `StoreReader`.
     pub fn get_store_reader(&self) -> io::Result<StoreReader> {
         StoreReader::open(self.store_file.clone())
@@ -148,65 +139,70 @@ impl SegmentReader {
 
     /// Open a new segment for reading.
     pub fn open(segment: &Segment) -> crate::Result<SegmentReader> {
-        let termdict_file = segment.open_read(SegmentComponent::TERMS)?;
+        Self::open_with_custom_alive_set(segment, None)
+    }
+
+    /// Open a new segment for reading.
+    pub fn open_with_custom_alive_set(
+        segment: &Segment,
+        custom_bitset: Option<AliveBitSet>,
+    ) -> crate::Result<SegmentReader> {
+        let termdict_file = segment.open_read(SegmentComponent::Terms)?;
         let termdict_composite = CompositeFile::open(&termdict_file)?;
 
-        let store_file = segment.open_read(SegmentComponent::STORE)?;
+        let store_file = segment.open_read(SegmentComponent::Store)?;
 
         fail_point!("SegmentReader::open#middle");
 
-        let postings_file = segment.open_read(SegmentComponent::POSTINGS)?;
+        let postings_file = segment.open_read(SegmentComponent::Postings)?;
         let postings_composite = CompositeFile::open(&postings_file)?;
 
         let positions_composite = {
-            if let Ok(positions_file) = segment.open_read(SegmentComponent::POSITIONS) {
+            if let Ok(positions_file) = segment.open_read(SegmentComponent::Positions) {
                 CompositeFile::open(&positions_file)?
             } else {
                 CompositeFile::empty()
             }
         };
 
-        let positions_idx_composite = {
-            if let Ok(positions_skip_file) = segment.open_read(SegmentComponent::POSITIONSSKIP) {
-                CompositeFile::open(&positions_skip_file)?
-            } else {
-                CompositeFile::empty()
-            }
-        };
-
         let schema = segment.schema();
 
-        let fast_fields_data = segment.open_read(SegmentComponent::FASTFIELDS)?;
+        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::new(
-            schema.clone(),
-            fast_fields_composite,
-        )?);
-
-        let fieldnorm_data = segment.open_read(SegmentComponent::FIELDNORMS)?;
+        let fast_field_readers =
+            Arc::new(FastFieldReaders::new(schema.clone(), fast_fields_composite));
+        let fieldnorm_data = segment.open_read(SegmentComponent::FieldNorms)?;
         let fieldnorm_readers = FieldNormReaders::open(fieldnorm_data)?;
 
-        let delete_bitset_opt = if segment.meta().has_deletes() {
-            let delete_data = segment.open_read(SegmentComponent::DELETE)?;
-            let delete_bitset = DeleteBitSet::open(delete_data)?;
-            Some(delete_bitset)
+        let original_bitset = if segment.meta().has_deletes() {
+            let delete_file_slice = segment.open_read(SegmentComponent::Delete)?;
+            let delete_data = delete_file_slice.read_bytes()?;
+            Some(AliveBitSet::open(delete_data))
         } else {
             None
         };
 
+        let alive_bitset_opt = intersect_alive_bitset(original_bitset, custom_bitset);
+
+        let max_doc = segment.meta().max_doc();
+        let num_docs = alive_bitset_opt
+            .as_ref()
+            .map(|alive_bitset| alive_bitset.num_alive_docs() as u32)
+            .unwrap_or(max_doc);
+
         Ok(SegmentReader {
             inv_idx_reader_cache: Default::default(),
-            max_doc: segment.meta().max_doc(),
-            num_docs: segment.meta().num_docs(),
+            num_docs,
+            max_doc,
             termdict_composite,
             postings_composite,
             fast_fields_readers: fast_field_readers,
             fieldnorm_readers,
             segment_id: segment.id(),
+            delete_opstamp: segment.meta().delete_opstamp(),
             store_file,
-            delete_bitset_opt,
+            alive_bitset_opt,
             positions_composite,
-            positions_idx_composite,
             schema,
         })
     }
@@ -254,26 +250,28 @@ impl SegmentReader {
         let record_option = record_option_opt.unwrap();
         let postings_file = postings_file_opt.unwrap();
 
-        let termdict_file: FileSlice = self.termdict_composite.open_read(field)
-            .ok_or_else(||
-               DataCorruption::comment_only(format!("Failed to open field {:?}'s term dictionary in the composite file. Has the schema been modified?", field_entry.name()))
-            )?;
+        let termdict_file: FileSlice =
+            self.termdict_composite.open_read(field).ok_or_else(|| {
+                DataCorruption::comment_only(format!(
+                    "Failed to open field {:?}'s term dictionary in the composite file. Has the \
+                     schema been modified?",
+                    field_entry.name()
+                ))
+            })?;
 
-        let positions_file = self
-            .positions_composite
-            .open_read(field)
-            .expect("Index corrupted. Failed to open field positions in composite file.");
-
-        let positions_idx_file = self
-            .positions_idx_composite
-            .open_read(field)
-            .expect("Index corrupted. Failed to open field positions in composite file.");
+        let positions_file = self.positions_composite.open_read(field).ok_or_else(|| {
+            let error_msg = format!(
+                "Failed to open field {:?}'s positions in the composite file. Has the schema been \
+                 modified?",
+                field_entry.name()
+            );
+            DataCorruption::comment_only(error_msg)
+        })?;
 
         let inv_idx_reader = Arc::new(InvertedIndexReader::new(
             TermDictionary::open(termdict_file)?,
             postings_file,
             positions_file,
-            positions_idx_file,
             record_option,
         )?);
 
@@ -292,23 +290,32 @@ impl SegmentReader {
         self.segment_id
     }
 
-    /// Returns the bitset representing
-    /// the documents that have been deleted.
-    pub fn delete_bitset(&self) -> Option<&DeleteBitSet> {
-        self.delete_bitset_opt.as_ref()
+    /// Returns the delete opstamp
+    pub fn delete_opstamp(&self) -> Option<Opstamp> {
+        self.delete_opstamp
     }
 
-    /// Returns true iff the `doc` is marked
+    /// Returns the bitset representing
+    /// the documents that have been deleted.
+    pub fn alive_bitset(&self) -> Option<&AliveBitSet> {
+        self.alive_bitset_opt.as_ref()
+    }
+
+    /// Returns true if the `doc` is marked
     /// as deleted.
     pub fn is_deleted(&self, doc: DocId) -> bool {
-        self.delete_bitset()
+        self.alive_bitset()
             .map(|delete_set| delete_set.is_deleted(doc))
             .unwrap_or(false)
     }
 
     /// Returns an iterator that will iterate over the alive document ids
-    pub fn doc_ids_alive(&self) -> impl Iterator<Item = DocId> + '_ {
-        (0u32..self.max_doc).filter(move |doc| !self.is_deleted(*doc))
+    pub fn doc_ids_alive(&self) -> Box<dyn Iterator<Item = DocId> + '_> {
+        if let Some(alive_bitset) = &self.alive_bitset_opt {
+            Box::new(alive_bitset.iter_alive())
+        } else {
+            Box::new(0u32..self.max_doc)
+        }
     }
 
     /// Summarize total space usage of this segment.
@@ -318,18 +325,32 @@ impl SegmentReader {
             self.termdict_composite.space_usage(),
             self.postings_composite.space_usage(),
             self.positions_composite.space_usage(),
-            self.positions_idx_composite.space_usage(),
             self.fast_fields_readers.space_usage(),
             self.fieldnorm_readers.space_usage(),
             self.get_store_reader()?.space_usage(),
-            self.delete_bitset_opt
+            self.alive_bitset_opt
                 .as_ref()
-                .map(DeleteBitSet::space_usage)
+                .map(AliveBitSet::space_usage)
                 .unwrap_or(0),
         ))
     }
 }
 
+fn intersect_alive_bitset(
+    left_opt: Option<AliveBitSet>,
+    right_opt: Option<AliveBitSet>,
+) -> Option<AliveBitSet> {
+    match (left_opt, right_opt) {
+        (Some(left), Some(right)) => {
+            assert_eq!(left.bitset().max_value(), right.bitset().max_value());
+            Some(intersect_alive_bitsets(left, right))
+        }
+        (Some(left), None) => Some(left),
+        (None, Some(right)) => Some(right),
+        (None, None) => None,
+    }
+}
+
 impl fmt::Debug for SegmentReader {
     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         write!(f, "SegmentReader({:?})", self.segment_id)
@@ -342,6 +363,32 @@ mod test {
     use crate::schema::{Schema, Term, STORED, TEXT};
     use crate::DocId;
 
+    #[test]
+    fn test_num_alive() -> crate::Result<()> {
+        let mut schema_builder = Schema::builder();
+        schema_builder.add_text_field("name", TEXT | STORED);
+        let schema = schema_builder.build();
+        let index = Index::create_in_ram(schema.clone());
+        let name = schema.get_field("name").unwrap();
+
+        {
+            let mut index_writer = index.writer_for_tests()?;
+            index_writer.add_document(doc!(name => "tantivy"))?;
+            index_writer.add_document(doc!(name => "horse"))?;
+            index_writer.add_document(doc!(name => "jockey"))?;
+            index_writer.add_document(doc!(name => "cap"))?;
+            // we should now have one segment with two docs
+            index_writer.delete_term(Term::from_field_text(name, "horse"));
+            index_writer.delete_term(Term::from_field_text(name, "cap"));
+
+            // ok, now we should have a deleted doc
+            index_writer.commit()?;
+        }
+        let searcher = index.reader()?.searcher();
+        assert_eq!(2, searcher.segment_reader(0).num_docs());
+        assert_eq!(4, searcher.segment_reader(0).max_doc());
+        Ok(())
+    }
     #[test]
     fn test_alive_docs_iterator() -> crate::Result<()> {
         let mut schema_builder = Schema::builder();
@@ -352,10 +399,10 @@ mod test {
 
         {
             let mut index_writer = index.writer_for_tests()?;
-            index_writer.add_document(doc!(name => "tantivy"));
-            index_writer.add_document(doc!(name => "horse"));
-            index_writer.add_document(doc!(name => "jockey"));
-            index_writer.add_document(doc!(name => "cap"));
+            index_writer.add_document(doc!(name => "tantivy"))?;
+            index_writer.add_document(doc!(name => "horse"))?;
+            index_writer.add_document(doc!(name => "jockey"))?;
+            index_writer.add_document(doc!(name => "cap"))?;
             // we should now have one segment with two docs
             index_writer.commit()?;
         }

src/directory/composite_file.rs

@@ -1,15 +1,13 @@
-use crate::common::BinarySerializable;
-use crate::common::CountingWriter;
-use crate::common::VInt;
-use crate::directory::FileSlice;
-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::{self, Read, Write};
+use std::iter::ExactSizeIterator;
+use std::ops::Range;
 
-use super::HasLen;
+use common::{BinarySerializable, CountingWriter, HasLen, VInt};
+
+use crate::directory::{FileSlice, TerminatingWrite, WritePtr};
+use crate::schema::Field;
+use crate::space_usage::{FieldUsage, PerFieldSpaceUsage};
 
 #[derive(Eq, PartialEq, Hash, Copy, Ord, PartialOrd, Clone, Debug)]
 pub struct FileAddr {
@@ -105,7 +103,7 @@ impl<W: TerminatingWrite + Write> CompositeWrite<W> {
 #[derive(Clone)]
 pub struct CompositeFile {
     data: FileSlice,
-    offsets_index: HashMap<FileAddr, (usize, usize)>,
+    offsets_index: HashMap<FileAddr, Range<usize>>,
 }
 
 impl CompositeFile {
@@ -117,7 +115,7 @@ impl CompositeFile {
         let footer_len = u32::deserialize(&mut footer_len_data.as_slice())? as usize;
         let footer_start = end - 4 - footer_len;
         let footer_data = data
-            .slice(footer_start, footer_start + footer_len)
+            .slice(footer_start..footer_start + footer_len)
             .read_bytes()?;
         let mut footer_buffer = footer_data.as_slice();
         let num_fields = VInt::deserialize(&mut footer_buffer)?.0 as usize;
@@ -138,7 +136,7 @@ impl CompositeFile {
             let file_addr = file_addrs[i];
             let start_offset = offsets[i];
             let end_offset = offsets[i + 1];
-            field_index.insert(file_addr, (start_offset, end_offset));
+            field_index.insert(file_addr, start_offset..end_offset);
         }
 
         Ok(CompositeFile {
@@ -167,16 +165,16 @@ impl CompositeFile {
     pub fn open_read_with_idx(&self, field: Field, idx: usize) -> Option<FileSlice> {
         self.offsets_index
             .get(&FileAddr { field, idx })
-            .map(|&(from, to)| self.data.slice(from, to))
+            .map(|byte_range| self.data.slice(byte_range.clone()))
     }
 
     pub fn space_usage(&self) -> PerFieldSpaceUsage {
         let mut fields = HashMap::new();
-        for (&field_addr, &(start, end)) in self.offsets_index.iter() {
+        for (&field_addr, byte_range) in &self.offsets_index {
             fields
                 .entry(field_addr.field)
                 .or_insert_with(|| FieldUsage::empty(field_addr.field))
-                .add_field_idx(field_addr.idx, end - start);
+                .add_field_idx(field_addr.idx, byte_range.len());
         }
         PerFieldSpaceUsage::new(fields)
     }
@@ -185,18 +183,19 @@ impl CompositeFile {
 #[cfg(test)]
 mod test {
 
-    use super::{CompositeFile, CompositeWrite};
-    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;
 
+    use common::{BinarySerializable, VInt};
+
+    use super::{CompositeFile, CompositeWrite};
+    use crate::directory::{Directory, RamDirectory};
+    use crate::schema::Field;
+
     #[test]
     fn test_composite_file() -> crate::Result<()> {
         let path = Path::new("test_path");
-        let directory = RAMDirectory::create();
+        let directory = RamDirectory::create();
         {
             let w = directory.open_write(path).unwrap();
             let mut composite_write = CompositeWrite::wrap(w);

src/directory/directory.rs

@@ -1,18 +1,12 @@
-use crate::directory::directory_lock::Lock;
-use crate::directory::error::LockError;
-use crate::directory::error::{DeleteError, OpenReadError, OpenWriteError};
-use crate::directory::WatchHandle;
-use crate::directory::{FileHandle, WatchCallback};
-use crate::directory::{FileSlice, 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::thread;
+use std::marker::{Send, Sync};
+use std::path::{Path, PathBuf};
 use std::time::Duration;
+use std::{fmt, io, thread};
+
+use crate::directory::directory_lock::Lock;
+use crate::directory::error::{DeleteError, LockError, OpenReadError, OpenWriteError};
+use crate::directory::{FileHandle, FileSlice, WatchCallback, WatchHandle, WritePtr};
 
 /// Retry the logic of acquiring locks is pretty simple.
 /// We just retry `n` times after a given `duratio`, both
@@ -43,10 +37,8 @@ impl RetryPolicy {
 }
 
 /// 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`.
+/// It is associated to a lock file, that gets deleted on `Drop.`
 pub struct DirectoryLock(Box<dyn Send + Sync + 'static>);
 
 struct DirectoryLockGuard {
@@ -70,7 +62,7 @@ impl Drop for DirectoryLockGuard {
 
 enum TryAcquireLockError {
     FileExists,
-    IOError(io::Error),
+    IoError(io::Error),
 }
 
 fn try_acquire_lock(
@@ -79,9 +71,9 @@ fn try_acquire_lock(
 ) -> 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),
+        OpenWriteError::IoError { io_error, .. } => TryAcquireLockError::IoError(io_error),
     })?;
-    write.flush().map_err(TryAcquireLockError::IOError)?;
+    write.flush().map_err(TryAcquireLockError::IoError)?;
     Ok(DirectoryLock::from(Box::new(DirectoryLockGuard {
         directory: directory.box_clone(),
         path: filepath.to_owned(),
@@ -104,9 +96,9 @@ fn retry_policy(is_blocking: bool) -> RetryPolicy {
 ///
 /// There are currently two implementations of `Directory`
 ///
-/// - The [`MMapDirectory`](struct.MmapDirectory.html), this
+/// - The [`MMapDirectory`][crate::directory::MmapDirectory], this
 /// should be your default choice.
-/// - The [`RAMDirectory`](struct.RAMDirectory.html), which
+/// - The [`RamDirectory`][crate::directory::RamDirectory], which
 /// should be used mostly for tests.
 pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
     /// Opens a file and returns a boxed `FileHandle`.
@@ -136,16 +128,22 @@ pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
     /// `DeleteError::DoesNotExist`.
     fn delete(&self, path: &Path) -> Result<(), DeleteError>;
 
-    /// Returns true iff the file exists
+    /// Returns true if and only if the file exists
     fn exists(&self, path: &Path) -> Result<bool, OpenReadError>;
 
     /// Opens a writer for the *virtual file* associated with
     /// a Path.
     ///
-    /// Right after this call, the file should be created
-    /// and any subsequent call to `open_read` for the
+    /// Right after this call, for the span of the execution of the program
+    /// the file should be created and any subsequent call to `open_read` for the
     /// same path should return a `FileSlice`.
     ///
+    /// However, depending on the directory implementation,
+    /// it might be required to call `sync_directory` to ensure
+    /// that the file is durably created.
+    /// (The semantics here are the same when dealing with
+    /// a posix filesystem.)
+    ///
     /// Write operations may be aggressively buffered.
     /// The client of this trait is responsible for calling flush
     /// to ensure that subsequent `read` operations
@@ -154,7 +152,7 @@ pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
     /// Flush operation should also be persistent.
     ///
     /// The user shall not rely on `Drop` triggering `flush`.
-    /// Note that `RAMDirectory` will panic! if `flush`
+    /// Note that `RamDirectory` will panic! if `flush`
     /// was not called.
     ///
     /// The file may not previously exist.
@@ -176,6 +174,12 @@ pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
     /// The file may or may not previously exist.
     fn atomic_write(&self, path: &Path, data: &[u8]) -> io::Result<()>;
 
+    /// Sync the directory.
+    ///
+    /// This call is required to ensure that newly created files are
+    /// effectively stored durably.
+    fn sync_directory(&self) -> io::Result<()>;
+
     /// Acquire a lock in the given directory.
     ///
     /// The method is blocking or not depending on the `Lock` object.
@@ -192,8 +196,8 @@ pub trait Directory: DirectoryClone + fmt::Debug + Send + Sync + 'static {
                         return Err(LockError::LockBusy);
                     }
                 }
-                Err(TryAcquireLockError::IOError(io_error)) => {
-                    return Err(LockError::IOError(io_error));
+                Err(TryAcquireLockError::IoError(io_error)) => {
+                    return Err(LockError::IoError(io_error));
                 }
             }
         }
@@ -223,10 +227,21 @@ pub trait DirectoryClone {
 }
 
 impl<T> DirectoryClone for T
-where
-    T: 'static + Directory + Clone,
+where T: 'static + Directory + Clone
 {
     fn box_clone(&self) -> Box<dyn Directory> {
         Box::new(self.clone())
     }
 }
+
+impl Clone for Box<dyn Directory> {
+    fn clone(&self) -> Self {
+        self.box_clone()
+    }
+}
+
+impl<T: Directory + 'static> From<T> for Box<dyn Directory> {
+    fn from(t: T) -> Self {
+        Box::new(t)
+    }
+}

src/directory/directory_lock.rs

@@ -1,17 +1,17 @@
-use once_cell::sync::Lazy;
 use std::path::PathBuf;
 
+use once_cell::sync::Lazy;
+
 /// 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)
+/// - [INDEX_WRITER_LOCK]
+/// - [META_LOCK]
 ///
 /// Check out these locks documentation for more information.
-///
 #[derive(Debug)]
 pub struct Lock {
     /// The lock needs to be associated with its own file `path`.