Introducing a column trait

And renaming the `Serializer` trait `FastFieldCodec`.

.github/FUNDING.yml

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

.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/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,19 @@
+---
+name: Bug report
+about: Create a report to help us improve
+
+---
+
+**Describe the bug**
+- What did you do?
+- What happened?
+- What was expected?
+
+**Which version of tantivy are you using?**
+If "master",  ideally give the specific sha1 revision.
+
+**To Reproduce**
+
+If your bug is deterministic, can you give a minimal reproducing code?
+Some bugs are not deterministic. Can you describe with precision in which context it happened?
+If this is possible, can you share your code?

.github/ISSUE_TEMPLATE/feature_request.md

@@ -0,0 +1,14 @@
+---
+name: Feature request
+about: Suggest an idea for this project
+
+---
+
+**Is your feature request related to a problem? Please describe.**
+A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
+
+**Describe the solution you'd like**
+A clear and concise description of what you want to happen.
+
+**[Optional] describe alternatives you've considered**
+A clear and concise description of any alternative solutions or features you've considered.

.github/ISSUE_TEMPLATE/question.md

@@ -0,0 +1,7 @@
+---
+name: Question
+about: Ask any question about tantivy's usage...
+
+---
+
+Try to be specific about your use case...

.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 +1,26 @@
+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 --profile minimal --component llvm-tools-preview
+      - uses: Swatinem/rust-cache@v2
+      - uses: taiki-e/install-action@cargo-llvm-cov
+      - name: Generate code coverage
+        run: cargo +nightly llvm-cov --all-features --workspace --lcov --output-path lcov.info
+      - name: Upload coverage to Codecov
+        uses: codecov/codecov-action@v3
+        continue-on-error: true
+        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 +1,28 @@
+name: Long running tests
+
+on:
+  push:
+    branches: [ main ]
+
+env:
+  CARGO_TERM_COLOR: always
+  NUM_FUNCTIONAL_TEST_ITERATIONS: 20000
+
+jobs:
+  test:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v3
+    - name: Install stable
+      uses: actions-rs/toolchain@v1
+      with:
+          toolchain: stable
+          profile: minimal
+          override: true
+
+    - name: Run indexing_unsorted
+      run: cargo test indexing_unsorted -- --ignored
+    - name: Run indexing_sorted
+      run: cargo test indexing_sorted -- --ignored

.github/workflows/test.yml

@@ -0,0 +1,74 @@
+name: Unit tests
+
+on:
+  push:
+    branches: [ main ]
+  pull_request:
+    branches: [ main ]
+
+env:
+  CARGO_TERM_COLOR: always
+
+jobs:
+  check:
+
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v3
+
+    - name: Install nightly
+      uses: actions-rs/toolchain@v1
+      with:
+            toolchain: nightly
+            profile: minimal
+            components: rustfmt
+    - name: Install stable
+      uses: actions-rs/toolchain@v1
+      with:
+            toolchain: stable
+            profile: minimal
+            components: clippy
+
+    - uses: Swatinem/rust-cache@v2
+
+    - name: Check Formatting
+      run: cargo +nightly fmt --all -- --check
+
+    - uses: actions-rs/clippy-check@v1
+      with:
+        toolchain: stable
+        token: ${{ secrets.GITHUB_TOKEN }}
+        args: --tests
+
+  test:
+
+    runs-on: ubuntu-latest
+
+    strategy:
+      matrix:
+        features: [
+            { label: "all", flags: "mmap,brotli-compression,lz4-compression,snappy-compression,zstd-compression,failpoints" },
+            { label: "quickwit", flags: "mmap,quickwit,failpoints" }
+        ]
+
+    name: test-${{ matrix.features.label}}
+
+    steps:
+    - uses: actions/checkout@v3
+
+    - name: Install stable
+      uses: actions-rs/toolchain@v1
+      with:
+            toolchain: stable
+            profile: minimal
+            override: true
+
+    - uses: taiki-e/install-action@nextest
+    - uses: Swatinem/rust-cache@v2
+
+    - name: Run tests
+      run: cargo +stable nextest run --features ${{ matrix.features.flags }} --verbose --workspace
+
+    - name: Run doctests
+      run: cargo +stable test --doc --features ${{ matrix.features.flags }} --verbose --workspace

.gitignore

@@ -1,3 +1,7 @@
+tantivy.iml
+.cargo
+proptest-regressions
+*.swp
 target
 target/debug
 .vscode
@@ -8,4 +12,5 @@ benchmark
 cpp/simdcomp/bitpackingbenchmark
 *.bk
 .idea
-trace.dat
+trace.dat
+cargo-timing*

.travis.yml

@@ -1,34 +0,0 @@
-language: rust
-cache: cargo
-rust:
-  - nightly
-env:
-  global:
-    - CC=gcc-4.8
-    - CXX=g++-4.8
-    - TRAVIS_CARGO_NIGHTLY_FEATURE=""
-    - secure: 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
-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
-before_script:
-  - |
-      cargo install cargo-travis || echo "cargo-travis already installed"
-      export PATH=$HOME/.cargo/bin:$PATH
-script:
-  - cargo build
-  - cargo test
-  - cargo run --example simple_search
-after_success:
-  - cargo coveralls --exclude-pattern cpp/
-  - travis-cargo doc-upload

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 remove from [alive bitset file](src/fastfield/alive_bitset.rs) that represents the bitset of the alive document ids.
+Like all segment files, this file is immutable. Because it is possible to have more than one alive bitset file at a given instant, the alive bitset 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
+
+```rust
+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 through 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.

AUTHORS

@@ -0,0 +1,11 @@
+# This is the list of authors of tantivy for copyright purposes.
+Paul Masurel
+Laurentiu Nicola
+Dru Sellers
+Ashley Mannix
+Michael J. Curry
+Jason Wolfe
+# As an employee of Google I am required to add Google LLC
+# in the list of authors, but this project is not affiliated to Google
+# in any other way.
+Google LLC 

CHANGELOG.md

@@ -1,22 +1,435 @@
+Tantivy 0.19
+================================
+
+- Updated [Date Field Type](https://github.com/quickwit-oss/tantivy/pull/1396)
+  The `DateTime` type has been updated to hold timestamps with microseconds precision.
+  `DateOptions` and `DatePrecision` have been added to configure Date fields. The precision is used to hint on fast values compression. Otherwise, seconds precision is used everywhere else (i.e terms, indexing).
+- Remove Searcher pool and make `Searcher` cloneable.
+
+Tantivy 0.18
+================================
+
+- For date values `chrono` has been replaced with `time` (@uklotzde) #1304 :
+  - The `time` crate is re-exported as `tantivy::time` instead of `tantivy::chrono`.
+  - The type alias `tantivy::DateTime` has been removed.
+  - `Value::Date` wraps `time::PrimitiveDateTime` without time zone information.
+  - Internally date/time values are stored as seconds since UNIX epoch in UTC.
+  - Converting a `time::OffsetDateTime` to `Value::Date` implicitly converts the value into UTC.
+    If this is not desired do the time zone conversion yourself and use `time::PrimitiveDateTime`
+    directly instead.
+- Add [histogram](https://github.com/quickwit-oss/tantivy/pull/1306) aggregation (@PSeitz)
+- Add support for fastfield on text fields (@PSeitz)
+- Add terms aggregation (@PSeitz)
+- Add support for zstd compression (@kryesh)
+
+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 for range, average and stats 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. (transposition_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).
+- Migrated tantivy error from the now deprecated `failure` crate to `thiserror` #760. (@hirevo)
+- API Change. Accessing the typed value off a `Schema::Value` now returns an Option instead of panicking if the type does not match.
+- Large API Change in the Directory API. Tantivy used to assume that all files could be somehow memory mapped. After this change, Directory return a `FileSlice` that can be reduced and eventually read into an `OwnedBytes` object. Long and blocking io operation are still required by they do not span over the entire file.
+- Added support for Brotli compression in the DocStore. (@ppodolsky)
+- Added helper for building intersections and unions in BooleanQuery (@guilload)
+- Bugfix in `Query::explain`
+- Removed dependency on `notify` #924. Replaced with `FileWatcher` struct that polls meta file every 500ms in background thread. (@halvorboe @guilload)
+- Added `FilterCollector`, which wraps another collector and filters docs using a predicate over a fast field (@barrotsteindev)
+- 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.
+
+Tantivy 0.13.2
+===================
+
+Bugfix. Acquiring a facet reader on a segment that does not contain any
+doc with this facet returns `None`. (#896)
+
+Tantivy 0.13.1
+===================
+
+Made `Query` and `Collector` `Send + Sync`.
+Updated misc dependency versions.
+
+Tantivy 0.13.0
+======================
+
+Tantivy 0.13 introduce a change in the index format that will require
+you to reindex your index (BlockWAND information are added in the skiplist).
+The index size increase is minor as this information is only added for
+full blocks.
+If you have a massive index for which reindexing is not an option, please contact me
+so that we can discuss possible solutions.
+
+- Bugfix in `FuzzyTermQuery` not matching terms by prefix when it should (@Peachball)
+- Relaxed constraints on the custom/tweak score functions. At the segment level, they can be mut, and they are not required to be Sync + Send.
+- `MMapDirectory::open` does not return a `Result` anymore.
+- Change in the DocSet and Scorer API. (@fulmicoton).
+A freshly created DocSet point directly to their first doc. A sentinel value called TERMINATED marks the end of a DocSet.
+`.advance()` returns the new DocId. `Scorer::skip(target)` has been replaced by `Scorer::seek(target)` and returns the resulting DocId.
+As a result, iterating through DocSet now looks as follows
+
+```rust
+let mut doc = docset.doc();
+while doc != TERMINATED {
+   // ...
+   doc = docset.advance();
+}
+```
+
+The change made it possible to greatly simplify a lot of the docset's code.
+
+- Misc internal optimization and introduction of the `Scorer::for_each_pruning` function. (@fulmicoton)
+- Added an offset option to the Top(.*)Collectors. (@robyoung)
+- Added Block WAND. Performance on TOP-K on term-unions should be greatly increased. (@fulmicoton, and special thanks
+to the PISA team for answering all my questions!)
+
+Tantivy 0.12.0
+======================
+
+- Removing static dispatch in tokenizers for simplicity. (#762)
+- Added backward iteration for `TermDictionary` stream. (@halvorboe)
+- Fixed a performance issue when searching for the posting lists of a missing term (@audunhalland)
+- Added a configurable maximum number of docs (10M by default) for a segment to be considered for merge (@hntd187, landed by @halvorboe #713)
+- Important Bugfix #777, causing tantivy to retain memory mapping. (diagnosed by @poljar)
+- Added support for field boosting. (#547, @fulmicoton)
+
+## How to update?
+
+Crates relying on custom tokenizer, or registering tokenizer in the manager will require some
+minor changes. Check <https://github.com/quickwit-oss/tantivy/blob/main/examples/custom_tokenizer.rs>
+to check for some code sample.
+
+Tantivy 0.11.3
+=======================
+
+- Fixed DateTime as a fast field (#735)
+
+Tantivy 0.11.2
+=======================
+
+- The future returned by `IndexWriter::merge` does not borrow `self` mutably anymore (#732)
+- Exposing a constructor for `WatchHandle` (#731)
+
+Tantivy 0.11.1
+=====================
+
+- Bug fix #729
+
+Tantivy 0.11.0
+=====================
+
+- Added f64 field. Internally reuse u64 code the same way i64 does (@fdb-hiroshima)
+- Various bugfixes in the query parser.
+  - Better handling of hyphens in query parser. (#609)
+  - Better handling of whitespaces.
+- Closes #498 - add support for Elastic-style unbounded range queries for alphanumeric types eg. "title:>hello", "weight:>=70.5", "height:<200" (@petr-tik)
+- API change around `Box<BoxableTokenizer>`. See detail in #629
+- Avoid rebuilding Regex automaton whenever a regex query is reused. #639 (@brainlock)
+- Add footer with some metadata to index files. #605 (@fdb-hiroshima)
+- Add a method to check the compatibility of the footer in the index with the running version of tantivy (@petr-tik)
+- TopDocs collector: ensure stable sorting on equal score. #671 (@brainlock)
+- Added handling of pre-tokenized text fields (#642), which will enable users to
+  load tokens created outside tantivy. See usage in examples/pre_tokenized_text. (@kkoziara)
+- Fix crash when committing multiple times with deleted documents. #681 (@brainlock)
+
+## How to update?
+
+- The index format is changed. You are required to reindex your data to use tantivy 0.11.
+- `Box<dyn BoxableTokenizer>` has been replaced by a `BoxedTokenizer` struct.
+- Regex are now compiled when the `RegexQuery` instance is built. As a result, it can now return
+an error and handling the `Result` is required.
+- `tantivy::version()` now returns a `Version` object. This object implements `ToString()`
+
+Tantivy 0.10.2
+=====================
+
+- Closes #656. Solving memory leak.
+
+Tantivy 0.10.1
+=====================
+
+- Closes #544.  A few users experienced problems with the directory watching system.
+Avoid watching the mmap directory until someone effectively creates a reader that uses
+this functionality.
+
+Tantivy 0.10.0
+=====================
+
+*Tantivy 0.10.0 index format is compatible with the index format in 0.9.0.*
+
+- Added an API to easily tweak or entirely replace the
+ default score. See `TopDocs::tweak_score`and `TopScore::custom_score` (@fulmicoton)
+- Added an ASCII folding filter (@drusellers)
+- Bugfix in `query.count` in presence of deletes (@fulmicoton)
+- Added `.explain(...)` in `Query` and `Weight` to (@fulmicoton)
+- Added an efficient way to `delete_all_documents` in `IndexWriter` (@petr-tik).
+  All segments are simply removed.
+
+Minor
+---------
+
+- Switched to Rust 2018 (@uvd)
+- Small simplification of the code.
+Calling .freq() or .doc() when .advance() has never been called
+on segment postings should panic from now on.
+- Tokens exceeding `u16::max_value() - 4` chars are discarded silently instead of panicking.
+- Fast fields are now preloaded when the `SegmentReader` is created.
+- `IndexMeta` is now public.  (@hntd187)
+- `IndexWriter` `add_document`, `delete_term`. `IndexWriter` is `Sync`, making it possible to use it with a `Arc<RwLock<IndexWriter>>`. `add_document` and `delete_term` can
+only require a read lock. (@fulmicoton)
+- Introducing `Opstamp` as an expressive type alias for `u64`. (@petr-tik)
+- Stamper now relies on `AtomicU64` on all platforms (@petr-tik)
+- Bugfix - Files get deleted slightly earlier
+- Compilation resources improved (@fdb-hiroshima)
+
+## How to update?
+
+Your program should be usable as is.
+
+### Fast fields
+
+Fast fields used to be accessed directly from the `SegmentReader`.
+The API changed, you are now required to acquire your fast field reader via the
+`segment_reader.fast_fields()`, and use one of the typed method:
+
+- `.u64()`, `.i64()` if your field is single-valued ;
+- `.u64s()`, `.i64s()` if your field is multi-valued ;
+- `.bytes()` if your field is bytes fast field.
+
+Tantivy 0.9.0
+=====================
+
+*0.9.0 index format is not compatible with the
+previous index format.*
+
+- MAJOR BUGFIX :
+  Some `Mmap` objects were being leaked, and would never get released. (@fulmicoton)
+- Removed most unsafe (@fulmicoton)
+- Indexer memory footprint improved. (VInt comp, inlining the first block. (@fulmicoton)
+- Stemming in other language possible (@pentlander)
+- Segments with no docs are deleted earlier (@barrotsteindev)
+- Added grouped add and delete operations.
+  They are guaranteed to happen together (i.e. they cannot be split by a commit).
+  In addition, adds are guaranteed to happen on the same segment. (@elbow-jason)
+- Removed `INT_STORED` and `INT_INDEXED`. It is now possible to use `STORED` and `INDEXED`
+  for int fields. (@fulmicoton)
+- Added DateTime field (@barrotsteindev)
+- Added IndexReader. By default, index is reloaded automatically upon new commits (@fulmicoton)
+- SIMD linear search within blocks (@fulmicoton)
+
+## How to update ?
+
+tantivy 0.9 brought some API breaking change.
+To update from tantivy 0.8, you will need to go through the following steps.
+
+- `schema::INT_INDEXED` and `schema::INT_STORED`  should be replaced by `schema::INDEXED` and `schema::INT_STORED`.
+- The index now does not hold the pool of searcher anymore. You are required to create an intermediary object called
+`IndexReader` for this.
+
+    ```rust
+    // create the reader. You typically need to create 1 reader for the entire
+    // lifetime of you program.
+    let reader = index.reader()?;
+
+    // Acquire a searcher (previously `index.searcher()`) is now written:
+    let searcher = reader.searcher();
+
+    // With the default setting of the reader, you are not required to
+    // call `index.load_searchers()` anymore.
+    //
+    // The IndexReader will pick up that change automatically, regardless
+    // of whether the update was done in a different process or not.
+    // If this behavior is not wanted, you can create your reader with
+    // the `ReloadPolicy::Manual`, and manually decide when to reload the index
+    // by calling `reader.reload()?`.
+
+    ```
+
+Tantivy 0.8.2
+=====================
+
+Fixing build for x86_64 platforms. (#496)
+No need to update from 0.8.1 if tantivy
+is building on your platform.
+
+Tantivy 0.8.1
+=====================
+
+Hotfix of #476.
+
+Merge was reflecting deletes before commit was passed.
+Thanks @barrotsteindev  for reporting the bug.
+
+Tantivy 0.8.0
+=====================
+
+*No change in the index format*
+
+- API Breaking change in the collector API. (@jwolfe, @fulmicoton)
+- Multithreaded search (@jwolfe, @fulmicoton)
+
+Tantivy 0.7.1
+=====================
+
+*No change in the index format*
+
+- Bugfix: NGramTokenizer panics on non ascii chars
+- Added a space usage API
+
+Tantivy 0.7
+=====================
+
+- Skip data for doc ids and positions (@fulmicoton),
+  greatly improving performance
+- Tantivy error now rely on the failure crate (@drusellers)
+- Added support for `AND`, `OR`, `NOT` syntax in addition to the `+`,`-` syntax
+- Added a snippet generator with highlight (@vigneshsarma, @fulmicoton)
+- Added a `TopFieldCollector` (@pentlander)
+
+Tantivy 0.6.1
+=========================
+
+- Bugfix #324. GC removing was removing file that were still in useful
+- Added support for parsing AllQuery and RangeQuery via QueryParser
+  - AllQuery: `*`
+  - RangeQuery:
+    - Inclusive `field:[startIncl to endIncl]`
+    - Exclusive `field:{startExcl to endExcl}`
+    - Mixed `field:[startIncl to endExcl}` and vice versa
+    - Unbounded `field:[start to *]`, `field:[* to end]`
+
+Tantivy 0.6
+==========================
+
+Special thanks to @drusellers and @jason-wolfe for their contributions
+to this release!
+
+- Removed C code. Tantivy is now pure Rust. (@fulmicoton)
+- BM25 (@fulmicoton)
+- Approximate field norms encoded over 1 byte. (@fulmicoton)
+- Compiles on stable rust (@fulmicoton)
+- Add &[u8] fastfield for associating arbitrary bytes to each document (@jason-wolfe) (#270)
+  - Completely uncompressed
+  - Internally: One u64 fast field for indexes, one fast field for the bytes themselves.
+- Add NGram token support (@drusellers)
+- Add Stopword Filter support (@drusellers)
+- Add a FuzzyTermQuery (@drusellers)
+- Add a RegexQuery (@drusellers)
+- Various performance improvements (@fulmicoton)_
+
+Tantivy 0.5.2
+===========================
+
+- bugfix #274
+- bugfix #280
+- bugfix #289
+
+Tantivy 0.5.1
+==========================
+
+- bugfix #254 : tantivy failed if no documents in a segment contained a specific field.
+
+Tantivy 0.5
+==========================
+
+- Faceting
+- RangeQuery
+- Configurable tokenization pipeline
+- Bugfix in PhraseQuery
+- Various query optimisation
+- Allowing very large indexes
+  - 64 bits file address
+  - Smarter encoding of the `TermInfo` objects
 
 Tantivy 0.4.3
 ==========================
 
 - Bugfix race condition when deleting files. (#198)
 
-
 Tantivy 0.4.2
 ==========================
 
 - Prevent usage of AVX2 instructions (#201)
 
-
 Tantivy 0.4.1
 ==========================
 
 - Bugfix for non-indexed fields. (#199)
 
-
 Tantivy 0.4.0
 ==========================
 
@@ -31,58 +444,49 @@ Tantivy 0.4.0
   - Searching for a non-indexed field returns an explicit Error
   - Phrase query for non-tokenized field are not tokenized by the query parser.
 - Faster/Better indexing (@fulmicoton)
-    - using murmurhash2
-    - faster merging
-    - more memory efficient fast field writer (@lnicola )
-    - better handling of collisions
-    - lesser memory usage
+  - using murmurhash2
+  - faster merging
+  - more memory efficient fast field writer (@lnicola )
+  - better handling of collisions
+  - lesser memory usage
 - Added API, most notably to iterate over ranges of terms (@fulmicoton)
 - Bugfix that was preventing to unmap segment files, on index drop (@fulmicoton)
 - Made the doc! macro public (@fulmicoton)
 - Added an alternative implementation of the streaming dictionary (@fulmicoton)
 
-
-
 Tantivy 0.3.1
 ==========================
 
 - Expose a method to trigger files garbage collection
 
-
-
 Tantivy 0.3
 ==========================
 
-
 Special thanks to @Kodraus @lnicola @Ameobea @manuel-woelker @celaus
 for their contribution to this release.
 
-Thanks also to everyone in tantivy gitter chat 
+Thanks also to everyone in tantivy gitter chat
 for their advise and company :)
 
-https://gitter.im/tantivy-search/tantivy
-
+<https://gitter.im/tantivy-search/tantivy>
 
 Warning:
 
-Tantivy 0.3 is NOT backward compatible with tantivy 0.2 
+Tantivy 0.3 is NOT backward compatible with tantivy 0.2
 code and index format.
-You should not expect backward compatibility before 
+You should not expect backward compatibility before
 tantivy 1.0.
 
-
-
 New Features
 ------------
 
 - Delete. You can now delete documents from an index.
 - Support for windows (Thanks to @lnicola)
 
-
 Various Bugfixes & small improvements
 ----------------------------------------
 
-- Added CI for Windows (https://ci.appveyor.com/project/fulmicoton/tantivy)
+- Added CI for Windows (<https://ci.appveyor.com/project/fulmicoton/tantivy>)
 Thanks to @KodrAus ! (#108)
 - Various dependy version update (Thanks to @Ameobea) #76
 - Fixed several race conditions in `Index.wait_merge_threads`
@@ -93,8 +497,4 @@ Thanks to @KodrAus ! (#108)
   the natural ordering.
 - Building binary targets for tantivy-cli (Thanks to @KodrAus)
 - Misc invisible bug fixes, and code cleanup.
-- Use 
-
-
-
-
+- Use

Cargo.toml

@@ -1,71 +1,127 @@
 [package]
 name = "tantivy"
-version = "0.5.0-dev"
+version = "0.18.0"
 authors = ["Paul Masurel <paul.masurel@gmail.com>"]
-build = "build.rs"
 license = "MIT"
 categories = ["database-implementations", "data-structures"]
-description = """Tantivy is a 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"
+description = """Search engine library"""
+documentation = "https://docs.rs/tantivy/"
+homepage = "https://github.com/quickwit-oss/tantivy"
+repository = "https://github.com/quickwit-oss/tantivy"
 readme = "README.md"
 keywords = ["search", "information", "retrieval"]
+edition = "2021"
 
 [dependencies]
-byteorder = "1.0"
-lazy_static = "0.2.1"
-tinysegmenter = "0.1.0"
-regex = "0.2"
-fst = "0.2"
-atomicwrites = "0.1.3"
-tempfile = "2.1"
-log = "0.3.6"
-combine = "2.2"
-tempdir = "0.3"
-serde = "1.0"
-serde_derive = "1.0"
-serde_json = "1.0"
-bincode = "0.8"
-libc = {version = "0.2.20", optional=true}
-num_cpus = "1.2"
-itertools = "0.5.9"
-lz4 = "1.20"
-bit-set = "0.4.0"
-time = "0.1"
-uuid = { version = "0.5", features = ["v4", "serde"] }
-chan = "0.1"
-version = "2"
-crossbeam = "0.3"
-futures = "0.1"
-futures-cpupool = "0.1"
-error-chain = "0.8"
-owning_ref = "0.3"
-stable_deref_trait = "1.0.0"
-rust-stemmers = "0.1.0"
+oneshot = "0.1.3"
+base64 = "0.13.0"
+byteorder = "1.4.3"
+crc32fast = "1.3.2"
+once_cell = "1.10.0"
+regex = { version = "1.5.5", default-features = false, features = ["std", "unicode"] }
+tantivy-fst = "0.3.0"
+memmap2 = { version = "0.5.3", optional = true }
+lz4_flex = { version = "0.9.2", default-features = false, features = ["checked-decode"], optional = true }
+brotli = { version = "3.3.4", optional = true }
+zstd = { version = "0.11", optional = true }
+snap = { version = "1.0.5", optional = true }
+tempfile = { version = "3.3.0", optional = true }
+log = "0.4.16"
+serde = { version = "1.0.136", features = ["derive"] }
+serde_json = "1.0.79"
+num_cpus = "1.13.1"
+fs2={ version = "0.4.3", optional = true }
+levenshtein_automata = "0.2.1"
+uuid = { version = "1.0.0", features = ["v4", "serde"] }
+crossbeam-channel = "0.5.4"
+tantivy-query-grammar = { version="0.18.0", path="./query-grammar" }
+tantivy-bitpacker = { version="0.2", path="./bitpacker" }
+common = { version = "0.3", path = "./common/", package = "tantivy-common" }
+fastfield_codecs = { version="0.2", path="./fastfield_codecs", default-features = false }
+ownedbytes = { version="0.3", path="./ownedbytes" }
+stable_deref_trait = "1.2.0"
+rust-stemmers = "1.2.0"
+downcast-rs = "1.2.0"
+bitpacking = { version = "0.8.4", default-features = false, features = ["bitpacker4x"] }
+census = "0.4.0"
+fnv = "1.0.7"
+thiserror = "1.0.30"
+htmlescape = "0.3.1"
+fail = "0.5.0"
+murmurhash32 = "0.2.0"
+time = { version = "0.3.10", features = ["serde-well-known"] }
+smallvec = "1.8.0"
+rayon = "1.5.2"
+lru = "0.7.5"
+fastdivide = "0.4.0"
+itertools = "0.10.3"
+measure_time = "0.8.2"
+pretty_assertions = "1.2.1"
+serde_cbor = { version = "0.11.2", optional = true }
+async-trait = "0.1.53"
+arc-swap = "1.5.0"
 
 [target.'cfg(windows)'.dependencies]
-winapi = "0.2"
+winapi = "0.3.9"
 
 [dev-dependencies]
-rand = "0.3"
-env_logger = "0.4"
+rand = "0.8.5"
+maplit = "1.0.2"
+matches = "0.1.9"
+proptest = "1.0.0"
+criterion = "0.3.5"
+test-log = "0.2.10"
+env_logger = "0.9.0"
+pprof = { version = "0.10.0", features = ["flamegraph", "criterion"] }
+futures = "0.3.21"
 
-[build-dependencies]
-cc = {version = "1.0.0", optional=true}
+[dev-dependencies.fail]
+version = "0.5.0"
+features = ["failpoints"]
 
 [profile.release]
 opt-level = 3
 debug = false
-lto = true
 debug-assertions = false
 
+[profile.test]
+debug-assertions = true
+overflow-checks = true
 
 [features]
-default = ["simdcompression"]
-simdcompression = ["libc", "cc"]
-streamdict = []
+default = ["mmap", "lz4-compression" ]
+mmap = ["fs2", "tempfile", "memmap2"]
 
+brotli-compression = ["brotli"]
+lz4-compression = ["lz4_flex"]
+snappy-compression = ["snap"]
+zstd-compression = ["zstd"]
+
+failpoints = ["fail/failpoints"]
+unstable = [] # useful for benches.
+
+quickwit = ["serde_cbor"]
+
+[workspace]
+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
+# in a different binary.
+#
+# We do that because, fail rely on a global definition of
+# failpoints behavior and hence, it is incompatible with
+# multithreading.
+[[test]]
+name = "failpoints"
+path = "tests/failpoints/mod.rs"
+required-features = ["fail/failpoints"]
+
+[[bench]]
+name = "analyzer"
+harness = false
+
+[[bench]]
+name = "index-bench"
+harness = false
 
-[badges]
-travis-ci = { repository = "tantivy-search/tantivy" }

LICENSE

@@ -1,4 +1,4 @@
-Copyright (c) 2016 Paul Masurel
+Copyright (c) 2018 by the project authors, as listed in the AUTHORS file. 
 
 Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
 

Makefile

@@ -0,0 +1,6 @@
+test:
+	echo "Run test only... No examples."
+	cargo test --tests --lib
+
+fmt:
+	cargo +nightly fmt --all

README.md

@@ -1,60 +1,168 @@
+[![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)
+[![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)
 
-[![Build Status](https://travis-ci.org/tantivy-search/tantivy.svg?branch=master)](https://travis-ci.org/tantivy-search/tantivy)
-[![Coverage Status](https://coveralls.io/repos/github/tantivy-search/tantivy/badge.svg?branch=master&refresh1)](https://coveralls.io/github/tantivy-search/tantivy?branch=master)
-[![Join the chat at https://gitter.im/tantivy-search/tantivy](https://badges.gitter.im/tantivy-search/tantivy.svg)](https://gitter.im/tantivy-search/tantivy?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=badge)
-[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
-[![Build status](https://ci.appveyor.com/api/projects/status/r7nb13kj23u8m9pj?svg=true)](https://ci.appveyor.com/project/fulmicoton/tantivy)
-![beacon for google analytics](https://ga-beacon.appspot.com/UA-88834340-1/tantivy/README)
+**Tantivy** is a **full-text search engine library** written in Rust.
 
-**Tantivy** is a **full text search engine library** written in rust.
+It is closer to [Apache Lucene](https://lucene.apache.org/) than to [Elasticsearch](https://www.elastic.co/products/elasticsearch) or [Apache Solr](https://lucene.apache.org/solr/) in the sense it is not
+an off-the-shelf search engine server, but rather a crate that can be used
+to build such a search engine.
 
-It is strongly inspired by Lucene's design.
+Tantivy is, in fact, strongly inspired by Lucene's design.
 
+If you are looking for an alternative to Elasticsearch or Apache Solr, check out [Quickwit](https://github.com/quickwit-oss/quickwit), our search engine built on top of Tantivy.
+
+# Benchmark
+
+The following [benchmark](https://tantivy-search.github.io/bench/) breakdowns
+performance for different types of queries/collections.
+
+Your mileage WILL vary depending on the nature of queries and their load.
+
+<img src="doc/assets/images/searchbenchmark.png">
 
 # Features
 
-- configurable indexing (optional term frequency and position indexing)
-- tf-idf scoring
-- Basic query language
-- Phrase queries
+- 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), [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)
+- Natural query language (e.g. `(michael AND jackson) OR "king of pop"`)
+- Phrase queries search (e.g. `"michael jackson"`)
 - Incremental indexing
 - Multithreaded indexing (indexing English Wikipedia takes < 3 minutes on my desktop)
-- mmap based
-- optional SIMD integer compression
-- u64 and i64 fast fields (equivalent of doc values in Lucene)
+- Mmap directory
+- SIMD integer compression when the platform/CPU includes the SSE2 instruction set
+- Single valued and multivalued u64, i64, and f64 fast fields (equivalent of doc values in Lucene)
+- `&[u8]` fast fields
+- Text, i64, u64, f64, dates, and hierarchical facet fields
 - LZ4 compressed document store
+- Range queries
+- Faceted search
+- Configurable indexing (optional term frequency and position indexing)
+- JSON Field
+- Aggregation Collector: range buckets, average, and stats metrics
+- LogMergePolicy with deletes
+- Searcher Warmer API
 - Cheesy logo with a horse
 
-Tantivy supports Linux, MacOS and Windows.
+## Non-features
 
+Distributed search is out of the scope of Tantivy, but if you are looking for this feature, check out [Quickwit](https://github.com/quickwit-oss/quickwit/).
 
 # Getting started
 
-- [tantivy's usage example](http://fulmicoton.com/tantivy-examples/simple_search.html)
-- [tantivy-cli and its tutorial](https://github.com/tantivy-search/tantivy-cli).
-It will walk you through getting a wikipedia search engine up and running in a few minutes.
-- [reference doc]
-    - [For the last released version](https://docs.rs/tantivy/)
-    - [For the last master branch](https://tantivy-search.github.io/tantivy/tantivy/index.html)
+Tantivy works on stable Rust (>= 1.27) and supports Linux, macOS, and Windows.
 
-# Compiling
+- [Tantivy's simple search example](https://tantivy-search.github.io/examples/basic_search.html)
+- [tantivy-cli and its tutorial](https://github.com/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/)
 
-Tantivy requires Rust Nightly because it uses requires the features [`box_syntax`](https://doc.rust-lang.org/stable/book/box-syntax-and-patterns.html), [`optin_builtin_traits`](https://github.com/rust-lang/rfcs/blob/master/text/0019-opt-in-builtin-traits.md), and [`conservative_impl_trait`](https://github.com/rust-lang/rfcs/blob/master/text/1522-conservative-impl-trait.md).
-The project can then be built using `cargo`.
+# How can I support this project?
 
-    git clone git@github.com:tantivy-search/tantivy.git
+There are many ways to support this project.
+
+- 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 Discord server](https://discord.gg/MT27AG5EVE))
+- Talk about Tantivy around you
+
+# Contributing code
+
+We use the GitHub Pull Request workflow: reference a GitHub ticket and/or include a comprehensive commit message when opening a PR.
+
+## Clone and build locally
+
+Tantivy compiles on stable Rust but requires `Rust >= 1.27`.
+To check out and run tests, you can simply run:
+
+```bash
+    git clone https://github.com/quickwit-oss/tantivy.git
     cd tantivy
     cargo build
+```
 
+## Run tests
 
-Alternatively, if you are trying to compile `tantivy` without simd compression,
-you can disable this functionality. In this case, this submodule is not required
-and you can compile tantivy by using the `--no-default-features` flag.
+Some tests will not run with just `cargo test` because of `fail-rs`.
+To run the tests exhaustively, run `./run-tests.sh`.
 
-    cargo build --no-default-features
+## Debug
 
+You might find it useful to step through the programme with a debugger.
 
-# Contribute
+### A failing test
 
-Send me an email (paul.masurel at gmail.com) if you want to contribute to tantivy.
+Make sure you haven't run `cargo clean` after the most recent `cargo test` or `cargo build` to guarantee that the `target/` directory exists. Use this bash script to find the name of the most recent debug build of Tantivy and run it under `rust-gdb`:
+
+```bash
+find target/debug/ -maxdepth 1 -executable -type f -name "tantivy*" -printf '%TY-%Tm-%Td %TT %p\n' | sort -r | cut -d " " -f 3 | xargs -I RECENT_DBG_TANTIVY rust-gdb RECENT_DBG_TANTIVY
+```
+
+Now that you are in `rust-gdb`, you can set breakpoints on lines and methods that match your source code and run the debug executable with flags that you normally pass to `cargo test` like this:
+
+```bash
+$gdb run --test-threads 1 --test $NAME_OF_TEST
+```
+
+### An example
+
+By default, `rustc` compiles everything in the `examples/` directory in debug mode. This makes it easy for you to make examples to reproduce bugs:
+
+```bash
+rust-gdb target/debug/examples/$EXAMPLE_NAME
+$ gdb run
+```
+
+# Companies Using Tantivy
+
+<p align="left">
+<img align="center" src="doc/assets/images/Nuclia.png#gh-light-mode-only" alt="Nuclia" height="25" width="auto" /> &nbsp;
+<img align="center" src="doc/assets/images/humanfirst.png#gh-light-mode-only" alt="Humanfirst.ai" height="30" width="auto" />
+<img align="center" src="doc/assets/images/element.io.svg#gh-light-mode-only" alt="Element.io" height="25" width="auto" />
+<img align="center" src="doc/assets/images/nuclia-dark-theme.png#gh-dark-mode-only" alt="Nuclia" height="35" width="auto" /> &nbsp;
+<img align="center" src="doc/assets/images/humanfirst.ai-dark-theme.png#gh-dark-mode-only" alt="Humanfirst.ai" height="25" width="auto" />&nbsp; &nbsp;
+<img align="center" src="doc/assets/images/element-dark-theme.png#gh-dark-mode-only" alt="Element.io" height="25" width="auto" />
+</p>
+
+# FAQ
+
+### Can I use Tantivy in other languages?
+
+- Python → [tantivy-py](https://github.com/quickwit-oss/tantivy-py)
+- Ruby → [tantiny](https://github.com/baygeldin/tantiny)
+
+You can also find other bindings on [GitHub](https://github.com/search?q=tantivy) but they may be less maintained.
+
+### What are some examples of Tantivy use?
+
+- [seshat](https://github.com/matrix-org/seshat/): A matrix message database/indexer
+- [tantiny](https://github.com/baygeldin/tantiny): Tiny full-text search for Ruby
+- [lnx](https://github.com/lnx-search/lnx): adaptable, typo tolerant search engine with a REST API
+- and [more](https://github.com/search?q=tantivy)!
+
+### On average, how much faster is Tantivy compared to Lucene?
+
+- According to our [search latency benchmark](https://tantivy-search.github.io/bench/), Tantivy is approximately 2x faster than Lucene.
+
+### Does tantivy support incremental indexing?
+
+- Yes.
+
+### How can I edit documents?
+
+- Data in tantivy is immutable. To edit a document, the document needs to be deleted and reindexed.
+
+### When will my documents be searchable during indexing?
+
+- Documents will be searchable after a `commit` is called on an `IndexWriter`. Existing `IndexReader`s will also need to be reloaded in order to reflect the changes. Finally, changes are only visible to newly acquired `Searcher`.

appveyor.yml

@@ -4,11 +4,8 @@
 os: Visual Studio 2015
 environment:
   matrix:
-    - channel: nightly
+    - channel: stable
       target: x86_64-pc-windows-msvc
-    - channel: nightly
-      target: x86_64-pc-windows-gnu
-      msys_bits: 64
 
 install:
   - appveyor DownloadFile https://win.rustup.rs/ -FileName rustup-init.exe
@@ -21,5 +18,6 @@ install:
 build: false
 
 test_script:
-  - REM SET RUST_LOG=tantivy,test & cargo test --verbose
-  - REM SET RUST_BACKTRACE=1 & cargo run --example simple_search
+  - 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

@@ -0,0 +1,22 @@
+use criterion::{criterion_group, criterion_main, Criterion};
+use tantivy::tokenizer::TokenizerManager;
+
+const ALICE_TXT: &str = include_str!("alice.txt");
+
+pub fn criterion_benchmark(c: &mut Criterion) {
+    let tokenizer_manager = TokenizerManager::default();
+    let tokenizer = tokenizer_manager.get("default").unwrap();
+    c.bench_function("default-tokenize-alice", |b| {
+        b.iter(|| {
+            let mut word_count = 0;
+            let mut token_stream = tokenizer.token_stream(ALICE_TXT);
+            while token_stream.advance() {
+                word_count += 1;
+            }
+            assert_eq!(word_count, 30_731);
+        })
+    });
+}
+
+criterion_group!(benches, criterion_benchmark);
+criterion_main!(benches);

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 = "0.2.0"
+edition = "2021"
+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

@@ -0,0 +1,145 @@
+use std::convert::TryInto;
+use std::io;
+
+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 {
+            mini_buffer: 0u64,
+            mini_buffer_written: 0,
+        }
+    }
+
+    #[inline]
+    pub fn write<TWrite: io::Write>(
+        &mut self,
+        val: u64,
+        num_bits: u8,
+        output: &mut TWrite,
+    ) -> io::Result<()> {
+        let val_u64 = val as u64;
+        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_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_all(self.mini_buffer.to_le_bytes().as_ref())?;
+                self.mini_buffer_written = 0;
+                self.mini_buffer = 0u64;
+            }
+        }
+        Ok(())
+    }
+
+    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 bytes = self.mini_buffer.to_le_bytes();
+            output.write_all(&bytes[..num_bytes])?;
+            self.mini_buffer_written = 0;
+            self.mini_buffer = 0;
+        }
+        Ok(())
+    }
+
+    pub fn close<TWrite: io::Write>(&mut self, output: &mut TWrite) -> io::Result<()> {
+        self.flush(output)?;
+        // Padding the write file to simplify reads.
+        output.write_all(&[0u8; 7])?;
+        Ok(())
+    }
+}
+
+#[derive(Clone, Debug, Default)]
+pub struct BitUnpacker {
+    num_bits: u64,
+    mask: u64,
+}
+
+impl BitUnpacker {
+    pub fn new(num_bits: u8) -> BitUnpacker {
+        let mask: u64 = if num_bits == 64 {
+            !0u64
+        } else {
+            (1u64 << num_bits) - 1u64
+        };
+        BitUnpacker {
+            num_bits: u64::from(num_bits),
+            mask,
+        }
+    }
+
+    pub fn bit_width(&self) -> u8 {
+        self.num_bits as u8
+    }
+
+    #[inline]
+    pub fn get(&self, idx: u64, data: &[u8]) -> u64 {
+        if self.num_bits == 0 {
+            return 0u64;
+        }
+        let addr_in_bits = idx * self.num_bits;
+        let addr = addr_in_bits >> 3;
+        let bit_shift = addr_in_bits & 7;
+        debug_assert!(
+            addr + 8 <= data.len() as u64,
+            "The fast field field should have been padded with 7 bytes."
+        );
+        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 & self.mask
+    }
+}
+
+#[cfg(test)]
+mod test {
+    use super::{BitPacker, BitUnpacker};
+
+    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;
+        let vals: Vec<u64> = (0u64..len as u64)
+            .map(|i| if max_val == 0 { 0 } else { i % max_val })
+            .collect();
+        for &val in &vals {
+            bitpacker.write(val, num_bits, &mut data).unwrap();
+        }
+        bitpacker.close(&mut data).unwrap();
+        assert_eq!(data.len(), ((num_bits as usize) * len + 7) / 8 + 7);
+        let bitunpacker = BitUnpacker::new(num_bits);
+        (bitunpacker, vals, data)
+    }
+
+    fn test_bitpacker_util(len: usize, num_bits: u8) {
+        let (bitunpacker, vals, data) = create_fastfield_bitpacker(len, num_bits);
+        for (i, val) in vals.iter().enumerate() {
+            assert_eq!(bitunpacker.get(i as u64, &data), *val);
+        }
+    }
+
+    #[test]
+    fn test_bitpacker() {
+        test_bitpacker_util(10, 3);
+        test_bitpacker_util(10, 0);
+        test_bitpacker_util(10, 1);
+        test_bitpacker_util(6, 14);
+        test_bitpacker_util(1000, 14);
+    }
+}

bitpacker/src/blocked_bitpacker.rs

@@ -0,0 +1,179 @@
+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);
+}
+
+fn mem_usage<T>(items: &Vec<T>) -> usize {
+    items.capacity() * std::mem::size_of::<T>()
+}
+
+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()
+            + mem_usage(&self.offset_and_bits)
+            + mem_usage(&self.buffer)
+    }
+
+    #[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)));
+}

build.rs

@@ -1,61 +0,0 @@
-#[cfg(feature = "simdcompression")]
-mod build {
-    extern crate cc;
-
-    pub fn build() {
-        let mut config = cc::Build::new();
-        config
-            .include("./cpp/simdcomp/include")
-            .file("cpp/simdcomp/src/avxbitpacking.c")
-            .file("cpp/simdcomp/src/simdintegratedbitpacking.c")
-            .file("cpp/simdcomp/src/simdbitpacking.c")
-            .file("cpp/simdcomp/src/simdpackedsearch.c")
-            .file("cpp/simdcomp/src/simdcomputil.c")
-            .file("cpp/simdcomp/src/simdpackedselect.c")
-            .file("cpp/simdcomp/src/simdfor.c")
-            .file("cpp/simdcomp_wrapper.c");
-
-        if !cfg!(debug_assertions) {
-            config.opt_level(3);
-
-            if cfg!(target_env = "msvc") {
-                config
-                    .define("NDEBUG", None)
-                    .flag("/Gm-")
-                    .flag("/GS-")
-                    .flag("/Gy")
-                    .flag("/Oi")
-                    .flag("/GL");
-            }
-        }
-
-        if !cfg!(target_env = "msvc") {
-            config
-                .include("./cpp/streamvbyte/include")
-                .file("cpp/streamvbyte/src/streamvbyte.c")
-                .file("cpp/streamvbyte/src/streamvbytedelta.c")
-                .flag("-msse4.1")
-                .flag("-march=native")
-                .flag("-std=c99");
-        }
-
-        config.compile("libsimdcomp.a");
-
-        // Workaround for linking static libraries built with /GL
-        // https://github.com/rust-lang/rust/issues/26003
-        if !cfg!(debug_assertions) && cfg!(target_env = "msvc") {
-            println!("cargo:rustc-link-lib=dylib=simdcomp");
-        }
-
-        println!("cargo:rerun-if-changed=cpp");
-    }
-}
-
-#[cfg(not(feature = "simdcompression"))]
-mod build {
-    pub fn build() {}
-}
-
-fn main() {
-    build::build();
-}

ci/before_deploy.ps1

@@ -0,0 +1,23 @@
+# This script takes care of packaging the build artifacts that will go in the
+# release zipfile
+
+$SRC_DIR = $PWD.Path
+$STAGE = [System.Guid]::NewGuid().ToString()
+
+Set-Location $ENV:Temp
+New-Item -Type Directory -Name $STAGE
+Set-Location $STAGE
+
+$ZIP = "$SRC_DIR\$($Env:CRATE_NAME)-$($Env:APPVEYOR_REPO_TAG_NAME)-$($Env:TARGET).zip"
+
+# TODO Update this to package the right artifacts
+Copy-Item "$SRC_DIR\target\$($Env:TARGET)\release\hello.exe" '.\'
+
+7z a "$ZIP" *
+
+Push-AppveyorArtifact "$ZIP"
+
+Remove-Item *.* -Force
+Set-Location ..
+Remove-Item $STAGE
+Set-Location $SRC_DIR

ci/before_deploy.sh

@@ -0,0 +1,33 @@
+# This script takes care of building your crate and packaging it for release
+
+set -ex
+
+main() {
+    local src=$(pwd) \
+          stage=
+
+    case $TRAVIS_OS_NAME in
+        linux)
+            stage=$(mktemp -d)
+            ;;
+        osx)
+            stage=$(mktemp -d -t tmp)
+            ;;
+    esac
+
+    test -f Cargo.lock || cargo generate-lockfile
+
+    # TODO Update this to build the artifacts that matter to you
+    cross rustc --bin hello --target $TARGET --release -- -C lto
+
+    # TODO Update this to package the right artifacts
+    cp target/$TARGET/release/hello $stage/
+
+    cd $stage
+    tar czf $src/$CRATE_NAME-$TRAVIS_TAG-$TARGET.tar.gz *
+    cd $src
+
+    rm -rf $stage
+}
+
+main

ci/install.sh

@@ -0,0 +1,47 @@
+set -ex
+
+main() {
+    local target=
+    if [ $TRAVIS_OS_NAME = linux ]; then
+        target=x86_64-unknown-linux-musl
+        sort=sort
+    else
+        target=x86_64-apple-darwin
+        sort=gsort  # for `sort --sort-version`, from brew's coreutils.
+    fi
+
+    # Builds for iOS are done on OSX, but require the specific target to be
+    # installed.
+    case $TARGET in
+        aarch64-apple-ios)
+            rustup target install aarch64-apple-ios
+            ;;
+        armv7-apple-ios)
+            rustup target install armv7-apple-ios
+            ;;
+        armv7s-apple-ios)
+            rustup target install armv7s-apple-ios
+            ;;
+        i386-apple-ios)
+            rustup target install i386-apple-ios
+            ;;
+        x86_64-apple-ios)
+            rustup target install x86_64-apple-ios
+            ;;
+    esac
+
+    # This fetches latest stable release
+    local tag=$(git ls-remote --tags --refs --exit-code https://github.com/japaric/cross \
+                       | cut -d/ -f3 \
+                       | grep -E '^v[0.1.0-9.]+$' \
+                       | $sort --version-sort \
+                       | tail -n1)
+    curl -LSfs https://japaric.github.io/trust/install.sh | \
+        sh -s -- \
+           --force \
+           --git japaric/cross \
+           --tag $tag \
+           --target $target
+}
+
+main

ci/script.sh

@@ -0,0 +1,30 @@
+#!/usr/bin/env bash
+
+# This script takes care of testing your crate
+
+set -ex
+
+main() {
+    if [ ! -z $CODECOV ]; then
+        echo "Codecov"
+        cargo build --verbose && cargo coverage --verbose --all && bash <(curl -s https://codecov.io/bash) -s target/kcov
+    else
+        echo "Build"
+        cross build --target $TARGET
+        if [ ! -z $DISABLE_TESTS ]; then
+            return
+        fi
+        echo "Test"
+        cross test --target $TARGET --no-default-features --features mmap
+        cross test --target $TARGET --no-default-features --features mmap query-grammar
+    fi
+    for example in $(ls examples/*.rs)
+    do
+        cargo run --example  $(basename $example .rs)
+    done
+}
+
+# we don't run the "test phase" when doing deploys
+if [ -z $TRAVIS_TAG ]; then
+    main
+fi

common/Cargo.toml

@@ -0,0 +1,17 @@
+[package]
+name = "tantivy-common"
+version = "0.3.0"
+authors = ["Paul Masurel <paul@quickwit.io>", "Pascal Seitz <pascal@quickwit.io>"]
+license = "MIT"
+edition = "2021"
+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="0.3", 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

@@ -0,0 +1,168 @@
+#![allow(clippy::len_without_is_empty)]
+
+use std::ops::Deref;
+
+pub use byteorder::LittleEndian as Endianness;
+
+mod bitset;
+mod serialize;
+mod vint;
+mod writer;
+
+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 {
+    /// Return length
+    fn len(&self) -> usize;
+
+    /// Returns true iff empty.
+    fn is_empty(&self) -> bool {
+        self.len() == 0
+    }
+}
+
+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`
+///
+/// For simplicity, tantivy internally handles `i64` as `u64`.
+/// The mapping is defined by this function.
+///
+/// Maps `i64` to `u64` so that
+/// `-2^63 .. 2^63-1` is mapped
+///     to
+/// `0 .. 2^64-1`
+/// in that order.
+///
+/// This is more suited than simply casting (`val as u64`)
+/// because of bitpacking.
+///
+/// Imagine a list of `i64` ranging from -10 to 10.
+/// When casting negative values, the negative values are projected
+/// to values over 2^63, and all values end up requiring 64 bits.
+///
+/// # See also
+/// The [reverse mapping is `u64_to_i64`](./fn.u64_to_i64.html).
+#[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]
+pub fn u64_to_i64(val: u64) -> i64 {
+    (val ^ HIGHEST_BIT) as i64
+}
+
+/// Maps a `f64` to `u64`
+///
+/// For simplicity, tantivy internally handles `f64` as `u64`.
+/// The mapping is defined by this function.
+///
+/// Maps `f64` to `u64` in a monotonic manner, so that bytes lexical order is preserved.
+///
+/// This is more suited than simply casting (`val as u64`)
+/// which would truncate the result
+///
+/// # Reference
+///
+/// Daniel Lemire's [blog post](https://lemire.me/blog/2020/12/14/converting-floating-point-numbers-to-integers-while-preserving-order/)
+/// explains the mapping in a clear manner.
+///
+/// # See also
+/// The [reverse mapping is `u64_to_f64`](./fn.u64_to_f64.html).
+#[inline]
+pub fn f64_to_u64(val: f64) -> u64 {
+    let bits = val.to_bits();
+    if val.is_sign_positive() {
+        bits ^ HIGHEST_BIT
+    } else {
+        !bits
+    }
+}
+
+/// Reverse the mapping given by [`i64_to_u64`](./fn.i64_to_u64.html).
+#[inline]
+pub fn u64_to_f64(val: u64) -> f64 {
+    f64::from_bits(if val & HIGHEST_BIT != 0 {
+        val ^ HIGHEST_BIT
+    } else {
+        !val
+    })
+}
+
+#[cfg(test)]
+pub mod test {
+
+    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);
+    }
+
+    fn test_f64_converter_helper(val: f64) {
+        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)) {
+            let left_u64 = f64_to_u64(left);
+            let right_u64 = f64_to_u64(right);
+            assert_eq!(left_u64 < right_u64,  left < right);
+        }
+    }
+
+    #[test]
+    fn test_i64_converter() {
+        assert_eq!(i64_to_u64(i64::MIN), u64::MIN);
+        assert_eq!(i64_to_u64(i64::MAX), u64::MAX);
+        test_i64_converter_helper(0i64);
+        test_i64_converter_helper(i64::MIN);
+        test_i64_converter_helper(i64::MAX);
+        for i in -1000i64..1000i64 {
+            test_i64_converter_helper(i);
+        }
+    }
+
+    #[test]
+    fn test_f64_converter() {
+        test_f64_converter_helper(f64::INFINITY);
+        test_f64_converter_helper(f64::NEG_INFINITY);
+        test_f64_converter_helper(0.0);
+        test_f64_converter_helper(-0.0);
+        test_f64_converter_helper(1.0);
+        test_f64_converter_helper(-1.0);
+    }
+
+    #[test]
+    fn test_f64_order() {
+        assert!(!(f64_to_u64(f64::NEG_INFINITY)..f64_to_u64(f64::INFINITY))
+            .contains(&f64_to_u64(f64::NAN))); // nan is not a number
+        assert!(f64_to_u64(1.5) > f64_to_u64(1.0)); // same exponent, different mantissa
+        assert!(f64_to_u64(2.0) > f64_to_u64(1.0)); // same mantissa, different exponent
+        assert!(f64_to_u64(2.0) > f64_to_u64(1.5)); // different exponent and mantissa
+        assert!(f64_to_u64(1.0) > f64_to_u64(-1.0)); // pos > neg
+        assert!(f64_to_u64(-1.5) < f64_to_u64(-1.0));
+        assert!(f64_to_u64(-2.0) < f64_to_u64(1.0));
+        assert!(f64_to_u64(-2.0) < f64_to_u64(-1.5));
+    }
+}

common/src/serialize.rs

@@ -0,0 +1,282 @@
+use std::io::{Read, Write};
+use std::{fmt, io};
+
+use byteorder::{ReadBytesExt, WriteBytesExt};
+
+use crate::{Endianness, VInt};
+
+/// Trait for a simple binary serialization.
+pub trait BinarySerializable: fmt::Debug + Sized {
+    /// Serialize
+    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()>;
+    /// Deserialize
+    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 preferable, 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 {
+    const SIZE_IN_BYTES: usize;
+}
+
+impl BinarySerializable for () {
+    fn serialize<W: Write>(&self, _: &mut W) -> io::Result<()> {
+        Ok(())
+    }
+    fn deserialize<R: Read>(_: &mut R) -> io::Result<Self> {
+        Ok(())
+    }
+}
+
+impl FixedSize for () {
+    const SIZE_IN_BYTES: usize = 0;
+}
+
+impl<T: BinarySerializable> BinarySerializable for Vec<T> {
+    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
+        VInt(self.len() as u64).serialize(writer)?;
+        for it in self {
+            it.serialize(writer)?;
+        }
+        Ok(())
+    }
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<Vec<T>> {
+        let num_items = VInt::deserialize(reader)?.val();
+        let mut items: Vec<T> = Vec::with_capacity(num_items as usize);
+        for _ in 0..num_items {
+            let item = T::deserialize(reader)?;
+            items.push(item);
+        }
+        Ok(items)
+    }
+}
+
+impl<Left: BinarySerializable, Right: BinarySerializable> BinarySerializable for (Left, Right) {
+    fn serialize<W: Write>(&self, write: &mut W) -> io::Result<()> {
+        self.0.serialize(write)?;
+        self.1.serialize(write)
+    }
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
+        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<()> {
+        writer.write_u32::<Endianness>(*self)
+    }
+
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<u32> {
+        reader.read_u32::<Endianness>()
+    }
+}
+
+impl FixedSize for u32 {
+    const SIZE_IN_BYTES: usize = 4;
+}
+
+impl BinarySerializable for u64 {
+    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
+        writer.write_u64::<Endianness>(*self)
+    }
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
+        reader.read_u64::<Endianness>()
+    }
+}
+
+impl FixedSize for u64 {
+    const SIZE_IN_BYTES: usize = 8;
+}
+
+impl BinarySerializable for f32 {
+    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
+        writer.write_f32::<Endianness>(*self)
+    }
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
+        reader.read_f32::<Endianness>()
+    }
+}
+
+impl FixedSize for f32 {
+    const SIZE_IN_BYTES: usize = 4;
+}
+
+impl BinarySerializable for i64 {
+    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
+        writer.write_i64::<Endianness>(*self)
+    }
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
+        reader.read_i64::<Endianness>()
+    }
+}
+
+impl FixedSize for i64 {
+    const SIZE_IN_BYTES: usize = 8;
+}
+
+impl BinarySerializable for f64 {
+    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
+        writer.write_f64::<Endianness>(*self)
+    }
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
+        reader.read_f64::<Endianness>()
+    }
+}
+
+impl FixedSize for f64 {
+    const SIZE_IN_BYTES: usize = 8;
+}
+
+impl BinarySerializable for u8 {
+    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
+        writer.write_u8(*self)
+    }
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<u8> {
+        reader.read_u8()
+    }
+}
+
+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();
+        VInt(data.len() as u64).serialize(writer)?;
+        writer.write_all(data)
+    }
+
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<String> {
+        let string_length = VInt::deserialize(reader)?.val() as usize;
+        let mut result = String::with_capacity(string_length);
+        reader
+            .take(string_length as u64)
+            .read_to_string(&mut result)?;
+        Ok(result)
+    }
+}
+
+#[cfg(test)]
+pub mod test {
+
+    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();
+        assert_eq!(buffer.len(), O::SIZE_IN_BYTES);
+    }
+
+    fn serialize_test<T: BinarySerializable + Eq>(v: T) -> usize {
+        let mut buffer: Vec<u8> = Vec::new();
+        v.serialize(&mut buffer).unwrap();
+        let num_bytes = buffer.len();
+        let mut cursor = &buffer[..];
+        let deser = T::deserialize(&mut cursor).unwrap();
+        assert_eq!(deser, v);
+        num_bytes
+    }
+
+    #[test]
+    fn test_serialize_u8() {
+        fixed_size_test::<u8>();
+    }
+
+    #[test]
+    fn test_serialize_u32() {
+        fixed_size_test::<u32>();
+        assert_eq!(4, serialize_test(3u32));
+        assert_eq!(4, serialize_test(5u32));
+        assert_eq!(4, serialize_test(u32::MAX));
+    }
+
+    #[test]
+    fn test_serialize_i64() {
+        fixed_size_test::<i64>();
+    }
+
+    #[test]
+    fn test_serialize_f64() {
+        fixed_size_test::<f64>();
+    }
+
+    #[test]
+    fn test_serialize_u64() {
+        fixed_size_test::<u64>();
+    }
+
+    #[test]
+    fn test_serialize_bool() {
+        fixed_size_test::<bool>();
+    }
+
+    #[test]
+    fn test_serialize_string() {
+        assert_eq!(serialize_test(String::from("")), 1);
+        assert_eq!(serialize_test(String::from("ぽよぽよ")), 1 + 3 * 4);
+        assert_eq!(serialize_test(String::from("富士さん見える。")), 1 + 3 * 8);
+    }
+
+    #[test]
+    fn test_serialize_vec() {
+        assert_eq!(serialize_test(Vec::<u8>::new()), 1);
+        assert_eq!(serialize_test(vec![1u32, 3u32]), 1 + 4 * 2);
+    }
+
+    #[test]
+    fn test_serialize_vint() {
+        for i in 0..10_000 {
+            serialize_test(VInt(i as u64));
+        }
+        assert_eq!(serialize_test(VInt(7u64)), 1);
+        assert_eq!(serialize_test(VInt(127u64)), 1);
+        assert_eq!(serialize_test(VInt(128u64)), 2);
+        assert_eq!(serialize_test(VInt(129u64)), 2);
+        assert_eq!(serialize_test(VInt(1234u64)), 2);
+        assert_eq!(serialize_test(VInt(16_383u64)), 2);
+        assert_eq!(serialize_test(VInt(16_384u64)), 3);
+        assert_eq!(serialize_test(VInt(u64::MAX)), 10);
+    }
+}

common/src/vint.rs

@@ -0,0 +1,233 @@
+use std::io;
+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)]
+pub struct VInt(pub u64);
+
+const STOP_BIT: u8 = 128;
+
+pub fn serialize_vint_u32(val: u32, buf: &mut [u8; 8]) -> &[u8] {
+    const START_2: u64 = 1 << 7;
+    const START_3: u64 = 1 << 14;
+    const START_4: u64 = 1 << 21;
+    const START_5: u64 = 1 << 28;
+
+    const STOP_1: u64 = START_2 - 1;
+    const STOP_2: u64 = START_3 - 1;
+    const STOP_3: u64 = START_4 - 1;
+    const STOP_4: u64 = START_5 - 1;
+
+    const MASK_1: u64 = 127;
+    const MASK_2: u64 = MASK_1 << 7;
+    const MASK_3: u64 = MASK_2 << 7;
+    const MASK_4: u64 = MASK_3 << 7;
+    const MASK_5: u64 = MASK_4 << 7;
+
+    let val = u64::from(val);
+    const STOP_BIT: u64 = 128u64;
+    let (res, num_bytes) = match val {
+        0..=STOP_1 => (val | STOP_BIT, 1),
+        START_2..=STOP_2 => (
+            (val & MASK_1) | ((val & MASK_2) << 1) | (STOP_BIT << (8)),
+            2,
+        ),
+        START_3..=STOP_3 => (
+            (val & MASK_1) | ((val & MASK_2) << 1) | ((val & MASK_3) << 2) | (STOP_BIT << (8 * 2)),
+            3,
+        ),
+        START_4..=STOP_4 => (
+            (val & MASK_1)
+                | ((val & MASK_2) << 1)
+                | ((val & MASK_3) << 2)
+                | ((val & MASK_4) << 3)
+                | (STOP_BIT << (8 * 3)),
+            4,
+        ),
+        _ => (
+            (val & MASK_1)
+                | ((val & MASK_2) << 1)
+                | ((val & MASK_3) << 2)
+                | ((val & MASK_4) << 3)
+                | ((val & MASK_5) << 4)
+                | (STOP_BIT << (8 * 4)),
+            5,
+        ),
+    };
+    LittleEndian::write_u64(&mut buf[..], res);
+    &buf[0..num_bytes]
+}
+
+/// Returns the number of bytes covered by a
+/// serialized vint `u32`.
+///
+/// Expects a buffer data that starts
+/// by the serialized `vint`, scans at most 5 bytes ahead until
+/// it finds the vint final byte.
+///
+/// # May Panic
+/// If the payload does not start by a valid `vint`
+fn vint_len(data: &[u8]) -> usize {
+    for (i, &val) in data.iter().enumerate().take(5) {
+        if val >= STOP_BIT {
+            return i + 1;
+        }
+    }
+    panic!("Corrupted data. Invalid VInt 32");
+}
+
+/// Reads a vint `u32` from a buffer, and
+/// consumes its payload data.
+///
+/// # Panics
+///
+/// If the buffer does not start by a valid
+/// vint payload
+pub fn read_u32_vint(data: &mut &[u8]) -> u32 {
+    let (result, vlen) = read_u32_vint_no_advance(*data);
+    *data = &data[vlen..];
+    result
+}
+
+pub fn read_u32_vint_no_advance(data: &[u8]) -> (u32, usize) {
+    let vlen = vint_len(data);
+    let mut result = 0u32;
+    let mut shift = 0u64;
+    for &b in &data[..vlen] {
+        result |= u32::from(b & 127u8) << shift;
+        shift += 7;
+    }
+    (result, vlen)
+}
+/// Write a `u32` as a vint payload.
+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)
+}
+
+impl VInt {
+    pub fn val(&self) -> u64 {
+        self.0
+    }
+
+    pub fn deserialize_u64<R: Read>(reader: &mut R) -> io::Result<u64> {
+        VInt::deserialize(reader).map(|vint| vint.0)
+    }
+
+    pub fn serialize_into_vec(&self, output: &mut Vec<u8>) {
+        let mut buffer = [0u8; 10];
+        let num_bytes = self.serialize_into(&mut buffer);
+        output.extend(&buffer[0..num_bytes]);
+    }
+
+    pub fn serialize_into(&self, buffer: &mut [u8; 10]) -> usize {
+        let mut remaining = self.0;
+        for (i, b) in buffer.iter_mut().enumerate() {
+            let next_byte: u8 = (remaining % 128u64) as u8;
+            remaining /= 128u64;
+            if remaining == 0u64 {
+                *b = next_byte | STOP_BIT;
+                return i + 1;
+            } else {
+                *b = next_byte;
+            }
+        }
+        unreachable!();
+    }
+}
+
+impl BinarySerializable for VInt {
+    fn serialize<W: Write>(&self, writer: &mut W) -> io::Result<()> {
+        let mut buffer = [0u8; 10];
+        let num_bytes = self.serialize_into(&mut buffer);
+        writer.write_all(&buffer[0..num_bytes])
+    }
+
+    fn deserialize<R: Read>(reader: &mut R) -> io::Result<Self> {
+        let mut bytes = reader.bytes();
+        let mut result = 0u64;
+        let mut shift = 0u64;
+        loop {
+            match bytes.next() {
+                Some(Ok(b)) => {
+                    result |= u64::from(b % 128u8) << shift;
+                    if b >= STOP_BIT {
+                        return Ok(VInt(result));
+                    }
+                    shift += 7;
+                }
+                _ => {
+                    return Err(io::Error::new(
+                        io::ErrorKind::InvalidData,
+                        "Reach end of buffer while reading VInt",
+                    ));
+                }
+            }
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+
+    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 el in &v[num_bytes..10] {
+            assert_eq!(el, &14u8);
+        }
+        assert!(num_bytes > 0);
+        if num_bytes < 10 {
+            assert!(1u64 << (7 * num_bytes) > val);
+        }
+        if num_bytes > 1 {
+            assert!(1u64 << (7 * (num_bytes - 1)) <= val);
+        }
+        let serdeser_val = VInt::deserialize(&mut &v[..]).unwrap();
+        assert_eq!(val, serdeser_val.0);
+    }
+
+    #[test]
+    fn test_vint() {
+        aux_test_vint(0);
+        aux_test_vint(1);
+        aux_test_vint(5);
+        aux_test_vint(u64::MAX);
+        for i in 1..9 {
+            let power_of_128 = 1u64 << (7 * i);
+            aux_test_vint(power_of_128 - 1u64);
+            aux_test_vint(power_of_128);
+            aux_test_vint(power_of_128 + 1u64);
+        }
+        aux_test_vint(10);
+    }
+
+    fn aux_test_serialize_vint_u32(val: u32) {
+        let mut buffer = [0u8; 10];
+        let mut buffer2 = [0u8; 8];
+        let len_vint = VInt(val as u64).serialize_into(&mut buffer);
+        let res2 = serialize_vint_u32(val, &mut buffer2);
+        assert_eq!(&buffer[..len_vint], res2, "array wrong for {}", val);
+    }
+
+    #[test]
+    fn test_vint_u32() {
+        aux_test_serialize_vint_u32(0);
+        aux_test_serialize_vint_u32(1);
+        aux_test_serialize_vint_u32(5);
+        for i in 1..3 {
+            let power_of_128 = 1u32 << (7 * i);
+            aux_test_serialize_vint_u32(power_of_128 - 1u32);
+            aux_test_serialize_vint_u32(power_of_128);
+            aux_test_serialize_vint_u32(power_of_128 + 1u32);
+        }
+        aux_test_serialize_vint_u32(u32::MAX);
+    }
+}

common/src/writer.rs

@@ -0,0 +1,114 @@
+use std::io::{self, BufWriter, Write};
+
+pub struct CountingWriter<W> {
+    underlying: W,
+    written_bytes: u64,
+}
+
+impl<W: Write> CountingWriter<W> {
+    pub fn wrap(underlying: W) -> CountingWriter<W> {
+        CountingWriter {
+            underlying,
+            written_bytes: 0,
+        }
+    }
+
+    #[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 + Send {
+    /// 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 std::io::Write;
+
+    use super::CountingWriter;
+
+    #[test]
+    fn test_counting_writer() {
+        let buffer: Vec<u8> = vec![];
+        let mut counting_writer = CountingWriter::wrap(buffer);
+        let bytes = (0u8..10u8).collect::<Vec<u8>>();
+        counting_writer.write_all(&bytes).unwrap();
+        let len = counting_writer.written_bytes();
+        let buffer_restituted: Vec<u8> = counting_writer.finish();
+        assert_eq!(len, 10u64);
+        assert_eq!(buffer_restituted.len(), 10);
+    }
+}

cpp/simdcomp/.gitignore

@@ -1,9 +0,0 @@
-Makefile.in
-lib*
-unit*
-*.o
-src/*.lo
-src/*.o
-src/.deps
-src/.dirstamp
-src/.libs

cpp/simdcomp/.travis.yml

@@ -1,11 +0,0 @@
-language: c
-sudo: false
-compiler:
-  - gcc
-  - clang
-
-branches:
-  only:
-    - master
-
-script: make && ./unit

cpp/simdcomp/CHANGELOG

@@ -1,9 +0,0 @@
-Upcoming
-  - added missing include
-  - improved portability (MSVC)
-  - implemented C89 compatibility
-Version 0.0.3 (19 May 2014)
-  - improved documentation
-Version 0.0.2 (6 February 2014)
-  - added go demo
-Version 0.0.1  (5 February 2014)

cpp/simdcomp/LICENSE

@@ -1,27 +0,0 @@
-Copyright (c) 2014--, The authors
-All rights reserved.
-
-Redistribution and use in source and binary forms, with or without modification,
-are permitted provided that the following conditions are met:
-
-* Redistributions of source code must retain the above copyright notice, this
-  list of conditions and the following disclaimer.
-
-* Redistributions in binary form must reproduce the above copyright notice, this
-  list of conditions and the following disclaimer in the documentation and/or
-  other materials provided with the distribution.
-
-* Neither the name of the {organization} nor the names of its
-  contributors may be used to endorse or promote products derived from
-  this software without specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
-ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
-WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
-DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR
-ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
-(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
-LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
-ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

cpp/simdcomp/README.md

@@ -1,137 +0,0 @@
-The SIMDComp library
-====================
-[![Build Status](https://travis-ci.org/lemire/simdcomp.png)](https://travis-ci.org/lemire/simdcomp)
-
-A simple C library for compressing lists of integers using binary packing and SIMD instructions.
-The assumption is either that you have a list of 32-bit integers where most of them are small, or a list of 32-bit integers where differences between successive integers are small. No software is able to reliably compress an array of 32-bit random numbers.
-
-This library can decode at least 4 billions of compressed integers per second on most
-desktop or laptop processors. That is, it can decompress data at a rate of 15 GB/s.
-This is significantly faster than generic codecs like gzip, LZO, Snappy or LZ4.
-
-On a Skylake Intel processor, it can decode integers at a rate 0.3 cycles per integer,
-which can easily translate into more than 8 decoded billions integers per second.
-
-Contributors: Daniel Lemire, Nathan Kurz, Christoph Rupp, Anatol Belski, Nick White and others
-
-What is it for?
--------------
-
-This is a low-level library for fast integer compression. By design it does not define a compressed
-format. It is up to the (sophisticated) user to create a compressed format.
-
-Requirements
--------------
-
-- Your processor should support SSE4.1 (It is supported by most Intel and AMD processors released since 2008.)
-- It is possible to build the core part of the code if your processor support SSE2 (Pentium4 or better)
-- C99 compliant compiler (GCC is assumed)
-- A Linux-like distribution is assumed by the makefile
-
-For a plain C version that does not use SIMD instructions, see https://github.com/lemire/LittleIntPacker
-
-Usage
--------
-
-Compression works over blocks of 128 integers.
-
-For a complete working example, see example.c (you can build it and
-run it with "make example; ./example").
-
-
-
-1) Lists of integers in random order.
-
-```C            
-const uint32_t b = maxbits(datain);// computes bit width
-simdpackwithoutmask(datain, buffer, b);//compressed to buffer, compressing 128 32-bit integers down to b*32 bytes
-simdunpack(buffer, backbuffer, b);//uncompressed to backbuffer
-```
-
-While 128 32-bit integers are read, only b 128-bit words are written. Thus, the compression ratio is 32/b.
-
-2) Sorted lists of integers.
-
-We used differential coding: we store the difference between successive integers. For this purpose, we need an initial value (called offset).
-
-```C            
-uint32_t offset = 0;
-uint32_t b1 = simdmaxbitsd1(offset,datain); // bit width
-simdpackwithoutmaskd1(offset, datain, buffer, b1);//compressing 128 32-bit integers down to b1*32 bytes
-simdunpackd1(offset, buffer, backbuffer, b1);//uncompressed
-```
-
-General example for arrays of arbitrary length:
-```C
-int compress_decompress_demo() {
-  size_t k, N = 9999;
-  __m128i * endofbuf;
-  uint32_t * datain = malloc(N * sizeof(uint32_t));
-  uint8_t * buffer;
-  uint32_t * backbuffer = malloc(N * sizeof(uint32_t));
-  uint32_t b;
-
-  for (k = 0; k < N; ++k){        /* start with k=0, not k=1! */
-    datain[k] = k;
-  }
-
-  b = maxbits_length(datain, N);
-  buffer = malloc(simdpack_compressedbytes(N,b)); // allocate just enough memory
-  endofbuf = simdpack_length(datain, N, (__m128i *)buffer, b);
-  /* compressed data is stored between buffer and endofbuf using (endofbuf-buffer)*sizeof(__m128i) bytes */
-  /* would be safe to do : buffer = realloc(buffer,(endofbuf-(__m128i *)buffer)*sizeof(__m128i)); */
-  simdunpack_length((const __m128i *)buffer, N, backbuffer, b);
-
-  for (k = 0; k < N; ++k){
-    if(datain[k] != backbuffer[k]) {
-      printf("bug\n");
-      return -1;
-    }
-  }
-  return 0;
-}
-```
-
-
-3) Frame-of-Reference 
-
-We also have frame-of-reference (FOR) functions (see simdfor.h header). They work like the bit packing
-routines, but do not use differential coding so they allow faster search in some cases, at the expense
-of compression.
-
-Setup
----------
-
-
-make
-make test
-
-and if you are daring:
-
-make install
-
-Go
---------
-
-If you are a go user, there is a "go" folder where you will find a simple demo.
-
-Other libraries
-----------------
-
-* Fast decoder for VByte-compressed integers https://github.com/lemire/MaskedVByte
-* Fast integer compression in C using StreamVByte https://github.com/lemire/streamvbyte
-* FastPFOR is a C++ research library well suited to compress unsorted arrays: https://github.com/lemire/FastPFor
-* SIMDCompressionAndIntersection is a C++ research library well suited for sorted arrays (differential coding)
-and computing intersections: https://github.com/lemire/SIMDCompressionAndIntersection
-* TurboPFor is a C library that offers lots of interesting optimizations. Well worth checking! (GPL license) https://github.com/powturbo/TurboPFor
-* Oroch is a C++ library that offers a usable API (MIT license) https://github.com/ademakov/Oroch
-
-
-References
-------------
-
-* Daniel Lemire, Leonid Boytsov, Nathan Kurz, SIMD Compression and the Intersection of Sorted Integers, Software Practice & Experience 46 (6) 2016. http://arxiv.org/abs/1401.6399
-* Daniel Lemire and Leonid Boytsov, Decoding billions of integers per second through vectorization, Software Practice & Experience 45 (1), 2015.  http://arxiv.org/abs/1209.2137 http://onlinelibrary.wiley.com/doi/10.1002/spe.2203/abstract
-* Jeff Plaisance, Nathan Kurz, Daniel Lemire, Vectorized VByte Decoding, International Symposium on Web Algorithms 2015, 2015. http://arxiv.org/abs/1503.07387
-* Wayne Xin Zhao, Xudong Zhang, Daniel Lemire, Dongdong Shan, Jian-Yun Nie, Hongfei Yan, Ji-Rong Wen, A General SIMD-based Approach to Accelerating Compression Algorithms, ACM Transactions on Information Systems 33 (3), 2015. http://arxiv.org/abs/1502.01916
-* T. D. Wu, Bitpacking techniques for indexing genomes: I. Hash tables, Algorithms for Molecular Biology 11 (5), 2016. http://almob.biomedcentral.com/articles/10.1186/s13015-016-0069-5

cpp/simdcomp/benchmarks/benchmark.c

@@ -1,235 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-#include <assert.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <time.h>
-
-#include "simdcomp.h"
-
-#ifdef _MSC_VER
-# include <windows.h>
-
-__int64 freq;
-
-typedef __int64 time_snap_t;
-
-static time_snap_t time_snap(void)
-{
-	__int64 now;
-
-	QueryPerformanceCounter((LARGE_INTEGER *)&now);
-
-	return (__int64)((now*1000000)/freq);
-}
-# define TIME_SNAP_FMT "%I64d"
-#else
-# define time_snap clock
-# define TIME_SNAP_FMT "%lu"
-typedef clock_t time_snap_t;
-#endif
-
-
-void benchmarkSelect() {
-    uint32_t buffer[128];
-    uint32_t backbuffer[128];
-    uint32_t initial = 33;
-    uint32_t b;
-    time_snap_t S1, S2, S3;
-    int i;
-    printf("benchmarking select \n");
-
-    /* this test creates delta encoded buffers with different bits, then
-     * performs lower bound searches for each key */
-    for (b = 0; b <= 32; b++) {
-        uint32_t prev = initial;
-        uint32_t out[128];
-        /* initialize the buffer */
-        for (i = 0; i < 128; i++) {
-            buffer[i] =  ((uint32_t)(1655765 * i )) ;
-            if(b < 32) buffer[i] %= (1<<b);
-        }
-        for (i = 0; i < 128; i++) {
-            buffer[i] = buffer[i] + prev;
-            prev = buffer[i];
-        }
-
-        for (i = 1; i < 128; i++) {
-            if(buffer[i] < buffer[i-1] )
-                buffer[i] = buffer[i-1];
-        }
-        assert(simdmaxbitsd1(initial, buffer)<=b);
-
-        for (i = 0; i < 128; i++) {
-            out[i] = 0; /* memset would do too */
-        }
-
-        /* delta-encode to 'i' bits */
-        simdpackwithoutmaskd1(initial, buffer, (__m128i *)out, b);
-
-        S1 = time_snap();
-        for (i = 0; i < 128 * 10; i++) {
-            uint32_t valretrieved = simdselectd1(initial, (__m128i *)out, b, (uint32_t)i % 128);
-            assert(valretrieved == buffer[i%128]);
-        }
-        S2 = time_snap();
-        for (i = 0; i < 128 * 10; i++) {
-            simdunpackd1(initial,  (__m128i *)out, backbuffer, b);
-            assert(backbuffer[i % 128] == buffer[i % 128]);
-        }
-        S3 = time_snap();
-        printf("bit width = %d, fast select function time = " TIME_SNAP_FMT ", naive time = " TIME_SNAP_FMT "  \n", b, (S2-S1), (S3-S2));
-    }
-}
-
-int uint32_cmp(const void *a, const void *b)
-{
-    const uint32_t *ia = (const uint32_t *)a;
-    const uint32_t *ib = (const uint32_t *)b;
-    if(*ia < *ib)
-        return -1;
-    else if (*ia > *ib)
-        return 1;
-    return 0;
-}
-
-/* adapted from wikipedia */
-int binary_search(uint32_t * A, uint32_t key, int imin, int imax)
-{
-    int imid;
-    imax --;
-    while(imin + 1 < imax) {
-        imid = imin + ((imax - imin) / 2);
-
-        if (A[imid] > key) {
-            imax = imid;
-        } else if (A[imid] < key) {
-            imin = imid;
-        } else {
-            return imid;
-        }
-    }
-    return imax;
-}
-
-
-/* adapted from wikipedia */
-int lower_bound(uint32_t * A, uint32_t key, int imin, int imax)
-{
-    int imid;
-    imax --;
-    while(imin + 1 < imax) {
-        imid = imin + ((imax - imin) / 2);
-
-        if (A[imid] >= key) {
-            imax = imid;
-        } else if (A[imid] < key) {
-            imin = imid;
-        }
-    }
-    if(A[imin] >= key) return imin;
-    return imax;
-}
-
-void benchmarkSearch() {
-    uint32_t buffer[128];
-    uint32_t backbuffer[128];
-    uint32_t out[128];
-    uint32_t result, initial = 0;
-    uint32_t b, i;
-    time_snap_t S1, S2, S3, S4;
-
-    printf("benchmarking search \n");
-
-    /* this test creates delta encoded buffers with different bits, then
-     * performs lower bound searches for each key */
-    for (b = 0; b <= 32; b++) {
-        uint32_t prev = initial;
-        /* initialize the buffer */
-        for (i = 0; i < 128; i++) {
-            buffer[i] =  ((uint32_t)rand()) ;
-            if(b < 32) buffer[i] %= (1<<b);
-        }
-
-        qsort(buffer,128, sizeof(uint32_t), uint32_cmp);
-
-        for (i = 0; i < 128; i++) {
-            buffer[i] = buffer[i] + prev;
-            prev = buffer[i];
-        }
-        for (i = 1; i < 128; i++) {
-            if(buffer[i] < buffer[i-1] )
-                buffer[i] = buffer[i-1];
-        }
-        assert(simdmaxbitsd1(initial, buffer)<=b);
-        for (i = 0; i < 128; i++) {
-            out[i] = 0; /* memset would do too */
-        }
-
-        /* delta-encode to 'i' bits */
-        simdpackwithoutmaskd1(initial, buffer, (__m128i *)out, b);
-        simdunpackd1(initial,  (__m128i *)out, backbuffer, b);
-
-        for (i = 0; i < 128; i++) {
-            assert(buffer[i] == backbuffer[i]);
-         }
-        S1 = time_snap();
-        for (i = 0; i < 128 * 10; i++) {
-
-            int pos;
-            uint32_t pseudorandomkey  =  buffer[i%128];
-            __m128i vecinitial = _mm_set1_epi32(initial);
-            pos = simdsearchd1(&vecinitial, (__m128i *)out, b,
-                               pseudorandomkey, &result);
-            if((result < pseudorandomkey) || (buffer[pos] != result)) {
-                printf("bug A.\n");
-            } else if (pos > 0) {
-                if(buffer[pos-1] >= pseudorandomkey)
-                    printf("bug B.\n");
-            }
-        }
-        S2 = time_snap();
-        for (i = 0; i < 128 * 10; i++) {
-            int pos;
-            uint32_t pseudorandomkey  =  buffer[i%128];
-            simdunpackd1(initial,  (__m128i *)out, backbuffer, b);
-            pos =  lower_bound(backbuffer, pseudorandomkey, 0, 128);
-            result = backbuffer[pos];
-
-            if((result < pseudorandomkey) || (buffer[pos] != result)) {
-                printf("bug C.\n");
-            } else if (pos > 0) {
-                if(buffer[pos-1] >= pseudorandomkey)
-                    printf("bug D.\n");
-            }
-        }
-        S3 = time_snap();
-        for (i = 0; i < 128 * 10; i++) {
-
-            int pos;
-            uint32_t pseudorandomkey  =  buffer[i%128];
-            pos = simdsearchwithlengthd1(initial, (__m128i *)out, b, 128,
-                               pseudorandomkey, &result);
-            if((result < pseudorandomkey) || (buffer[pos] != result)) {
-                printf("bug A.\n");
-            } else if (pos > 0) {
-                if(buffer[pos-1] >= pseudorandomkey)
-                    printf("bug B.\n");
-            }
-        }
-        S4 = time_snap();
-
-        printf("bit width = %d, fast search function time = " TIME_SNAP_FMT ", naive time = " TIME_SNAP_FMT " , fast with length time = " TIME_SNAP_FMT "  \n", b, (S2-S1), (S3-S2), (S4-S3) );
-    }
-}
-
-
-int main() {
-#ifdef _MSC_VER
-    QueryPerformanceFrequency((LARGE_INTEGER *)&freq);
-#endif
-    benchmarkSearch();
-    benchmarkSelect();
-    return 0;
-}

cpp/simdcomp/benchmarks/bitpackingbenchmark.c

@@ -1,205 +0,0 @@
-#include <stdio.h>
-
-#include "simdcomp.h"
-
-
-#define RDTSC_START(cycles)                                                   \
-    do {                                                                      \
-        register unsigned cyc_high, cyc_low;                                  \
-        __asm volatile(                                                       \
-            "cpuid\n\t"                                                       \
-            "rdtsc\n\t"                                                       \
-            "mov %%edx, %0\n\t"                                               \
-            "mov %%eax, %1\n\t"                                               \
-            : "=r"(cyc_high), "=r"(cyc_low)::"%rax", "%rbx", "%rcx", "%rdx"); \
-        (cycles) = ((uint64_t)cyc_high << 32) | cyc_low;                      \
-    } while (0)
-
-#define RDTSC_FINAL(cycles)                                                   \
-    do {                                                                      \
-        register unsigned cyc_high, cyc_low;                                  \
-        __asm volatile(                                                       \
-            "rdtscp\n\t"                                                      \
-            "mov %%edx, %0\n\t"                                               \
-            "mov %%eax, %1\n\t"                                               \
-            "cpuid\n\t"                                                       \
-            : "=r"(cyc_high), "=r"(cyc_low)::"%rax", "%rbx", "%rcx", "%rdx"); \
-        (cycles) = ((uint64_t)cyc_high << 32) | cyc_low;                      \
-    } while (0)
-
-
-
-
-uint32_t * get_random_array_from_bit_width(uint32_t length, uint32_t bit) {
-    uint32_t * answer = malloc(sizeof(uint32_t) * length);
-    uint32_t mask = (uint32_t) ((UINT64_C(1) << bit) - 1);
-    uint32_t i;
-    for(i = 0; i < length; ++i) {
-        answer[i] = rand() & mask;
-    }
-    return answer;
-}
-
-uint32_t * get_random_array_from_bit_width_d1(uint32_t length, uint32_t bit) {
-    uint32_t * answer = malloc(sizeof(uint32_t) * length);
-    uint32_t mask = (uint32_t) ((UINT64_C(1) << bit) - 1);
-    uint32_t i;
-    answer[0] = rand() & mask;
-    for(i = 1; i < length; ++i) {
-        answer[i] = answer[i-1] + (rand() & mask);
-    }
-    return answer;
-}
-
-
-void demo128() {
-    const uint32_t length = 128;
-    uint32_t bit;
-    printf("# --- %s\n", __func__);
-    printf("# compressing %d integers\n",length);
-    printf("# format: bit width, pack in cycles per int, unpack in cycles per int\n");
-    for(bit = 1; bit <= 32; ++bit) {
-        uint32_t i;
-
-        uint32_t * data = get_random_array_from_bit_width(length, bit);
-        __m128i * buffer = malloc(length * sizeof(uint32_t));
-        uint32_t * backdata = malloc(length * sizeof(uint32_t));
-        uint32_t repeat = 500;
-        uint64_t min_diff;
-        printf("%d\t",bit);
-        min_diff = (uint64_t)-1;
-        for (i = 0; i < repeat; i++) {
-            uint64_t cycles_start, cycles_final, cycles_diff;
-            __asm volatile("" ::: /* pretend to clobber */ "memory");
-            RDTSC_START(cycles_start);
-            simdpackwithoutmask(data,buffer, bit);
-            RDTSC_FINAL(cycles_final);
-            cycles_diff = (cycles_final - cycles_start);
-            if (cycles_diff < min_diff) min_diff = cycles_diff;
-        }
-        printf("%.2f\t",min_diff*1.0/length);
-        min_diff = (uint64_t)-1;
-        for (i = 0; i < repeat; i++) {
-            uint64_t cycles_start, cycles_final, cycles_diff;
-            __asm volatile("" ::: /* pretend to clobber */ "memory");
-            RDTSC_START(cycles_start);
-            simdunpack(buffer, backdata,bit);
-            RDTSC_FINAL(cycles_final);
-            cycles_diff = (cycles_final - cycles_start);
-            if (cycles_diff < min_diff) min_diff = cycles_diff;
-        }
-        printf("%.2f\t",min_diff*1.0/length);
-
-        free(data);
-        free(buffer);
-        free(backdata);
-        printf("\n");
-    }
-    printf("\n\n"); /* two blank lines are required by gnuplot */
-}
-
-void demo128_d1() {
-    const uint32_t length = 128;
-    uint32_t bit;
-    printf("# --- %s\n", __func__);
-    printf("# compressing %d integers\n",length);
-    printf("# format: bit width, pack in cycles per int, unpack in cycles per int\n");
-    for(bit = 1; bit <= 32; ++bit) {
-        uint32_t i;
-
-        uint32_t * data = get_random_array_from_bit_width_d1(length, bit);
-        __m128i * buffer = malloc(length * sizeof(uint32_t));
-        uint32_t * backdata = malloc(length * sizeof(uint32_t));
-        uint32_t repeat = 500;
-        uint64_t min_diff;
-        printf("%d\t",bit);
-        min_diff = (uint64_t)-1;
-        for (i = 0; i < repeat; i++) {
-            uint64_t cycles_start, cycles_final, cycles_diff;
-            __asm volatile("" ::: /* pretend to clobber */ "memory");
-            RDTSC_START(cycles_start);
-            simdpackwithoutmaskd1(0,data,buffer, bit);
-            RDTSC_FINAL(cycles_final);
-            cycles_diff = (cycles_final - cycles_start);
-            if (cycles_diff < min_diff) min_diff = cycles_diff;
-        }
-        printf("%.2f\t",min_diff*1.0/length);
-        min_diff = (uint64_t)-1;
-        for (i = 0; i < repeat; i++) {
-            uint64_t cycles_start, cycles_final, cycles_diff;
-            __asm volatile("" ::: /* pretend to clobber */ "memory");
-            RDTSC_START(cycles_start);
-            simdunpackd1(0,buffer, backdata,bit);
-            RDTSC_FINAL(cycles_final);
-            cycles_diff = (cycles_final - cycles_start);
-            if (cycles_diff < min_diff) min_diff = cycles_diff;
-        }
-        printf("%.2f\t",min_diff*1.0/length);
-
-        free(data);
-        free(buffer);
-        free(backdata);
-        printf("\n");
-    }
-    printf("\n\n"); /* two blank lines are required by gnuplot */
-}
-
-#ifdef __AVX2__
-void demo256() {
-    const uint32_t length = 256;
-    uint32_t bit;
-    printf("# --- %s\n", __func__);
-    printf("# compressing %d integers\n",length);
-    printf("# format: bit width, pack in cycles per int, unpack in cycles per int\n");
-    for(bit = 1; bit <= 32; ++bit) {
-        uint32_t i;
-
-        uint32_t * data = get_random_array_from_bit_width(length, bit);
-        __m256i * buffer = malloc(length * sizeof(uint32_t));
-        uint32_t * backdata = malloc(length * sizeof(uint32_t));
-        uint32_t repeat = 500;
-        uint64_t min_diff;
-        printf("%d\t",bit);
-        min_diff = (uint64_t)-1;
-        for (i = 0; i < repeat; i++) {
-            uint64_t cycles_start, cycles_final, cycles_diff;
-            __asm volatile("" ::: /* pretend to clobber */ "memory");
-            RDTSC_START(cycles_start);
-            avxpackwithoutmask(data,buffer, bit);
-            RDTSC_FINAL(cycles_final);
-            cycles_diff = (cycles_final - cycles_start);
-            if (cycles_diff < min_diff) min_diff = cycles_diff;
-        }
-        printf("%.2f\t",min_diff*1.0/length);
-        min_diff = (uint64_t)-1;
-        for (i = 0; i < repeat; i++) {
-            uint64_t cycles_start, cycles_final, cycles_diff;
-            __asm volatile("" ::: /* pretend to clobber */ "memory");
-            RDTSC_START(cycles_start);
-            avxunpack(buffer, backdata,bit);
-            RDTSC_FINAL(cycles_final);
-            cycles_diff = (cycles_final - cycles_start);
-            if (cycles_diff < min_diff) min_diff = cycles_diff;
-        }
-        printf("%.2f\t",min_diff*1.0/length);
-
-        free(data);
-        free(buffer);
-        free(backdata);
-        printf("\n");
-    }
-    printf("\n\n"); /* two blank lines are required by gnuplot */
-}
-#endif /* avx 2 */
-
-
-int main() {
-    demo128();
-    demo128_d1();
-#ifdef __AVX2__
-    demo256();
-#endif
-    return 0;
-
-
-}

cpp/simdcomp/example.c

@@ -1,195 +0,0 @@
-/* Type "make example" to build this example program. */
-#include <stdio.h>
-#include <time.h>
-#include <stdlib.h>
-#include "simdcomp.h"
-
-/**
-We provide several different code examples.
-**/
-
-
-/* very simple test to illustrate a simple application */
-int compress_decompress_demo() {
-    size_t k, N = 9999;
-    __m128i * endofbuf;
-    int howmanybytes;
-    float compratio;
-    uint32_t * datain = malloc(N * sizeof(uint32_t));
-    uint8_t * buffer;
-    uint32_t * backbuffer = malloc(N * sizeof(uint32_t));
-    uint32_t b;
-    printf("== simple test\n");
-
-    for (k = 0; k < N; ++k) {       /* start with k=0, not k=1! */
-        datain[k] = k;
-    }
-
-    b = maxbits_length(datain, N);
-    buffer = malloc(simdpack_compressedbytes(N,b));
-    endofbuf = simdpack_length(datain, N, (__m128i *)buffer, b);
-    howmanybytes = (endofbuf-(__m128i *)buffer)*sizeof(__m128i); /* number of compressed bytes */
-    compratio = N*sizeof(uint32_t) * 1.0 / howmanybytes;
-    /* endofbuf points to the end of the compressed data */
-    buffer = realloc(buffer,(endofbuf-(__m128i *)buffer)*sizeof(__m128i)); /* optional but safe. */
-    printf("Compressed %d integers down to %d bytes (comp. ratio = %f).\n",(int)N,howmanybytes,compratio);
-    /* in actual applications b must be stored and retrieved: caller is responsible for that. */
-    simdunpack_length((const __m128i *)buffer, N, backbuffer, b); /* will return a pointer to endofbuf */ 
-
-    for (k = 0; k < N; ++k) {
-        if(datain[k] != backbuffer[k]) {
-            printf("bug at %lu \n",(unsigned long)k);
-            return -1;
-        }
-    }
-    printf("Code works!\n");
-    free(datain);
-    free(buffer);
-    free(backbuffer);
-    return 0;
-}
-
-
-
-/* compresses data from datain to buffer, returns how many bytes written
-used below in simple_demo */
-size_t compress(uint32_t * datain, size_t length, uint8_t * buffer) {
-    uint32_t offset;
-    uint8_t * initout;
-    size_t k;
-    if(length/SIMDBlockSize*SIMDBlockSize != length) {
-        printf("Data length should be a multiple of %i \n",SIMDBlockSize);
-    }
-    offset = 0;
-    initout = buffer;
-    for(k = 0; k < length / SIMDBlockSize; ++k) {
-        uint32_t b = simdmaxbitsd1(offset,
-                                   datain + k * SIMDBlockSize);
-        *buffer++ = b;
-        simdpackwithoutmaskd1(offset, datain + k * SIMDBlockSize, (__m128i *) buffer,
-                              b);
-        offset = datain[k * SIMDBlockSize + SIMDBlockSize - 1];
-        buffer += b * sizeof(__m128i);
-    }
-    return buffer - initout;
-}
-
-/* Another illustration ... */
-void simple_demo() {
-    size_t REPEAT = 10, gap;
-    size_t N = 1000 * SIMDBlockSize;/* SIMDBlockSize is 128 */
-    uint32_t * datain = malloc(N * sizeof(uint32_t));
-    size_t compsize;
-    clock_t start, end;
-    uint8_t * buffer = malloc(N * sizeof(uint32_t) + N / SIMDBlockSize); /* output buffer */
-    uint32_t * backbuffer = malloc(SIMDBlockSize * sizeof(uint32_t));
-    printf("== simple demo\n");
-    for (gap = 1; gap <= 243; gap *= 3) {
-        size_t k, repeat;
-        uint32_t offset = 0;
-        uint32_t bogus = 0;
-        double numberofseconds;
-
-        printf("\n");
-        printf(" gap = %lu \n", (unsigned long) gap);
-        datain[0] = 0;
-        for (k = 1; k < N; ++k)
-            datain[k] = datain[k-1] + ( rand() % (gap + 1) );
-        compsize = compress(datain,N,buffer);
-        printf("compression ratio = %f \n",  (N * sizeof(uint32_t))/ (compsize * 1.0 ));
-        start = clock();
-        for(repeat = 0; repeat < REPEAT; ++repeat) {
-            uint8_t * decbuffer = buffer;
-            for (k = 0; k * SIMDBlockSize < N; ++k) {
-                uint8_t b = *decbuffer++;
-                simdunpackd1(offset, (__m128i *) decbuffer, backbuffer, b);
-                /* do something here with backbuffer */
-                bogus += backbuffer[3];
-                decbuffer += b * sizeof(__m128i);
-                offset = backbuffer[SIMDBlockSize - 1];
-            }
-        }
-        end = clock();
-        numberofseconds = (end-start)/(double)CLOCKS_PER_SEC;
-        printf("decoding speed in million of integers per second %f \n",N*REPEAT/(numberofseconds*1000.0*1000.0));
-        start = clock();
-        for(repeat = 0; repeat < REPEAT; ++repeat) {
-            uint8_t * decbuffer = buffer;
-            for (k = 0; k * SIMDBlockSize < N; ++k) {
-                memcpy(backbuffer,decbuffer+k*SIMDBlockSize,SIMDBlockSize*sizeof(uint32_t));
-                bogus += backbuffer[3] - backbuffer[100];
-            }
-        }
-        end = clock();
-        numberofseconds = (end-start)/(double)CLOCKS_PER_SEC;
-        printf("memcpy speed in million of integers per second %f \n",N*REPEAT/(numberofseconds*1000.0*1000.0));
-        printf("ignore me %i \n",bogus);
-        printf("All tests are in CPU cache. Avoid out-of-cache decoding in applications.\n");
-    }
-    free(buffer);
-    free(datain);
-    free(backbuffer);
-}
-
-/* Used below in more_sophisticated_demo ... */
-size_t varying_bit_width_compress(uint32_t * datain, size_t length, uint8_t * buffer) {
-    uint8_t * initout;
-    size_t k;
-    if(length/SIMDBlockSize*SIMDBlockSize != length) {
-        printf("Data length should be a multiple of %i \n",SIMDBlockSize);
-    }
-    initout = buffer;
-    for(k = 0; k < length / SIMDBlockSize; ++k) {
-        uint32_t b = maxbits(datain);
-        *buffer++ = b;
-        simdpackwithoutmask(datain, (__m128i *)buffer, b);
-        datain += SIMDBlockSize;
-        buffer += b * sizeof(__m128i);
-    }
-    return buffer - initout;
-}
-
-/* Here we compress the data in blocks of 128 integers with varying bit width */
-int varying_bit_width_demo() {
-    size_t nn = 128 * 2;
-    uint32_t * datainn = malloc(nn * sizeof(uint32_t));
-    uint8_t * buffern = malloc(nn * sizeof(uint32_t) + nn / SIMDBlockSize);
-    uint8_t * initbuffern = buffern;
-    uint32_t * backbuffern = malloc(nn * sizeof(uint32_t));
-    size_t k, compsize;
-    printf("== varying bit-width demo\n");
-
-    for(k=0; k<nn; ++k) {
-        datainn[k] = rand() % (k + 1);
-    }
-
-    compsize = varying_bit_width_compress(datainn,nn,buffern);
-    printf("encoded size: %u (original size: %u)\n", (unsigned)compsize,
-           (unsigned)(nn * sizeof(uint32_t)));
-
-    for (k = 0; k * SIMDBlockSize < nn; ++k) {
-        uint32_t b = *buffern;
-        buffern++;
-        simdunpack((const __m128i *)buffern, backbuffern + k * SIMDBlockSize, b);
-        buffern += b * sizeof(__m128i);
-    }
-
-    for (k = 0; k < nn; ++k) {
-        if(backbuffern[k] != datainn[k]) {
-            printf("bug\n");
-            return -1;
-        }
-    }
-    printf("Code works!\n");
-    free(datainn);
-    free(initbuffern);
-    free(backbuffern);
-    return 0;
-}
-
-int main() {
-    if(compress_decompress_demo() != 0) return -1;
-    if(varying_bit_width_demo() != 0) return -1;
-    simple_demo();
-    return 0;
-}

cpp/simdcomp/go/README.md

@@ -1,13 +0,0 @@
-Simple Go demo
-==============
-
-Setup
-======
-
-Start by installing the simdcomp library (make && make install).
-
-Then type:
-
-go run test.go
-
-

cpp/simdcomp/go/test.go

@@ -1,71 +0,0 @@
-/////////
-// This particular file is in the public domain.
-// Author: Daniel Lemire
-////////
-
-package main 
-
-/*
-#cgo LDFLAGS: -lsimdcomp
-#include <simdcomp.h>
-*/
-import "C"
-import "fmt"
-
-//////////
-// For this demo, we pack and unpack blocks of 128 integers
-/////////
-func main() {
-        // I am going to use C types. Alternative might be to use unsafe.Pointer calls, see http://bit.ly/1ndw3W3
-        // this is our original data
-        var data [128]C.uint32_t
-        for i := C.uint32_t(0); i < C.uint32_t(128); i++ {
-            data[i] = i
-        }
-
-
-
-
-
-        ////////////
-        // We first pack without differential coding
-        ///////////
-        // computing how many bits per int. is needed
-        b  := C.maxbits(&data[0])
-        ratio := 32.0/float64(b)
-        fmt.Println("Bit width  ", b)
-        fmt.Println(fmt.Sprintf("Compression ratio %f ", ratio))
-         // we are now going to create a buffer to receive the packed data (each __m128i uses 128 bits)
-        out := make([] C.__m128i,b)       
-        C.simdpackwithoutmask( &data[0],&out[0],b);
-        var recovereddata [128]C.uint32_t
-        C.simdunpack(&out[0],&recovereddata[0],b)
-        for i := 0; i < 128; i++ {
-            if data[i] != recovereddata[i]  {
-                  fmt.Println("Bug ")
-                  return
-            }
-        } 
-
-        ///////////
-        // Next, we use differential coding
-        //////////
-        offset := C.uint32_t(0) // if you pack data from K to K + 128, offset should be the value at K-1. When K = 0, choose a default
-        b1  := C.simdmaxbitsd1(offset,&data[0])
-        ratio1 := 32.0/float64(b1)
-        fmt.Println("Bit width  ", b1)
-        fmt.Println(fmt.Sprintf("Compression ratio %f ", ratio1))
-         // we are now going to create a buffer to receive the packed data (each __m128i uses 128 bits)
-        out = make([] C.__m128i,b1)       
-        C.simdpackwithoutmaskd1(offset, &data[0],&out[0],b1);
-        C.simdunpackd1(offset,&out[0],&recovereddata[0],b1)
-        for i := 0; i < 128; i++ {
-            if data[i] != recovereddata[i]  {
-                  fmt.Println("Bug ")
-                  return
-            }
-        } 
-
-        fmt.Println("test succesful.")
-      
-}

cpp/simdcomp/include/avxbitpacking.h

@@ -1,40 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-
-#ifndef INCLUDE_AVXBITPACKING_H_
-#define INCLUDE_AVXBITPACKING_H_
-
-
-#ifdef __AVX2__
-
-#include "portability.h"
-
-
-/* AVX2 is required */
-#include <immintrin.h>
-/* for memset */
-#include <string.h>
-
-#include "simdcomputil.h"
-
-enum{ AVXBlockSize = 256};
-
-/* max integer logarithm over a range of AVXBlockSize integers (256 integer) */
-uint32_t avxmaxbits(const uint32_t * begin);
-
-/* reads 256 values from "in", writes  "bit" 256-bit vectors to "out" */
-void avxpack(const uint32_t *  in,__m256i *  out, const uint32_t bit);
-
-/* reads 256 values from "in", writes  "bit" 256-bit vectors to "out" */
-void avxpackwithoutmask(const uint32_t *  in,__m256i *  out, const uint32_t bit);
-
-/* reads  "bit" 256-bit vectors from "in", writes  256 values to "out" */
-void avxunpack(const __m256i *  in,uint32_t *  out, const uint32_t bit);
-
-
-
-
-#endif /* __AVX2__ */
-
-#endif /* INCLUDE_AVXBITPACKING_H_ */

cpp/simdcomp/include/portability.h

@@ -1,81 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-#ifndef SIMDBITCOMPAT_H_
-#define SIMDBITCOMPAT_H_
-
-#include <iso646.h> /* mostly for Microsoft compilers */
-#include <string.h>
-
-#if SIMDCOMP_DEBUG
-# define SIMDCOMP_ALWAYS_INLINE inline
-# define SIMDCOMP_NEVER_INLINE
-# define SIMDCOMP_PURE
-#else
-# if defined(__GNUC__)
-#  if __GNUC__ >= 3
-#   define SIMDCOMP_ALWAYS_INLINE inline __attribute__((always_inline))
-#   define SIMDCOMP_NEVER_INLINE __attribute__((noinline))
-#   define SIMDCOMP_PURE __attribute__((pure))
-#  else
-#   define SIMDCOMP_ALWAYS_INLINE inline
-#   define SIMDCOMP_NEVER_INLINE
-#   define SIMDCOMP_PURE
-#  endif
-# elif defined(_MSC_VER)
-#  define SIMDCOMP_ALWAYS_INLINE __forceinline
-#  define SIMDCOMP_NEVER_INLINE
-#  define SIMDCOMP_PURE
-# else
-#  if __has_attribute(always_inline)
-#   define SIMDCOMP_ALWAYS_INLINE inline __attribute__((always_inline))
-#  else
-#   define SIMDCOMP_ALWAYS_INLINE inline
-#  endif
-#  if __has_attribute(noinline)
-#   define SIMDCOMP_NEVER_INLINE __attribute__((noinline))
-#  else
-#   define SIMDCOMP_NEVER_INLINE
-#  endif
-#  if __has_attribute(pure)
-#   define SIMDCOMP_PURE __attribute__((pure))
-#  else
-#   define SIMDCOMP_PURE
-#  endif
-# endif
-#endif
-
-#if defined(_MSC_VER) && _MSC_VER < 1600
-typedef unsigned int uint32_t;
-typedef unsigned char uint8_t;
-typedef signed char int8_t;
-#else
-#include <stdint.h> /* part of Visual Studio 2010 and better, others likely anyway */
-#endif
-
-#if defined(_MSC_VER)
-#define SIMDCOMP_ALIGNED(x) __declspec(align(x))
-#else
-#if defined(__GNUC__)
-#define SIMDCOMP_ALIGNED(x) __attribute__ ((aligned(x)))
-#endif
-#endif
-
-#if defined(_MSC_VER)
-# include <intrin.h>
-/* 64-bit needs extending */
-# define SIMDCOMP_CTZ(result, mask) do { \
-		unsigned long index; \
-		if (!_BitScanForward(&(index), (mask))) { \
-			(result) = 32U; \
-		} else { \
-			(result) = (uint32_t)(index); \
-		} \
-	} while (0)
-#else
-# define SIMDCOMP_CTZ(result, mask) \
-	result = __builtin_ctz(mask)
-#endif
-
-#endif /* SIMDBITCOMPAT_H_ */
-

cpp/simdcomp/include/simdbitpacking.h

@@ -1,72 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-#ifndef SIMDBITPACKING_H_
-#define SIMDBITPACKING_H_
-
-#include "portability.h"
-
-/* SSE2 is required */
-#include <emmintrin.h>
-/* for memset */
-#include <string.h>
-
-#include "simdcomputil.h"
-
-/***
-* Please see example.c for various examples on how to make good use
-* of these functions.
-*/
-
-
-
-/* reads 128 values from "in", writes  "bit" 128-bit vectors to "out".
- * The input values are masked so that only the least significant "bit" bits are used. */
-void simdpack(const uint32_t *  in,__m128i *  out, const uint32_t bit);
-
-/* reads 128 values from "in", writes  "bit" 128-bit vectors to "out".
- * The input values are assumed to be less than 1<<bit. */
-void simdpackwithoutmask(const uint32_t *  in,__m128i *  out, const uint32_t bit);
-
-/* reads  "bit" 128-bit vectors from "in", writes  128 values to "out" */
-void simdunpack(const __m128i *  in,uint32_t *  out, const uint32_t bit);
-
-
-
-/* how many compressed bytes are needed to compressed length integers using a bit width of bit with 
-the  simdpackFOR_length function. */
-int simdpack_compressedbytes(int length, const uint32_t bit);
-
-/* like simdpack, but supports an undetermined number of inputs.
- * This is useful if you need to unpack an array of integers that is not divisible by 128 integers.
- * Returns a pointer to the (advanced) compressed array. Compressed data is stored in the memory location between 
- the provided (out) pointer and the returned pointer. */
-__m128i * simdpack_length(const uint32_t *   in, size_t length, __m128i *    out, const uint32_t bit);
-
-/* like simdunpack, but supports an undetermined number of inputs.
- * This is useful if you need to unpack an array of integers that is not divisible by 128 integers.
- * Returns a pointer to the (advanced) compressed array. The read compressed data is between the provided 
- (in) pointer and the returned pointer. */
-const __m128i * simdunpack_length(const __m128i *   in, size_t length, uint32_t * out, const uint32_t bit);
-
-
-
-
-/* like simdpack, but supports an undetermined small number of inputs. This is useful if you need to pack less 
-than 128 integers.
- * Note that this function is much slower.
- * Returns a pointer to the (advanced) compressed array. Compressed data is stored in the memory location 
- between the provided (out) pointer and the returned pointer. */
-__m128i * simdpack_shortlength(const uint32_t *   in, int length, __m128i *    out, const uint32_t bit);
-
-/* like simdunpack, but supports an undetermined small number of inputs. This is useful if you need to unpack less
- than 128 integers.
- * Note that this function is much slower.
- * Returns a pointer to the (advanced) compressed array. The read compressed data is between the provided (in) 
- pointer and the returned pointer. */
-const __m128i * simdunpack_shortlength(const __m128i *   in, int length, uint32_t * out, const uint32_t bit);
-
-/* given a block of 128 packed values, this function sets the value at index "index" to "value" */
-void simdfastset(__m128i * in128, uint32_t b, uint32_t value, size_t index);
-
-#endif /* SIMDBITPACKING_H_ */

cpp/simdcomp/include/simdcomp.h

@@ -1,22 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-
-#ifndef SIMDCOMP_H_
-#define SIMDCOMP_H_
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-#include "simdbitpacking.h"
-#include "simdcomputil.h"
-#include "simdfor.h"
-#include "simdintegratedbitpacking.h"
-#include "avxbitpacking.h"
-
-#ifdef __cplusplus
-} // extern "C"
-#endif
-
-#endif 

cpp/simdcomp/include/simdcomputil.h

@@ -1,54 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-
-#ifndef SIMDCOMPUTIL_H_
-#define SIMDCOMPUTIL_H_
-
-#include "portability.h"
-
-/* SSE2 is required */
-#include <emmintrin.h>
-
-
-
-
-/* returns the integer logarithm of v (bit width) */
-uint32_t bits(const uint32_t v);
-
-/* max integer logarithm over a range of SIMDBlockSize integers (128 integer) */
-uint32_t maxbits(const uint32_t * begin);
-
-/* same as maxbits, but we specify the number of integers */
-uint32_t maxbits_length(const uint32_t * in,uint32_t length);
-
-enum{ SIMDBlockSize = 128};
-
-
-/* computes (quickly) the minimal value of 128 values */
-uint32_t simdmin(const uint32_t * in);
-
-/* computes (quickly) the minimal value of the specified number of values */
-uint32_t simdmin_length(const uint32_t * in, uint32_t length);
-
-#ifdef __SSE4_1__
-/* computes (quickly) the minimal and maximal value of the specified number of values */
-void simdmaxmin_length(const uint32_t * in, uint32_t length, uint32_t * getmin, uint32_t * getmax);
-
-/* computes (quickly) the minimal and maximal value of the 128 values */
-void simdmaxmin(const uint32_t * in, uint32_t * getmin, uint32_t * getmax);
-
-#endif
-
-/* like maxbit over 128 integers (SIMDBlockSize) with provided initial value 
-   and using differential coding */
-uint32_t simdmaxbitsd1(uint32_t initvalue, const uint32_t * in);
-
-/* like simdmaxbitsd1, but calculates maxbits over |length| integers 
-   with provided initial value. |length| can be any arbitrary value. */
-uint32_t simdmaxbitsd1_length(uint32_t initvalue, const uint32_t * in,
-                uint32_t length);
-
-
-
-#endif /* SIMDCOMPUTIL_H_ */

cpp/simdcomp/include/simdfor.h

@@ -1,72 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-#ifndef INCLUDE_SIMDFOR_H_
-#define INCLUDE_SIMDFOR_H_
-
-#include "portability.h"
-
-/* SSE2 is required */
-#include <emmintrin.h>
-
-#include "simdcomputil.h"
-#include "simdbitpacking.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/* reads 128 values from "in", writes  "bit" 128-bit vectors to "out" */
-void simdpackFOR(uint32_t initvalue, const uint32_t *  in,__m128i *  out, const uint32_t bit);
-
-
-/* reads "bit" 128-bit vectors from "in", writes  128 values to "out" */
-void simdunpackFOR(uint32_t initvalue, const __m128i *  in,uint32_t *  out, const uint32_t bit);
-
-
-/* how many compressed bytes are needed to compressed length integers using a bit width of bit with 
-the  simdpackFOR_length function. */
-int simdpackFOR_compressedbytes(int length, const uint32_t bit);
-
-/* like simdpackFOR, but supports an undetermined number of inputs. 
-This is useful if you need to pack less than 128 integers. Note that this function is much slower. 
- Compressed data is stored in the memory location between 
- the provided (out) pointer and the returned pointer. */
-__m128i * simdpackFOR_length(uint32_t initvalue, const uint32_t *   in, int length, __m128i *    out, const uint32_t bit);
-
-/* like simdunpackFOR, but supports an undetermined number of inputs. 
-This is useful if you need to unpack less than 128 integers. Note that this function is much slower. 
- The read compressed data is between the provided 
- (in) pointer and the returned pointer.  */
-const __m128i * simdunpackFOR_length(uint32_t initvalue, const __m128i *   in, int length, uint32_t * out, const uint32_t bit);
-
-
-/* returns the value stored at the specified "slot".
-* */
-uint32_t simdselectFOR(uint32_t initvalue, const __m128i *in, uint32_t bit,
-                int slot);
-
-/* given a block of 128 packed values, this function sets the value at index "index" to "value" */
-void simdfastsetFOR(uint32_t initvalue, __m128i * in, uint32_t bit, uint32_t value, size_t index);
-
-
-/* searches "bit" 128-bit vectors from "in" (= length<=128 encoded integers) for the first encoded uint32 value
- * which is >= |key|, and returns its position. It is assumed that the values
- * stored are in sorted order.
- * The encoded key is stored in "*presult".
- * The first length decoded integers, ignoring others. If no value is larger or equal to the key,
- * length is returned. Length should be no larger than 128.
- *
- * If no value is larger or equal to the key,
-* length is returned */
-int simdsearchwithlengthFOR(uint32_t initvalue, const __m128i *in, uint32_t bit,
-                int length, uint32_t key, uint32_t *presult);
-
-#ifdef __cplusplus
-} // extern "C"
-#endif
-
-
-
-
-#endif /* INCLUDE_SIMDFOR_H_ */

cpp/simdcomp/include/simdintegratedbitpacking.h

@@ -1,98 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-
-#ifndef SIMD_INTEGRATED_BITPACKING_H
-#define SIMD_INTEGRATED_BITPACKING_H
-
-#include "portability.h"
-
-/* SSE2 is required */
-#include <emmintrin.h>
-
-#include "simdcomputil.h"
-#include "simdbitpacking.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/* reads 128 values from "in", writes  "bit" 128-bit vectors to "out"
-   integer values should be in sorted order (for best results).
-   The differences are masked so that only the least significant "bit" bits are used. */
-void simdpackd1(uint32_t initvalue, const uint32_t *  in,__m128i *  out, const uint32_t bit);
-
-
-/* reads 128 values from "in", writes  "bit" 128-bit vectors to "out"
-   integer values should be in sorted order (for best results).
-   The difference values are assumed to be less than 1<<bit. */
-void simdpackwithoutmaskd1(uint32_t initvalue, const uint32_t *  in,__m128i *  out, const uint32_t bit);
-
-
-/* reads "bit" 128-bit vectors from "in", writes  128 values to "out" */
-void simdunpackd1(uint32_t initvalue, const __m128i *  in,uint32_t *  out, const uint32_t bit);
-
-
-/* searches "bit" 128-bit vectors from "in" (= 128 encoded integers) for the first encoded uint32 value
- * which is >= |key|, and returns its position. It is assumed that the values
- * stored are in sorted order.
- * The encoded key is stored in "*presult". If no value is larger or equal to the key,
-* 128 is returned. The pointer initOffset is a pointer to the last four value decoded
-* (when starting out, this can be a zero vector or initialized with _mm_set1_epi32(init)),
-* and the vector gets updated.
-**/
-int
-simdsearchd1(__m128i * initOffset, const __m128i *in, uint32_t bit,
-                uint32_t key, uint32_t *presult);
-
-
-/* searches "bit" 128-bit vectors from "in" (= length<=128 encoded integers) for the first encoded uint32 value
- * which is >= |key|, and returns its position. It is assumed that the values
- * stored are in sorted order.
- * The encoded key is stored in "*presult".
- * The first length decoded integers, ignoring others. If no value is larger or equal to the key,
- * length is returned. Length should be no larger than 128.
- *
- * If no value is larger or equal to the key,
-* length is returned */
-int simdsearchwithlengthd1(uint32_t initvalue, const __m128i *in, uint32_t bit,
-                int length, uint32_t key, uint32_t *presult);
-
-
-
-/* returns the value stored at the specified "slot".
-* */
-uint32_t simdselectd1(uint32_t initvalue, const __m128i *in, uint32_t bit,
-                int slot);
-
-/* given a block of 128 packed values, this function sets the value at index "index" to "value",
- * you must somehow know the previous value.
- * Because of differential coding, all following values are incremented by the offset between this new
- * value and the old value... 
- * This functions is useful if you want to modify the last value. 
- */
-void simdfastsetd1fromprevious( __m128i * in, uint32_t bit, uint32_t previousvalue, uint32_t value, size_t index);
-
-/* given a block of 128 packed values, this function sets the value at index "index" to "value",
- * This function computes the previous value if needed.
- * Because of differential coding, all following values are incremented by the offset between this new
- * value and the old value...
- * This functions is useful if you want to modify the last value. 
- */
-void simdfastsetd1(uint32_t initvalue, __m128i * in, uint32_t bit, uint32_t value, size_t index);
-
-
-/*Simply scan the data
-* The pointer initOffset is a pointer to the last four value decoded
-* (when starting out, this can be a zero vector or initialized with _mm_set1_epi32(init);),
-* and the vector gets updated.
-* */
-
-void
-simdscand1(__m128i * initOffset, const __m128i *in, uint32_t bit);
-
-#ifdef __cplusplus
-} // extern "C"
-#endif
-
-#endif

cpp/simdcomp/makefile

@@ -1,79 +0,0 @@
-# minimalist makefile
-.SUFFIXES:
-#
-.SUFFIXES: .cpp .o .c .h
-ifeq ($(DEBUG),1)
-CFLAGS = -fPIC  -std=c89 -ggdb -msse4.1 -march=native -Wall -Wextra -Wshadow -fsanitize=undefined  -fno-omit-frame-pointer -fsanitize=address
-else
-CFLAGS = -fPIC -std=c89 -O3 -msse4.1  -march=native -Wall -Wextra -Wshadow
-endif # debug
-LDFLAGS = -shared
-LIBNAME=libsimdcomp.so.0.0.3
-all:  unit unit_chars bitpackingbenchmark $(LIBNAME)
-test:
-	./unit
-	./unit_chars
-install: $(OBJECTS)
-	cp $(LIBNAME) /usr/local/lib
-	ln -s /usr/local/lib/$(LIBNAME) /usr/local/lib/libsimdcomp.so
-	ldconfig
-	cp $(HEADERS) /usr/local/include
-
-
-
-HEADERS=./include/simdbitpacking.h ./include/simdcomputil.h ./include/simdintegratedbitpacking.h ./include/simdcomp.h ./include/simdfor.h ./include/avxbitpacking.h
-
-uninstall:
-	for h in $(HEADERS) ; do rm  /usr/local/$$h; done
-	rm  /usr/local/lib/$(LIBNAME)
-	rm /usr/local/lib/libsimdcomp.so
-	ldconfig
-
-
-OBJECTS= simdbitpacking.o simdintegratedbitpacking.o simdcomputil.o \
-		 simdpackedsearch.o simdpackedselect.o simdfor.o avxbitpacking.o
-
-$(LIBNAME): $(OBJECTS)
-	$(CC) $(CFLAGS) -o $(LIBNAME) $(OBJECTS)  $(LDFLAGS)
-
-
-avxbitpacking.o: ./src/avxbitpacking.c $(HEADERS)
-	$(CC) $(CFLAGS) -c ./src/avxbitpacking.c -Iinclude
-
-
-simdfor.o: ./src/simdfor.c $(HEADERS)
-	$(CC) $(CFLAGS) -c ./src/simdfor.c -Iinclude
-
-
-simdcomputil.o: ./src/simdcomputil.c $(HEADERS)
-	$(CC) $(CFLAGS) -c ./src/simdcomputil.c -Iinclude
-
-simdbitpacking.o: ./src/simdbitpacking.c $(HEADERS)
-	$(CC) $(CFLAGS) -c ./src/simdbitpacking.c -Iinclude
-
-simdintegratedbitpacking.o: ./src/simdintegratedbitpacking.c  $(HEADERS)
-	$(CC) $(CFLAGS) -c ./src/simdintegratedbitpacking.c -Iinclude
-
-simdpackedsearch.o: ./src/simdpackedsearch.c $(HEADERS)
-	$(CC) $(CFLAGS) -c ./src/simdpackedsearch.c -Iinclude
-
-simdpackedselect.o: ./src/simdpackedselect.c $(HEADERS)
-	$(CC) $(CFLAGS) -c ./src/simdpackedselect.c -Iinclude
-
-example: ./example.c    $(HEADERS) $(OBJECTS)
-	$(CC) $(CFLAGS) -o example ./example.c -Iinclude  $(OBJECTS)
-
-unit: ./tests/unit.c    $(HEADERS) $(OBJECTS)
-	$(CC) $(CFLAGS) -o unit ./tests/unit.c -Iinclude  $(OBJECTS)
-
-bitpackingbenchmark: ./benchmarks/bitpackingbenchmark.c    $(HEADERS) $(OBJECTS)
-	$(CC) $(CFLAGS) -o bitpackingbenchmark ./benchmarks/bitpackingbenchmark.c -Iinclude  $(OBJECTS)
-benchmark: ./benchmarks/benchmark.c    $(HEADERS) $(OBJECTS)
-	$(CC) $(CFLAGS) -o benchmark ./benchmarks/benchmark.c -Iinclude  $(OBJECTS)
-dynunit: ./tests/unit.c    $(HEADERS) $(LIBNAME)
-	$(CC) $(CFLAGS) -o dynunit ./tests/unit.c -Iinclude  -lsimdcomp
-
-unit_chars: ./tests/unit_chars.c    $(HEADERS) $(OBJECTS)
-	$(CC) $(CFLAGS) -o unit_chars ./tests/unit_chars.c -Iinclude  $(OBJECTS)
-clean:
-	rm -f unit *.o $(LIBNAME) example benchmark bitpackingbenchmark dynunit unit_chars

cpp/simdcomp/makefile.vc

@@ -1,104 +0,0 @@
-
-!IFNDEF MACHINE
-!IF "$(PROCESSOR_ARCHITECTURE)"=="AMD64"
-MACHINE=x64
-!ELSE
-MACHINE=x86
-!ENDIF
-!ENDIF
-
-!IFNDEF DEBUG
-DEBUG=no
-!ENDIF
-
-!IFNDEF CC
-CC=cl.exe
-!ENDIF
-
-!IFNDEF AR
-AR=lib.exe
-!ENDIF
-
-!IFNDEF LINK
-LINK=link.exe
-!ENDIF
-
-!IFNDEF PGO
-PGO=no
-!ENDIF
-
-!IFNDEF PGI
-PGI=no
-!ENDIF
-
-INC = /Iinclude
-
-!IF "$(DEBUG)"=="yes"
-CFLAGS = /nologo /MDd /LDd /Od /Zi /D_DEBUG /RTC1 /W3 /GS /Gm
-ARFLAGS = /nologo
-LDFLAGS = /nologo /debug /nodefaultlib:msvcrt
-!ELSE
-CFLAGS = /nologo /MD /O2 /Zi /DNDEBUG /W3 /Gm- /GS /Gy /Oi /GL /MP
-ARFLAGS = /nologo /LTCG
-LDFLAGS = /nologo /LTCG /DYNAMICBASE /incremental:no /debug /opt:ref,icf
-!ENDIF
-
-!IF "$(PGI)"=="yes"
-LDFLAGS = $(LDFLAGS) /ltcg:pgi
-!ENDIF
-
-!IF "$(PGO)"=="yes"
-LDFLAGS = $(LDFLAGS) /ltcg:pgo
-!ENDIF
-
-LIB_OBJS = simdbitpacking.obj simdintegratedbitpacking.obj simdcomputil.obj \
-	simdpackedsearch.obj simdpackedselect.obj simdfor.obj
-
-
-all: lib dll dynunit unit_chars example benchmark
-# need some good use case scenario to train the instrumented build
-	@if "$(PGI)"=="yes" echo Running PGO training
-	@if "$(PGI)"=="yes" benchmark.exe >nul 2>&1
-	@if "$(PGI)"=="yes" example.exe >nul 2>&1
-
-
-$(LIB_OBJS):
-	$(CC) $(INC) $(CFLAGS) /c src/simdbitpacking.c src/simdintegratedbitpacking.c src/simdcomputil.c \
-		src/simdpackedsearch.c src/simdpackedselect.c src/simdfor.c
-
-lib: $(LIB_OBJS)
-	$(AR) $(ARFLAGS) /OUT:simdcomp_a.lib $(LIB_OBJS)
-
-dll: $(LIB_OBJS)
-	$(LINK) /DLL $(LDFLAGS) /OUT:simdcomp.dll /IMPLIB:simdcomp.lib /DEF:simdcomp.def $(LIB_OBJS)
-
-unit: lib
-	$(CC) $(INC) $(CFLAGS) /c src/unit.c 
-	$(LINK) $(LDFLAGS) /OUT:unit.exe unit.obj simdcomp_a.lib
-
-dynunit: dll
-	$(CC) $(INC) $(CFLAGS) /c src/unit.c 
-	$(LINK) $(LDFLAGS) /OUT:unit.exe unit.obj simdcomp.lib
-
-unit_chars: lib
-	$(CC) $(INC) $(CFLAGS) /c src/unit_chars.c
-	$(LINK) $(LDFLAGS) /OUT:unit_chars.exe unit_chars.obj simdcomp.lib
-
-
-example: lib
-	$(CC) $(INC) $(CFLAGS) /c example.c
-	$(LINK) $(LDFLAGS) /OUT:example.exe example.obj simdcomp.lib
-
-benchmark: lib
-	$(CC) $(INC) $(CFLAGS) /c src/benchmark.c
-	$(LINK) $(LDFLAGS) /OUT:benchmark.exe benchmark.obj simdcomp.lib
-
-clean:
-	del /Q *.obj
-	del /Q *.lib
-	del /Q *.exe
-	del /Q *.dll
-	del /Q *.pgc
-	del /Q *.pgd
-	del /Q *.pdb
-

cpp/simdcomp/package.json

@@ -1,16 +0,0 @@
-{
-  "name": "simdcomp",
-  "version": "0.0.3",
-  "repo": "lemire/simdcomp",
-  "description": "A simple C library for compressing lists of integers",
-  "license": "BSD-3-Clause",
-  "src": [
-    "src/simdbitpacking.c",
-    "src/simdcomputil.c",
-    "src/simdintegratedbitpacking.c",
-    "include/simdbitpacking.h",
-    "include/simdcomp.h",
-    "include/simdcomputil.h",
-    "include/simdintegratedbitpacking.h"
-  ]
-}

cpp/simdcomp/scripts/avxpacking.py

@@ -1,182 +0,0 @@
-#!/usr/bin/env python
-import sys
-def howmany(bit):
-    """ how many values are we going to pack? """
-    return 256
-
-def howmanywords(bit):
-    return (howmany(bit) * bit + 255)/256
-
-def howmanybytes(bit):
-    return howmanywords(bit) * 16
-
-print("""
-/** code generated by avxpacking.py starts here **/
-""")
-
-print("""typedef void (*avxpackblockfnc)(const uint32_t * pin, __m256i * compressed);""")
-print("""typedef void (*avxunpackblockfnc)(const __m256i * compressed, uint32_t * pout);""")
-
-
-
-
-
-
-def plurial(number):
-    if(number <> 1):
-        return "s"
-    else :
-        return ""
-
-print("")
-print("static void avxpackblock0(const uint32_t * pin, __m256i * compressed) {");
-print("  (void)compressed;");
-print("  (void) pin; /* we consumed {0} 32-bit integer{1} */ ".format(howmany(0),plurial(howmany(0))));
-print("}");
-print("")
-
-for bit in range(1,33):
-    print("")
-    print("/* we are going to pack {0} {1}-bit values, touching {2} 256-bit words, using {3} bytes */ ".format(howmany(bit),bit,howmanywords(bit),howmanybytes(bit)))
-    print("static void avxpackblock{0}(const uint32_t * pin, __m256i * compressed) {{".format(bit));
-    print("  const __m256i * in = (const __m256i *)  pin;");
-    print("  /* we are going to touch  {0} 256-bit word{1} */ ".format(howmanywords(bit),plurial(howmanywords(bit))));
-    if(howmanywords(bit) == 1):
-      print("  __m256i w0;")
-    else:
-      print("  __m256i w0, w1;")
-    if( (bit & (bit-1)) <> 0) : print("  __m256i tmp; /* used to store inputs at word boundary */")
-    oldword = 0
-    for j in range(howmany(bit)/8):
-      firstword = j * bit / 32
-      if(firstword > oldword):
-        print("  _mm256_storeu_si256(compressed + {0}, w{1});".format(oldword,oldword%2))
-        oldword = firstword
-      secondword = (j * bit + bit - 1)/32
-      firstshift = (j*bit) % 32
-      if( firstword == secondword):
-          if(firstshift == 0):
-            print("  w{0} = _mm256_lddqu_si256 (in + {1});".format(firstword%2,j))
-          else:
-            print("  w{0} = _mm256_or_si256(w{0},_mm256_slli_epi32(_mm256_lddqu_si256 (in + {1}) , {2}));".format(firstword%2,j,firstshift))
-      else:
-          print("  tmp = _mm256_lddqu_si256 (in + {0});".format(j))
-          print("  w{0} = _mm256_or_si256(w{0},_mm256_slli_epi32(tmp , {2}));".format(firstword%2,j,firstshift))
-          secondshift = 32-firstshift
-          print("  w{0} = _mm256_srli_epi32(tmp,{2});".format(secondword%2,j,secondshift))
-    print("  _mm256_storeu_si256(compressed + {0}, w{1});".format(secondword,secondword%2))
-    print("}");
-    print("")
-
-
-print("")
-print("static void avxpackblockmask0(const uint32_t * pin, __m256i * compressed) {");
-print("  (void)compressed;");
-print("  (void) pin; /* we consumed {0} 32-bit integer{1} */ ".format(howmany(0),plurial(howmany(0))));
-print("}");
-print("")
-
-for bit in range(1,33):
-    print("")
-    print("/* we are going to pack {0} {1}-bit values, touching {2} 256-bit words, using {3} bytes */ ".format(howmany(bit),bit,howmanywords(bit),howmanybytes(bit)))
-    print("static void avxpackblockmask{0}(const uint32_t * pin, __m256i * compressed) {{".format(bit));
-    print("  /* we are going to touch  {0} 256-bit word{1} */ ".format(howmanywords(bit),plurial(howmanywords(bit))));
-    if(howmanywords(bit) == 1):
-      print("  __m256i w0;")
-    else:
-      print("  __m256i w0, w1;")
-    print("  const __m256i * in = (const __m256i *) pin;");
-    if(bit < 32): print("  const __m256i mask = _mm256_set1_epi32({0});".format((1<<bit)-1));
-    def maskfnc(x):
-        if(bit == 32): return x
-        return " _mm256_and_si256 ( mask, {0}) ".format(x)
-    if( (bit & (bit-1)) <> 0) : print("  __m256i tmp; /* used to store inputs at word boundary */")
-    oldword = 0
-    for j in range(howmany(bit)/8):
-      firstword = j * bit / 32
-      if(firstword > oldword):
-        print("  _mm256_storeu_si256(compressed + {0}, w{1});".format(oldword,oldword%2))
-        oldword = firstword
-      secondword = (j * bit + bit - 1)/32
-      firstshift = (j*bit) % 32
-      loadstr = maskfnc(" _mm256_lddqu_si256 (in + {0}) ".format(j))
-      if( firstword == secondword):
-          if(firstshift == 0):
-            print("  w{0} = {1};".format(firstword%2,loadstr))
-          else:
-            print("  w{0} = _mm256_or_si256(w{0},_mm256_slli_epi32({1} , {2}));".format(firstword%2,loadstr,firstshift))
-      else:
-          print("  tmp = {0};".format(loadstr))
-          print("  w{0} = _mm256_or_si256(w{0},_mm256_slli_epi32(tmp , {2}));".format(firstword%2,j,firstshift))
-          secondshift = 32-firstshift
-          print("  w{0} = _mm256_srli_epi32(tmp,{2});".format(secondword%2,j,secondshift))
-    print("  _mm256_storeu_si256(compressed + {0}, w{1});".format(secondword,secondword%2))
-    print("}");
-    print("")
-
-
-print("static void avxunpackblock0(const __m256i * compressed, uint32_t * pout) {");
-print("  (void) compressed;");
-print("  memset(pout,0,{0});".format(howmany(0)));
-print("}");
-print("")
-
-for bit in range(1,33):
-    print("")
-    print("/* we packed {0} {1}-bit values, touching {2} 256-bit words, using {3} bytes */ ".format(howmany(bit),bit,howmanywords(bit),howmanybytes(bit)))
-    print("static void avxunpackblock{0}(const __m256i * compressed, uint32_t * pout) {{".format(bit));
-    print("  /* we are going to access  {0} 256-bit word{1} */ ".format(howmanywords(bit),plurial(howmanywords(bit))));
-    if(howmanywords(bit) == 1):
-      print("  __m256i w0;")
-    else:
-      print("  __m256i w0, w1;")
-    print("  __m256i * out = (__m256i *) pout;");
-    if(bit < 32): print("  const __m256i mask = _mm256_set1_epi32({0});".format((1<<bit)-1));
-    maskstr = " _mm256_and_si256 ( mask, {0}) "
-    if (bit == 32) : maskstr = " {0} " # no need
-    oldword = 0
-    print("  w0 = _mm256_lddqu_si256 (compressed);")
-    for j in range(howmany(bit)/8):
-      firstword = j * bit / 32
-      secondword = (j * bit + bit - 1)/32
-      if(secondword > oldword):
-        print("  w{0} = _mm256_lddqu_si256 (compressed + {1});".format(secondword%2,secondword))
-        oldword = secondword
-      firstshift = (j*bit) % 32
-      firstshiftstr = "_mm256_srli_epi32( w{0} , "+str(firstshift)+") "
-      if(firstshift == 0):
-          firstshiftstr =" w{0} " # no need
-      wfirst = firstshiftstr.format(firstword%2)
-      if( firstword == secondword):
-          if(firstshift + bit <> 32):
-            wfirst  = maskstr.format(wfirst)
-          print("  _mm256_storeu_si256(out + {0}, {1});".format(j,wfirst))
-      else:
-          secondshift = (32-firstshift)
-          wsecond = "_mm256_slli_epi32( w{0} , {1} ) ".format((firstword+1)%2,secondshift)
-          wfirstorsecond = " _mm256_or_si256 ({0},{1}) ".format(wfirst,wsecond)
-          wfirstorsecond = maskstr.format(wfirstorsecond)
-          print("  _mm256_storeu_si256(out + {0},\n    {1});".format(j,wfirstorsecond))
-    print("}");
-    print("")
-
-
-print("static avxpackblockfnc avxfuncPackArr[] = {")
-for bit in range(0,32):
-  print("&avxpackblock{0},".format(bit))
-print("&avxpackblock32")
-print("};")
-
-print("static avxpackblockfnc avxfuncPackMaskArr[] = {")
-for bit in range(0,32):
-  print("&avxpackblockmask{0},".format(bit))
-print("&avxpackblockmask32")
-print("};")
-
-
-print("static avxunpackblockfnc avxfuncUnpackArr[] = {")
-for bit in range(0,32):
-  print("&avxunpackblock{0},".format(bit))
-print("&avxunpackblock32")
-print("};")
-print("/** code generated by avxpacking.py ends here **/")

cpp/simdcomp/scripts/simdfor.py

@@ -1,152 +0,0 @@
-#!/usr/bin/env python3
-
-
-from math import ceil
-
-print("""
-/**
-* Blablabla
-*
-*/
-
-""");
-
-def mask(bit):
-  return str((1 << bit) - 1)
-
-for length in [32]:
-  print("""
-static __m128i  iunpackFOR0(__m128i initOffset, const __m128i *   _in , uint32_t *    _out) {
-    __m128i       *out = (__m128i*)(_out);
-    int i;
-    (void) _in;
-    for (i = 0; i < 8; ++i) {
-        _mm_store_si128(out++, initOffset);
-    	_mm_store_si128(out++, initOffset);
-        _mm_store_si128(out++, initOffset);
-        _mm_store_si128(out++, initOffset);
-    }
-
-    return initOffset;
-}
-
-  """)
-  print("""
-
-static void ipackFOR0(__m128i initOffset , const uint32_t *   _in , __m128i *  out  ) {
-    (void) initOffset;
-    (void) _in;
-    (void) out;
-}
-""") 
-  for bit in range(1,33):
-    offsetVar = " initOffset";
-    print("""  
-static void ipackFOR"""+str(bit)+"""(__m128i """+offsetVar+""", const uint32_t *   _in, __m128i *   out) {
-    const __m128i       *in = (const __m128i*)(_in);
-    __m128i    OutReg;
-
-      """);
-    
-    if (bit != 32):
-      print("    __m128i CurrIn = _mm_load_si128(in);");
-      print("    __m128i InReg = _mm_sub_epi32(CurrIn, initOffset);");
-    else:
-      print("    __m128i InReg = _mm_load_si128(in);");
-      print("    (void) initOffset;");
-
-
-    inwordpointer = 0
-    valuecounter = 0
-    for k in range(ceil((length * bit) / 32)):
-      if(valuecounter == length): break
-      for x in range(inwordpointer,32,bit):
-        if(x!=0) :
-          print("    OutReg = _mm_or_si128(OutReg, _mm_slli_epi32(InReg, " + str(x) + "));");
-        else:
-          print("    OutReg = InReg; ");
-        if((x+bit>=32) ):
-          while(inwordpointer<32):
-            inwordpointer += bit
-          print("    _mm_store_si128(out, OutReg);");
-          print("");
-
-          if(valuecounter + 1 < length):
-            print("    ++out;")
-          inwordpointer -= 32;
-          if(inwordpointer>0):
-            print("    OutReg = _mm_srli_epi32(InReg, " + str(bit) + " - " + str(inwordpointer) + ");");
-        if(valuecounter + 1 < length):
-          print("    ++in;") 
-
-          if (bit != 32):
-            print("    CurrIn = _mm_load_si128(in);");
-            print("    InReg = _mm_sub_epi32(CurrIn, initOffset);");
-          else:
-            print("    InReg = _mm_load_si128(in);");
-          print("");
-        valuecounter = valuecounter + 1
-        if(valuecounter == length): break
-    assert(valuecounter == length)
-    print("\n}\n\n""")
-
-  for bit in range(1,32):
-    offsetVar = " initOffset";
-    print("""\n
-static __m128i iunpackFOR"""+str(bit)+"""(__m128i """+offsetVar+""", const  __m128i*   in, uint32_t *   _out) {
-      """);
-    print("""    __m128i*   out = (__m128i*)(_out);
-    __m128i    InReg = _mm_load_si128(in);
-    __m128i    OutReg;    
-    __m128i     tmp;
-    const __m128i mask =  _mm_set1_epi32((1U<<"""+str(bit)+""")-1);
-
-    """);
-
-    MainText = "";
-
-    MainText += "\n";
-    inwordpointer = 0
-    valuecounter = 0
-    for k in range(ceil((length * bit) / 32)):
-      for x in range(inwordpointer,32,bit):
-        if(valuecounter == length): break
-        if (x > 0):
-          MainText += "    tmp = _mm_srli_epi32(InReg," + str(x) +");\n"; 
-        else:
-          MainText += "    tmp = InReg;\n"; 
-        if(x+bit<32):
-          MainText += "    OutReg = _mm_and_si128(tmp, mask);\n";
-        else:
-          MainText += "    OutReg = tmp;\n";        
-        if((x+bit>=32) ):      
-          while(inwordpointer<32):
-            inwordpointer += bit
-          if(valuecounter + 1 < length):
-             MainText += "    ++in;"
-             MainText += "    InReg = _mm_load_si128(in);\n";
-          inwordpointer -= 32;
-          if(inwordpointer>0):
-            MainText += "    OutReg = _mm_or_si128(OutReg, _mm_and_si128(_mm_slli_epi32(InReg, " + str(bit) + "-" + str(inwordpointer) + "), mask));\n\n";
-        if (bit != 32):
-          MainText += "    OutReg = _mm_add_epi32(OutReg, initOffset);\n"; 
-        MainText += "    _mm_store_si128(out++, OutReg);\n\n"; 
-        MainText += "";
-        valuecounter = valuecounter + 1
-        if(valuecounter == length): break
-    assert(valuecounter == length)
-    print(MainText)
-    print("    return initOffset;");
-    print("\n}\n\n")
-  print("""
-static __m128i iunpackFOR32(__m128i initvalue , const  __m128i*   in, uint32_t *    _out) {
-	__m128i * mout = (__m128i *)_out;
-	__m128i invec;
-	size_t k;
-	for(k = 0; k < 128/4; ++k) {
-		invec =  _mm_load_si128(in++);
-	    _mm_store_si128(mout++, invec);
-	}
-	return invec;
-}
-  """)

cpp/simdcomp/simdcomp.def

@@ -1,40 +0,0 @@
-EXPORTS
-	simdpack
-	simdpackwithoutmask
-	simdunpack
-	bits
-	maxbits
-	maxbits_length
-	simdmin
-	simdmin_length
-	simdmaxmin
-	simdmaxmin_length
-	simdmaxbitsd1
-	simdmaxbitsd1_length
-	simdpackd1
-	simdpackwithoutmaskd1
-	simdunpackd1
-	simdsearchd1
-	simdsearchwithlengthd1
-	simdselectd1
-	simdpackFOR
-	simdselectFOR
-	simdsearchwithlengthFOR
-	simdunpackFOR
-	simdmin_length
-	simdmaxmin
-	simdmaxmin_length
-	simdpack_length
-	simdpackFOR_length
-	simdunpackFOR_length
-	simdpack_shortlength
-	simdfastsetFOR
-	simdfastset
-	simdfastsetd1
-	simdunpack_length
-	simdunpack_shortlength
-	simdsearchwithlengthFOR
-	simdscand1
-	simdfastsetd1fromprevious
-	simdfastsetd1
-

cpp/simdcomp/src/simdcomputil.c

@@ -1,234 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-
-#include "simdcomputil.h"
-#ifdef __SSE4_1__
-#include <smmintrin.h>
-#endif
-#include <assert.h>
-
-#define Delta(curr, prev) \
-    _mm_sub_epi32(curr, \
-            _mm_or_si128(_mm_slli_si128(curr, 4), _mm_srli_si128(prev, 12)))
-
-/* returns the integer logarithm of v (bit width) */
-uint32_t bits(const uint32_t v) {
-#ifdef _MSC_VER
-    unsigned long answer;
-    if (v == 0) {
-        return 0;
-    }
-    _BitScanReverse(&answer, v);
-    return answer + 1;
-#else
-    return v == 0 ? 0 : 32 - __builtin_clz(v); /* assume GCC-like compiler if not microsoft */
-#endif
-}
-
-
-
-static uint32_t maxbitas32int(const __m128i accumulator) {
-	const __m128i _tmp1 = _mm_or_si128(_mm_srli_si128(accumulator, 8), accumulator); /* (A,B,C,D) xor (0,0,A,B) = (A,B,C xor A,D xor B)*/
-	const __m128i _tmp2 = _mm_or_si128(_mm_srli_si128(_tmp1, 4), _tmp1); /*  (A,B,C xor A,D xor B) xor  (0,0,0,C xor A)*/
-	uint32_t ans =  _mm_cvtsi128_si32(_tmp2);
-	return bits(ans);
-}
-
-SIMDCOMP_PURE uint32_t maxbits(const uint32_t * begin) {
-	    const __m128i* pin = (const __m128i*)(begin);
-	    __m128i accumulator = _mm_loadu_si128(pin);
-	    uint32_t k = 1;
-	    for(; 4*k < SIMDBlockSize; ++k) {
-	    	__m128i newvec = _mm_loadu_si128(pin+k);
-	        accumulator = _mm_or_si128(accumulator,newvec);
-	    }
-	    return maxbitas32int(accumulator);
-}
-static uint32_t orasint(const __m128i accumulator) {
-	const __m128i _tmp1 = _mm_or_si128(_mm_srli_si128(accumulator, 8), accumulator); /* (A,B,C,D) xor (0,0,A,B) = (A,B,C xor A,D xor B)*/
-	const __m128i _tmp2 = _mm_or_si128(_mm_srli_si128(_tmp1, 4), _tmp1); /*  (A,B,C xor A,D xor B) xor  (0,0,0,C xor A)*/
-	return  _mm_cvtsi128_si32(_tmp2);
-}
-
-#ifdef __SSE4_1__
-
-static uint32_t minasint(const __m128i accumulator) {
-	const __m128i _tmp1 = _mm_min_epu32(_mm_srli_si128(accumulator, 8), accumulator); /* (A,B,C,D) xor (0,0,A,B) = (A,B,C xor A,D xor B)*/
-	const __m128i _tmp2 = _mm_min_epu32(_mm_srli_si128(_tmp1, 4), _tmp1); /*  (A,B,C xor A,D xor B) xor  (0,0,0,C xor A)*/
-	return  _mm_cvtsi128_si32(_tmp2);
-}
-
-static uint32_t maxasint(const __m128i accumulator) {
-	const __m128i _tmp1 = _mm_max_epu32(_mm_srli_si128(accumulator, 8), accumulator); /* (A,B,C,D) xor (0,0,A,B) = (A,B,C xor A,D xor B)*/
-	const __m128i _tmp2 = _mm_max_epu32(_mm_srli_si128(_tmp1, 4), _tmp1); /*  (A,B,C xor A,D xor B) xor  (0,0,0,C xor A)*/
-	return  _mm_cvtsi128_si32(_tmp2);
-}
-
-uint32_t simdmin(const uint32_t * in) {
-    const __m128i* pin = (const __m128i*)(in);
-    __m128i accumulator =  _mm_loadu_si128(pin);
-     uint32_t k = 1;
-     for(; 4*k < SIMDBlockSize; ++k) {
-    	 __m128i newvec = _mm_loadu_si128(pin+k);
-         accumulator = _mm_min_epu32(accumulator,newvec);
-     }
-     return minasint(accumulator);
-}
-
-void simdmaxmin(const uint32_t * in, uint32_t * getmin, uint32_t * getmax) {
-    const __m128i* pin = (const __m128i*)(in);
-    __m128i minaccumulator =  _mm_loadu_si128(pin);
-    __m128i maxaccumulator =  minaccumulator;
-    uint32_t k = 1;
-     for(; 4*k < SIMDBlockSize; ++k) {
-    	 __m128i newvec = _mm_loadu_si128(pin+k);
-         minaccumulator = _mm_min_epu32(minaccumulator,newvec);
-         maxaccumulator = _mm_max_epu32(maxaccumulator,newvec);
-     }
-     *getmin = minasint(minaccumulator);
-     *getmax = maxasint(maxaccumulator);
-}
-
-
-uint32_t simdmin_length(const uint32_t * in, uint32_t length) {
-	uint32_t currentmin = 0xFFFFFFFF;
-	uint32_t lengthdividedby4 = length / 4;
-	uint32_t offset = lengthdividedby4 * 4;
-	uint32_t k;
-	if (lengthdividedby4 > 0) {
-		const __m128i* pin = (const __m128i*)(in);
-		__m128i accumulator = _mm_loadu_si128(pin);
-		k = 1;
-		for(; 4*k < lengthdividedby4 * 4; ++k) {
-			__m128i newvec = _mm_loadu_si128(pin+k);
-			accumulator = _mm_min_epu32(accumulator,newvec);
-		}
-		currentmin = minasint(accumulator);
-	}
-	for (k = offset; k < length; ++k)
-		if (in[k] < currentmin)
-			currentmin = in[k];
-	return currentmin;
-}
-
-void simdmaxmin_length(const uint32_t * in, uint32_t length, uint32_t * getmin, uint32_t * getmax) {
-	uint32_t lengthdividedby4 = length / 4;
-	uint32_t offset = lengthdividedby4 * 4;
-	uint32_t k;
-	*getmin = 0xFFFFFFFF;
-	*getmax = 0;
-	if (lengthdividedby4 > 0) {
-		const __m128i* pin = (const __m128i*)(in);
-		__m128i minaccumulator = _mm_loadu_si128(pin);
-		__m128i maxaccumulator = minaccumulator;
-		k = 1;
-		for(; 4*k < lengthdividedby4 * 4; ++k) {
-			__m128i newvec = _mm_loadu_si128(pin+k);
-			minaccumulator = _mm_min_epu32(minaccumulator,newvec);
-			maxaccumulator = _mm_max_epu32(maxaccumulator,newvec);
-		}
-		*getmin = minasint(minaccumulator);
-		*getmax = maxasint(maxaccumulator);
-	}
-	for (k = offset; k < length; ++k) {
-		if (in[k] < *getmin)
-			*getmin = in[k];
-		if (in[k] > *getmax)
-			*getmax = in[k];
-	}
-}
-
-#endif
-
-SIMDCOMP_PURE uint32_t maxbits_length(const uint32_t * in,uint32_t length) {
-	  uint32_t k;
-	  uint32_t lengthdividedby4 = length / 4;
-	  uint32_t offset = lengthdividedby4 * 4;
-	  uint32_t bigxor = 0;
-	  if(lengthdividedby4 > 0) {
-		    const __m128i* pin = (const __m128i*)(in);
-		    __m128i accumulator = _mm_loadu_si128(pin);
-		    k = 1;
-		    for(; 4*k < 4*lengthdividedby4; ++k) {
-		    	__m128i newvec = _mm_loadu_si128(pin+k);
-		        accumulator = _mm_or_si128(accumulator,newvec);
-		    }
-		    bigxor = orasint(accumulator);
-	  }
-	  for(k = offset; k < length; ++k)
-		  bigxor |= in[k];
-	  return bits(bigxor);
-}
-
-
-/* maxbit over 128 integers (SIMDBlockSize) with provided initial value */
-uint32_t simdmaxbitsd1(uint32_t initvalue, const uint32_t * in) {
-    __m128i  initoffset = _mm_set1_epi32 (initvalue);
-    const __m128i* pin = (const __m128i*)(in);
-    __m128i newvec = _mm_loadu_si128(pin);
-    __m128i accumulator = Delta(newvec , initoffset);
-    __m128i oldvec = newvec;
-    uint32_t k = 1;
-    for(; 4*k < SIMDBlockSize; ++k) {
-        newvec = _mm_loadu_si128(pin+k);
-        accumulator = _mm_or_si128(accumulator,Delta(newvec , oldvec));
-        oldvec = newvec;
-    }
-    initoffset = oldvec;
-    return maxbitas32int(accumulator);
-}
-
-
-/* maxbit over |length| integers with provided initial value */
-uint32_t simdmaxbitsd1_length(uint32_t initvalue, const uint32_t * in,
-                uint32_t length) {
-    __m128i newvec;
-    __m128i oldvec;
-    __m128i initoffset;
-    __m128i accumulator;
-    const __m128i *pin;
-    uint32_t tmparray[4];
-    uint32_t k = 1;
-    uint32_t acc;
-
-    assert(length > 0);
-
-    pin = (const __m128i *)(in);
-    initoffset = _mm_set1_epi32(initvalue);
-    switch (length) {
-      case 1:
-        newvec = _mm_set1_epi32(in[0]);
-        break;
-      case 2:
-        newvec = _mm_setr_epi32(in[0], in[1], in[1], in[1]);
-        break;
-      case 3:
-        newvec = _mm_setr_epi32(in[0], in[1], in[2], in[2]);
-        break;
-      default:
-        newvec = _mm_loadu_si128(pin);
-        break;
-    }
-    accumulator = Delta(newvec, initoffset);
-    oldvec = newvec;
-
-    /* process 4 integers and build an accumulator */
-    while (k * 4 + 4 <= length) {
-        newvec = _mm_loadu_si128(pin + k);
-        accumulator = _mm_or_si128(accumulator, Delta(newvec, oldvec));
-        oldvec = newvec;
-        k++;
-    }
-
-    /* extract the accumulator as an integer */
-    _mm_storeu_si128((__m128i *)(tmparray), accumulator);
-    acc = tmparray[0] | tmparray[1] | tmparray[2] | tmparray[3];
-
-    /* now process the remaining integers */
-    for (k *= 4; k < length; k++)
-        acc |= in[k] - (k == 0 ? initvalue : in[k - 1]);
-
-    /* return the number of bits */
-    return bits(acc);
-}

cpp/simdcomp/tests/unit.c

@@ -1,900 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-#include <assert.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include "simdcomp.h"
-
-
-
-int testshortpack() {
-	int bit;
-	size_t i;
-	size_t length;
-	__m128i * bb;
-	srand(0);
-	printf("testshortpack\n");
-	for (bit = 0; bit < 32; ++bit) {
-		const size_t N = 128;
-		uint32_t * data = malloc(N * sizeof(uint32_t));
-		uint32_t * backdata = malloc(N * sizeof(uint32_t));
-		uint32_t * buffer = malloc((2 * N + 1024) * sizeof(uint32_t));
-
-		for (i = 0; i < N; ++i) {
-			data[i] = rand() & ((1 << bit) - 1);
-		}
-		for (length = 0; length <= N; ++length) {
-			for (i = 0; i < N; ++i) {
-				backdata[i] = 0;
-			}
-			bb = simdpack_shortlength(data, length, (__m128i *) buffer,
-					bit);
-			if((bb - (__m128i *) buffer) * sizeof(__m128i) != (unsigned) simdpack_compressedbytes(length,bit)) {
-			 printf("bug\n");
-			 return -1;
-			}
-			simdunpack_shortlength((__m128i *) buffer, length,
-					backdata, bit);
-			for (i = 0; i < length; ++i) {
-
-				if (data[i] != backdata[i]) {
-				    printf("bug\n");
-					return -1;
-				}
-			}
-		}
-		free(data);
-		free(backdata);
-		free(buffer);
-	}
-	return 0;
-}
-
-int testlongpack() {
-	int bit;
-	size_t i;
-	size_t length;
-	__m128i * bb;
-	srand(0);
-	printf("testlongpack\n");
-	for (bit = 0; bit < 32; ++bit) {
-		const size_t N = 2048;
-		uint32_t * data = malloc(N * sizeof(uint32_t));
-		uint32_t * backdata = malloc(N * sizeof(uint32_t));
-		uint32_t * buffer = malloc((2 * N + 1024) * sizeof(uint32_t));
-
-		for (i = 0; i < N; ++i) {
-			data[i] = rand() & ((1 << bit) - 1);
-		}
-		for (length = 0; length <= N; ++length) {
-			for (i = 0; i < N; ++i) {
-				backdata[i] = 0;
-			}
-			bb = simdpack_length(data, length, (__m128i *) buffer,
-					bit);
-			if((bb - (__m128i *) buffer) * sizeof(__m128i) != (unsigned) simdpack_compressedbytes(length,bit)) {
-			 printf("bug\n");
-			 return -1;
-			}
-			simdunpack_length((__m128i *) buffer, length,
-					backdata, bit);
-			for (i = 0; i < length; ++i) {
-
-				if (data[i] != backdata[i]) {
-				    printf("bug\n");
-					return -1;
-				}
-			}
-		}
-		free(data);
-		free(backdata);
-		free(buffer);
-	}
-	return 0;
-}
-
-
-
-int testset() {
-	int bit;
-	size_t i;
-	const size_t N = 128;
-	uint32_t * data = malloc(N * sizeof(uint32_t));
-	uint32_t * backdata = malloc(N * sizeof(uint32_t));
-	uint32_t * buffer = malloc((2 * N + 1024) * sizeof(uint32_t));
-
-	srand(0);
-
-	for (bit = 0; bit < 32; ++bit) {
-		printf("simple set %d \n",bit);
-
-		for (i = 0; i < N; ++i) {
-			data[i] = rand() & ((1 << bit) - 1);
-		}
-		for (i = 0; i < N; ++i) {
-			backdata[i] = 0;
-		}
-		simdpack(data, (__m128i *) buffer, bit);
-		simdunpack((__m128i *) buffer, backdata, bit);
-		for (i = 0; i < N; ++i) {
-			if (data[i] != backdata[i]) {
-			    printf("bug\n");
-				return -1;
-			}
-		}
-
-		for(i = N  ; i > 0; i--) {
-			simdfastset((__m128i *) buffer, bit, data[N - i], i - 1);
-		}
-		simdunpack((__m128i *) buffer, backdata, bit);
-		for (i = 0; i < N; ++i) {
-			if (data[i] != backdata[N - i - 1]) {
-			    printf("bug\n");
-				return -1;
-			}
-		}
-		simdpack(data, (__m128i *) buffer, bit);
-		for(i = 1  ; i <= N; i++) {
-			simdfastset((__m128i *) buffer, bit, data[i - 1], i - 1);
-		}
-		simdunpack((__m128i *) buffer, backdata, bit);
-		for (i = 0; i < N; ++i) {
-			if (data[i] != backdata[i]) {
-			    printf("bug\n");
-				return -1;
-			}
-		}
-
-	}
-	free(data);
-	free(backdata);
-	free(buffer);
-
-	return 0;
-}
-
-#ifdef __SSE4_1__
-
-int testsetd1() {
-	int bit;
-	size_t i;
-	uint32_t newvalue;
-	const size_t N = 128;
-	uint32_t * data = malloc(N * sizeof(uint32_t));
-	uint32_t * datazeroes = malloc(N * sizeof(uint32_t));
-
-	uint32_t * backdata = malloc(N * sizeof(uint32_t));
-	uint32_t * buffer = malloc((2 * N + 1024) * sizeof(uint32_t));
-
-	srand(0);
-	for (bit = 0; bit < 32; ++bit) {
-		printf("simple set d1 %d \n",bit);
-		data[0] = rand() & ((1 << bit) - 1);
-		datazeroes[0] = 0;
-
-		for (i = 1; i < N; ++i) {
-			data[i] = data[i - 1] + (rand() & ((1 << bit) - 1));
-			datazeroes[i] = 0;
-		}
-		for (i = 0; i < N; ++i) {
-			backdata[i] = 0;
-		}
-		simdpackd1(0,datazeroes, (__m128i *) buffer, bit);
- 	    for(i = 1  ; i <= N; i++) {
-			simdfastsetd1(0,(__m128i *) buffer, bit, data[i - 1], i - 1);
-			newvalue = simdselectd1(0, (const __m128i *) buffer, bit,i - 1);
-			if( newvalue != data[i-1] ) {
-				printf("bad set-select\n");
-				return -1;
-			}
-		}
-		simdunpackd1(0,(__m128i *) buffer, backdata, bit);
-		for (i = 0; i < N; ++i) {
-			if (data[i] != backdata[i])
-				return -1;
-		}
-	}
-	free(data);
-	free(backdata);
-	free(buffer);
-        free(datazeroes);
-	return 0;
-}
-#endif
-
-int testsetFOR() {
-	int bit;
-	size_t i;
-	uint32_t newvalue;
-	const size_t N = 128;
-	uint32_t * data = malloc(N * sizeof(uint32_t));
-	uint32_t * datazeroes = malloc(N * sizeof(uint32_t));
-
-	uint32_t * backdata = malloc(N * sizeof(uint32_t));
-	uint32_t * buffer = malloc((2 * N + 1024) * sizeof(uint32_t));
-
-	srand(0);
-	for (bit = 0; bit < 32; ++bit) {
-		printf("simple set FOR %d \n",bit);
-		for (i = 0; i < N; ++i) {
-			data[i] = (rand() & ((1 << bit) - 1));
-			datazeroes[i] = 0;
-		}
-		for (i = 0; i < N; ++i) {
-			backdata[i] = 0;
-		}
-		simdpackFOR(0,datazeroes, (__m128i *) buffer, bit);
- 	    for(i = 1  ; i <= N; i++) {
- 	    	simdfastsetFOR(0,(__m128i *) buffer, bit, data[i - 1], i - 1);
-			newvalue = simdselectFOR(0, (const __m128i *) buffer, bit,i - 1);
-			if( newvalue != data[i-1] ) {
-				printf("bad set-select\n");
-				return -1;
-			}
-		}
-		simdunpackFOR(0,(__m128i *) buffer, backdata, bit);
-		for (i = 0; i < N; ++i) {
-			if (data[i] != backdata[i])
-				return -1;
-		}
-	}
-	free(data);
-	free(backdata);
-	free(buffer);
-        free(datazeroes);
-	return 0;
-}
-
-int testshortFORpack() {
-	int bit;
-	size_t i;
-	__m128i * rb;
-	size_t length;
-	uint32_t offset = 7;
-	srand(0);
-	for (bit = 0; bit < 32; ++bit) {
-		const size_t N = 128;
-		uint32_t * data = malloc(N * sizeof(uint32_t));
-		uint32_t * backdata = malloc(N * sizeof(uint32_t));
-		uint32_t * buffer = malloc((2 * N + 1024) * sizeof(uint32_t));
-
-		for (i = 0; i < N; ++i) {
-			data[i] = (rand() & ((1 << bit) - 1)) + offset;
-		}
-		for (length = 0; length <= N; ++length) {
-			for (i = 0; i < N; ++i) {
-				backdata[i] = 0;
-			}
-			rb = simdpackFOR_length(offset,data, length, (__m128i *) buffer,
-					bit);
-		    if(((rb - (__m128i *) buffer)*sizeof(__m128i)) != (unsigned) simdpackFOR_compressedbytes(length,bit)) {
-		      return -1;
-		    }
-			simdunpackFOR_length(offset,(__m128i *) buffer, length,
-					backdata, bit);
-			for (i = 0; i < length; ++i) {
-
-				if (data[i] != backdata[i])
-					return -1;
-			}
-		}
-		free(data);
-		free(backdata);
-		free(buffer);
-	}
-	return 0;
-}
-
-
-#ifdef __AVX2__
-
-int testbabyavx() {
-	int bit;
-	int trial;
-	unsigned int i,j;
-	const size_t N = AVXBlockSize;
-	srand(0);
-	printf("testbabyavx\n");
-	printf("bit = ");
-	for (bit = 0; bit < 32; ++bit) {
-		printf(" %d ",bit);
-		fflush(stdout);
-		for(trial = 0; trial < 100; ++trial) {
-			uint32_t * data = malloc(N * sizeof(uint32_t)+ 64 * sizeof(uint32_t));
-			uint32_t * backdata = malloc(N * sizeof(uint32_t) + 64 * sizeof(uint32_t) );
-			__m256i * buffer = malloc((2 * N + 1024) * sizeof(uint32_t) + 32);
-
-			for (i = 0; i < N; ++i) {
-				data[i] = rand() & ((uint32_t)(1 << bit) - 1);
-			}
-			for (i = 0; i < N; ++i) {
-				backdata[i] = 0;
-			}
-            if(avxmaxbits(data) != maxbits_length(data,N)) {
-            	printf("avxmaxbits is buggy\n");
-				return -1;
-            }
-
-			avxpackwithoutmask(data, buffer, bit);
-			avxunpack(buffer, backdata, bit);
-			for (i = 0; i < AVXBlockSize; ++i) {
-				if (data[i] != backdata[i]) {
-					printf("bug\n");
-					for (j = 0; j < N; ++j) {
-						if (data[j] != backdata[j]) {
-							printf("data[%d]=%d v.s. backdata[%d]=%d\n",j,data[j],j,backdata[j]);
-						} else {
-							printf("data[%d]=%d\n",j,data[j]);
-						}
-					}
-					return -1;
-				}
-			}
-			free(data);
-			free(backdata);
-			free(buffer);
-		}
-	}
-	printf("\n");
-	return 0;
-}
-
-int testavx2() {
-    int N = 5000 * AVXBlockSize, gap;
-    __m256i * buffer = malloc(AVXBlockSize * sizeof(uint32_t));
-    uint32_t * datain = malloc(N * sizeof(uint32_t));
-    uint32_t * backbuffer = malloc(AVXBlockSize * sizeof(uint32_t));
-    for (gap = 1; gap <= 387420489; gap *= 3) {
-        int k;
-        printf(" gap = %u \n", gap);
-        for (k = 0; k < N; ++k)
-            datain[k] = k * gap;
-        for (k = 0; k * AVXBlockSize < N; ++k) {
-            /*
-               First part works for general arrays (sorted or unsorted)
-            */
-            int j;
-       	    /* we compute the bit width */
-            const uint32_t b = avxmaxbits(datain + k * AVXBlockSize);
-            if(avxmaxbits(datain + k * AVXBlockSize) != maxbits_length(datain + k * AVXBlockSize,AVXBlockSize)) {
-            	printf("avxmaxbits is buggy %d %d \n",
-            			avxmaxbits(datain + k * AVXBlockSize),
-						maxbits_length(datain + k * AVXBlockSize,AVXBlockSize));
-				return -1;
-            }
-            printf("bit width = %d\n",b);
-
-
-            /* we read 256 integers at "datain + k * AVXBlockSize" and
-               write b 256-bit vectors at "buffer" */
-            avxpackwithoutmask(datain + k * AVXBlockSize, buffer, b);
-            /* we read back b1 128-bit vectors at "buffer" and write 128 integers at backbuffer */
-			avxunpack(buffer, backbuffer, b);/* uncompressed */
-			for (j = 0; j < AVXBlockSize; ++j) {
-				if (backbuffer[j] != datain[k * AVXBlockSize + j]) {
-					int i;
-					printf("bug in avxpack\n");
-					for(i = 0; i < AVXBlockSize; ++i) {
-						printf("data[%d]=%d got back %d %s\n",i,
-								datain[k * AVXBlockSize + i],backbuffer[i],
-								datain[k * AVXBlockSize + i]!=backbuffer[i]?"bug":"");
-					}
-					return -2;
-				}
-			}
-        }
-    }
-    free(buffer);
-    free(datain);
-    free(backbuffer);
-    printf("Code looks good.\n");
-    return 0;
-}
-#endif /* avx2 */
-
-int test() {
-    int N = 5000 * SIMDBlockSize, gap;
-    __m128i * buffer = malloc(SIMDBlockSize * sizeof(uint32_t));
-    uint32_t * datain = malloc(N * sizeof(uint32_t));
-    uint32_t * backbuffer = malloc(SIMDBlockSize * sizeof(uint32_t));
-    for (gap = 1; gap <= 387420489; gap *= 3) {
-        int k;
-        printf(" gap = %u \n", gap);
-        for (k = 0; k < N; ++k)
-            datain[k] = k * gap;
-        for (k = 0; k * SIMDBlockSize < N; ++k) {
-            /*
-               First part works for general arrays (sorted or unsorted)
-            */
-            int j;
-       	    /* we compute the bit width */
-            const uint32_t b = maxbits(datain + k * SIMDBlockSize);
-            /* we read 128 integers at "datain + k * SIMDBlockSize" and
-               write b 128-bit vectors at "buffer" */
-            simdpackwithoutmask(datain + k * SIMDBlockSize, buffer, b);
-            /* we read back b1 128-bit vectors at "buffer" and write 128 integers at backbuffer */
-            simdunpack(buffer, backbuffer, b);/* uncompressed */
-            for (j = 0; j < SIMDBlockSize; ++j) {
-                if (backbuffer[j] != datain[k * SIMDBlockSize + j]) {
-                    printf("bug in simdpack\n");
-                    return -2;
-                }
-            }
-
-	    {
-                /*
-                 next part assumes that the data is sorted (uses differential coding)
-                */
-                uint32_t offset = 0;
-                /* we compute the bit width */
-                const uint32_t b1 = simdmaxbitsd1(offset,
-                    datain + k * SIMDBlockSize);
-               /* we read 128 integers at "datain + k * SIMDBlockSize" and
-                  write b1 128-bit vectors at "buffer" */
-               simdpackwithoutmaskd1(offset, datain + k * SIMDBlockSize, buffer,
-                    b1);
-               /* we read back b1 128-bit vectors at "buffer" and write 128 integers at backbuffer */
-               simdunpackd1(offset, buffer, backbuffer, b1);
-               for (j = 0; j < SIMDBlockSize; ++j) {
-                   if (backbuffer[j] != datain[k * SIMDBlockSize + j]) {
-                       printf("bug in simdpack d1\n");
-                       return -3;
-                   }
-               }
-               offset = datain[k * SIMDBlockSize + SIMDBlockSize - 1];
-	    }
-        }
-    }
-    free(buffer);
-    free(datain);
-    free(backbuffer);
-    printf("Code looks good.\n");
-    return 0;
-}
-
-#ifdef __SSE4_1__
-int testFOR() {
-    int N = 5000 * SIMDBlockSize, gap;
-    __m128i * buffer = malloc(SIMDBlockSize * sizeof(uint32_t));
-    uint32_t * datain = malloc(N * sizeof(uint32_t));
-    uint32_t * backbuffer = malloc(SIMDBlockSize * sizeof(uint32_t));
-    uint32_t tmax, tmin, tb;
-    for (gap = 1; gap <= 387420489; gap *= 2) {
-        int k;
-        printf(" gap = %u \n", gap);
-        for (k = 0; k < N; ++k)
-            datain[k] = k * gap;
-        for (k = 0; k * SIMDBlockSize < N; ++k) {
-            int j;
-            simdmaxmin_length(datain + k * SIMDBlockSize,SIMDBlockSize,&tmin,&tmax);
-       	    /* we compute the bit width */
-            tb  = bits(tmax - tmin);
-
-
-            /* we read 128 integers at "datain + k * SIMDBlockSize" and
-               write b 128-bit vectors at "buffer" */
-            simdpackFOR(tmin,datain + k * SIMDBlockSize, buffer, tb);
-
-            for (j = 0; j < SIMDBlockSize; ++j) {
-                        uint32_t selectedvalue = simdselectFOR(tmin,buffer,tb,j);
-                    	if (selectedvalue != datain[k * SIMDBlockSize + j]) {
-                            printf("bug in simdselectFOR\n");
-                            return -3;
-                        }
-            }
-            /* we read back b1 128-bit vectors at "buffer" and write 128 integers at backbuffer */
-            simdunpackFOR(tmin,buffer, backbuffer, tb);/* uncompressed */
-            for (j = 0; j < SIMDBlockSize; ++j) {
-            	if (backbuffer[j] != datain[k * SIMDBlockSize + j]) {
-                    printf("bug in simdpackFOR\n");
-                    return -2;
-                }
-            }
-        }
-    }
-    free(buffer);
-    free(datain);
-    free(backbuffer);
-    printf("Code looks good.\n");
-    return 0;
-}
-#endif
-
-#define MAX 300
-int test_simdmaxbitsd1_length() {
-    uint32_t result, buffer[MAX + 1];
-    int i, j;
-
-    memset(&buffer[0], 0xff, sizeof(buffer));
-
-    /* this test creates buffers of different length; each buffer is
-     * initialized to result in the following deltas:
-     * length 1: 2
-     * length 2: 1 2
-     * length 3: 1 1 2
-     * length 4: 1 1 1 2
-     * length 5: 1 1 1 1 2
-     * etc. Each sequence's "maxbits" is 2. */
-    for (i = 0; i < MAX; i++) {
-      for (j = 0; j < i; j++)
-        buffer[j] = j + 1;
-      buffer[i] = i + 2;
-
-      result = simdmaxbitsd1_length(0, &buffer[0], i + 1);
-      if (result != 2) {
-        printf("simdmaxbitsd1_length: unexpected result %u in loop %d\n",
-                result, i);
-        return -1;
-      }
-    }
-    printf("simdmaxbitsd1_length: ok\n");
-    return 0;
-}
-
-int uint32_cmp(const void *a, const void *b)
-{
-    const uint32_t *ia = (const uint32_t *)a;
-    const uint32_t *ib = (const uint32_t *)b;
-    if(*ia < *ib)
-    	return -1;
-    else if (*ia > *ib)
-    	return 1;
-    return 0;
-}
-
-#ifdef __SSE4_1__
-int test_simdpackedsearch() {
-    uint32_t buffer[128];
-    uint32_t result = 0;
-    int b, i;
-    uint32_t init = 0;
-    __m128i initial = _mm_set1_epi32(init);
-
-    /* initialize the buffer */
-    for (i = 0; i < 128; i++)
-        buffer[i] = (uint32_t)(i + 1);
-
-    /* this test creates delta encoded buffers with different bits, then
-     * performs lower bound searches for each key */
-    for (b = 1; b <= 32; b++) {
-        uint32_t out[128];
-        /* delta-encode to 'i' bits */
-        simdpackwithoutmaskd1(init, buffer, (__m128i *)out, b);
-        initial = _mm_setzero_si128();
-        printf("simdsearchd1: %d bits\n", b);
-
-        /* now perform the searches */
-        initial = _mm_set1_epi32(init);
-        assert(simdsearchd1(&initial, (__m128i *)out, b, 0, &result) == 0);
-        assert(result > 0);
-
-        for (i = 1; i <= 128; i++) {
-        	initial = _mm_set1_epi32(init);
-            assert(simdsearchd1(&initial, (__m128i *)out, b,
-                                    (uint32_t)i, &result) == i - 1);
-            assert(result == (unsigned)i);
-        }
-        initial = _mm_set1_epi32(init);
-        assert(simdsearchd1(&initial, (__m128i *)out, b, 200, &result)
-                        == 128);
-        assert(result > 200);
-    }
-    printf("simdsearchd1: ok\n");
-    return 0;
-}
-
-int test_simdpackedsearchFOR() {
-    uint32_t buffer[128];
-    uint32_t result = 0;
-    int b;
-    uint32_t i;
-    uint32_t maxv, tmin, tmax, tb;
-    uint32_t out[128];
-
-    /* this test creates delta encoded buffers with different bits, then
-     * performs lower bound searches for each key */
-    for (b = 1; b <= 32; b++) {
-        /* initialize the buffer */
-    	maxv = (b == 32)
-    			? 0xFFFFFFFF
-    					: ((1U<<b) - 1);
-        for (i = 0; i < 128; i++)
-            buffer[i] = maxv * (i + 1) / 128;
-        simdmaxmin_length(buffer,SIMDBlockSize,&tmin,&tmax);
-   	    /* we compute the bit width */
-        tb  = bits(tmax - tmin);
-        /* delta-encode to 'i' bits */
-        simdpackFOR(tmin, buffer, (__m128i *)out, tb);
-        printf("simdsearchd1: %d bits\n", b);
-
-        /* now perform the searches */
-        for (i = 0; i < 128; i++) {
-        	assert(buffer[i] == simdselectFOR(tmin, (__m128i *)out, tb,i));
-        }
-        for (i = 0; i < 128; i++) {
-            int x = simdsearchwithlengthFOR(tmin, (__m128i *)out, tb,
-                                    128,buffer[i], &result) ;
-            assert(simdselectFOR(tmin, (__m128i *)out, tb,x) == buffer[x]);
-            assert(simdselectFOR(tmin, (__m128i *)out, tb,x) == result);
-            assert(buffer[x] == result);
-            assert(result == buffer[i]);
-            assert(buffer[x] == buffer[i]);
-        }
-    }
-    printf("simdsearchFOR: ok\n");
-    return 0;
-}
-
-int test_simdpackedsearch_advanced() {
-    uint32_t buffer[128];
-    uint32_t backbuffer[128];
-	uint32_t out[128];
-    uint32_t result = 0;
-    uint32_t b, i;
-    uint32_t init = 0;
-    __m128i initial = _mm_set1_epi32(init);
-
-
-    /* this test creates delta encoded buffers with different bits, then
-     * performs lower bound searches for each key */
-    for (b = 0; b <= 32; b++) {
-    	uint32_t prev = init;
-        /* initialize the buffer */
-        for (i = 0; i < 128; i++) {
-            buffer[i] =  ((uint32_t)(1431655765 * i + 0xFFFFFFFF)) ;
-            if(b < 32) buffer[i] %= (1<<b);
-        }
-
-        qsort(buffer,128, sizeof(uint32_t), uint32_cmp);
-
-        for (i = 0; i < 128; i++) {
-           buffer[i] = buffer[i] + prev;
-           prev = buffer[i];
-        }
-        for (i = 1; i < 128; i++) {
-        	if(buffer[i] < buffer[i-1] )
-        		buffer[i] = buffer[i-1];
-        }
-        assert(simdmaxbitsd1(init, buffer)<=b);
-        for (i = 0; i < 128; i++) {
-        	out[i] = 0; /* memset would do too */
-        }
-
-        /* delta-encode to 'i' bits */
-        simdpackwithoutmaskd1(init, buffer, (__m128i *)out, b);
-        simdunpackd1(init,  (__m128i *)out, backbuffer, b);
-
-        for (i = 0; i < 128; i++) {
-        	assert(buffer[i] == backbuffer[i]);
-        }
-
-        printf("advanced simdsearchd1: %d bits\n", b);
-
-        for (i = 0; i < 128; i++) {
-        	int pos;
-            initial = _mm_set1_epi32(init);
-        	pos = simdsearchd1(&initial, (__m128i *)out, b,
-                    buffer[i], &result);
-        	assert(pos == simdsearchwithlengthd1(init, (__m128i *)out, b, 128,
-                    buffer[i], &result));
-        	assert(buffer[pos] == buffer[i]);
-            if(pos > 0)
-            	assert(buffer[pos - 1] < buffer[i]);
-            assert(result == buffer[i]);
-        }
-        for (i = 0; i < 128; i++) {
-        	int pos;
-        	if(buffer[i] == 0) continue;
-        	initial = _mm_set1_epi32(init);
-        	pos = simdsearchd1(&initial, (__m128i *)out, b,
-                    buffer[i] - 1, &result);
-        	assert(pos == simdsearchwithlengthd1(init, (__m128i *)out, b, 128,
-                    buffer[i] - 1, &result));
-        	assert(buffer[pos] >= buffer[i]  - 1);
-            if(pos > 0)
-            	assert(buffer[pos - 1] < buffer[i]  - 1);
-            assert(result == buffer[pos]);
-        }
-		for (i = 0; i < 128; i++) {
-			int pos;
-			if (buffer[i] + 1 == 0)
-				continue;
-			initial = _mm_set1_epi32(init);
-			pos = simdsearchd1(&initial, (__m128i *) out, b,
-					buffer[i] + 1, &result);
-			assert(pos == simdsearchwithlengthd1(init, (__m128i *)out, b, 128,
-                    buffer[i] + 1, &result));
-			if(pos == 128) {
-				assert(buffer[i] == buffer[127]);
-			} else {
-			  assert(buffer[pos] >= buffer[i] + 1);
-			  if (pos > 0)
-				assert(buffer[pos - 1] < buffer[i] + 1);
-			  assert(result == buffer[pos]);
-			}
-		}
-    }
-    printf("advanced simdsearchd1: ok\n");
-    return 0;
-}
-
-int test_simdpackedselect() {
-    uint32_t buffer[128];
-    uint32_t initial = 33;
-    int b, i;
-
-    /* initialize the buffer */
-    for (i = 0; i < 128; i++)
-        buffer[i] = (uint32_t)(initial + i);
-
-    /* this test creates delta encoded buffers with different bits, then
-     * performs lower bound searches for each key */
-    for (b = 1; b <= 32; b++) {
-        uint32_t out[128];
-        /* delta-encode to 'i' bits */
-        simdpackwithoutmaskd1(initial, buffer, (__m128i *)out, b);
-
-        printf("simdselectd1: %d bits\n", b);
-
-        /* now perform the searches */
-        for (i = 0; i < 128; i++) {
-            assert(simdselectd1(initial, (__m128i *)out, b, (uint32_t)i)
-                            == initial + i);
-        }
-    }
-    printf("simdselectd1: ok\n");
-    return 0;
-}
-
-int test_simdpackedselect_advanced() {
-    uint32_t buffer[128];
-    uint32_t initial = 33;
-    uint32_t b;
-    int i;
-
-    /* this test creates delta encoded buffers with different bits, then
-     * performs lower bound searches for each key */
-    for (b = 0; b <= 32; b++) {
-        uint32_t prev = initial;
-    	uint32_t out[128];
-        /* initialize the buffer */
-        for (i = 0; i < 128; i++) {
-            buffer[i] =  ((uint32_t)(165576 * i)) ;
-            if(b < 32) buffer[i] %= (1<<b);
-        }
-        for (i = 0; i < 128; i++) {
-           buffer[i] = buffer[i] + prev;
-           prev = buffer[i];
-        }
-
-        for (i = 1; i < 128; i++) {
-        	if(buffer[i] < buffer[i-1] )
-        		buffer[i] = buffer[i-1];
-        }
-        assert(simdmaxbitsd1(initial, buffer)<=b);
-
-        for (i = 0; i < 128; i++) {
-        	out[i] = 0; /* memset would do too */
-        }
-
-        /* delta-encode to 'i' bits */
-        simdpackwithoutmaskd1(initial, buffer, (__m128i *)out, b);
-
-        printf("simdselectd1: %d bits\n", b);
-
-        /* now perform the searches */
-        for (i = 0; i < 128; i++) {
-        	uint32_t valretrieved = simdselectd1(initial, (__m128i *)out, b, (uint32_t)i);
-            assert(valretrieved == buffer[i]);
-        }
-    }
-    printf("advanced simdselectd1: ok\n");
-    return 0;
-}
-#endif
-
-
-int main() {
-    int r;
-    r =  testsetFOR();
-    if (r) {
-         printf("test failure 1\n");
-         return r;
-    }
-
-#ifdef __SSE4_1__
-    r =  testsetd1();
-    if (r) {
-         printf("test failure 2\n");
-         return r;
-    }
-#endif
-    r =  testset();
-    if (r) {
-         printf("test failure 3\n");
-         return r;
-    }
-
-    r = testshortFORpack();
-    if (r) {
-         printf("test failure 4\n");
-         return r;
-    }
-    r = testshortpack();
-    if (r) {
-         printf("test failure 5\n");
-         return r;
-    }
-    r = testlongpack();
-    if (r) {
-         printf("test failure 6\n");
-         return r;
-    }
-#ifdef __SSE4_1__
-    r = test_simdpackedsearchFOR();
-    if (r) {
-         printf("test failure 7\n");
-         return r;
-    }
-
-    r = testFOR();
-    if (r) {
-         printf("test failure 8\n");
-         return r;
-    }
-#endif
-#ifdef __AVX2__
-    r= testbabyavx();
-    if (r) {
-         printf("test failure baby avx\n");
-         return r;
-    }
-
-    r = testavx2();
-    if (r) {
-         printf("test failure 9 avx\n");
-         return r;
-    }
-#endif
-    r = test();
-    if (r) {
-         printf("test failure 9\n");
-         return r;
-    }
-
-    r = test_simdmaxbitsd1_length();
-    if (r) {
-         printf("test failure 10\n");
-         return r;
-    }
-#ifdef __SSE4_1__
-    r = test_simdpackedsearch();
-    if (r) {
-         printf("test failure 11\n");
-         return r;
-    }
-
-    r = test_simdpackedsearch_advanced();
-    if (r) {
-         printf("test failure 12\n");
-         return r;
-    }
-
-    r = test_simdpackedselect();
-    if (r) {
-         printf("test failure 13\n");
-         return r;
-    }
-
-    r = test_simdpackedselect_advanced();
-    if (r) {
-         printf("test failure 14\n");
-         return r;
-    }
-#endif
-    printf("All tests OK!\n");
-
-
-    return 0;
-}

cpp/simdcomp/tests/unit_chars.c

@@ -1,102 +0,0 @@
-/**
- * This code is released under a BSD License.
- */
-#include <stdio.h>
-#include <stdlib.h>
-#include <time.h>
-#include "simdcomp.h"
-
-
-#define get_random_char() (uint8_t)(rand() % 256);
-
-
-int main() {
-    int N = 5000 * SIMDBlockSize, gap;
-    __m128i * buffer = malloc(SIMDBlockSize * sizeof(uint32_t));
-    uint32_t * datain = malloc(N * sizeof(uint32_t));
-    uint32_t * backbuffer = malloc(SIMDBlockSize * sizeof(uint32_t));
-
-    srand(time(NULL));
-
-    for (gap = 1; gap <= 387420489; gap *= 3) {
-        int k;
-        printf(" gap = %u \n", gap);
-
-    /* simulate some random character string, don't care about endiannes */
-        for (k = 0; k < N; ++k) {
-        uint8_t _tmp[4];
- 
-            _tmp[0] = get_random_char();
-            _tmp[1] = get_random_char();
-            _tmp[2] = get_random_char();
-            _tmp[3] = get_random_char();
-
-            memmove(&datain[k], _tmp, 4);
-        }
-        for (k = 0; k * SIMDBlockSize < N; ++k) {
-            /*
-               First part works for general arrays (sorted or unsorted)
-            */
-            int j;
-               /* we compute the bit width */
-            const uint32_t b = maxbits(datain + k * SIMDBlockSize);
-            /* we read 128 integers at "datain + k * SIMDBlockSize" and
-               write b 128-bit vectors at "buffer" */
-            simdpackwithoutmask(datain + k * SIMDBlockSize, buffer, b);
-            /* we read back b1 128-bit vectors at "buffer" and write 128 integers at backbuffer */
-            simdunpack(buffer, backbuffer, b);/* uncompressed */
-            for (j = 0; j < SIMDBlockSize; ++j) {
-                uint8_t chars_back[4];
-                uint8_t chars_in[4];
-
-                memmove(chars_back, &backbuffer[j], 4);
-                memmove(chars_in, &datain[k * SIMDBlockSize + j], 4);
-
-                if (chars_in[0] != chars_back[0]
-                    || chars_in[1] != chars_back[1]
-                    || chars_in[2] != chars_back[2]
-                    || chars_in[3] != chars_back[3]) {
-                    printf("bug in simdpack\n");
-                    return -2;
-                }
-            }
-
-            {
-                /*
-                 next part assumes that the data is sorted (uses differential coding)
-                */
-                uint32_t offset = 0;
-                /* we compute the bit width */
-                const uint32_t b1 = simdmaxbitsd1(offset,
-                datain + k * SIMDBlockSize);
-                   /* we read 128 integers at "datain + k * SIMDBlockSize" and
-                  write b1 128-bit vectors at "buffer" */
-                   simdpackwithoutmaskd1(offset, datain + k * SIMDBlockSize, buffer,
-                b1);
-                   /* we read back b1 128-bit vectors at "buffer" and write 128 integers at backbuffer */
-                   simdunpackd1(offset, buffer, backbuffer, b1);
-                for (j = 0; j < SIMDBlockSize; ++j) {
-                    uint8_t chars_back[4];
-                    uint8_t chars_in[4];
-
-                    memmove(chars_back, &backbuffer[j], 4);
-                    memmove(chars_in, &datain[k * SIMDBlockSize + j], 4);
-
-                    if (chars_in[0] != chars_back[0]
-                        || chars_in[1] != chars_back[1]
-                        || chars_in[2] != chars_back[2]
-                        || chars_in[3] != chars_back[3]) {
-                        printf("bug in simdpack\n");
-                        return -3;
-                    }
-                }
-                offset = datain[k * SIMDBlockSize + SIMDBlockSize - 1];
-            }
-        }
-    }
-    free(buffer);
-    free(datain);
-    free(backbuffer);
-    printf("Code looks good.\n");
-    return 0;
-}

cpp/simdcomp_wrapper.c

@@ -1,42 +0,0 @@
-#include "simdcomp.h"
-#include "simdcomputil.h"
-
-// assumes datain has a size of 128 uint32
-// and that buffer is large enough to host the data.
-size_t compress_sorted(
-        const uint32_t* datain,
-        uint8_t* output,
-        const uint32_t offset) {
-    const uint32_t b = simdmaxbitsd1(offset, datain);
-    *output++ = b;
-    simdpackwithoutmaskd1(offset, datain, (__m128i *) output,  b);
-    return 1 + b * sizeof(__m128i);
-}
-
-// assumes datain has a size of 128 uint32
-// and that buffer is large enough to host the data.
-size_t uncompress_sorted(
-        const uint8_t* compressed_data, 
-        uint32_t* output, 
-        uint32_t offset) {
-    const uint32_t b = *compressed_data++;
-    simdunpackd1(offset, (__m128i *)compressed_data, output, b);
-    return 1 + b * sizeof(__m128i);
-}
-
-size_t compress_unsorted(
-        const uint32_t* datain,
-        uint8_t* output) {
-    const uint32_t b = maxbits(datain);
-    *output++ = b;
-    simdpackwithoutmask(datain, (__m128i *) output,  b);
-    return 1 + b * sizeof(__m128i);
-}
-
-size_t uncompress_unsorted(
-        const uint8_t* compressed_data, 
-        uint32_t* output) {
-    const uint32_t b = *compressed_data++;
-    simdunpack((__m128i *)compressed_data, output, b);
-    return 1 + b * sizeof(__m128i);
-}

cpp/streamvbyte/.gitignore

@@ -1,32 +0,0 @@
-# Object files
-*.o
-*.ko
-*.obj
-*.elf
-
-# Precompiled Headers
-*.gch
-*.pch
-
-# Libraries
-*.lib
-*.a
-*.la
-*.lo
-
-# Shared objects (inc. Windows DLLs)
-*.dll
-*.so
-*.so.*
-*.dylib
-
-# Executables
-*.exe
-*.out
-*.app
-*.i*86
-*.x86_64
-*.hex
-
-# Debug files
-*.dSYM/

cpp/streamvbyte/.travis.yml

@@ -1,7 +0,0 @@
-language: c
-sudo: false
-compiler:
-  - gcc
-  - clang
-
-script: make && ./unit

cpp/streamvbyte/LICENSE

@@ -1,202 +0,0 @@
-Apache License
-                           Version 2.0, January 2004
-                        http://www.apache.org/licenses/
-
-   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
-
-   1. Definitions.
-
-      "License" shall mean the terms and conditions for use, reproduction,
-      and distribution as defined by Sections 1 through 9 of this document.
-
-      "Licensor" shall mean the copyright owner or entity authorized by
-      the copyright owner that is granting the License.
-
-      "Legal Entity" shall mean the union of the acting entity and all
-      other entities that control, are controlled by, or are under common
-      control with that entity. For the purposes of this definition,
-      "control" means (i) the power, direct or indirect, to cause the
-      direction or management of such entity, whether by contract or
-      otherwise, or (ii) ownership of fifty percent (50%) or more of the
-      outstanding shares, or (iii) beneficial ownership of such entity.
-
-      "You" (or "Your") shall mean an individual or Legal Entity
-      exercising permissions granted by this License.
-
-      "Source" form shall mean the preferred form for making modifications,
-      including but not limited to software source code, documentation
-      source, and configuration files.
-
-      "Object" form shall mean any form resulting from mechanical
-      transformation or translation of a Source form, including but
-      not limited to compiled object code, generated documentation,
-      and conversions to other media types.
-
-      "Work" shall mean the work of authorship, whether in Source or
-      Object form, made available under the License, as indicated by a
-      copyright notice that is included in or attached to the work
-      (an example is provided in the Appendix below).
-
-      "Derivative Works" shall mean any work, whether in Source or Object
-      form, that is based on (or derived from) the Work and for which the
-      editorial revisions, annotations, elaborations, or other modifications
-      represent, as a whole, an original work of authorship. For the purposes
-      of this License, Derivative Works shall not include works that remain
-      separable from, or merely link (or bind by name) to the interfaces of,
-      the Work and Derivative Works thereof.
-
-      "Contribution" shall mean any work of authorship, including
-      the original version of the Work and any modifications or additions
-      to that Work or Derivative Works thereof, that is intentionally
-      submitted to Licensor for inclusion in the Work by the copyright owner
-      or by an individual or Legal Entity authorized to submit on behalf of
-      the copyright owner. For the purposes of this definition, "submitted"
-      means any form of electronic, verbal, or written communication sent
-      to the Licensor or its representatives, including but not limited to
-      communication on electronic mailing lists, source code control systems,
-      and issue tracking systems that are managed by, or on behalf of, the
-      Licensor for the purpose of discussing and improving the Work, but
-      excluding communication that is conspicuously marked or otherwise
-      designated in writing by the copyright owner as "Not a Contribution."
-
-      "Contributor" shall mean Licensor and any individual or Legal Entity
-      on behalf of whom a Contribution has been received by Licensor and
-      subsequently incorporated within the Work.
-
-   2. Grant of Copyright License. Subject to the terms and conditions of
-      this License, each Contributor hereby grants to You a perpetual,
-      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
-      copyright license to reproduce, prepare Derivative Works of,
-      publicly display, publicly perform, sublicense, and distribute the
-      Work and such Derivative Works in Source or Object form.
-
-   3. Grant of Patent License. Subject to the terms and conditions of
-      this License, each Contributor hereby grants to You a perpetual,
-      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
-      (except as stated in this section) patent license to make, have made,
-      use, offer to sell, sell, import, and otherwise transfer the Work,
-      where such license applies only to those patent claims licensable
-      by such Contributor that are necessarily infringed by their
-      Contribution(s) alone or by combination of their Contribution(s)
-      with the Work to which such Contribution(s) was submitted. If You
-      institute patent litigation against any entity (including a
-      cross-claim or counterclaim in a lawsuit) alleging that the Work
-      or a Contribution incorporated within the Work constitutes direct
-      or contributory patent infringement, then any patent licenses
-      granted to You under this License for that Work shall terminate
-      as of the date such litigation is filed.
-
-   4. Redistribution. You may reproduce and distribute copies of the
-      Work or Derivative Works thereof in any medium, with or without
-      modifications, and in Source or Object form, provided that You
-      meet the following conditions:
-
-      (a) You must give any other recipients of the Work or
-          Derivative Works a copy of this License; and
-
-      (b) You must cause any modified files to carry prominent notices
-          stating that You changed the files; and
-
-      (c) You must retain, in the Source form of any Derivative Works
-          that You distribute, all copyright, patent, trademark, and
-          attribution notices from the Source form of the Work,
-          excluding those notices that do not pertain to any part of
-          the Derivative Works; and
-
-      (d) If the Work includes a "NOTICE" text file as part of its
-          distribution, then any Derivative Works that You distribute must
-          include a readable copy of the attribution notices contained
-          within such NOTICE file, excluding those notices that do not
-          pertain to any part of the Derivative Works, in at least one
-          of the following places: within a NOTICE text file distributed
-          as part of the Derivative Works; within the Source form or
-          documentation, if provided along with the Derivative Works; or,
-          within a display generated by the Derivative Works, if and
-          wherever such third-party notices normally appear. The contents
-          of the NOTICE file are for informational purposes only and
-          do not modify the License. You may add Your own attribution
-          notices within Derivative Works that You distribute, alongside
-          or as an addendum to the NOTICE text from the Work, provided
-          that such additional attribution notices cannot be construed
-          as modifying the License.
-
-      You may add Your own copyright statement to Your modifications and
-      may provide additional or different license terms and conditions
-      for use, reproduction, or distribution of Your modifications, or
-      for any such Derivative Works as a whole, provided Your use,
-      reproduction, and distribution of the Work otherwise complies with
-      the conditions stated in this License.
-
-   5. Submission of Contributions. Unless You explicitly state otherwise,
-      any Contribution intentionally submitted for inclusion in the Work
-      by You to the Licensor shall be under the terms and conditions of
-      this License, without any additional terms or conditions.
-      Notwithstanding the above, nothing herein shall supersede or modify
-      the terms of any separate license agreement you may have executed
-      with Licensor regarding such Contributions.
-
-   6. Trademarks. This License does not grant permission to use the trade
-      names, trademarks, service marks, or product names of the Licensor,
-      except as required for reasonable and customary use in describing the
-      origin of the Work and reproducing the content of the NOTICE file.
-
-   7. Disclaimer of Warranty. Unless required by applicable law or
-      agreed to in writing, Licensor provides the Work (and each
-      Contributor provides its Contributions) on an "AS IS" BASIS,
-      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
-      implied, including, without limitation, any warranties or conditions
-      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
-      PARTICULAR PURPOSE. You are solely responsible for determining the
-      appropriateness of using or redistributing the Work and assume any
-      risks associated with Your exercise of permissions under this License.
-
-   8. Limitation of Liability. In no event and under no legal theory,
-      whether in tort (including negligence), contract, or otherwise,
-      unless required by applicable law (such as deliberate and grossly
-      negligent acts) or agreed to in writing, shall any Contributor be
-      liable to You for damages, including any direct, indirect, special,
-      incidental, or consequential damages of any character arising as a
-      result of this License or out of the use or inability to use the
-      Work (including but not limited to damages for loss of goodwill,
-      work stoppage, computer failure or malfunction, or any and all
-      other commercial damages or losses), even if such Contributor
-      has been advised of the possibility of such damages.
-
-   9. Accepting Warranty or Additional Liability. While redistributing
-      the Work or Derivative Works thereof, You may choose to offer,
-      and charge a fee for, acceptance of support, warranty, indemnity,
-      or other liability obligations and/or rights consistent with this
-      License. However, in accepting such obligations, You may act only
-      on Your own behalf and on Your sole responsibility, not on behalf
-      of any other Contributor, and only if You agree to indemnify,
-      defend, and hold each Contributor harmless for any liability
-      incurred by, or claims asserted against, such Contributor by reason
-      of your accepting any such warranty or additional liability.
-
-   END OF TERMS AND CONDITIONS
-
-   APPENDIX: How to apply the Apache License to your work.
-
-      To apply the Apache License to your work, attach the following
-      boilerplate notice, with the fields enclosed by brackets "{}"
-      replaced with your own identifying information. (Don't include
-      the brackets!)  The text should be enclosed in the appropriate
-      comment syntax for the file format. We also recommend that a
-      file or class name and description of purpose be included on the
-      same "printed page" as the copyright notice for easier
-      identification within third-party archives.
-
-   Copyright {yyyy} {name of copyright owner}
-
-   Licensed under the Apache License, Version 2.0 (the "License");
-   you may not use this file except in compliance with the License.
-   You may obtain a copy of the License at
-
-       http://www.apache.org/licenses/LICENSE-2.0
-
-   Unless required by applicable law or agreed to in writing, software
-   distributed under the License is distributed on an "AS IS" BASIS,
-   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-   See the License for the specific language governing permissions and
-   limitations under the License.
-

cpp/streamvbyte/README.md

@@ -1,60 +0,0 @@
-streamvbyte
-===========
-[![Build Status](https://travis-ci.org/lemire/streamvbyte.png)](https://travis-ci.org/lemire/streamvbyte)
-
-StreamVByte is a new integer compression technique that applies SIMD instructions (vectorization) to
-Google's Group Varint approach. The net result is faster than other byte-oriented compression
-techniques.
-
-The approach is patent-free, the code is available under the Apache License.
-
-
-It includes fast differential coding.
-
-It assumes a recent Intel processor (e.g., haswell or better) .
-
-The code should build using most standard-compliant C99 compilers. The provided makefile
-expects a Linux-like system.
-
-
-Usage:
-
-      make
-      ./unit
-
-See example.c for an example.
-
-Short code sample:
-```C
-// suppose that datain is an array of uint32_t integers
-size_t compsize = streamvbyte_encode(datain, N, compressedbuffer); // encoding
-// here the result is stored in compressedbuffer using compsize bytes
-streamvbyte_decode(compressedbuffer, recovdata, N); // decoding (fast)
-```
-
-If the values are sorted, then it might be preferable to use differential coding:
-```C
-// suppose that datain is an array of uint32_t integers
-size_t compsize = streamvbyte_delta_encode(datain, N, compressedbuffer,0); // encoding
-// here the result is stored in compressedbuffer using compsize bytes
-streamvbyte_delta_decode(compressedbuffer, recovdata, N,0); // decoding (fast)
-```
-You have to know how many integers were coded when you decompress. You can store this 
-information along with the compressed stream.
-
-See also
---------
-* SIMDCompressionAndIntersection: A C++ library to compress and intersect sorted lists of integers using SIMD instructions https://github.com/lemire/SIMDCompressionAndIntersect
-* The FastPFOR C++ library : Fast integer compression https://github.com/lemire/FastPFor
-* High-performance dictionary coding https://github.com/lemire/dictionary
-* LittleIntPacker: C library to pack and unpack short arrays of integers as fast as possible https://github.com/lemire/LittleIntPacker
-* The SIMDComp library: A simple C library for compressing lists of integers using binary packing https://github.com/lemire/simdcomp
-* MaskedVByte: Fast decoder for VByte-compressed integers https://github.com/lemire/MaskedVByte
-* CSharpFastPFOR: A C#  integer compression library  https://github.com/Genbox/CSharpFastPFOR
-* JavaFastPFOR: A java integer compression library https://github.com/lemire/JavaFastPFOR
-* Encoding: Integer Compression Libraries for Go https://github.com/zhenjl/encoding
-* FrameOfReference is a C++ library dedicated to frame-of-reference (FOR) compression: https://github.com/lemire/FrameOfReference
-* libvbyte: A fast implementation for varbyte 32bit/64bit integer compression https://github.com/cruppstahl/libvbyte
-* TurboPFor is a C library that offers lots of interesting optimizations. Well worth checking! (GPL license) https://github.com/powturbo/TurboPFor
-* Oroch is a C++ library that offers a usable API (MIT license) https://github.com/ademakov/Oroch
-

cpp/streamvbyte/example.c

@@ -1,24 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <assert.h>
-
-#include "streamvbyte.h"
-
-int main() {
-	int N = 5000;
-	uint32_t * datain = malloc(N * sizeof(uint32_t));
-	uint8_t * compressedbuffer = malloc(N * sizeof(uint32_t));
-	uint32_t * recovdata = malloc(N * sizeof(uint32_t));
-	for (int k = 0; k < N; ++k)
-		datain[k] = 120;
-	size_t compsize = streamvbyte_encode(datain, N, compressedbuffer); // encoding
-	// here the result is stored in compressedbuffer using compsize bytes
-	size_t compsize2 = streamvbyte_decode(compressedbuffer, recovdata,
-					N); // decoding (fast)
-	assert(compsize == compsize2);
-	free(datain);
-	free(compressedbuffer);
-	free(recovdata);
-	printf("Compressed %d integers down to %d bytes.\n",N,(int) compsize);
-	return 0;
-}

cpp/streamvbyte/include/streamvbyte.h

@@ -1,19 +0,0 @@
-
-#ifndef VARINTDECODE_H_
-#define VARINTDECODE_H_
-#define __STDC_FORMAT_MACROS
-#include <inttypes.h>
-#include <stdint.h>// please use a C99-compatible compiler
-#include <stddef.h>
-
-
-// Encode an array of a given length read from in to bout in varint format.
-// Returns the number of bytes written.
-size_t streamvbyte_encode(const uint32_t *in, uint32_t length, uint8_t *out);
-
-// Read "length" 32-bit integers in varint format from in, storing the result in out.
-// Returns the number of bytes read.
-size_t streamvbyte_decode(const uint8_t* in, uint32_t* out, uint32_t length);
-
-
-#endif /* VARINTDECODE_H_ */

cpp/streamvbyte/include/streamvbytedelta.h

@@ -1,24 +0,0 @@
-/*
- * streamvbytedelta.h
- *
- *  Created on: Apr 14, 2016
- *      Author: lemire
- */
-
-#ifndef INCLUDE_STREAMVBYTEDELTA_H_
-#define INCLUDE_STREAMVBYTEDELTA_H_
-
-
-// Encode an array of a given length read from in to bout in StreamVByte format.
-// Returns the number of bytes written.
-// this version uses differential coding (coding differences between values) starting at prev (you can often set prev to zero)
-size_t streamvbyte_delta_encode(const uint32_t *in, uint32_t length, uint8_t *out, uint32_t  prev);
-
-// Read "length" 32-bit integers in StreamVByte format from in, storing the result in out.
-// Returns the number of bytes read.
-// this version uses differential coding (coding differences between values) starting at prev (you can often set prev to zero)
-size_t streamvbyte_delta_decode(const uint8_t* in, uint32_t* out, uint32_t length, uint32_t  prev);
-
-
-
-#endif /* INCLUDE_STREAMVBYTEDELTA_H_ */

cpp/streamvbyte/makefile

@@ -1,58 +0,0 @@
-# minimalist makefile
-.SUFFIXES:
-#
-.SUFFIXES: .cpp .o .c .h
-
-CFLAGS = -fPIC -march=native -std=c99 -O3 -Wall -Wextra -pedantic -Wshadow
-LDFLAGS = -shared
-LIBNAME=libstreamvbyte.so.0.0.1
-all:  unit $(LIBNAME)
-test:
-	./unit
-install: $(OBJECTS)
-	cp $(LIBNAME) /usr/local/lib
-	ln -s /usr/local/lib/$(LIBNAME) /usr/local/lib/libstreamvbyte.so
-	ldconfig
-	cp $(HEADERS) /usr/local/include
-
-
-
-HEADERS=./include/streamvbyte.h ./include/streamvbytedelta.h 
-
-uninstall:
-	for h in $(HEADERS) ; do rm  /usr/local/$$h; done
-	rm  /usr/local/lib/$(LIBNAME)
-	rm /usr/local/lib/libstreamvbyte.so
-	ldconfig
-
-
-OBJECTS= streamvbyte.o streamvbytedelta.o
-
-
-
-streamvbytedelta.o: ./src/streamvbytedelta.c $(HEADERS)
-	$(CC) $(CFLAGS) -c ./src/streamvbytedelta.c -Iinclude
-
-
-streamvbyte.o: ./src/streamvbyte.c $(HEADERS)
-	$(CC) $(CFLAGS) -c ./src/streamvbyte.c -Iinclude
-
-
-
-$(LIBNAME): $(OBJECTS)
-	$(CC) $(CFLAGS) -o $(LIBNAME) $(OBJECTS)  $(LDFLAGS)
-
-
-
-
-example: ./example.c    $(HEADERS) $(OBJECTS)
-	$(CC) $(CFLAGS) -o example ./example.c -Iinclude  $(OBJECTS)
-
-unit: ./tests/unit.c    $(HEADERS) $(OBJECTS)
-	$(CC) $(CFLAGS) -o unit ./tests/unit.c -Iinclude  $(OBJECTS)
-
-dynunit: ./tests/unit.c    $(HEADERS) $(LIBNAME)
-	$(CC) $(CFLAGS) -o dynunit ./tests/unit.c -Iinclude  -lstreamvbyte
-
-clean:
-	rm -f unit *.o $(LIBNAME) example

cpp/streamvbyte/src/streamvbyte.c

@@ -1,495 +0,0 @@
-#include "streamvbyte.h"
-#if defined(_MSC_VER)
-     /* Microsoft C/C++-compatible compiler */
-     #include <intrin.h>
-#elif defined(__GNUC__) && (defined(__x86_64__) || defined(__i386__))
-     /* GCC-compatible compiler, targeting x86/x86-64 */
-     #include <x86intrin.h>
-#elif defined(__GNUC__) && defined(__ARM_NEON__)
-     /* GCC-compatible compiler, targeting ARM with NEON */
-     #include <arm_neon.h>
-#elif defined(__GNUC__) && defined(__IWMMXT__)
-     /* GCC-compatible compiler, targeting ARM with WMMX */
-     #include <mmintrin.h>
-#elif (defined(__GNUC__) || defined(__xlC__)) && (defined(__VEC__) || defined(__ALTIVEC__))
-     /* XLC or GCC-compatible compiler, targeting PowerPC with VMX/VSX */
-     #include <altivec.h>
-#elif defined(__GNUC__) && defined(__SPE__)
-     /* GCC-compatible compiler, targeting PowerPC with SPE */
-     #include <spe.h>
-#endif
-
-static uint8_t lengthTable[256] = { 4, 5, 6, 7, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9,
-		10, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9, 10, 8, 9, 10, 11, 6, 7, 8, 9, 7, 8,
-		9, 10, 8, 9, 10, 11, 9, 10, 11, 12, 7, 8, 9, 10, 8, 9, 10, 11, 9, 10,
-		11, 12, 10, 11, 12, 13, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9, 10, 8, 9, 10,
-		11, 6, 7, 8, 9, 7, 8, 9, 10, 8, 9, 10, 11, 9, 10, 11, 12, 7, 8, 9, 10,
-		8, 9, 10, 11, 9, 10, 11, 12, 10, 11, 12, 13, 8, 9, 10, 11, 9, 10, 11,
-		12, 10, 11, 12, 13, 11, 12, 13, 14, 6, 7, 8, 9, 7, 8, 9, 10, 8, 9, 10,
-		11, 9, 10, 11, 12, 7, 8, 9, 10, 8, 9, 10, 11, 9, 10, 11, 12, 10, 11, 12,
-		13, 8, 9, 10, 11, 9, 10, 11, 12, 10, 11, 12, 13, 11, 12, 13, 14, 9, 10,
-		11, 12, 10, 11, 12, 13, 11, 12, 13, 14, 12, 13, 14, 15, 7, 8, 9, 10, 8,
-		9, 10, 11, 9, 10, 11, 12, 10, 11, 12, 13, 8, 9, 10, 11, 9, 10, 11, 12,
-		10, 11, 12, 13, 11, 12, 13, 14, 9, 10, 11, 12, 10, 11, 12, 13, 11, 12,
-		13, 14, 12, 13, 14, 15, 10, 11, 12, 13, 11, 12, 13, 14, 12, 13, 14, 15,
-		13, 14, 15, 16 };
-
-static uint8_t shuffleTable[256][16] = { { 0, -1, -1, -1, 1, -1, -1, -1, 2, -1,
-		-1, -1, 3, -1, -1, -1 }, // 1111
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, -1, -1, -1, 4, -1, -1, -1 },  // 2111
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, -1, -1, -1, 5, -1, -1, -1 },   // 3111
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, -1, -1, -1, 6, -1, -1, -1 },    // 4111
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, -1, -1, -1, 4, -1, -1, -1 },  // 1211
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, -1, -1, -1, 5, -1, -1, -1 },   // 2211
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, -1, -1, -1, 6, -1, -1, -1 },    // 3211
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, -1, -1, -1, 7, -1, -1, -1 },     // 4211
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, -1, -1, -1, 5, -1, -1, -1 },   // 1311
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, -1, -1, -1, 6, -1, -1, -1 },    // 2311
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, -1, -1, -1, 7, -1, -1, -1 },     // 3311
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, -1, -1, -1, 8, -1, -1, -1 },      // 4311
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, -1, -1, -1, 6, -1, -1, -1 },    // 1411
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, -1, -1, -1, 7, -1, -1, -1 },     // 2411
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, -1, -1, -1, 8, -1, -1, -1 },      // 3411
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, -1, -1, -1, 9, -1, -1, -1 },       // 4411
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, -1, -1, 4, -1, -1, -1 },  // 1121
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, -1, -1, 5, -1, -1, -1 },   // 2121
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, -1, -1, 6, -1, -1, -1 },    // 3121
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, -1, -1, 7, -1, -1, -1 },     // 4121
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, -1, -1, 5, -1, -1, -1 },   // 1221
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, -1, -1, 6, -1, -1, -1 },    // 2221
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, -1, -1, 7, -1, -1, -1 },     // 3221
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, -1, -1, 8, -1, -1, -1 },      // 4221
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, -1, -1, 6, -1, -1, -1 },    // 1321
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, -1, -1, 7, -1, -1, -1 },     // 2321
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, -1, -1, 8, -1, -1, -1 },      // 3321
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, -1, -1, 9, -1, -1, -1 },       // 4321
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, -1, -1, 7, -1, -1, -1 },     // 1421
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, -1, -1, 8, -1, -1, -1 },      // 2421
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, -1, -1, 9, -1, -1, -1 },       // 3421
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, 10, -1, -1, -1 },       // 4421
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, -1, 5, -1, -1, -1 },   // 1131
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, -1, 6, -1, -1, -1 },    // 2131
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, -1, 7, -1, -1, -1 },     // 3131
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, -1, 8, -1, -1, -1 },      // 4131
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, -1, 6, -1, -1, -1 },    // 1231
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, -1, 7, -1, -1, -1 },     // 2231
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, -1, 8, -1, -1, -1 },      // 3231
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, -1, 9, -1, -1, -1 },       // 4231
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, -1, 7, -1, -1, -1 },     // 1331
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, -1, 8, -1, -1, -1 },      // 2331
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, -1, 9, -1, -1, -1 },       // 3331
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, -1, 10, -1, -1, -1 },       // 4331
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, -1, 8, -1, -1, -1 },      // 1431
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, -1, 9, -1, -1, -1 },       // 2431
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, -1, 10, -1, -1, -1 },       // 3431
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, 11, -1, -1, -1 },       // 4431
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, 5, 6, -1, -1, -1 },    // 1141
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, 6, 7, -1, -1, -1 },     // 2141
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, 7, 8, -1, -1, -1 },      // 3141
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, 8, 9, -1, -1, -1 },       // 4141
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, 6, 7, -1, -1, -1 },     // 1241
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, 7, 8, -1, -1, -1 },      // 2241
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, 8, 9, -1, -1, -1 },       // 3241
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, 9, 10, -1, -1, -1 },       // 4241
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, 7, 8, -1, -1, -1 },      // 1341
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, 8, 9, -1, -1, -1 },       // 2341
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, 9, 10, -1, -1, -1 },       // 3341
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, 10, 11, -1, -1, -1 },       // 4341
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, -1 },       // 1441
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, -1, -1 },       // 2441
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, 10, 11, -1, -1, -1 },       // 3441
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, -1, -1, -1 },       // 4441
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, -1, -1, -1, 3, 4, -1, -1 },  // 1112
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, -1, -1, -1, 4, 5, -1, -1 },   // 2112
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, -1, -1, -1, 5, 6, -1, -1 },    // 3112
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, -1, -1, -1, 6, 7, -1, -1 },     // 4112
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, -1, -1, -1, 4, 5, -1, -1 },   // 1212
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, -1, -1, -1, 5, 6, -1, -1 },    // 2212
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, -1, -1, -1, 6, 7, -1, -1 },     // 3212
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, -1, -1, -1, 7, 8, -1, -1 },      // 4212
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, -1, -1, -1, 5, 6, -1, -1 },    // 1312
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, -1, -1, -1, 6, 7, -1, -1 },     // 2312
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, -1, -1, -1, 7, 8, -1, -1 },      // 3312
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, -1, -1, -1, 8, 9, -1, -1 },       // 4312
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, -1, -1, -1, 6, 7, -1, -1 },     // 1412
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, -1, -1, -1, 7, 8, -1, -1 },      // 2412
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, -1, -1, -1, 8, 9, -1, -1 },       // 3412
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, -1, -1, -1, 9, 10, -1, -1 },       // 4412
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, -1, -1, 4, 5, -1, -1 },   // 1122
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, -1, -1, 5, 6, -1, -1 },    // 2122
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, -1, -1, 6, 7, -1, -1 },     // 3122
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, -1, -1, 7, 8, -1, -1 },      // 4122
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, -1, -1, 5, 6, -1, -1 },    // 1222
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, -1, -1, 6, 7, -1, -1 },     // 2222
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, -1, -1, 7, 8, -1, -1 },      // 3222
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, -1, -1, 8, 9, -1, -1 },       // 4222
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, -1, -1, 6, 7, -1, -1 },     // 1322
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, -1, -1, 7, 8, -1, -1 },      // 2322
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, -1, -1, 8, 9, -1, -1 },       // 3322
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, -1, -1, 9, 10, -1, -1 },       // 4322
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, -1, -1, 7, 8, -1, -1 },      // 1422
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, -1, -1, 8, 9, -1, -1 },       // 2422
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, -1, -1, 9, 10, -1, -1 },       // 3422
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, 10, 11, -1, -1 },       // 4422
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, -1, 5, 6, -1, -1 },    // 1132
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, -1, 6, 7, -1, -1 },     // 2132
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, -1, 7, 8, -1, -1 },      // 3132
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, -1, 8, 9, -1, -1 },       // 4132
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, -1, 6, 7, -1, -1 },     // 1232
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, -1, 7, 8, -1, -1 },      // 2232
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, -1, 8, 9, -1, -1 },       // 3232
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, -1, 9, 10, -1, -1 },       // 4232
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, -1, 7, 8, -1, -1 },      // 1332
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, -1, 8, 9, -1, -1 },       // 2332
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, -1, 9, 10, -1, -1 },       // 3332
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, -1, 10, 11, -1, -1 },       // 4332
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, -1, 8, 9, -1, -1 },       // 1432
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, -1, 9, 10, -1, -1 },       // 2432
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, -1, 10, 11, -1, -1 },       // 3432
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, 11, 12, -1, -1 },       // 4432
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, 5, 6, 7, -1, -1 },     // 1142
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, 6, 7, 8, -1, -1 },      // 2142
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, 7, 8, 9, -1, -1 },       // 3142
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, 8, 9, 10, -1, -1 },       // 4142
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, 6, 7, 8, -1, -1 },      // 1242
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, 7, 8, 9, -1, -1 },       // 2242
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, 8, 9, 10, -1, -1 },       // 3242
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, 9, 10, 11, -1, -1 },       // 4242
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, 7, 8, 9, -1, -1 },       // 1342
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, 8, 9, 10, -1, -1 },       // 2342
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, 9, 10, 11, -1, -1 },       // 3342
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, 10, 11, 12, -1, -1 },       // 4342
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, -1 },       // 1442
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, -1, -1 },       // 2442
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, -1, -1 },       // 3442
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, -1, -1 },       // 4442
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, -1, -1, -1, 3, 4, 5, -1 },   // 1113
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, -1, -1, -1, 4, 5, 6, -1 },    // 2113
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, -1, -1, -1, 5, 6, 7, -1 },     // 3113
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, -1, -1, -1, 6, 7, 8, -1 },      // 4113
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, -1, -1, -1, 4, 5, 6, -1 },    // 1213
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, -1, -1, -1, 5, 6, 7, -1 },     // 2213
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, -1, -1, -1, 6, 7, 8, -1 },      // 3213
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, -1, -1, -1, 7, 8, 9, -1 },       // 4213
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, -1, -1, -1, 5, 6, 7, -1 },     // 1313
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, -1, -1, -1, 6, 7, 8, -1 },      // 2313
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, -1, -1, -1, 7, 8, 9, -1 },       // 3313
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, -1, -1, -1, 8, 9, 10, -1 },       // 4313
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, -1, -1, -1, 6, 7, 8, -1 },      // 1413
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, -1, -1, -1, 7, 8, 9, -1 },       // 2413
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, -1, -1, -1, 8, 9, 10, -1 },       // 3413
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, -1, -1, -1, 9, 10, 11, -1 },       // 4413
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, -1, -1, 4, 5, 6, -1 },    // 1123
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, -1, -1, 5, 6, 7, -1 },     // 2123
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, -1, -1, 6, 7, 8, -1 },      // 3123
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, -1, -1, 7, 8, 9, -1 },       // 4123
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, -1, -1, 5, 6, 7, -1 },     // 1223
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, -1, -1, 6, 7, 8, -1 },      // 2223
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, -1, -1, 7, 8, 9, -1 },       // 3223
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, -1, -1, 8, 9, 10, -1 },       // 4223
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, -1, -1, 6, 7, 8, -1 },      // 1323
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, -1, -1, 7, 8, 9, -1 },       // 2323
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, -1, -1, 8, 9, 10, -1 },       // 3323
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, -1, -1, 9, 10, 11, -1 },       // 4323
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, -1, -1, 7, 8, 9, -1 },       // 1423
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, -1, -1, 8, 9, 10, -1 },       // 2423
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, -1, -1, 9, 10, 11, -1 },       // 3423
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, 10, 11, 12, -1 },       // 4423
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, -1, 5, 6, 7, -1 },     // 1133
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, -1, 6, 7, 8, -1 },      // 2133
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, -1, 7, 8, 9, -1 },       // 3133
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, -1, 8, 9, 10, -1 },       // 4133
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, -1, 6, 7, 8, -1 },      // 1233
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, -1, 7, 8, 9, -1 },       // 2233
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, -1, 8, 9, 10, -1 },       // 3233
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, -1, 9, 10, 11, -1 },       // 4233
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, -1, 7, 8, 9, -1 },       // 1333
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, -1, 8, 9, 10, -1 },       // 2333
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, -1, 9, 10, 11, -1 },       // 3333
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, -1, 10, 11, 12, -1 },       // 4333
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, -1, 8, 9, 10, -1 },       // 1433
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, -1, 9, 10, 11, -1 },       // 2433
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, -1, 10, 11, 12, -1 },       // 3433
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, 11, 12, 13, -1 },       // 4433
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, 5, 6, 7, 8, -1 },      // 1143
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, 6, 7, 8, 9, -1 },       // 2143
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, 7, 8, 9, 10, -1 },       // 3143
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, 8, 9, 10, 11, -1 },       // 4143
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, 6, 7, 8, 9, -1 },       // 1243
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, 7, 8, 9, 10, -1 },       // 2243
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, 8, 9, 10, 11, -1 },       // 3243
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, 9, 10, 11, 12, -1 },       // 4243
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, 7, 8, 9, 10, -1 },       // 1343
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, 8, 9, 10, 11, -1 },       // 2343
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, 9, 10, 11, 12, -1 },       // 3343
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, 10, 11, 12, 13, -1 },       // 4343
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, -1 },       // 1443
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, -1 },       // 2443
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, -1 },       // 3443
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, -1 },       // 4443
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, -1, -1, -1, 3, 4, 5, 6 },    // 1114
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, -1, -1, -1, 4, 5, 6, 7 },     // 2114
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, -1, -1, -1, 5, 6, 7, 8 },      // 3114
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, -1, -1, -1, 6, 7, 8, 9 },       // 4114
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, -1, -1, -1, 4, 5, 6, 7 },     // 1214
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, -1, -1, -1, 5, 6, 7, 8 },      // 2214
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, -1, -1, -1, 6, 7, 8, 9 },       // 3214
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, -1, -1, -1, 7, 8, 9, 10 },       // 4214
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, -1, -1, -1, 5, 6, 7, 8 },      // 1314
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, -1, -1, -1, 6, 7, 8, 9 },       // 2314
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, -1, -1, -1, 7, 8, 9, 10 },       // 3314
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, -1, -1, -1, 8, 9, 10, 11 },       // 4314
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, -1, -1, -1, 6, 7, 8, 9 },       // 1414
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, -1, -1, -1, 7, 8, 9, 10 },       // 2414
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, -1, -1, -1, 8, 9, 10, 11 },       // 3414
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, -1, -1, -1, 9, 10, 11, 12 },       // 4414
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, -1, -1, 4, 5, 6, 7 },     // 1124
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, -1, -1, 5, 6, 7, 8 },      // 2124
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, -1, -1, 6, 7, 8, 9 },       // 3124
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, -1, -1, 7, 8, 9, 10 },       // 4124
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, -1, -1, 5, 6, 7, 8 },      // 1224
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, -1, -1, 6, 7, 8, 9 },       // 2224
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, -1, -1, 7, 8, 9, 10 },       // 3224
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, -1, -1, 8, 9, 10, 11 },       // 4224
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, -1, -1, 6, 7, 8, 9 },       // 1324
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, -1, -1, 7, 8, 9, 10 },       // 2324
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, -1, -1, 8, 9, 10, 11 },       // 3324
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, -1, -1, 9, 10, 11, 12 },       // 4324
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, -1, -1, 7, 8, 9, 10 },       // 1424
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, -1, -1, 8, 9, 10, 11 },       // 2424
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, -1, -1, 9, 10, 11, 12 },       // 3424
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, 10, 11, 12, 13 },       // 4424
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, -1, 5, 6, 7, 8 },      // 1134
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, -1, 6, 7, 8, 9 },       // 2134
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, -1, 7, 8, 9, 10 },       // 3134
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, -1, 8, 9, 10, 11 },       // 4134
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, -1, 6, 7, 8, 9 },       // 1234
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, -1, 7, 8, 9, 10 },       // 2234
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, -1, 8, 9, 10, 11 },       // 3234
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, -1, 9, 10, 11, 12 },       // 4234
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, -1, 7, 8, 9, 10 },       // 1334
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, -1, 8, 9, 10, 11 },       // 2334
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, -1, 9, 10, 11, 12 },       // 3334
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, -1, 10, 11, 12, 13 },       // 4334
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, -1, 8, 9, 10, 11 },       // 1434
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, -1, 9, 10, 11, 12 },       // 2434
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, -1, 10, 11, 12, 13 },       // 3434
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, 11, 12, 13, 14 },       // 4434
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, 5, 6, 7, 8, 9 },       // 1144
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, 6, 7, 8, 9, 10 },       // 2144
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, 7, 8, 9, 10, 11 },       // 3144
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, 8, 9, 10, 11, 12 },       // 4144
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, 6, 7, 8, 9, 10 },       // 1244
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, 7, 8, 9, 10, 11 },       // 2244
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, 8, 9, 10, 11, 12 },       // 3244
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, 9, 10, 11, 12, 13 },       // 4244
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, 7, 8, 9, 10, 11 },       // 1344
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, 8, 9, 10, 11, 12 },       // 2344
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, 9, 10, 11, 12, 13 },       // 3344
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, 10, 11, 12, 13, 14 },       // 4344
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 },       // 1444
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13 },       // 2444
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 },       // 3444
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 }        // 4444
-};
-
-static uint8_t _encode_data(uint32_t val, uint8_t *__restrict__ *dataPtrPtr) {
-	uint8_t *dataPtr = *dataPtrPtr;
-	uint8_t code;
-
-	if (val < (1 << 8)) { // 1 byte
-		*dataPtr = (uint8_t)(val);
-		*dataPtrPtr += 1;
-		code = 0;
-	} else if (val < (1 << 16)) { // 2 bytes
-		*(uint16_t *) dataPtr = (uint16_t)(val);
-		*dataPtrPtr += 2;
-		code = 1;
-	} else if (val < (1 << 24)) { // 3 bytes
-		*(uint16_t *) dataPtr = (uint16_t)(val);
-		*(dataPtr + 2) = (uint8_t)(val >> 16);
-		*dataPtrPtr += 3;
-		code = 2;
-	} else { // 4 bytes
-		*(uint32_t *) dataPtr = val;
-		*dataPtrPtr += 4;
-		code = 3;
-	}
-
-	return code;
-}
-
-static uint8_t *svb_encode_scalar(const uint32_t *in,
-		uint8_t *__restrict__ keyPtr, uint8_t *__restrict__ dataPtr,
-		uint32_t count) {
-	if (count == 0)
-		return dataPtr; // exit immediately if no data
-
-	uint8_t shift = 0; // cycles 0, 2, 4, 6, 0, 2, 4, 6, ...
-	uint8_t key = 0;
-	for (uint32_t c = 0; c < count; c++) {
-		if (shift == 8) {
-			shift = 0;
-			*keyPtr++ = key;
-			key = 0;
-		}
-		uint32_t val = in[c];
-		uint8_t code = _encode_data(val, &dataPtr);
-		key |= code << shift;
-		shift += 2;
-	}
-
-	*keyPtr = key;  // write last key (no increment needed)
-	return dataPtr; // pointer to first unused data byte
-}
-
-// Encode an array of a given length read from in to bout in streamvbyte format.
-// Returns the number of bytes written.
-size_t streamvbyte_encode(const uint32_t *in, uint32_t count, uint8_t *out) {
-	uint8_t *keyPtr = out;
-	uint32_t keyLen = (count + 3) / 4; // 2-bits rounded to full byte
-	uint8_t *dataPtr = keyPtr + keyLen; // variable byte data after all keys
-	return svb_encode_scalar(in, keyPtr, dataPtr, count) - out;
-}
-
-static inline __m128i _decode_avx(uint32_t key,
-		const uint8_t *__restrict__ *dataPtrPtr) {
-	uint8_t len = lengthTable[key];
-	__m128i Data = _mm_loadu_si128((__m128i *) *dataPtrPtr);
-	__m128i Shuf = *(__m128i *) &shuffleTable[key];
-
-	Data = _mm_shuffle_epi8(Data, Shuf);
-	*dataPtrPtr += len;
-	return Data;
-}
-
-static inline void _write_avx(uint32_t *out, __m128i Vec) {
-	_mm_storeu_si128((__m128i *) out, Vec);
-}
-
-static inline uint32_t _decode_data(const uint8_t **dataPtrPtr, uint8_t code) {
-	const uint8_t *dataPtr = *dataPtrPtr;
-	uint32_t val;
-
-	if (code == 0) { // 1 byte
-		val = (uint32_t) * dataPtr;
-		dataPtr += 1;
-	} else if (code == 1) { // 2 bytes
-		val = (uint32_t) * (uint16_t *) dataPtr;
-		dataPtr += 2;
-	} else if (code == 2) { // 3 bytes
-		val = (uint32_t) * (uint16_t *) dataPtr;
-		val |= *(dataPtr + 2) << 16;
-		dataPtr += 3;
-	} else {                      // code == 3
-		val = *(uint32_t *) dataPtr; // 4 bytes
-		dataPtr += 4;
-	}
-
-	*dataPtrPtr = dataPtr;
-	return val;
-}
-static const uint8_t *svb_decode_scalar(uint32_t *outPtr, const uint8_t *keyPtr,
-		const uint8_t *dataPtr, uint32_t count) {
-	if (count == 0)
-		return dataPtr; // no reads or writes if no data
-
-	uint8_t shift = 0;
-	uint32_t key = *keyPtr++;
-	for (uint32_t c = 0; c < count; c++) {
-		if (shift == 8) {
-			shift = 0;
-			key = *keyPtr++;
-		}
-		uint32_t val = _decode_data(&dataPtr, (key >> shift) & 0x3);
-		*outPtr++ = val;
-		shift += 2;
-	}
-
-	return dataPtr; // pointer to first unused byte after end
-}
-
-const uint8_t *svb_decode_avx_simple(uint32_t *out,
-		const uint8_t *__restrict__ keyPtr, const uint8_t *__restrict__ dataPtr,
-		uint64_t count) {
-
-	uint64_t keybytes = count / 4; // number of key bytes
-	__m128i Data;
-	if (keybytes >= 8) {
-
-		int64_t Offset = -(int64_t) keybytes / 8 + 1;
-
-		const uint64_t *keyPtr64 = (const uint64_t *) keyPtr - Offset;
-		uint64_t nextkeys = keyPtr64[Offset];
-		for (; Offset != 0; ++Offset) {
-			uint64_t keys = nextkeys;
-			nextkeys = keyPtr64[Offset + 1];
-
-			Data = _decode_avx((keys & 0xFF), &dataPtr);
-			_write_avx(out, Data);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			_write_avx(out + 4, Data);
-
-			keys >>= 16;
-			Data = _decode_avx((keys & 0xFF), &dataPtr);
-			_write_avx(out + 8, Data);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			_write_avx(out + 12, Data);
-
-			keys >>= 16;
-			Data = _decode_avx((keys & 0xFF), &dataPtr);
-			_write_avx(out + 16, Data);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			_write_avx(out + 20, Data);
-
-			keys >>= 16;
-			Data = _decode_avx((keys & 0xFF), &dataPtr);
-			_write_avx(out + 24, Data);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			_write_avx(out + 28, Data);
-
-			out += 32;
-		}
-		{
-			uint64_t keys = nextkeys;
-
-			Data = _decode_avx((keys & 0xFF), &dataPtr);
-			_write_avx(out, Data);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			_write_avx(out + 4, Data);
-
-			keys >>= 16;
-			Data = _decode_avx((keys & 0xFF), &dataPtr);
-			_write_avx(out + 8, Data);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			_write_avx(out + 12, Data);
-
-			keys >>= 16;
-			Data = _decode_avx((keys & 0xFF), &dataPtr);
-			_write_avx(out + 16, Data);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			_write_avx(out + 20, Data);
-
-			keys >>= 16;
-			Data = _decode_avx((keys & 0xFF), &dataPtr);
-			_write_avx(out + 24, Data);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			_write_avx(out + 28, Data);
-
-			out += 32;
-		}
-	}
-	uint64_t consumedkeys = keybytes - (keybytes & 7);
-	return svb_decode_scalar(out, keyPtr + consumedkeys, dataPtr, count & 31);
-}
-
-// Read count 32-bit integers in maskedvbyte format from in, storing the result in out.  Returns the number of bytes read.
-size_t streamvbyte_decode(const uint8_t* in, uint32_t* out, uint32_t count) {
-	if (count == 0)
-		return 0;
-	const uint8_t *keyPtr = in;            // full list of keys is next
-	uint32_t keyLen = ((count + 3) / 4); // 2-bits per key (rounded up)
-	const uint8_t *dataPtr = keyPtr + keyLen;  // data starts at end of keys
-	return svb_decode_avx_simple(out, keyPtr, dataPtr, count) - in;
-
-}

cpp/streamvbyte/src/streamvbytedelta.c

@@ -1,575 +0,0 @@
-#include "streamvbyte.h"
-#if defined(_MSC_VER)
-     /* Microsoft C/C++-compatible compiler */
-     #include <intrin.h>
-#elif defined(__GNUC__) && (defined(__x86_64__) || defined(__i386__))
-     /* GCC-compatible compiler, targeting x86/x86-64 */
-     #include <x86intrin.h>
-#elif defined(__GNUC__) && defined(__ARM_NEON__)
-     /* GCC-compatible compiler, targeting ARM with NEON */
-     #include <arm_neon.h>
-#elif defined(__GNUC__) && defined(__IWMMXT__)
-     /* GCC-compatible compiler, targeting ARM with WMMX */
-     #include <mmintrin.h>
-#elif (defined(__GNUC__) || defined(__xlC__)) && (defined(__VEC__) || defined(__ALTIVEC__))
-     /* XLC or GCC-compatible compiler, targeting PowerPC with VMX/VSX */
-     #include <altivec.h>
-#elif defined(__GNUC__) && defined(__SPE__)
-     /* GCC-compatible compiler, targeting PowerPC with SPE */
-     #include <spe.h>
-#endif
-
-static uint8_t lengthTable[256] = { 4, 5, 6, 7, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9,
-		10, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9, 10, 8, 9, 10, 11, 6, 7, 8, 9, 7, 8,
-		9, 10, 8, 9, 10, 11, 9, 10, 11, 12, 7, 8, 9, 10, 8, 9, 10, 11, 9, 10,
-		11, 12, 10, 11, 12, 13, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9, 10, 8, 9, 10,
-		11, 6, 7, 8, 9, 7, 8, 9, 10, 8, 9, 10, 11, 9, 10, 11, 12, 7, 8, 9, 10,
-		8, 9, 10, 11, 9, 10, 11, 12, 10, 11, 12, 13, 8, 9, 10, 11, 9, 10, 11,
-		12, 10, 11, 12, 13, 11, 12, 13, 14, 6, 7, 8, 9, 7, 8, 9, 10, 8, 9, 10,
-		11, 9, 10, 11, 12, 7, 8, 9, 10, 8, 9, 10, 11, 9, 10, 11, 12, 10, 11, 12,
-		13, 8, 9, 10, 11, 9, 10, 11, 12, 10, 11, 12, 13, 11, 12, 13, 14, 9, 10,
-		11, 12, 10, 11, 12, 13, 11, 12, 13, 14, 12, 13, 14, 15, 7, 8, 9, 10, 8,
-		9, 10, 11, 9, 10, 11, 12, 10, 11, 12, 13, 8, 9, 10, 11, 9, 10, 11, 12,
-		10, 11, 12, 13, 11, 12, 13, 14, 9, 10, 11, 12, 10, 11, 12, 13, 11, 12,
-		13, 14, 12, 13, 14, 15, 10, 11, 12, 13, 11, 12, 13, 14, 12, 13, 14, 15,
-		13, 14, 15, 16 };
-
-static uint8_t shuffleTable[256][16] = { { 0, -1, -1, -1, 1, -1, -1, -1, 2, -1,
-		-1, -1, 3, -1, -1, -1 }, // 1111
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, -1, -1, -1, 4, -1, -1, -1 },  // 2111
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, -1, -1, -1, 5, -1, -1, -1 },   // 3111
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, -1, -1, -1, 6, -1, -1, -1 },    // 4111
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, -1, -1, -1, 4, -1, -1, -1 },  // 1211
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, -1, -1, -1, 5, -1, -1, -1 },   // 2211
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, -1, -1, -1, 6, -1, -1, -1 },    // 3211
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, -1, -1, -1, 7, -1, -1, -1 },     // 4211
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, -1, -1, -1, 5, -1, -1, -1 },   // 1311
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, -1, -1, -1, 6, -1, -1, -1 },    // 2311
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, -1, -1, -1, 7, -1, -1, -1 },     // 3311
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, -1, -1, -1, 8, -1, -1, -1 },      // 4311
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, -1, -1, -1, 6, -1, -1, -1 },    // 1411
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, -1, -1, -1, 7, -1, -1, -1 },     // 2411
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, -1, -1, -1, 8, -1, -1, -1 },      // 3411
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, -1, -1, -1, 9, -1, -1, -1 },       // 4411
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, -1, -1, 4, -1, -1, -1 },  // 1121
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, -1, -1, 5, -1, -1, -1 },   // 2121
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, -1, -1, 6, -1, -1, -1 },    // 3121
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, -1, -1, 7, -1, -1, -1 },     // 4121
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, -1, -1, 5, -1, -1, -1 },   // 1221
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, -1, -1, 6, -1, -1, -1 },    // 2221
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, -1, -1, 7, -1, -1, -1 },     // 3221
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, -1, -1, 8, -1, -1, -1 },      // 4221
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, -1, -1, 6, -1, -1, -1 },    // 1321
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, -1, -1, 7, -1, -1, -1 },     // 2321
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, -1, -1, 8, -1, -1, -1 },      // 3321
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, -1, -1, 9, -1, -1, -1 },       // 4321
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, -1, -1, 7, -1, -1, -1 },     // 1421
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, -1, -1, 8, -1, -1, -1 },      // 2421
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, -1, -1, 9, -1, -1, -1 },       // 3421
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, 10, -1, -1, -1 },       // 4421
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, -1, 5, -1, -1, -1 },   // 1131
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, -1, 6, -1, -1, -1 },    // 2131
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, -1, 7, -1, -1, -1 },     // 3131
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, -1, 8, -1, -1, -1 },      // 4131
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, -1, 6, -1, -1, -1 },    // 1231
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, -1, 7, -1, -1, -1 },     // 2231
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, -1, 8, -1, -1, -1 },      // 3231
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, -1, 9, -1, -1, -1 },       // 4231
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, -1, 7, -1, -1, -1 },     // 1331
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, -1, 8, -1, -1, -1 },      // 2331
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, -1, 9, -1, -1, -1 },       // 3331
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, -1, 10, -1, -1, -1 },       // 4331
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, -1, 8, -1, -1, -1 },      // 1431
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, -1, 9, -1, -1, -1 },       // 2431
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, -1, 10, -1, -1, -1 },       // 3431
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, 11, -1, -1, -1 },       // 4431
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, 5, 6, -1, -1, -1 },    // 1141
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, 6, 7, -1, -1, -1 },     // 2141
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, 7, 8, -1, -1, -1 },      // 3141
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, 8, 9, -1, -1, -1 },       // 4141
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, 6, 7, -1, -1, -1 },     // 1241
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, 7, 8, -1, -1, -1 },      // 2241
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, 8, 9, -1, -1, -1 },       // 3241
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, 9, 10, -1, -1, -1 },       // 4241
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, 7, 8, -1, -1, -1 },      // 1341
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, 8, 9, -1, -1, -1 },       // 2341
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, 9, 10, -1, -1, -1 },       // 3341
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, 10, 11, -1, -1, -1 },       // 4341
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, -1 },       // 1441
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, -1, -1 },       // 2441
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, 10, 11, -1, -1, -1 },       // 3441
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, -1, -1, -1 },       // 4441
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, -1, -1, -1, 3, 4, -1, -1 },  // 1112
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, -1, -1, -1, 4, 5, -1, -1 },   // 2112
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, -1, -1, -1, 5, 6, -1, -1 },    // 3112
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, -1, -1, -1, 6, 7, -1, -1 },     // 4112
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, -1, -1, -1, 4, 5, -1, -1 },   // 1212
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, -1, -1, -1, 5, 6, -1, -1 },    // 2212
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, -1, -1, -1, 6, 7, -1, -1 },     // 3212
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, -1, -1, -1, 7, 8, -1, -1 },      // 4212
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, -1, -1, -1, 5, 6, -1, -1 },    // 1312
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, -1, -1, -1, 6, 7, -1, -1 },     // 2312
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, -1, -1, -1, 7, 8, -1, -1 },      // 3312
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, -1, -1, -1, 8, 9, -1, -1 },       // 4312
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, -1, -1, -1, 6, 7, -1, -1 },     // 1412
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, -1, -1, -1, 7, 8, -1, -1 },      // 2412
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, -1, -1, -1, 8, 9, -1, -1 },       // 3412
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, -1, -1, -1, 9, 10, -1, -1 },       // 4412
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, -1, -1, 4, 5, -1, -1 },   // 1122
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, -1, -1, 5, 6, -1, -1 },    // 2122
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, -1, -1, 6, 7, -1, -1 },     // 3122
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, -1, -1, 7, 8, -1, -1 },      // 4122
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, -1, -1, 5, 6, -1, -1 },    // 1222
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, -1, -1, 6, 7, -1, -1 },     // 2222
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, -1, -1, 7, 8, -1, -1 },      // 3222
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, -1, -1, 8, 9, -1, -1 },       // 4222
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, -1, -1, 6, 7, -1, -1 },     // 1322
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, -1, -1, 7, 8, -1, -1 },      // 2322
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, -1, -1, 8, 9, -1, -1 },       // 3322
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, -1, -1, 9, 10, -1, -1 },       // 4322
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, -1, -1, 7, 8, -1, -1 },      // 1422
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, -1, -1, 8, 9, -1, -1 },       // 2422
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, -1, -1, 9, 10, -1, -1 },       // 3422
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, 10, 11, -1, -1 },       // 4422
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, -1, 5, 6, -1, -1 },    // 1132
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, -1, 6, 7, -1, -1 },     // 2132
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, -1, 7, 8, -1, -1 },      // 3132
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, -1, 8, 9, -1, -1 },       // 4132
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, -1, 6, 7, -1, -1 },     // 1232
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, -1, 7, 8, -1, -1 },      // 2232
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, -1, 8, 9, -1, -1 },       // 3232
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, -1, 9, 10, -1, -1 },       // 4232
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, -1, 7, 8, -1, -1 },      // 1332
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, -1, 8, 9, -1, -1 },       // 2332
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, -1, 9, 10, -1, -1 },       // 3332
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, -1, 10, 11, -1, -1 },       // 4332
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, -1, 8, 9, -1, -1 },       // 1432
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, -1, 9, 10, -1, -1 },       // 2432
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, -1, 10, 11, -1, -1 },       // 3432
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, 11, 12, -1, -1 },       // 4432
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, 5, 6, 7, -1, -1 },     // 1142
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, 6, 7, 8, -1, -1 },      // 2142
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, 7, 8, 9, -1, -1 },       // 3142
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, 8, 9, 10, -1, -1 },       // 4142
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, 6, 7, 8, -1, -1 },      // 1242
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, 7, 8, 9, -1, -1 },       // 2242
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, 8, 9, 10, -1, -1 },       // 3242
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, 9, 10, 11, -1, -1 },       // 4242
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, 7, 8, 9, -1, -1 },       // 1342
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, 8, 9, 10, -1, -1 },       // 2342
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, 9, 10, 11, -1, -1 },       // 3342
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, 10, 11, 12, -1, -1 },       // 4342
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, -1 },       // 1442
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, -1, -1 },       // 2442
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, -1, -1 },       // 3442
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, -1, -1 },       // 4442
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, -1, -1, -1, 3, 4, 5, -1 },   // 1113
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, -1, -1, -1, 4, 5, 6, -1 },    // 2113
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, -1, -1, -1, 5, 6, 7, -1 },     // 3113
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, -1, -1, -1, 6, 7, 8, -1 },      // 4113
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, -1, -1, -1, 4, 5, 6, -1 },    // 1213
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, -1, -1, -1, 5, 6, 7, -1 },     // 2213
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, -1, -1, -1, 6, 7, 8, -1 },      // 3213
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, -1, -1, -1, 7, 8, 9, -1 },       // 4213
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, -1, -1, -1, 5, 6, 7, -1 },     // 1313
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, -1, -1, -1, 6, 7, 8, -1 },      // 2313
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, -1, -1, -1, 7, 8, 9, -1 },       // 3313
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, -1, -1, -1, 8, 9, 10, -1 },       // 4313
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, -1, -1, -1, 6, 7, 8, -1 },      // 1413
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, -1, -1, -1, 7, 8, 9, -1 },       // 2413
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, -1, -1, -1, 8, 9, 10, -1 },       // 3413
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, -1, -1, -1, 9, 10, 11, -1 },       // 4413
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, -1, -1, 4, 5, 6, -1 },    // 1123
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, -1, -1, 5, 6, 7, -1 },     // 2123
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, -1, -1, 6, 7, 8, -1 },      // 3123
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, -1, -1, 7, 8, 9, -1 },       // 4123
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, -1, -1, 5, 6, 7, -1 },     // 1223
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, -1, -1, 6, 7, 8, -1 },      // 2223
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, -1, -1, 7, 8, 9, -1 },       // 3223
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, -1, -1, 8, 9, 10, -1 },       // 4223
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, -1, -1, 6, 7, 8, -1 },      // 1323
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, -1, -1, 7, 8, 9, -1 },       // 2323
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, -1, -1, 8, 9, 10, -1 },       // 3323
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, -1, -1, 9, 10, 11, -1 },       // 4323
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, -1, -1, 7, 8, 9, -1 },       // 1423
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, -1, -1, 8, 9, 10, -1 },       // 2423
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, -1, -1, 9, 10, 11, -1 },       // 3423
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, 10, 11, 12, -1 },       // 4423
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, -1, 5, 6, 7, -1 },     // 1133
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, -1, 6, 7, 8, -1 },      // 2133
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, -1, 7, 8, 9, -1 },       // 3133
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, -1, 8, 9, 10, -1 },       // 4133
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, -1, 6, 7, 8, -1 },      // 1233
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, -1, 7, 8, 9, -1 },       // 2233
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, -1, 8, 9, 10, -1 },       // 3233
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, -1, 9, 10, 11, -1 },       // 4233
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, -1, 7, 8, 9, -1 },       // 1333
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, -1, 8, 9, 10, -1 },       // 2333
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, -1, 9, 10, 11, -1 },       // 3333
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, -1, 10, 11, 12, -1 },       // 4333
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, -1, 8, 9, 10, -1 },       // 1433
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, -1, 9, 10, 11, -1 },       // 2433
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, -1, 10, 11, 12, -1 },       // 3433
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, 11, 12, 13, -1 },       // 4433
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, 5, 6, 7, 8, -1 },      // 1143
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, 6, 7, 8, 9, -1 },       // 2143
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, 7, 8, 9, 10, -1 },       // 3143
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, 8, 9, 10, 11, -1 },       // 4143
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, 6, 7, 8, 9, -1 },       // 1243
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, 7, 8, 9, 10, -1 },       // 2243
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, 8, 9, 10, 11, -1 },       // 3243
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, 9, 10, 11, 12, -1 },       // 4243
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, 7, 8, 9, 10, -1 },       // 1343
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, 8, 9, 10, 11, -1 },       // 2343
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, 9, 10, 11, 12, -1 },       // 3343
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, 10, 11, 12, 13, -1 },       // 4343
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, -1 },       // 1443
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, -1 },       // 2443
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, -1 },       // 3443
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, -1 },       // 4443
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, -1, -1, -1, 3, 4, 5, 6 },    // 1114
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, -1, -1, -1, 4, 5, 6, 7 },     // 2114
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, -1, -1, -1, 5, 6, 7, 8 },      // 3114
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, -1, -1, -1, 6, 7, 8, 9 },       // 4114
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, -1, -1, -1, 4, 5, 6, 7 },     // 1214
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, -1, -1, -1, 5, 6, 7, 8 },      // 2214
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, -1, -1, -1, 6, 7, 8, 9 },       // 3214
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, -1, -1, -1, 7, 8, 9, 10 },       // 4214
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, -1, -1, -1, 5, 6, 7, 8 },      // 1314
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, -1, -1, -1, 6, 7, 8, 9 },       // 2314
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, -1, -1, -1, 7, 8, 9, 10 },       // 3314
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, -1, -1, -1, 8, 9, 10, 11 },       // 4314
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, -1, -1, -1, 6, 7, 8, 9 },       // 1414
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, -1, -1, -1, 7, 8, 9, 10 },       // 2414
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, -1, -1, -1, 8, 9, 10, 11 },       // 3414
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, -1, -1, -1, 9, 10, 11, 12 },       // 4414
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, -1, -1, 4, 5, 6, 7 },     // 1124
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, -1, -1, 5, 6, 7, 8 },      // 2124
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, -1, -1, 6, 7, 8, 9 },       // 3124
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, -1, -1, 7, 8, 9, 10 },       // 4124
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, -1, -1, 5, 6, 7, 8 },      // 1224
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, -1, -1, 6, 7, 8, 9 },       // 2224
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, -1, -1, 7, 8, 9, 10 },       // 3224
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, -1, -1, 8, 9, 10, 11 },       // 4224
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, -1, -1, 6, 7, 8, 9 },       // 1324
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, -1, -1, 7, 8, 9, 10 },       // 2324
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, -1, -1, 8, 9, 10, 11 },       // 3324
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, -1, -1, 9, 10, 11, 12 },       // 4324
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, -1, -1, 7, 8, 9, 10 },       // 1424
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, -1, -1, 8, 9, 10, 11 },       // 2424
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, -1, -1, 9, 10, 11, 12 },       // 3424
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, -1, -1, 10, 11, 12, 13 },       // 4424
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, -1, 5, 6, 7, 8 },      // 1134
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, -1, 6, 7, 8, 9 },       // 2134
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, -1, 7, 8, 9, 10 },       // 3134
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, -1, 8, 9, 10, 11 },       // 4134
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, -1, 6, 7, 8, 9 },       // 1234
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, -1, 7, 8, 9, 10 },       // 2234
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, -1, 8, 9, 10, 11 },       // 3234
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, -1, 9, 10, 11, 12 },       // 4234
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, -1, 7, 8, 9, 10 },       // 1334
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, -1, 8, 9, 10, 11 },       // 2334
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, -1, 9, 10, 11, 12 },       // 3334
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, -1, 10, 11, 12, 13 },       // 4334
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, -1, 8, 9, 10, 11 },       // 1434
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, -1, 9, 10, 11, 12 },       // 2434
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, -1, 10, 11, 12, 13 },       // 3434
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, -1, 11, 12, 13, 14 },       // 4434
-		{ 0, -1, -1, -1, 1, -1, -1, -1, 2, 3, 4, 5, 6, 7, 8, 9 },       // 1144
-		{ 0, 1, -1, -1, 2, -1, -1, -1, 3, 4, 5, 6, 7, 8, 9, 10 },       // 2144
-		{ 0, 1, 2, -1, 3, -1, -1, -1, 4, 5, 6, 7, 8, 9, 10, 11 },       // 3144
-		{ 0, 1, 2, 3, 4, -1, -1, -1, 5, 6, 7, 8, 9, 10, 11, 12 },       // 4144
-		{ 0, -1, -1, -1, 1, 2, -1, -1, 3, 4, 5, 6, 7, 8, 9, 10 },       // 1244
-		{ 0, 1, -1, -1, 2, 3, -1, -1, 4, 5, 6, 7, 8, 9, 10, 11 },       // 2244
-		{ 0, 1, 2, -1, 3, 4, -1, -1, 5, 6, 7, 8, 9, 10, 11, 12 },       // 3244
-		{ 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, 9, 10, 11, 12, 13 },       // 4244
-		{ 0, -1, -1, -1, 1, 2, 3, -1, 4, 5, 6, 7, 8, 9, 10, 11 },       // 1344
-		{ 0, 1, -1, -1, 2, 3, 4, -1, 5, 6, 7, 8, 9, 10, 11, 12 },       // 2344
-		{ 0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, 9, 10, 11, 12, 13 },       // 3344
-		{ 0, 1, 2, 3, 4, 5, 6, -1, 7, 8, 9, 10, 11, 12, 13, 14 },       // 4344
-		{ 0, -1, -1, -1, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 },       // 1444
-		{ 0, 1, -1, -1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13 },       // 2444
-		{ 0, 1, 2, -1, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 },       // 3444
-		{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 }        // 4444
-};
-
-static uint8_t _encode_data(uint32_t val, uint8_t *__restrict__ *dataPtrPtr) {
-	uint8_t *dataPtr = *dataPtrPtr;
-	uint8_t code;
-
-	if (val < (1 << 8)) { // 1 byte
-		*dataPtr = (uint8_t)(val);
-		*dataPtrPtr += 1;
-		code = 0;
-	} else if (val < (1 << 16)) { // 2 bytes
-		*(uint16_t *) dataPtr = (uint16_t)(val);
-		*dataPtrPtr += 2;
-		code = 1;
-	} else if (val < (1 << 24)) { // 3 bytes
-		*(uint16_t *) dataPtr = (uint16_t)(val);
-		*(dataPtr + 2) = (uint8_t)(val >> 16);
-		*dataPtrPtr += 3;
-		code = 2;
-	} else { // 4 bytes
-		*(uint32_t *) dataPtr = val;
-		*dataPtrPtr += 4;
-		code = 3;
-	}
-
-	return code;
-}
-
-static uint8_t *svb_encode_scalar_d1_init(const uint32_t *in,
-		uint8_t *__restrict__ keyPtr, uint8_t *__restrict__ dataPtr,
-		uint32_t count, uint32_t prev) {
-	if (count == 0)
-		return dataPtr; // exit immediately if no data
-
-	uint8_t shift = 0; // cycles 0, 2, 4, 6, 0, 2, 4, 6, ...
-	uint8_t key = 0;
-	for (uint32_t c = 0; c < count; c++) {
-		if (shift == 8) {
-			shift = 0;
-			*keyPtr++ = key;
-			key = 0;
-		}
-		uint32_t val = in[c] - prev;
-		prev = in[c];
-		uint8_t code = _encode_data(val, &dataPtr);
-		key |= code << shift;
-		shift += 2;
-	}
-
-	*keyPtr = key;  // write last key (no increment needed)
-	return dataPtr; // pointer to first unused data byte
-}
-
-size_t streamvbyte_delta_encode(const uint32_t *in, uint32_t count, uint8_t *out,
-		uint32_t prev) {
-	uint8_t *keyPtr = out;         // keys come immediately after 32-bit count
-	uint32_t keyLen = (count + 3) / 4; // 2-bits rounded to full byte
-	uint8_t *dataPtr = keyPtr + keyLen; // variable byte data after all keys
-
-	return svb_encode_scalar_d1_init(in, keyPtr, dataPtr, count, prev) - out;
-
-}
-
-static inline __m128i _decode_avx(uint32_t key, const uint8_t *__restrict__ *dataPtrPtr) {
-	uint8_t len = lengthTable[key];
-	__m128i Data = _mm_loadu_si128((__m128i *) *dataPtrPtr);
-	__m128i Shuf = *(__m128i *) &shuffleTable[key];
-
-	Data = _mm_shuffle_epi8(Data, Shuf);
-	*dataPtrPtr += len;
-
-	return Data;
-}
-#define BroadcastLastXMM 0xFF // bits 0-7 all set to choose highest element
-
-
-
-static inline void _write_avx(uint32_t *out, __m128i Vec) {
-	_mm_storeu_si128((__m128i *) out, Vec);
-}
-
-static __m128i _write_avx_d1(uint32_t *out, __m128i Vec, __m128i Prev) {
-	__m128i Add = _mm_slli_si128(Vec, 4); // Cycle 1: [- A B C] (already done)
-	Prev = _mm_shuffle_epi32(Prev, BroadcastLastXMM); // Cycle 2: [P P P P]
-	Vec = _mm_add_epi32(Vec, Add);                    // Cycle 2: [A AB BC CD]
-	Add = _mm_slli_si128(Vec, 8);                     // Cycle 3: [- - A AB]
-	Vec = _mm_add_epi32(Vec, Prev);                 // Cycle 3: [PA PAB PBC PCD]
-	Vec = _mm_add_epi32(Vec, Add); // Cycle 4: [PA PAB PABC PABCD]
-
-	_write_avx(out, Vec);
-	return Vec;
-}
-
-#ifndef _MSC_VER
-static __m128i High16To32 = {0xFFFF0B0AFFFF0908, 0xFFFF0F0EFFFF0D0C};
-#else
-static __m128i High16To32 = {8,  9,  -1, -1, 10, 11, -1, -1,
-                           12, 13, -1, -1, 14, 15, -1, -1};
-#endif
-
-static inline __m128i _write_16bit_avx_d1(uint32_t *out, __m128i Vec, __m128i Prev) {
-  // vec == [A B C D E F G H] (16 bit values)
-  __m128i Add = _mm_slli_si128(Vec, 2);               // [- A B C D E F G]
-  Prev = _mm_shuffle_epi32(Prev, BroadcastLastXMM); // [P P P P] (32-bit)
-  Vec = _mm_add_epi32(Vec, Add);                    // [A AB BC CD DE FG GH]
-  Add = _mm_slli_si128(Vec, 4);                     // [- - A AB BC CD DE EF]
-  Vec = _mm_add_epi32(Vec, Add);      // [A AB ABC ABCD BCDE CDEF DEFG EFGH]
-  __m128i V1 = _mm_cvtepu16_epi32(Vec); // [A AB ABC ABCD] (32-bit)
-  V1 = _mm_add_epi32(V1, Prev);       // [PA PAB PABC PABCD] (32-bit)
-  __m128i V2 =
-      _mm_shuffle_epi8(Vec, High16To32); // [BCDE CDEF DEFG EFGH] (32-bit)
-  V2 = _mm_add_epi32(V1, V2); // [PABCDE PABCDEF PABCDEFG PABCDEFGH] (32-bit)
-  _write_avx(out, V1);
-  _write_avx(out + 4, V2);
-  return V2;
-}
-
-static inline uint32_t _decode_data(const uint8_t **dataPtrPtr, uint8_t code) {
-	const uint8_t *dataPtr = *dataPtrPtr;
-	uint32_t val;
-
-	if (code == 0) { // 1 byte
-		val = (uint32_t) * dataPtr;
-		dataPtr += 1;
-	} else if (code == 1) { // 2 bytes
-		val = (uint32_t) * (uint16_t *) dataPtr;
-		dataPtr += 2;
-	} else if (code == 2) { // 3 bytes
-		val = (uint32_t) * (uint16_t *) dataPtr;
-		val |= *(dataPtr + 2) << 16;
-		dataPtr += 3;
-	} else {                      // code == 3
-		val = *(uint32_t *) dataPtr; // 4 bytes
-		dataPtr += 4;
-	}
-
-	*dataPtrPtr = dataPtr;
-	return val;
-}
-
-const uint8_t *svb_decode_scalar_d1_init(uint32_t *outPtr, const uint8_t *keyPtr,
-		const uint8_t *dataPtr, uint32_t count,
-                                   uint32_t prev) {
-  if (count == 0)
-    return dataPtr; // no reads or writes if no data
-
-  uint8_t shift = 0;
-  uint32_t key = *keyPtr++;
-
-  for (uint32_t c = 0; c < count; c++) {
-    if (shift == 8) {
-      shift = 0;
-      key = *keyPtr++;
-    }
-    uint32_t val = _decode_data(&dataPtr, (key >> shift) & 0x3);
-    val += prev;
-    *outPtr++ = val;
-    prev = val;
-    shift += 2;
-  }
-
-  return dataPtr; // pointer to first unused byte after end
-}
-
-const uint8_t *svb_decode_avx_d1_init(uint32_t *out, const uint8_t *__restrict__ keyPtr,
-		const uint8_t *__restrict__ dataPtr, uint64_t count, uint32_t prev) {
-	uint64_t keybytes = count / 4; // number of key bytes
-	if (keybytes >= 8) {
-		__m128i Prev = _mm_set1_epi32(prev);
-		__m128i Data;
-
-		int64_t Offset = -(int64_t) keybytes / 8 + 1;
-
-		const uint64_t *keyPtr64 = (const uint64_t *) keyPtr - Offset;
-		uint64_t nextkeys = keyPtr64[Offset];
-		for (; Offset != 0; ++Offset) {
-			uint64_t keys = nextkeys;
-			nextkeys = keyPtr64[Offset + 1];
-			// faster 16-bit delta since we only have 8-bit values
-			if (!keys) { // 32 1-byte ints in a row
-
-				Data = _mm_cvtepu8_epi16(_mm_lddqu_si128((__m128i *) (dataPtr)));
-				Prev = _write_16bit_avx_d1(out, Data, Prev);
-				Data = _mm_cvtepu8_epi16(
-						_mm_lddqu_si128((__m128i *) (dataPtr + 8)));
-				Prev = _write_16bit_avx_d1(out + 8, Data, Prev);
-				Data = _mm_cvtepu8_epi16(
-						_mm_lddqu_si128((__m128i *) (dataPtr + 16)));
-				Prev = _write_16bit_avx_d1(out + 16, Data, Prev);
-				Data = _mm_cvtepu8_epi16(
-						_mm_lddqu_si128((__m128i *) (dataPtr + 24)));
-				Prev = _write_16bit_avx_d1(out + 24, Data, Prev);
-				out += 32;
-				dataPtr += 32;
-				continue;
-			}
-
-			Data = _decode_avx(keys & 0x00FF, &dataPtr);
-			Prev = _write_avx_d1(out, Data, Prev);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			Prev = _write_avx_d1(out + 4, Data, Prev);
-
-			keys >>= 16;
-			Data = _decode_avx((keys & 0x00FF), &dataPtr);
-			Prev = _write_avx_d1(out + 8, Data, Prev);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			Prev = _write_avx_d1(out + 12, Data, Prev);
-
-			keys >>= 16;
-			Data = _decode_avx((keys & 0x00FF), &dataPtr);
-			Prev = _write_avx_d1(out + 16, Data, Prev);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			Prev = _write_avx_d1(out + 20, Data, Prev);
-
-			keys >>= 16;
-			Data = _decode_avx((keys & 0x00FF), &dataPtr);
-			Prev = _write_avx_d1(out + 24, Data, Prev);
-			Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-			Prev = _write_avx_d1(out + 28, Data, Prev);
-
-			out += 32;
-		}
-		{
-			uint64_t keys = nextkeys;
-			// faster 16-bit delta since we only have 8-bit values
-			if (!keys) { // 32 1-byte ints in a row
-				Data = _mm_cvtepu8_epi16(_mm_lddqu_si128((__m128i *) (dataPtr)));
-				Prev = _write_16bit_avx_d1(out, Data, Prev);
-				Data = _mm_cvtepu8_epi16(
-						_mm_lddqu_si128((__m128i *) (dataPtr + 8)));
-				Prev = _write_16bit_avx_d1(out + 8, Data, Prev);
-				Data = _mm_cvtepu8_epi16(
-						_mm_lddqu_si128((__m128i *) (dataPtr + 16)));
-				Prev = _write_16bit_avx_d1(out + 16, Data, Prev);
-				Data = _mm_cvtepu8_epi16(
-						_mm_loadl_epi64((__m128i *) (dataPtr + 24)));
-				Prev = _write_16bit_avx_d1(out + 24, Data, Prev);
-				out += 32;
-				dataPtr += 32;
-
-			} else {
-
-				Data = _decode_avx(keys & 0x00FF, &dataPtr);
-				Prev = _write_avx_d1(out, Data, Prev);
-				Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-				Prev = _write_avx_d1(out + 4, Data, Prev);
-
-				keys >>= 16;
-				Data = _decode_avx((keys & 0x00FF), &dataPtr);
-				Prev = _write_avx_d1(out + 8, Data, Prev);
-				Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-				Prev = _write_avx_d1(out + 12, Data, Prev);
-
-				keys >>= 16;
-				Data = _decode_avx((keys & 0x00FF), &dataPtr);
-				Prev = _write_avx_d1(out + 16, Data, Prev);
-				Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-				Prev = _write_avx_d1(out + 20, Data, Prev);
-
-				keys >>= 16;
-				Data = _decode_avx((keys & 0x00FF), &dataPtr);
-				Prev = _write_avx_d1(out + 24, Data, Prev);
-				Data = _decode_avx((keys & 0xFF00) >> 8, &dataPtr);
-				Prev = _write_avx_d1(out + 28, Data, Prev);
-
-				out += 32;
-			}
-		}
-		prev = out[-1];
-	}
-	uint64_t consumedkeys = keybytes - (keybytes & 7);
-	return svb_decode_scalar_d1_init(out, keyPtr + consumedkeys, dataPtr,
-			count & 31, prev);
-}
-
-size_t streamvbyte_delta_decode(const uint8_t* in, uint32_t* out,
-		uint32_t count, uint32_t prev) {
-	uint32_t keyLen = ((count + 3) / 4); // 2-bits per key (rounded up)
-	const uint8_t *keyPtr = in;
-	const uint8_t *dataPtr = keyPtr + keyLen;  // data starts at end of keys
-	return svb_decode_avx_d1_init(out, keyPtr, dataPtr, count, prev) - in;
-}

cpp/streamvbyte/tests/unit.c

@@ -1,73 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-
-#include "streamvbyte.h"
-#include "streamvbytedelta.h"
-
-int main() {
-	int N = 4096;
-	uint32_t * datain = malloc(N * sizeof(uint32_t));
-	uint8_t * compressedbuffer = malloc(2 * N * sizeof(uint32_t));
-	uint32_t * recovdata = malloc(N * sizeof(uint32_t));
-
-	for (int length = 0; length <= N;) {
-		printf("length = %d \n", length);
-		for (uint32_t gap = 1; gap <= 387420489; gap *= 3) {
-			for (int k = 0; k < length; ++k)
-				datain[k] = gap;
-			size_t compsize = streamvbyte_encode(datain, length,
-					compressedbuffer);
-			size_t usedbytes = streamvbyte_decode(compressedbuffer, recovdata,
-					length);
-			if (compsize != usedbytes) {
-				printf(
-						"[streamvbyte_decode] code is buggy gap = %d, size mismatch %d %d \n",
-						(int) gap, (int) compsize, (int) usedbytes);
-				return -1;
-			}
-			for (int k = 0; k < length; ++k) {
-				if (recovdata[k] != datain[k]) {
-					printf("[streamvbyte_decode] code is buggy gap = %d\n",
-							(int) gap);
-					return -1;
-				}
-			}
-		}
-
-		printf("Delta \n");
-		for (size_t gap = 1; gap <= 531441; gap *= 3) {
-			for (int k = 0; k < length; ++k)
-				datain[k] = gap * k;
-			size_t compsize = streamvbyte_delta_encode(datain, length,
-					compressedbuffer, 0);
-			size_t usedbytes = streamvbyte_delta_decode(compressedbuffer,
-					recovdata, length, 0);
-			if (compsize != usedbytes) {
-				printf(
-						"[streamvbyte_delta_decode] code is buggy gap = %d, size mismatch %d %d \n",
-						(int) gap, (int) compsize, (int) usedbytes);
-				return -1;
-			}
-			for (int k = 0; k < length; ++k) {
-				if (recovdata[k] != datain[k]) {
-					printf(
-							"[streamvbyte_delta_decode] code is buggy gap = %d\n",
-							(int) gap);
-					return -1;
-				}
-			}
-
-		}
-
-		if (length < 128)
-			++length;
-		else {
-			length *= 2;
-		}
-	}
-	free(datain);
-	free(compressedbuffer);
-	free(recovdata);
-	printf("Code looks good.\n");
-	return 0;
-}

doc/.gitignore

@@ -0,0 +1 @@
+book

doc/assets/images/element.io.svg

@@ -0,0 +1,8 @@
+<svg width="518" height="112" viewBox="0 0 518 112" fill="none" xmlns="http://www.w3.org/2000/svg">
+<path fill-rule="evenodd" clip-rule="evenodd" d="M56 112C86.9279 112 112 86.9279 112 56C112 25.0721 86.9279 0 56 0C25.0721 0 0 25.0721 0 56C0 86.9279 25.0721 112 56 112Z" fill="#0DBD8B"/>
+<path fill-rule="evenodd" clip-rule="evenodd" d="M45.7615 26.093C45.7615 23.8325 47.5977 22.0001 49.8629 22.0001C65.2154 22.0001 77.6611 34.4199 77.6611 49.7406C77.6611 52.001 75.8248 53.8335 73.5597 53.8335C71.2945 53.8335 69.4583 52.001 69.4583 49.7406C69.4583 38.9408 60.6851 30.1859 49.8629 30.1859C47.5977 30.1859 45.7615 28.3534 45.7615 26.093Z" fill="white"/>
+<path fill-rule="evenodd" clip-rule="evenodd" d="M85.8986 45.6477C88.1637 45.6477 89.9999 47.4801 89.9999 49.7406C89.9999 65.0612 77.5543 77.4811 62.2017 77.4811C59.9366 77.4811 58.1003 75.6486 58.1003 73.3882C58.1003 71.1277 59.9366 69.2953 62.2017 69.2953C73.024 69.2953 81.7972 60.5403 81.7972 49.7406C81.7972 47.4801 83.6334 45.6477 85.8986 45.6477Z" fill="white"/>
+<path fill-rule="evenodd" clip-rule="evenodd" d="M66.3031 85.907C66.3031 88.1675 64.4668 89.9999 62.2017 89.9999C46.8492 89.9999 34.4035 77.58 34.4035 62.2594C34.4035 59.9989 36.2398 58.1665 38.5049 58.1665C40.77 58.1665 42.6063 59.9989 42.6063 62.2594C42.6063 73.0592 51.3795 81.8141 62.2017 81.8141C64.4668 81.8141 66.3031 83.6466 66.3031 85.907Z" fill="white"/>
+<path fill-rule="evenodd" clip-rule="evenodd" d="M26.1014 66.3523C23.8363 66.3523 22.0001 64.5199 22.0001 62.2594C22 46.9388 34.4457 34.5189 49.7983 34.5189C52.0634 34.5189 53.8997 36.3514 53.8997 38.6118C53.8997 40.8723 52.0634 42.7047 49.7983 42.7047C38.976 42.7047 30.2028 51.4597 30.2028 62.2594C30.2028 64.5199 28.3666 66.3523 26.1014 66.3523Z" fill="white"/>
+<path d="M197 63.5H157.5C157.967 67.6333 159.467 70.9333 162 73.4C164.533 75.8 167.867 77 172 77C174.733 77 177.2 76.3333 179.4 75C181.6 73.6667 183.167 71.8667 184.1 69.6H196.1C194.5 74.8667 191.5 79.1333 187.1 82.4C182.767 85.6 177.633 87.2 171.7 87.2C163.967 87.2 157.7 84.6333 152.9 79.5C148.167 74.3667 145.8 67.8667 145.8 60C145.8 52.3333 148.2 45.9 153 40.7C157.8 35.5 164 32.9 171.6 32.9C179.2 32.9 185.333 35.4667 190 40.6C194.733 45.6667 197.1 52.0667 197.1 59.8L197 63.5ZM171.6 42.6C167.867 42.6 164.767 43.7 162.3 45.9C159.833 48.1 158.3 51.0333 157.7 54.7H185.3C184.767 51.0333 183.3 48.1 180.9 45.9C178.5 43.7 175.4 42.6 171.6 42.6ZM205.289 70.5V11H217.189V70.7C217.189 73.3667 218.656 74.7 221.589 74.7L223.689 74.6V85.9C222.556 86.1 221.356 86.2 220.089 86.2C214.956 86.2 211.189 84.9 208.789 82.3C206.456 79.7 205.289 75.7667 205.289 70.5ZM279.109 63.5H239.609C240.076 67.6333 241.576 70.9333 244.109 73.4C246.643 75.8 249.976 77 254.109 77C256.843 77 259.309 76.3333 261.509 75C263.709 73.6667 265.276 71.8667 266.209 69.6H278.209C276.609 74.8667 273.609 79.1333 269.209 82.4C264.876 85.6 259.743 87.2 253.809 87.2C246.076 87.2 239.809 84.6333 235.009 79.5C230.276 74.3667 227.909 67.8667 227.909 60C227.909 52.3333 230.309 45.9 235.109 40.7C239.909 35.5 246.109 32.9 253.709 32.9C261.309 32.9 267.443 35.4667 272.109 40.6C276.843 45.6667 279.209 52.0667 279.209 59.8L279.109 63.5ZM253.709 42.6C249.976 42.6 246.876 43.7 244.409 45.9C241.943 48.1 240.409 51.0333 239.809 54.7H267.409C266.876 51.0333 265.409 48.1 263.009 45.9C260.609 43.7 257.509 42.6 253.709 42.6ZM332.798 56.2V86H320.898V54.9C320.898 47.0333 317.632 43.1 311.098 43.1C307.565 43.1 304.732 44.2333 302.598 46.5C300.532 48.7667 299.498 51.8667 299.498 55.8V86H287.598V34.1H298.598V41C299.865 38.6667 301.798 36.7333 304.398 35.2C306.998 33.6667 310.232 32.9 314.098 32.9C321.298 32.9 326.498 35.6333 329.698 41.1C334.098 35.6333 339.965 32.9 347.298 32.9C353.365 32.9 358.032 34.8 361.298 38.6C364.565 42.3333 366.198 47.2667 366.198 53.4V86H354.298V54.9C354.298 47.0333 351.032 43.1 344.498 43.1C340.898 43.1 338.032 44.2667 335.898 46.6C333.832 48.8667 332.798 52.0667 332.798 56.2ZM425.379 63.5H385.879C386.346 67.6333 387.846 70.9333 390.379 73.4C392.912 75.8 396.246 77 400.379 77C403.112 77 405.579 76.3333 407.779 75C409.979 73.6667 411.546 71.8667 412.479 69.6H424.479C422.879 74.8667 419.879 79.1333 415.479 82.4C411.146 85.6 406.012 87.2 400.079 87.2C392.346 87.2 386.079 84.6333 381.279 79.5C376.546 74.3667 374.179 67.8667 374.179 60C374.179 52.3333 376.579 45.9 381.379 40.7C386.179 35.5 392.379 32.9 399.979 32.9C407.579 32.9 413.712 35.4667 418.379 40.6C423.112 45.6667 425.479 52.0667 425.479 59.8L425.379 63.5ZM399.979 42.6C396.246 42.6 393.146 43.7 390.679 45.9C388.212 48.1 386.679 51.0333 386.079 54.7H413.679C413.146 51.0333 411.679 48.1 409.279 45.9C406.879 43.7 403.779 42.6 399.979 42.6ZM444.868 34.1V41C446.068 38.7333 448.035 36.8333 450.768 35.3C453.568 33.7 456.935 32.9 460.868 32.9C467.001 32.9 471.735 34.7667 475.068 38.5C478.468 42.2333 480.168 47.2 480.168 53.4V86H468.268V54.9C468.268 51.2333 467.401 48.3667 465.668 46.3C464.001 44.1667 461.435 43.1 457.968 43.1C454.168 43.1 451.168 44.2333 448.968 46.5C446.835 48.7667 445.768 51.9 445.768 55.9V86H433.868V34.1H444.868ZM514.922 75.4V85.7C513.455 86.1 511.389 86.3 508.722 86.3C498.589 86.3 493.522 81.2 493.522 71V43.6H485.622V34.1H493.522V20.6H505.422V34.1H515.122V43.6H505.422V69.8C505.422 73.8667 507.355 75.9 511.222 75.9L514.922 75.4Z" fill="black"/>
+</svg>

doc/book.toml

@@ -0,0 +1,5 @@
+[book]
+authors = ["Paul Masurel"]
+multilingual = false
+src = "src"
+title = "Tantivy, the user guide"

doc/src/SUMMARY.md

@@ -0,0 +1,14 @@
+# Summary
+
+[Avant Propos](./avant-propos.md)
+
+- [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)
+
+[Frequently Asked Questions](./faq.md)
+[Examples](./examples.md)

doc/src/avant-propos.md

@@ -0,0 +1,34 @@
+# Foreword, what is the scope of tantivy?
+
+> Tantivy is a **search** engine **library** for Rust.
+
+If you are familiar with Lucene, it's an excellent approximation to consider tantivy as Lucene for rust. tantivy is heavily inspired by Lucene's design and
+they both have the same scope and targeted use cases.
+
+If you are not familiar with Lucene, let's break down our little tagline.
+
+- **Search** here means full-text search : fundamentally, tantivy is here to help you
+identify efficiently what are the documents matching a given query in your corpus.
+But modern search UI are so much more : text processing, facetting, autocomplete, fuzzy search, good
+relevancy, collapsing, highlighting, spatial search.
+
+  While some of these features are not available in tantivy yet, all of these are relevant
+  feature requests. Tantivy's objective is to offer a solid toolbox to create the best search
+  experience. But keep in mind this is just a toolbox.
+  Which bring us to the second keyword...
+
+- **Library** means that you will have to write code. tantivy is not an *all-in-one* server solution like elastic search for instance.
+
+  Sometimes a functionality will not be available in tantivy because it is too
+  specific to your use case. By design, tantivy should make it possible to extend
+  the available set of features using the existing rock-solid datastructures.
+
+  Most frequently this will mean writing your own `Collector`, your own `Scorer` or your own
+  `TokenFilter`... Some of your requirements may also be related to
+  something closer to architecture or operations. For instance, you may
+  want to build a large corpus on Hadoop, fine-tune the merge policy to keep your
+  index sharded in a time-wise fashion, or you may want to convert and existing
+  index from a different format.
+
+  Tantivy exposes a lot of low level API to do all of these things.
+  

doc/src/basis.md

@@ -0,0 +1,68 @@
+# Anatomy of an index
+
+## Straight from disk
+
+Tantivy accesses its data using an abstracting trait called `Directory`.
+In theory, one can come and override the data access logic. In practise, the
+trait somewhat assumes that your data can be mapped to memory, and tantivy
+seems deeply married to using `mmap` for its io [^1], and the only persisting
+directory shipped with tantivy is the `MmapDirectory`.
+
+While this design has some downsides, this greatly simplifies the source code of
+tantivy. Caching is also entirely delegated to the OS.
+
+`tantivy` works entirely (or almost) by directly reading the datastructures as they are laid on disk. As a result, the act of opening an indexing does not involve loading different datastructures from the disk into random access memory : starting a process, opening an index, and performing your first query can typically be done in a matter of milliseconds.
+
+This is an interesting property for a command line search engine, or for some multi-tenant log search engine : spawning a new process for each new query can be a perfectly sensible solution in some use case.
+
+In later chapters, we will discuss tantivy's inverted index data layout.
+One key take away is that to achieve great performance, search indexes are extremely compact.
+Of course this is crucial to reduce IO, and ensure that as much of our index can sit in RAM.
+
+Also, whenever possible its data is accessed sequentially. Of course, this is an amazing property when tantivy needs to access the data from your spinning hard disk, but this is also
+critical for performance, if your data is read from and an `SSD` or even already in your pagecache.
+
+## Segments, and the log method
+
+That kind of compact layout comes at one cost: it prevents our datastructures from being dynamic.
+In fact, the `Directory` trait does not even allow you to modify part of a file.
+
+To allow the addition / deletion of documents, and create the illusion that
+your index is dynamic (i.e.: adding and deleting documents), tantivy uses a common database trick sometimes referred to as the *log method*.
+
+Let's forget about deletes for a moment.
+
+As you add documents, these documents are processed and stored in a dedicated datastructure, in a `RAM` buffer. This datastructure is not ready for search, but it is useful to receive your data and rearrange it very rapidly.
+
+As you add documents, this buffer will reach its capacity and tantivy will transparently stop adding document to it and start converting this datastructure to its final read-only format on disk. Once written, an brand empty buffer is available to resume adding documents.
+
+The resulting chunk of index obtained after this serialization is called a `Segment`.
+
+> A segment is a self-contained atomic piece of index. It is identified with a UUID, and all of its files are identified using the naming scheme : `<UUID>.*`.
+
+Which brings us to the nature of a tantivy `Index`.
+
+> A tantivy `Index` is a collection of `Segments`.
+
+Physically, this really just means and index is a bunch of segment files in a given `Directory`,
+linked together by a `meta.json` file. This transparency can become extremely handy
+to get tantivy to fit your use case:
+
+*Example 1* You could for instance use hadoop to build a very large search index in a timely manner, copy all of the resulting segment files in the same directory and edit the `meta.json` to get a functional index.[^2]
+
+*Example 2* You could also disable your merge policy and enforce daily segments. Removing data after one week can then be done very efficiently by just editing the `meta.json` and deleting the files associated to segment `D-7`.
+
+## Merging
+
+As you index more and more data, your index will accumulate more and more segments.
+Having a lot of small segments is not really optimal. There is a bit of redundancy in having
+all these term dictionary. Also when searching, we will need to do term lookups as many times as we have segments.  It can hurt search performance a bit.
+
+That's where merging or compacting comes into place. Tantivy will continuously consider merge
+opportunities and start merging segments in the background.
+
+## Indexing throughput, number of indexing threads
+
+[^1]: This may eventually change.
+
+[^2]: Be careful however. By default these files will not be considered as *managed* by tantivy. This means they will never be garbage collected by tantivy, regardless of whether they become obsolete or not.

doc/src/examples.md

@@ -0,0 +1,3 @@
+# Examples
+
+- [Basic search](/examples/basic_search.html)

doc/src/facetting.md

@@ -0,0 +1,5 @@
+# Facetting
+
+wewew
+
+## weeewe

doc/src/index_sorting.md

@@ -0,0 +1,62 @@
+
+- [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 Tantivy 0.16 only fast fields are allowed to be used.
+
+```rust
+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).