diff options
| author | Chih-Hung Hsieh <chh@google.com> | 2020-11-09 15:57:01 -0800 |
|---|---|---|
| committer | Chih-Hung Hsieh <chh@google.com> | 2020-11-09 16:04:03 -0800 |
| commit | cdc803b2d8a708561464f3969b9c7ab2ce1ac885 (patch) | |
| tree | c1175c3534d94f38d3c79fa5b5adc2fac623b620 | |
| parent | 20b03fd894d8d6f04fb417d0d979dc6a12ebbb34 (diff) | |
| download | hashbrown-cdc803b2d8a708561464f3969b9c7ab2ce1ac885.tar.gz | |
Import hashbrown-0.9.1
* Add Android.bp, OWNERS, METADATA and other 3p package required files.
Bug: 171754295
Test: build all rust crates.
Change-Id: I87e5577f6683f412b22dc79b39fa06e40924fbf8
36 files changed, 13362 insertions, 0 deletions
diff --git a/.cargo_vcs_info.json b/.cargo_vcs_info.json new file mode 100644 index 0000000..845b5e7 --- /dev/null +++ b/.cargo_vcs_info.json @@ -0,0 +1,5 @@ +{ + "git": { + "sha1": "34c11891e13fa3c0d08b0540e869aace9d347c26" + } +} diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..6936990 --- /dev/null +++ b/.gitignore @@ -0,0 +1,3 @@ +/target +**/*.rs.bk +Cargo.lock diff --git a/Android.bp b/Android.bp new file mode 100644 index 0000000..8f2981a --- /dev/null +++ b/Android.bp @@ -0,0 +1,20 @@ +// This file is generated by cargo2android.py --device --run --dependencies. + +rust_library { + name: "libhashbrown", + host_supported: true, + crate_name: "hashbrown", + srcs: ["src/lib.rs"], + edition: "2018", + features: [ + "ahash", + "default", + "inline-more", + ], + rustlibs: [ + "libahash", + ], +} + +// dependent_library ["feature_list"] +// ahash-0.4.6 diff --git a/CHANGELOG.md b/CHANGELOG.md new file mode 100644 index 0000000..b6eb671 --- /dev/null +++ b/CHANGELOG.md @@ -0,0 +1,294 @@ +# Change Log + +All notable changes to this project will be documented in this file. + +The format is based on [Keep a Changelog](http://keepachangelog.com/) +and this project adheres to [Semantic Versioning](http://semver.org/). + +## [Unreleased] + +## [v0.9.1] - 2020-09-28 + +## Added +- Added safe methods to `RawTable` (#202): + - `get`: `find` and `as_ref` + - `get_mut`: `find` and `as_mut` + - `insert_entry`: `insert` and `as_mut` + - `remove_entry`: `find` and `remove` + - `erase_entry`: `find` and `erase` + +## Changed +- Removed `from_key_hashed_nocheck`'s `Q: Hash`. (#200) +- Made `RawTable::drain` safe. (#201) + +## [v0.9.0] - 2020-09-03 + +### Fixed +- `drain_filter` now removes and yields items that do match the predicate, + rather than items that don't. This is a **breaking change** to match the + behavior of the `drain_filter` methods in `std`. (#187) + +### Added +- Added `replace_entry_with` to `OccupiedEntry`, and `and_replace_entry_with` to `Entry`. (#190) +- Implemented `FusedIterator` and `size_hint` for `DrainFilter`. (#188) + +### Changed +- The minimum Rust version has been bumped to 1.36 (due to `crossbeam` dependency). (#193) +- Updated `ahash` dependency to 0.4. (#198) +- `HashMap::with_hasher` and `HashSet::with_hasher` are now `const fn`. (#195) +- Removed `T: Hash + Eq` and `S: BuildHasher` bounds on `HashSet::new`, + `with_capacity`, `with_hasher`, and `with_capacity_and_hasher`. (#185) + +## [v0.8.2] - 2020-08-08 + +### Changed +- Avoid closures to improve compile times. (#183) +- Do not iterate to drop if empty. (#182) + +## [v0.8.1] - 2020-07-16 + +### Added +- Added `erase` and `remove` to `RawTable`. (#171) +- Added `try_with_capacity` to `RawTable`. (#174) +- Added methods that allow re-using a `RawIter` for `RawDrain`, + `RawIntoIter`, and `RawParIter`. (#175) +- Added `reflect_remove` and `reflect_insert` to `RawIter`. (#175) +- Added a `drain_filter` function to `HashSet`. (#179) + +### Changed +- Deprecated `RawTable::erase_no_drop` in favor of `erase` and `remove`. (#176) +- `insert_no_grow` is now exposed under the `"raw"` feature. (#180) + +## [v0.8.0] - 2020-06-18 + +### Fixed +- Marked `RawTable::par_iter` as `unsafe`. (#157) + +### Changed +- Reduced the size of `HashMap`. (#159) +- No longer create tables with a capacity of 1 element. (#162) +- Removed `K: Eq + Hash` bounds on `retain`. (#163) +- Pulled in `HashMap` changes from rust-lang/rust (#164): + - `extend_one` support on nightly. + - `CollectionAllocErr` renamed to `TryReserveError`. + - Added `HashSet::get_or_insert_owned`. + - `Default` for `HashSet` no longer requires `T: Eq + Hash` and `S: BuildHasher`. + +## [v0.7.2] - 2020-04-27 + +### Added +- Added `or_insert_with_key` to `Entry`. (#152) + +### Fixed +- Partially reverted `Clone` optimization which was unsound. (#154) + +### Changed +- Disabled use of `const-random` by default, which prevented reproducible builds. (#155) +- Optimized `repeat` function. (#150) +- Use `NonNull` for buckets, which improves codegen for iterators. (#148) + +## [v0.7.1] - 2020-03-16 + +### Added +- Added `HashMap::get_key_value_mut`. (#145) + +### Changed +- Optimized `Clone` implementation. (#146) + +## [v0.7.0] - 2020-01-31 + +### Added +- Added a `drain_filter` function to `HashMap`. (#135) + +### Changed +- Updated `ahash` dependency to 0.3. (#141) +- Optimized set union and intersection. (#130) +- `raw_entry` can now be used without requiring `S: BuildHasher`. (#123) +- `RawTable::bucket_index` can now be used under the `raw` feature. (#128) + +## [v0.6.3] - 2019-10-31 + +### Added +- Added an `ahash-compile-time-rng` feature (enabled by default) which allows disabling the + `compile-time-rng` feature in `ahash` to work around a Cargo bug. (#125) + +## [v0.6.2] - 2019-10-23 + +### Added +- Added an `inline-more` feature (enabled by default) which allows choosing a tradeoff between + runtime performance and compilation time. (#119) + +## [v0.6.1] - 2019-10-04 + +### Added +- Added `Entry::insert` and `RawEntryMut::insert`. (#118) + +### Changed +- `Group::static_empty` was changed from a `const` to a `static` (#116). + +## [v0.6.0] - 2019-08-13 + +### Fixed +- Fixed AHash accidentally depending on `std`. (#110) + +### Changed +- The minimum Rust version has been bumped to 1.32 (due to `rand` dependency). + +## ~~[v0.5.1] - 2019-08-04~~ + +This release was _yanked_ due to a breaking change for users of `no-default-features`. + +### Added +- The experimental and unsafe `RawTable` API is available under the "raw" feature. (#108) +- Added entry-like methods for `HashSet`. (#98) + +### Changed +- Changed the default hasher from FxHash to AHash. (#97) +- `hashbrown` is now fully `no_std` on recent Rust versions (1.36+). (#96) + +### Fixed +- We now avoid growing the table during insertions when it wasn't necessary. (#106) +- `RawOccupiedEntryMut` now properly implements `Send` and `Sync`. (#100) +- Relaxed `lazy_static` version. (#92) + +## [v0.5.0] - 2019-06-12 + +### Fixed +- Resize with a more conservative amount of space after deletions. (#86) + +### Changed +- Exposed the Layout of the failed allocation in CollectionAllocErr::AllocErr. (#89) + +## [v0.4.0] - 2019-05-30 + +### Fixed +- Fixed `Send` trait bounds on `IterMut` not matching the libstd one. (#82) + +## [v0.3.1] - 2019-05-30 + +### Fixed +- Fixed incorrect use of slice in unsafe code. (#80) + +## [v0.3.0] - 2019-04-23 + +### Changed +- Changed shrink_to to not panic if min_capacity < capacity. (#67) + +### Fixed +- Worked around emscripten bug emscripten-core/emscripten-fastcomp#258. (#66) + +## [v0.2.2] - 2019-04-16 + +### Fixed +- Inlined non-nightly lowest_set_bit_nonzero. (#64) +- Fixed build on latest nightly. (#65) + +## [v0.2.1] - 2019-04-14 + +### Changed +- Use for_each in map Extend and FromIterator. (#58) +- Improved worst-case performance of HashSet.is_subset. (#61) + +### Fixed +- Removed incorrect debug_assert. (#60) + +## [v0.2.0] - 2019-03-31 + +### Changed +- The code has been updated to Rust 2018 edition. This means that the minimum + Rust version has been bumped to 1.31 (2018 edition). + +### Added +- Added `insert_with_hasher` to the raw_entry API to allow `K: !(Hash + Eq)`. (#54) +- Added support for using hashbrown as the hash table implementation in libstd. (#46) + +### Fixed +- Fixed cargo build with minimal-versions. (#45) +- Fixed `#[may_dangle]` attributes to match the libstd `HashMap`. (#46) +- ZST keys and values are now handled properly. (#46) + +## [v0.1.8] - 2019-01-14 + +### Added +- Rayon parallel iterator support (#37) +- `raw_entry` support (#31) +- `#[may_dangle]` on nightly (#31) +- `try_reserve` support (#31) + +### Fixed +- Fixed variance on `IterMut`. (#31) + +## [v0.1.7] - 2018-12-05 + +### Fixed +- Fixed non-SSE version of convert_special_to_empty_and_full_to_deleted. (#32) +- Fixed overflow in rehash_in_place. (#33) + +## [v0.1.6] - 2018-11-17 + +### Fixed +- Fixed compile error on nightly. (#29) + +## [v0.1.5] - 2018-11-08 + +### Fixed +- Fixed subtraction overflow in generic::Group::match_byte. (#28) + +## [v0.1.4] - 2018-11-04 + +### Fixed +- Fixed a bug in the `erase_no_drop` implementation. (#26) + +## [v0.1.3] - 2018-11-01 + +### Added +- Serde support. (#14) + +### Fixed +- Make the compiler inline functions more aggressively. (#20) + +## [v0.1.2] - 2018-10-31 + +### Fixed +- `clear` segfaults when called on an empty table. (#13) + +## [v0.1.1] - 2018-10-30 + +### Fixed +- `erase_no_drop` optimization not triggering in the SSE2 implementation. (#3) +- Missing `Send` and `Sync` for hash map and iterator types. (#7) +- Bug when inserting into a table smaller than the group width. (#5) + +## v0.1.0 - 2018-10-29 + +- Initial release + +[Unreleased]: https://github.com/rust-lang/hashbrown/compare/v0.9.1...HEAD +[v0.9.1]: https://github.com/rust-lang/hashbrown/compare/v0.9.0...v0.9.1 +[v0.9.0]: https://github.com/rust-lang/hashbrown/compare/v0.8.2...v0.9.0 +[v0.8.2]: https://github.com/rust-lang/hashbrown/compare/v0.8.1...v0.8.2 +[v0.8.1]: https://github.com/rust-lang/hashbrown/compare/v0.8.0...v0.8.1 +[v0.8.0]: https://github.com/rust-lang/hashbrown/compare/v0.7.2...v0.8.0 +[v0.7.2]: https://github.com/rust-lang/hashbrown/compare/v0.7.1...v0.7.2 +[v0.7.1]: https://github.com/rust-lang/hashbrown/compare/v0.7.0...v0.7.1 +[v0.7.0]: https://github.com/rust-lang/hashbrown/compare/v0.6.3...v0.7.0 +[v0.6.3]: https://github.com/rust-lang/hashbrown/compare/v0.6.2...v0.6.3 +[v0.6.2]: https://github.com/rust-lang/hashbrown/compare/v0.6.1...v0.6.2 +[v0.6.1]: https://github.com/rust-lang/hashbrown/compare/v0.6.0...v0.6.1 +[v0.6.0]: https://github.com/rust-lang/hashbrown/compare/v0.5.1...v0.6.0 +[v0.5.1]: https://github.com/rust-lang/hashbrown/compare/v0.5.0...v0.5.1 +[v0.5.0]: https://github.com/rust-lang/hashbrown/compare/v0.4.0...v0.5.0 +[v0.4.0]: https://github.com/rust-lang/hashbrown/compare/v0.3.1...v0.4.0 +[v0.3.1]: https://github.com/rust-lang/hashbrown/compare/v0.3.0...v0.3.1 +[v0.3.0]: https://github.com/rust-lang/hashbrown/compare/v0.2.2...v0.3.0 +[v0.2.2]: https://github.com/rust-lang/hashbrown/compare/v0.2.1...v0.2.2 +[v0.2.1]: https://github.com/rust-lang/hashbrown/compare/v0.2.0...v0.2.1 +[v0.2.0]: https://github.com/rust-lang/hashbrown/compare/v0.1.8...v0.2.0 +[v0.1.8]: https://github.com/rust-lang/hashbrown/compare/v0.1.7...v0.1.8 +[v0.1.7]: https://github.com/rust-lang/hashbrown/compare/v0.1.6...v0.1.7 +[v0.1.6]: https://github.com/rust-lang/hashbrown/compare/v0.1.5...v0.1.6 +[v0.1.5]: https://github.com/rust-lang/hashbrown/compare/v0.1.4...v0.1.5 +[v0.1.4]: https://github.com/rust-lang/hashbrown/compare/v0.1.3...v0.1.4 +[v0.1.3]: https://github.com/rust-lang/hashbrown/compare/v0.1.2...v0.1.3 +[v0.1.2]: https://github.com/rust-lang/hashbrown/compare/v0.1.1...v0.1.2 +[v0.1.1]: https://github.com/rust-lang/hashbrown/compare/v0.1.0...v0.1.1 diff --git a/Cargo.toml b/Cargo.toml new file mode 100644 index 0000000..7be0341 --- /dev/null +++ b/Cargo.toml @@ -0,0 +1,80 @@ +# THIS FILE IS AUTOMATICALLY GENERATED BY CARGO +# +# When uploading crates to the registry Cargo will automatically +# "normalize" Cargo.toml files for maximal compatibility +# with all versions of Cargo and also rewrite `path` dependencies +# to registry (e.g., crates.io) dependencies +# +# If you believe there's an error in this file please file an +# issue against the rust-lang/cargo repository. If you're +# editing this file be aware that the upstream Cargo.toml +# will likely look very different (and much more reasonable) + +[package] +edition = "2018" +name = "hashbrown" +version = "0.9.1" +authors = ["Amanieu d'Antras <amanieu@gmail.com>"] +exclude = [".travis.yml", "bors.toml", "/ci/*"] +description = "A Rust port of Google's SwissTable hash map" +readme = "README.md" +keywords = ["hash", "no_std", "hashmap", "swisstable"] +categories = ["data-structures", "no-std"] +license = "Apache-2.0/MIT" +repository = "https://github.com/rust-lang/hashbrown" +[package.metadata.docs.rs] +features = ["nightly", "rayon", "serde", "raw"] +[dependencies.ahash] +version = "0.4.4" +optional = true +default-features = false + +[dependencies.alloc] +version = "1.0.0" +optional = true +package = "rustc-std-workspace-alloc" + +[dependencies.compiler_builtins] +version = "0.1.2" +optional = true + +[dependencies.core] +version = "1.0.0" +optional = true +package = "rustc-std-workspace-core" + +[dependencies.rayon] +version = "1.0" +optional = true + +[dependencies.serde] +version = "1.0.25" +optional = true +default-features = false +[dev-dependencies.doc-comment] +version = "0.3.1" + +[dev-dependencies.lazy_static] +version = "1.2" + +[dev-dependencies.rand] +version = "0.7.3" +features = ["small_rng"] + +[dev-dependencies.rayon] +version = "1.0" + +[dev-dependencies.rustc-hash] +version = "=1.0" + +[dev-dependencies.serde_test] +version = "1.0" + +[features] +ahash-compile-time-rng = ["ahash/compile-time-rng"] +default = ["ahash", "inline-more"] +inline-more = [] +nightly = [] +raw = [] +rustc-dep-of-std = ["nightly", "core", "compiler_builtins", "alloc", "rustc-internal-api"] +rustc-internal-api = [] diff --git a/Cargo.toml.orig b/Cargo.toml.orig new file mode 100644 index 0000000..21bd5c2 --- /dev/null +++ b/Cargo.toml.orig @@ -0,0 +1,56 @@ +[package] +name = "hashbrown" +version = "0.9.1" +authors = ["Amanieu d'Antras <amanieu@gmail.com>"] +description = "A Rust port of Google's SwissTable hash map" +license = "Apache-2.0/MIT" +repository = "https://github.com/rust-lang/hashbrown" +readme = "README.md" +keywords = ["hash", "no_std", "hashmap", "swisstable"] +categories = ["data-structures", "no-std"] +exclude = [".travis.yml", "bors.toml", "/ci/*"] +edition = "2018" + +[dependencies] +# For the default hasher +ahash = { version = "0.4.4", optional = true, default-features = false } + +# For external trait impls +rayon = { version = "1.0", optional = true } +serde = { version = "1.0.25", default-features = false, optional = true } + +# When built as part of libstd +core = { version = "1.0.0", optional = true, package = "rustc-std-workspace-core" } +compiler_builtins = { version = "0.1.2", optional = true } +alloc = { version = "1.0.0", optional = true, package = "rustc-std-workspace-alloc" } + +[dev-dependencies] +lazy_static = "1.2" +rand = { version = "0.7.3", features = ["small_rng"] } +rayon = "1.0" +rustc-hash = "=1.0" +serde_test = "1.0" +doc-comment = "0.3.1" + +[features] +default = ["ahash", "inline-more"] + +ahash-compile-time-rng = ["ahash/compile-time-rng"] +nightly = [] +rustc-internal-api = [] +rustc-dep-of-std = [ + "nightly", + "core", + "compiler_builtins", + "alloc", + "rustc-internal-api", +] +raw = [] + +# Enables usage of `#[inline]` on far more functions than by default in this +# crate. This may lead to a performance increase but often comes at a compile +# time cost. +inline-more = [] + +[package.metadata.docs.rs] +features = ["nightly", "rayon", "serde", "raw"] @@ -0,0 +1 @@ +LICENSE-APACHE
\ No newline at end of file diff --git a/LICENSE-APACHE b/LICENSE-APACHE new file mode 100644 index 0000000..16fe87b --- /dev/null +++ b/LICENSE-APACHE @@ -0,0 +1,201 @@ + 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. diff --git a/LICENSE-MIT b/LICENSE-MIT new file mode 100644 index 0000000..5afc2a7 --- /dev/null +++ b/LICENSE-MIT @@ -0,0 +1,25 @@ +Copyright (c) 2016 Amanieu d'Antras + +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: + +The above copyright notice and this permission notice +shall be included in all copies or substantial portions +of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF +ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED +TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A +PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT +SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY +CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR +IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER +DEALINGS IN THE SOFTWARE. diff --git a/METADATA b/METADATA new file mode 100644 index 0000000..b3d1a7c --- /dev/null +++ b/METADATA @@ -0,0 +1,19 @@ +name: "hashbrown" +description: "A Rust port of Google\'s SwissTable hash map" +third_party { + url { + type: HOMEPAGE + value: "https://crates.io/crates/hashbrown" + } + url { + type: ARCHIVE + value: "https://static.crates.io/crates/hashbrown/hashbrown-0.9.1.crate" + } + version: "0.9.1" + license_type: NOTICE + last_upgrade_date { + year: 2020 + month: 11 + day: 9 + } +} diff --git a/MODULE_LICENSE_APACHE2 b/MODULE_LICENSE_APACHE2 new file mode 100644 index 0000000..e69de29 --- /dev/null +++ b/MODULE_LICENSE_APACHE2 @@ -0,0 +1 @@ +include platform/prebuilts/rust:/OWNERS diff --git a/README.md b/README.md new file mode 100644 index 0000000..2e43171 --- /dev/null +++ b/README.md @@ -0,0 +1,126 @@ +hashbrown +========= + +[](https://travis-ci.com/rust-lang/hashbrown) +[](https://crates.io/crates/hashbrown) +[](https://docs.rs/hashbrown) +[](https://github.com/rust-lang/hashbrown) + +This crate is a Rust port of Google's high-performance [SwissTable] hash +map, adapted to make it a drop-in replacement for Rust's standard `HashMap` +and `HashSet` types. + +The original C++ version of SwissTable can be found [here], and this +[CppCon talk] gives an overview of how the algorithm works. + +Since Rust 1.36, this is now the `HashMap` implementation for the Rust standard +library. However you may still want to use this crate instead since it works +in environments without `std`, such as embedded systems and kernels. + +[SwissTable]: https://abseil.io/blog/20180927-swisstables +[here]: https://github.com/abseil/abseil-cpp/blob/master/absl/container/internal/raw_hash_set.h +[CppCon talk]: https://www.youtube.com/watch?v=ncHmEUmJZf4 + +## [Change log](CHANGELOG.md) + +## Features + +- Drop-in replacement for the standard library `HashMap` and `HashSet` types. +- Uses `AHash` as the default hasher, which is much faster than SipHash. +- Around 2x faster than the previous standard library `HashMap`. +- Lower memory usage: only 1 byte of overhead per entry instead of 8. +- Compatible with `#[no_std]` (but requires a global allocator with the `alloc` crate). +- Empty hash maps do not allocate any memory. +- SIMD lookups to scan multiple hash entries in parallel. + +## Performance + +Compared to the previous implementation of `std::collections::HashMap` (Rust 1.35). + +With the hashbrown default AHash hasher (not HashDoS-resistant): + +```text + name oldstdhash ns/iter hashbrown ns/iter diff ns/iter diff % speedup + insert_ahash_highbits 20,846 7,397 -13,449 -64.52% x 2.82 + insert_ahash_random 20,515 7,796 -12,719 -62.00% x 2.63 + insert_ahash_serial 21,668 7,264 -14,404 -66.48% x 2.98 + insert_erase_ahash_highbits 29,570 17,498 -12,072 -40.83% x 1.69 + insert_erase_ahash_random 39,569 17,474 -22,095 -55.84% x 2.26 + insert_erase_ahash_serial 32,073 17,332 -14,741 -45.96% x 1.85 + iter_ahash_highbits 1,572 2,087 515 32.76% x 0.75 + iter_ahash_random 1,609 2,074 465 28.90% x 0.78 + iter_ahash_serial 2,293 2,120 -173 -7.54% x 1.08 + lookup_ahash_highbits 3,460 4,403 943 27.25% x 0.79 + lookup_ahash_random 6,377 3,911 -2,466 -38.67% x 1.63 + lookup_ahash_serial 3,629 3,586 -43 -1.18% x 1.01 + lookup_fail_ahash_highbits 5,286 3,411 -1,875 -35.47% x 1.55 + lookup_fail_ahash_random 12,365 4,171 -8,194 -66.27% x 2.96 + lookup_fail_ahash_serial 4,902 3,240 -1,662 -33.90% x 1.51 +``` + +With the libstd default SipHash hasher (HashDoS-resistant): + +```text + name oldstdhash ns/iter hashbrown ns/iter diff ns/iter diff % speedup + insert_std_highbits 32,598 20,199 -12,399 -38.04% x 1.61 + insert_std_random 29,824 20,760 -9,064 -30.39% x 1.44 + insert_std_serial 33,151 17,256 -15,895 -47.95% x 1.92 + insert_erase_std_highbits 74,731 48,735 -25,996 -34.79% x 1.53 + insert_erase_std_random 73,828 47,649 -26,179 -35.46% x 1.55 + insert_erase_std_serial 73,864 40,147 -33,717 -45.65% x 1.84 + iter_std_highbits 1,518 2,264 746 49.14% x 0.67 + iter_std_random 1,502 2,414 912 60.72% x 0.62 + iter_std_serial 6,361 2,118 -4,243 -66.70% x 3.00 + lookup_std_highbits 21,705 16,962 -4,743 -21.85% x 1.28 + lookup_std_random 21,654 17,158 -4,496 -20.76% x 1.26 + lookup_std_serial 18,726 14,509 -4,217 -22.52% x 1.29 + lookup_fail_std_highbits 25,852 17,323 -8,529 -32.99% x 1.49 + lookup_fail_std_random 25,913 17,760 -8,153 -31.46% x 1.46 + lookup_fail_std_serial 22,648 14,839 -7,809 -34.48% x 1.53 +``` + +## Usage + +Add this to your `Cargo.toml`: + +```toml +[dependencies] +hashbrown = "0.9" +``` + +Then: + +```rust +use hashbrown::HashMap; + +let mut map = HashMap::new(); +map.insert(1, "one"); +``` + +This crate has the following Cargo features: + +- `nightly`: Enables nightly-only features: `#[may_dangle]`. +- `serde`: Enables serde serialization support. +- `rayon`: Enables rayon parallel iterator support. +- `raw`: Enables access to the experimental and unsafe `RawTable` API. +- `inline-more`: Adds inline hints to most functions, improving run-time performance at the cost + of compilation time. (enabled by default) +- `ahash`: Compiles with ahash as default hasher. (enabled by default) +- `ahash-compile-time-rng`: Activates the `compile-time-rng` feature of ahash, to increase the + DOS-resistance, but can result in issues for `no_std` builds. More details in + [issue#124](https://github.com/rust-lang/hashbrown/issues/124). (enabled by default) + +## License + +Licensed under either of: + + * Apache License, Version 2.0, ([LICENSE-APACHE](LICENSE-APACHE) or http://www.apache.org/licenses/LICENSE-2.0) + * MIT license ([LICENSE-MIT](LICENSE-MIT) or http://opensource.org/licenses/MIT) + +at your option. + +### Contribution + +Unless you explicitly state otherwise, any contribution intentionally submitted +for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any +additional terms or conditions. diff --git a/benches/bench.rs b/benches/bench.rs new file mode 100644 index 0000000..771e716 --- /dev/null +++ b/benches/bench.rs @@ -0,0 +1,260 @@ +// This benchmark suite contains some benchmarks along a set of dimensions: +// Hasher: std default (SipHash) and crate default (AHash). +// Int key distribution: low bit heavy, top bit heavy, and random. +// Task: basic functionality: insert, insert_erase, lookup, lookup_fail, iter +#![feature(test)] + +extern crate test; + +use test::{black_box, Bencher}; + +use hashbrown::hash_map::DefaultHashBuilder; +use hashbrown::HashMap; +use std::collections::hash_map::RandomState; + +const SIZE: usize = 1000; + +// The default hashmap when using this crate directly. +type AHashMap<K, V> = HashMap<K, V, DefaultHashBuilder>; +// This uses the hashmap from this crate with the default hasher of the stdlib. +type StdHashMap<K, V> = HashMap<K, V, RandomState>; + +// A random key iterator. +#[derive(Clone, Copy)] +struct RandomKeys { + state: usize, +} + +impl RandomKeys { + fn new() -> Self { + RandomKeys { state: 0 } + } +} + +impl Iterator for RandomKeys { + type Item = usize; + fn next(&mut self) -> Option<usize> { + // Add 1 then multiply by some 32 bit prime. + self.state = self.state.wrapping_add(1).wrapping_mul(3787392781); + Some(self.state) + } +} + +macro_rules! bench_suite { + ($bench_macro:ident, $bench_ahash_serial:ident, $bench_std_serial:ident, + $bench_ahash_highbits:ident, $bench_std_highbits:ident, + $bench_ahash_random:ident, $bench_std_random:ident) => { + $bench_macro!($bench_ahash_serial, AHashMap, 0..); + $bench_macro!($bench_std_serial, StdHashMap, 0..); + $bench_macro!( + $bench_ahash_highbits, + AHashMap, + (0..).map(usize::swap_bytes) + ); + $bench_macro!( + $bench_std_highbits, + StdHashMap, + (0..).map(usize::swap_bytes) + ); + $bench_macro!($bench_ahash_random, AHashMap, RandomKeys::new()); + $bench_macro!($bench_std_random, StdHashMap, RandomKeys::new()); + }; +} + +macro_rules! bench_insert { + ($name:ident, $maptype:ident, $keydist:expr) => { + #[bench] + fn $name(b: &mut Bencher) { + let mut m = $maptype::with_capacity_and_hasher(SIZE, Default::default()); + b.iter(|| { + m.clear(); + for i in ($keydist).take(SIZE) { + m.insert(i, i); + } + black_box(&mut m); + }) + } + }; +} + +bench_suite!( + bench_insert, + insert_ahash_serial, + insert_std_serial, + insert_ahash_highbits, + insert_std_highbits, + insert_ahash_random, + insert_std_random +); + +macro_rules! bench_insert_erase { + ($name:ident, $maptype:ident, $keydist:expr) => { + #[bench] + fn $name(b: &mut Bencher) { + let mut base = $maptype::default(); + for i in ($keydist).take(SIZE) { + base.insert(i, i); + } + let skip = $keydist.skip(SIZE); + b.iter(|| { + let mut m = base.clone(); + let mut add_iter = skip.clone(); + let mut remove_iter = $keydist; + // While keeping the size constant, + // replace the first keydist with the second. + for (add, remove) in (&mut add_iter).zip(&mut remove_iter).take(SIZE) { + m.insert(add, add); + black_box(m.remove(&remove)); + } + black_box(m); + }) + } + }; +} + +bench_suite!( + bench_insert_erase, + insert_erase_ahash_serial, + insert_erase_std_serial, + insert_erase_ahash_highbits, + insert_erase_std_highbits, + insert_erase_ahash_random, + insert_erase_std_random +); + +macro_rules! bench_lookup { + ($name:ident, $maptype:ident, $keydist:expr) => { + #[bench] + fn $name(b: &mut Bencher) { + let mut m = $maptype::default(); + for i in $keydist.take(SIZE) { + m.insert(i, i); + } + + b.iter(|| { + for i in $keydist.take(SIZE) { + black_box(m.get(&i)); + } + }) + } + }; +} + +bench_suite!( + bench_lookup, + lookup_ahash_serial, + lookup_std_serial, + lookup_ahash_highbits, + lookup_std_highbits, + lookup_ahash_random, + lookup_std_random +); + +macro_rules! bench_lookup_fail { + ($name:ident, $maptype:ident, $keydist:expr) => { + #[bench] + fn $name(b: &mut Bencher) { + let mut m = $maptype::default(); + let mut iter = $keydist; + for i in (&mut iter).take(SIZE) { + m.insert(i, i); + } + + b.iter(|| { + for i in (&mut iter).take(SIZE) { + black_box(m.get(&i)); + } + }) + } + }; +} + +bench_suite!( + bench_lookup_fail, + lookup_fail_ahash_serial, + lookup_fail_std_serial, + lookup_fail_ahash_highbits, + lookup_fail_std_highbits, + lookup_fail_ahash_random, + lookup_fail_std_random +); + +macro_rules! bench_iter { + ($name:ident, $maptype:ident, $keydist:expr) => { + #[bench] + fn $name(b: &mut Bencher) { + let mut m = $maptype::default(); + for i in ($keydist).take(SIZE) { + m.insert(i, i); + } + + b.iter(|| { + for i in &m { + black_box(i); + } + }) + } + }; +} + +bench_suite!( + bench_iter, + iter_ahash_serial, + iter_std_serial, + iter_ahash_highbits, + iter_std_highbits, + iter_ahash_random, + iter_std_random +); + +#[bench] +fn clone_small(b: &mut Bencher) { + let mut m = HashMap::new(); + for i in 0..10 { + m.insert(i, i); + } + + b.iter(|| { + black_box(m.clone()); + }) +} + +#[bench] +fn clone_from_small(b: &mut Bencher) { + let mut m = HashMap::new(); + let mut m2 = HashMap::new(); + for i in 0..10 { + m.insert(i, i); + } + + b.iter(|| { + m2.clone_from(&m); + black_box(&mut m2); + }) +} + +#[bench] +fn clone_large(b: &mut Bencher) { + let mut m = HashMap::new(); + for i in 0..1000 { + m.insert(i, i); + } + + b.iter(|| { + black_box(m.clone()); + }) +} + +#[bench] +fn clone_from_large(b: &mut Bencher) { + let mut m = HashMap::new(); + let mut m2 = HashMap::new(); + for i in 0..1000 { + m.insert(i, i); + } + + b.iter(|| { + m2.clone_from(&m); + black_box(&mut m2); + }) +} diff --git a/clippy.toml b/clippy.toml new file mode 100644 index 0000000..d98bf2c --- /dev/null +++ b/clippy.toml @@ -0,0 +1 @@ +doc-valid-idents = [ "CppCon", "SwissTable", "SipHash", "HashDoS" ] diff --git a/src/external_trait_impls/mod.rs b/src/external_trait_impls/mod.rs new file mode 100644 index 0000000..ef49783 --- /dev/null +++ b/src/external_trait_impls/mod.rs @@ -0,0 +1,4 @@ +#[cfg(feature = "rayon")] +pub(crate) mod rayon; +#[cfg(feature = "serde")] +mod serde; diff --git a/src/external_trait_impls/rayon/helpers.rs b/src/external_trait_impls/rayon/helpers.rs new file mode 100644 index 0000000..9382007 --- /dev/null +++ b/src/external_trait_impls/rayon/helpers.rs @@ -0,0 +1,26 @@ +use alloc::collections::LinkedList; +use alloc::vec::Vec; + +use rayon::iter::{IntoParallelIterator, ParallelIterator}; + +/// Helper for collecting parallel iterators to an intermediary +pub(super) fn collect<I: IntoParallelIterator>(iter: I) -> (LinkedList<Vec<I::Item>>, usize) { + let list = iter + .into_par_iter() + .fold(Vec::new, |mut vec, elem| { + vec.push(elem); + vec + }) + .map(|vec| { + let mut list = LinkedList::new(); + list.push_back(vec); + list + }) + .reduce(LinkedList::new, |mut list1, mut list2| { + list1.append(&mut list2); + list1 + }); + + let len = list.iter().map(Vec::len).sum(); + (list, len) +} diff --git a/src/external_trait_impls/rayon/map.rs b/src/external_trait_impls/rayon/map.rs new file mode 100644 index 0000000..334f8bb --- /dev/null +++ b/src/external_trait_impls/rayon/map.rs @@ -0,0 +1,666 @@ +//! Rayon extensions for `HashMap`. + +use crate::hash_map::HashMap; +use core::fmt; +use core::hash::{BuildHasher, Hash}; +use rayon::iter::plumbing::UnindexedConsumer; +use rayon::iter::{FromParallelIterator, IntoParallelIterator, ParallelExtend, ParallelIterator}; + +/// Parallel iterator over shared references to entries in a map. +/// +/// This iterator is created by the [`par_iter`] method on [`HashMap`] +/// (provided by the [`IntoParallelRefIterator`] trait). +/// See its documentation for more. +/// +/// [`par_iter`]: /hashbrown/struct.HashMap.html#method.par_iter +/// [`HashMap`]: /hashbrown/struct.HashMap.html +/// [`IntoParallelRefIterator`]: https://docs.rs/rayon/1.0/rayon/iter/trait.IntoParallelRefIterator.html +pub struct ParIter<'a, K, V, S> { + map: &'a HashMap<K, V, S>, +} + +impl<'a, K: Sync, V: Sync, S: Sync> ParallelIterator for ParIter<'a, K, V, S> { + type Item = (&'a K, &'a V); + + #[cfg_attr(feature = "inline-more", inline)] + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + unsafe { self.map.table.par_iter() } + .map(|x| unsafe { + let r = x.as_ref(); + (&r.0, &r.1) + }) + .drive_unindexed(consumer) + } +} + +impl<K, V, S> Clone for ParIter<'_, K, V, S> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + ParIter { map: self.map } + } +} + +impl<K: fmt::Debug + Eq + Hash, V: fmt::Debug, S: BuildHasher> fmt::Debug for ParIter<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + self.map.iter().fmt(f) + } +} + +/// Parallel iterator over shared references to keys in a map. +/// +/// This iterator is created by the [`par_keys`] method on [`HashMap`]. +/// See its documentation for more. +/// +/// [`par_keys`]: /hashbrown/struct.HashMap.html#method.par_keys +/// [`HashMap`]: /hashbrown/struct.HashMap.html +pub struct ParKeys<'a, K, V, S> { + map: &'a HashMap<K, V, S>, +} + +impl<'a, K: Sync, V: Sync, S: Sync> ParallelIterator for ParKeys<'a, K, V, S> { + type Item = &'a K; + + #[cfg_attr(feature = "inline-more", inline)] + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + unsafe { self.map.table.par_iter() } + .map(|x| unsafe { &x.as_ref().0 }) + .drive_unindexed(consumer) + } +} + +impl<K, V, S> Clone for ParKeys<'_, K, V, S> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + ParKeys { map: self.map } + } +} + +impl<K: fmt::Debug + Eq + Hash, V, S: BuildHasher> fmt::Debug for ParKeys<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + self.map.keys().fmt(f) + } +} + +/// Parallel iterator over shared references to values in a map. +/// +/// This iterator is created by the [`par_values`] method on [`HashMap`]. +/// See its documentation for more. +/// +/// [`par_values`]: /hashbrown/struct.HashMap.html#method.par_values +/// [`HashMap`]: /hashbrown/struct.HashMap.html +pub struct ParValues<'a, K, V, S> { + map: &'a HashMap<K, V, S>, +} + +impl<'a, K: Sync, V: Sync, S: Sync> ParallelIterator for ParValues<'a, K, V, S> { + type Item = &'a V; + + #[cfg_attr(feature = "inline-more", inline)] + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + unsafe { self.map.table.par_iter() } + .map(|x| unsafe { &x.as_ref().1 }) + .drive_unindexed(consumer) + } +} + +impl<K, V, S> Clone for ParValues<'_, K, V, S> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + ParValues { map: self.map } + } +} + +impl<K: Eq + Hash, V: fmt::Debug, S: BuildHasher> fmt::Debug for ParValues<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + self.map.values().fmt(f) + } +} + +/// Parallel iterator over mutable references to entries in a map. +/// +/// This iterator is created by the [`par_iter_mut`] method on [`HashMap`] +/// (provided by the [`IntoParallelRefMutIterator`] trait). +/// See its documentation for more. +/// +/// [`par_iter_mut`]: /hashbrown/struct.HashMap.html#method.par_iter_mut +/// [`HashMap`]: /hashbrown/struct.HashMap.html +/// [`IntoParallelRefMutIterator`]: https://docs.rs/rayon/1.0/rayon/iter/trait.IntoParallelRefMutIterator.html +pub struct ParIterMut<'a, K, V, S> { + map: &'a mut HashMap<K, V, S>, +} + +impl<'a, K: Send + Sync, V: Send, S: Send> ParallelIterator for ParIterMut<'a, K, V, S> { + type Item = (&'a K, &'a mut V); + + #[cfg_attr(feature = "inline-more", inline)] + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + unsafe { self.map.table.par_iter() } + .map(|x| unsafe { + let r = x.as_mut(); + (&r.0, &mut r.1) + }) + .drive_unindexed(consumer) + } +} + +impl<K: fmt::Debug + Eq + Hash, V: fmt::Debug, S: BuildHasher> fmt::Debug + for ParIterMut<'_, K, V, S> +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + self.map.iter().fmt(f) + } +} + +/// Parallel iterator over mutable references to values in a map. +/// +/// This iterator is created by the [`par_values_mut`] method on [`HashMap`]. +/// See its documentation for more. +/// +/// [`par_values_mut`]: /hashbrown/struct.HashMap.html#method.par_values_mut +/// [`HashMap`]: /hashbrown/struct.HashMap.html +pub struct ParValuesMut<'a, K, V, S> { + map: &'a mut HashMap<K, V, S>, +} + +impl<'a, K: Send, V: Send, S: Send> ParallelIterator for ParValuesMut<'a, K, V, S> { + type Item = &'a mut V; + + #[cfg_attr(feature = "inline-more", inline)] + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + unsafe { self.map.table.par_iter() } + .map(|x| unsafe { &mut x.as_mut().1 }) + .drive_unindexed(consumer) + } +} + +impl<K: Eq + Hash, V: fmt::Debug, S: BuildHasher> fmt::Debug for ParValuesMut<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + self.map.values().fmt(f) + } +} + +/// Parallel iterator over entries of a consumed map. +/// +/// This iterator is created by the [`into_par_iter`] method on [`HashMap`] +/// (provided by the [`IntoParallelIterator`] trait). +/// See its documentation for more. +/// +/// [`into_par_iter`]: /hashbrown/struct.HashMap.html#method.into_par_iter +/// [`HashMap`]: /hashbrown/struct.HashMap.html +/// [`IntoParallelIterator`]: https://docs.rs/rayon/1.0/rayon/iter/trait.IntoParallelIterator.html +pub struct IntoParIter<K, V, S> { + map: HashMap<K, V, S>, +} + +impl<K: Send, V: Send, S: Send> ParallelIterator for IntoParIter<K, V, S> { + type Item = (K, V); + + #[cfg_attr(feature = "inline-more", inline)] + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + self.map.table.into_par_iter().drive_unindexed(consumer) + } +} + +impl<K: fmt::Debug + Eq + Hash, V: fmt::Debug, S: BuildHasher> fmt::Debug for IntoParIter<K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + self.map.iter().fmt(f) + } +} + +/// Parallel draining iterator over entries of a map. +/// +/// This iterator is created by the [`par_drain`] method on [`HashMap`]. +/// See its documentation for more. +/// +/// [`par_drain`]: /hashbrown/struct.HashMap.html#method.par_drain +/// [`HashMap`]: /hashbrown/struct.HashMap.html +pub struct ParDrain<'a, K, V, S> { + map: &'a mut HashMap<K, V, S>, +} + +impl<K: Send, V: Send, S: Send> ParallelIterator for ParDrain<'_, K, V, S> { + type Item = (K, V); + + #[cfg_attr(feature = "inline-more", inline)] + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + self.map.table.par_drain().drive_unindexed(consumer) + } +} + +impl<K: fmt::Debug + Eq + Hash, V: fmt::Debug, S: BuildHasher> fmt::Debug + for ParDrain<'_, K, V, S> +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + self.map.iter().fmt(f) + } +} + +impl<K: Sync, V: Sync, S: Sync> HashMap<K, V, S> { + /// Visits (potentially in parallel) immutably borrowed keys in an arbitrary order. + #[cfg_attr(feature = "inline-more", inline)] + pub fn par_keys(&self) -> ParKeys<'_, K, V, S> { + ParKeys { map: self } + } + + /// Visits (potentially in parallel) immutably borrowed values in an arbitrary order. + #[cfg_attr(feature = "inline-more", inline)] + pub fn par_values(&self) -> ParValues<'_, K, V, S> { + ParValues { map: self } + } +} + +impl<K: Send, V: Send, S: Send> HashMap<K, V, S> { + /// Visits (potentially in parallel) mutably borrowed values in an arbitrary order. + #[cfg_attr(feature = "inline-more", inline)] + pub fn par_values_mut(&mut self) -> ParValuesMut<'_, K, V, S> { + ParValuesMut { map: self } + } + + /// Consumes (potentially in parallel) all values in an arbitrary order, + /// while preserving the map's allocated memory for reuse. + #[cfg_attr(feature = "inline-more", inline)] + pub fn par_drain(&mut self) -> ParDrain<'_, K, V, S> { + ParDrain { map: self } + } +} + +impl<K, V, S> HashMap<K, V, S> +where + K: Eq + Hash + Sync, + V: PartialEq + Sync, + S: BuildHasher + Sync, +{ + /// Returns `true` if the map is equal to another, + /// i.e. both maps contain the same keys mapped to the same values. + /// + /// This method runs in a potentially parallel fashion. + pub fn par_eq(&self, other: &Self) -> bool { + self.len() == other.len() + && self + .into_par_iter() + .all(|(key, value)| other.get(key).map_or(false, |v| *value == *v)) + } +} + +impl<K: Send, V: Send, S: Send> IntoParallelIterator for HashMap<K, V, S> { + type Item = (K, V); + type Iter = IntoParIter<K, V, S>; + + #[cfg_attr(feature = "inline-more", inline)] + fn into_par_iter(self) -> Self::Iter { + IntoParIter { map: self } + } +} + +impl<'a, K: Sync, V: Sync, S: Sync> IntoParallelIterator for &'a HashMap<K, V, S> { + type Item = (&'a K, &'a V); + type Iter = ParIter<'a, K, V, S>; + + #[cfg_attr(feature = "inline-more", inline)] + fn into_par_iter(self) -> Self::Iter { + ParIter { map: self } + } +} + +impl<'a, K: Send + Sync, V: Send, S: Send> IntoParallelIterator for &'a mut HashMap<K, V, S> { + type Item = (&'a K, &'a mut V); + type Iter = ParIterMut<'a, K, V, S>; + + #[cfg_attr(feature = "inline-more", inline)] + fn into_par_iter(self) -> Self::Iter { + ParIterMut { map: self } + } +} + +/// Collect (key, value) pairs from a parallel iterator into a +/// hashmap. If multiple pairs correspond to the same key, then the +/// ones produced earlier in the parallel iterator will be +/// overwritten, just as with a sequential iterator. +impl<K, V, S> FromParallelIterator<(K, V)> for HashMap<K, V, S> +where + K: Eq + Hash + Send, + V: Send, + S: BuildHasher + Default, +{ + fn from_par_iter<P>(par_iter: P) -> Self + where + P: IntoParallelIterator<Item = (K, V)>, + { + let mut map = HashMap::default(); + map.par_extend(par_iter); + map + } +} + +/// Extend a hash map with items from a parallel iterator. +impl<K, V, S> ParallelExtend<(K, V)> for HashMap<K, V, S> +where + K: Eq + Hash + Send, + V: Send, + S: BuildHasher, +{ + fn par_extend<I>(&mut self, par_iter: I) + where + I: IntoParallelIterator<Item = (K, V)>, + { + extend(self, par_iter); + } +} + +/// Extend a hash map with copied items from a parallel iterator. +impl<'a, K, V, S> ParallelExtend<(&'a K, &'a V)> for HashMap<K, V, S> +where + K: Copy + Eq + Hash + Sync, + V: Copy + Sync, + S: BuildHasher, +{ + fn par_extend<I>(&mut self, par_iter: I) + where + I: IntoParallelIterator<Item = (&'a K, &'a V)>, + { + extend(self, par_iter); + } +} + +// This is equal to the normal `HashMap` -- no custom advantage. +fn extend<K, V, S, I>(map: &mut HashMap<K, V, S>, par_iter: I) +where + K: Eq + Hash, + S: BuildHasher, + I: IntoParallelIterator, + HashMap<K, V, S>: Extend<I::Item>, +{ + let (list, len) = super::helpers::collect(par_iter); + + // Keys may be already present or show multiple times in the iterator. + // Reserve the entire length if the map is empty. + // Otherwise reserve half the length (rounded up), so the map + // will only resize twice in the worst case. + let reserve = if map.is_empty() { len } else { (len + 1) / 2 }; + map.reserve(reserve); + for vec in list { + map.extend(vec); + } +} + +#[cfg(test)] +mod test_par_map { + use alloc::vec::Vec; + use core::hash::{Hash, Hasher}; + use core::sync::atomic::{AtomicUsize, Ordering}; + + use rayon::prelude::*; + + use crate::hash_map::HashMap; + + struct Dropable<'a> { + k: usize, + counter: &'a AtomicUsize, + } + + impl Dropable<'_> { + fn new(k: usize, counter: &AtomicUsize) -> Dropable<'_> { + counter.fetch_add(1, Ordering::Relaxed); + + Dropable { k, counter } + } + } + + impl Drop for Dropable<'_> { + fn drop(&mut self) { + self.counter.fetch_sub(1, Ordering::Relaxed); + } + } + + impl Clone for Dropable<'_> { + fn clone(&self) -> Self { + Dropable::new(self.k, self.counter) + } + } + + impl Hash for Dropable<'_> { + fn hash<H>(&self, state: &mut H) + where + H: Hasher, + { + self.k.hash(state) + } + } + + impl PartialEq for Dropable<'_> { + fn eq(&self, other: &Self) -> bool { + self.k == other.k + } + } + + impl Eq for Dropable<'_> {} + + #[test] + fn test_into_iter_drops() { + let key = AtomicUsize::new(0); + let value = AtomicUsize::new(0); + + let hm = { + let mut hm = HashMap::new(); + + assert_eq!(key.load(Ordering::Relaxed), 0); + assert_eq!(value.load(Ordering::Relaxed), 0); + + for i in 0..100 { + let d1 = Dropable::new(i, &key); + let d2 = Dropable::new(i + 100, &value); + hm.insert(d1, d2); + } + + assert_eq!(key.load(Ordering::Relaxed), 100); + assert_eq!(value.load(Ordering::Relaxed), 100); + + hm + }; + + // By the way, ensure that cloning doesn't screw up the dropping. + drop(hm.clone()); + + assert_eq!(key.load(Ordering::Relaxed), 100); + assert_eq!(value.load(Ordering::Relaxed), 100); + + // Ensure that dropping the iterator does not leak anything. + drop(hm.clone().into_par_iter()); + + { + assert_eq!(key.load(Ordering::Relaxed), 100); + assert_eq!(value.load(Ordering::Relaxed), 100); + + // retain only half + let _v: Vec<_> = hm + .into_par_iter() + .filter(|&(ref key, _)| key.k < 50) + .collect(); + + assert_eq!(key.load(Ordering::Relaxed), 50); + assert_eq!(value.load(Ordering::Relaxed), 50); + }; + + assert_eq!(key.load(Ordering::Relaxed), 0); + assert_eq!(value.load(Ordering::Relaxed), 0); + } + + #[test] + fn test_drain_drops() { + let key = AtomicUsize::new(0); + let value = AtomicUsize::new(0); + + let mut hm = { + let mut hm = HashMap::new(); + + assert_eq!(key.load(Ordering::Relaxed), 0); + assert_eq!(value.load(Ordering::Relaxed), 0); + + for i in 0..100 { + let d1 = Dropable::new(i, &key); + let d2 = Dropable::new(i + 100, &value); + hm.insert(d1, d2); + } + + assert_eq!(key.load(Ordering::Relaxed), 100); + assert_eq!(value.load(Ordering::Relaxed), 100); + + hm + }; + + // By the way, ensure that cloning doesn't screw up the dropping. + drop(hm.clone()); + + assert_eq!(key.load(Ordering::Relaxed), 100); + assert_eq!(value.load(Ordering::Relaxed), 100); + + // Ensure that dropping the drain iterator does not leak anything. + drop(hm.clone().par_drain()); + + { + assert_eq!(key.load(Ordering::Relaxed), 100); + assert_eq!(value.load(Ordering::Relaxed), 100); + + // retain only half + let _v: Vec<_> = hm.drain().filter(|&(ref key, _)| key.k < 50).collect(); + assert!(hm.is_empty()); + + assert_eq!(key.load(Ordering::Relaxed), 50); + assert_eq!(value.load(Ordering::Relaxed), 50); + }; + + assert_eq!(key.load(Ordering::Relaxed), 0); + assert_eq!(value.load(Ordering::Relaxed), 0); + } + + #[test] + fn test_empty_iter() { + let mut m: HashMap<isize, bool> = HashMap::new(); + assert_eq!(m.par_drain().count(), 0); + assert_eq!(m.par_keys().count(), 0); + assert_eq!(m.par_values().count(), 0); + assert_eq!(m.par_values_mut().count(), 0); + assert_eq!(m.par_iter().count(), 0); + assert_eq!(m.par_iter_mut().count(), 0); + assert_eq!(m.len(), 0); + assert!(m.is_empty()); + assert_eq!(m.into_par_iter().count(), 0); + } + + #[test] + fn test_iterate() { + let mut m = HashMap::with_capacity(4); + for i in 0..32 { + assert!(m.insert(i, i * 2).is_none()); + } + assert_eq!(m.len(), 32); + + let observed = AtomicUsize::new(0); + + m.par_iter().for_each(|(k, v)| { + assert_eq!(*v, *k * 2); + observed.fetch_or(1 << *k, Ordering::Relaxed); + }); + assert_eq!(observed.into_inner(), 0xFFFF_FFFF); + } + + #[test] + fn test_keys() { + let vec = vec![(1, 'a'), (2, 'b'), (3, 'c')]; + let map: HashMap<_, _> = vec.into_par_iter().collect(); + let keys: Vec<_> = map.par_keys().cloned().collect(); + assert_eq!(keys.len(), 3); + assert!(keys.contains(&1)); + assert!(keys.contains(&2)); + assert!(keys.contains(&3)); + } + + #[test] + fn test_values() { + let vec = vec![(1, 'a'), (2, 'b'), (3, 'c')]; + let map: HashMap<_, _> = vec.into_par_iter().collect(); + let values: Vec<_> = map.par_values().cloned().collect(); + assert_eq!(values.len(), 3); + assert!(values.contains(&'a')); + assert!(values.contains(&'b')); + assert!(values.contains(&'c')); + } + + #[test] + fn test_values_mut() { + let vec = vec![(1, 1), (2, 2), (3, 3)]; + let mut map: HashMap<_, _> = vec.into_par_iter().collect(); + map.par_values_mut().for_each(|value| *value = (*value) * 2); + let values: Vec<_> = map.par_values().cloned().collect(); + assert_eq!(values.len(), 3); + assert!(values.contains(&2)); + assert!(values.contains(&4)); + assert!(values.contains(&6)); + } + + #[test] + fn test_eq() { + let mut m1 = HashMap::new(); + m1.insert(1, 2); + m1.insert(2, 3); + m1.insert(3, 4); + + let mut m2 = HashMap::new(); + m2.insert(1, 2); + m2.insert(2, 3); + + assert!(!m1.par_eq(&m2)); + + m2.insert(3, 4); + + assert!(m1.par_eq(&m2)); + } + + #[test] + fn test_from_iter() { + let xs = [(1, 1), (2, 2), (3, 3), (4, 4), (5, 5), (6, 6)]; + + let map: HashMap<_, _> = xs.par_iter().cloned().collect(); + + for &(k, v) in &xs { + assert_eq!(map.get(&k), Some(&v)); + } + } + + #[test] + fn test_extend_ref() { + let mut a = HashMap::new(); + a.insert(1, "one"); + let mut b = HashMap::new(); + b.insert(2, "two"); + b.insert(3, "three"); + + a.par_extend(&b); + + assert_eq!(a.len(), 3); + assert_eq!(a[&1], "one"); + assert_eq!(a[&2], "two"); + assert_eq!(a[&3], "three"); + } +} diff --git a/src/external_trait_impls/rayon/mod.rs b/src/external_trait_impls/rayon/mod.rs new file mode 100644 index 0000000..99337a1 --- /dev/null +++ b/src/external_trait_impls/rayon/mod.rs @@ -0,0 +1,4 @@ +mod helpers; +pub(crate) mod map; +pub(crate) mod raw; +pub(crate) mod set; diff --git a/src/external_trait_impls/rayon/raw.rs b/src/external_trait_impls/rayon/raw.rs new file mode 100644 index 0000000..1bd2c17 --- /dev/null +++ b/src/external_trait_impls/rayon/raw.rs @@ -0,0 +1,199 @@ +use crate::raw::Bucket; +use crate::raw::{RawIter, RawIterRange, RawTable}; +use crate::scopeguard::guard; +use alloc::alloc::dealloc; +use core::marker::PhantomData; +use core::mem; +use core::ptr::NonNull; +use rayon::iter::{ + plumbing::{self, Folder, UnindexedConsumer, UnindexedProducer}, + ParallelIterator, +}; + +/// Parallel iterator which returns a raw pointer to every full bucket in the table. +pub struct RawParIter<T> { + iter: RawIterRange<T>, +} + +impl<T> From<RawIter<T>> for RawParIter<T> { + fn from(it: RawIter<T>) -> Self { + RawParIter { iter: it.iter } + } +} + +impl<T> ParallelIterator for RawParIter<T> { + type Item = Bucket<T>; + + #[cfg_attr(feature = "inline-more", inline)] + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + let producer = ParIterProducer { iter: self.iter }; + plumbing::bridge_unindexed(producer, consumer) + } +} + +/// Producer which returns a `Bucket<T>` for every element. +struct ParIterProducer<T> { + iter: RawIterRange<T>, +} + +impl<T> UnindexedProducer for ParIterProducer<T> { + type Item = Bucket<T>; + + #[cfg_attr(feature = "inline-more", inline)] + fn split(self) -> (Self, Option<Self>) { + let (left, right) = self.iter.split(); + let left = ParIterProducer { iter: left }; + let right = right.map(|right| ParIterProducer { iter: right }); + (left, right) + } + + #[cfg_attr(feature = "inline-more", inline)] + fn fold_with<F>(self, folder: F) -> F + where + F: Folder<Self::Item>, + { + folder.consume_iter(self.iter) + } +} + +/// Parallel iterator which consumes a table and returns elements. +pub struct RawIntoParIter<T> { + table: RawTable<T>, +} + +impl<T: Send> ParallelIterator for RawIntoParIter<T> { + type Item = T; + + #[cfg_attr(feature = "inline-more", inline)] + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + let iter = unsafe { self.table.iter().iter }; + let _guard = guard(self.table.into_alloc(), |alloc| { + if let Some((ptr, layout)) = *alloc { + unsafe { + dealloc(ptr.as_ptr(), layout); + } + } + }); + let producer = ParDrainProducer { iter }; + plumbing::bridge_unindexed(producer, consumer) + } +} + +/// Parallel iterator which consumes elements without freeing the table storage. +pub struct RawParDrain<'a, T> { + // We don't use a &'a mut RawTable<T> because we want RawParDrain to be + // covariant over T. + table: NonNull<RawTable<T>>, + marker: PhantomData<&'a RawTable<T>>, +} + +unsafe impl<T> Send for RawParDrain<'_, T> {} + +impl<T: Send> ParallelIterator for RawParDrain<'_, T> { + type Item = T; + + #[cfg_attr(feature = "inline-more", inline)] + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + let _guard = guard(self.table, |table| unsafe { + table.as_mut().clear_no_drop() + }); + let iter = unsafe { self.table.as_ref().iter().iter }; + mem::forget(self); + let producer = ParDrainProducer { iter }; + plumbing::bridge_unindexed(producer, consumer) + } +} + +impl<T> Drop for RawParDrain<'_, T> { + fn drop(&mut self) { + // If drive_unindexed is not called then simply clear the table. + unsafe { self.table.as_mut().clear() } + } +} + +/// Producer which will consume all elements in the range, even if it is dropped +/// halfway through. +struct ParDrainProducer<T> { + iter: RawIterRange<T>, +} + +impl<T: Send> UnindexedProducer for ParDrainProducer<T> { + type Item = T; + + #[cfg_attr(feature = "inline-more", inline)] + fn split(self) -> (Self, Option<Self>) { + let (left, right) = self.iter.clone().split(); + mem::forget(self); + let left = ParDrainProducer { iter: left }; + let right = right.map(|right| ParDrainProducer { iter: right }); + (left, right) + } + + #[cfg_attr(feature = "inline-more", inline)] + fn fold_with<F>(mut self, mut folder: F) -> F + where + F: Folder<Self::Item>, + { + // Make sure to modify the iterator in-place so that any remaining + // elements are processed in our Drop impl. + while let Some(item) = self.iter.next() { + folder = folder.consume(unsafe { item.read() }); + if folder.full() { + return folder; + } + } + + // If we processed all elements then we don't need to run the drop. + mem::forget(self); + folder + } +} + +impl<T> Drop for ParDrainProducer<T> { + #[cfg_attr(feature = "inline-more", inline)] + fn drop(&mut self) { + // Drop all remaining elements + if mem::needs_drop::<T>() { + while let Some(item) = self.iter.next() { + unsafe { + item.drop(); + } + } + } + } +} + +impl<T> RawTable<T> { + /// Returns a parallel iterator over the elements in a `RawTable`. + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn par_iter(&self) -> RawParIter<T> { + RawParIter { + iter: self.iter().iter, + } + } + + /// Returns a parallel iterator over the elements in a `RawTable`. + #[cfg_attr(feature = "inline-more", inline)] + pub fn into_par_iter(self) -> RawIntoParIter<T> { + RawIntoParIter { table: self } + } + + /// Returns a parallel iterator which consumes all elements of a `RawTable` + /// without freeing its memory allocation. + #[cfg_attr(feature = "inline-more", inline)] + pub fn par_drain(&mut self) -> RawParDrain<'_, T> { + RawParDrain { + table: NonNull::from(self), + marker: PhantomData, + } + } +} diff --git a/src/external_trait_impls/rayon/set.rs b/src/external_trait_impls/rayon/set.rs new file mode 100644 index 0000000..53d2660 --- /dev/null +++ b/src/external_trait_impls/rayon/set.rs @@ -0,0 +1,646 @@ +//! Rayon extensions for `HashSet`. + +use crate::hash_set::HashSet; +use core::hash::{BuildHasher, Hash}; +use rayon::iter::plumbing::UnindexedConsumer; +use rayon::iter::{FromParallelIterator, IntoParallelIterator, ParallelExtend, ParallelIterator}; + +/// Parallel iterator over elements of a consumed set. +/// +/// This iterator is created by the [`into_par_iter`] method on [`HashSet`] +/// (provided by the [`IntoParallelIterator`] trait). +/// See its documentation for more. +/// +/// [`into_par_iter`]: /hashbrown/struct.HashSet.html#method.into_par_iter +/// [`HashSet`]: /hashbrown/struct.HashSet.html +/// [`IntoParallelIterator`]: https://docs.rs/rayon/1.0/rayon/iter/trait.IntoParallelIterator.html +pub struct IntoParIter<T, S> { + set: HashSet<T, S>, +} + +impl<T: Send, S: Send> ParallelIterator for IntoParIter<T, S> { + type Item = T; + + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + self.set + .map + .into_par_iter() + .map(|(k, _)| k) + .drive_unindexed(consumer) + } +} + +/// Parallel draining iterator over entries of a set. +/// +/// This iterator is created by the [`par_drain`] method on [`HashSet`]. +/// See its documentation for more. +/// +/// [`par_drain`]: /hashbrown/struct.HashSet.html#method.par_drain +/// [`HashSet`]: /hashbrown/struct.HashSet.html +pub struct ParDrain<'a, T, S> { + set: &'a mut HashSet<T, S>, +} + +impl<T: Send, S: Send> ParallelIterator for ParDrain<'_, T, S> { + type Item = T; + + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + self.set + .map + .par_drain() + .map(|(k, _)| k) + .drive_unindexed(consumer) + } +} + +/// Parallel iterator over shared references to elements in a set. +/// +/// This iterator is created by the [`par_iter`] method on [`HashSet`] +/// (provided by the [`IntoParallelRefIterator`] trait). +/// See its documentation for more. +/// +/// [`par_iter`]: /hashbrown/struct.HashSet.html#method.par_iter +/// [`HashSet`]: /hashbrown/struct.HashSet.html +/// [`IntoParallelRefIterator`]: https://docs.rs/rayon/1.0/rayon/iter/trait.IntoParallelRefIterator.html +pub struct ParIter<'a, T, S> { + set: &'a HashSet<T, S>, +} + +impl<'a, T: Sync, S: Sync> ParallelIterator for ParIter<'a, T, S> { + type Item = &'a T; + + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + self.set.map.par_keys().drive_unindexed(consumer) + } +} + +/// Parallel iterator over shared references to elements in the difference of +/// sets. +/// +/// This iterator is created by the [`par_difference`] method on [`HashSet`]. +/// See its documentation for more. +/// +/// [`par_difference`]: /hashbrown/struct.HashSet.html#method.par_difference +/// [`HashSet`]: /hashbrown/struct.HashSet.html +pub struct ParDifference<'a, T, S> { + a: &'a HashSet<T, S>, + b: &'a HashSet<T, S>, +} + +impl<'a, T, S> ParallelIterator for ParDifference<'a, T, S> +where + T: Eq + Hash + Sync, + S: BuildHasher + Sync, +{ + type Item = &'a T; + + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + self.a + .into_par_iter() + .filter(|&x| !self.b.contains(x)) + .drive_unindexed(consumer) + } +} + +/// Parallel iterator over shared references to elements in the symmetric +/// difference of sets. +/// +/// This iterator is created by the [`par_symmetric_difference`] method on +/// [`HashSet`]. +/// See its documentation for more. +/// +/// [`par_symmetric_difference`]: /hashbrown/struct.HashSet.html#method.par_symmetric_difference +/// [`HashSet`]: /hashbrown/struct.HashSet.html +pub struct ParSymmetricDifference<'a, T, S> { + a: &'a HashSet<T, S>, + b: &'a HashSet<T, S>, +} + +impl<'a, T, S> ParallelIterator for ParSymmetricDifference<'a, T, S> +where + T: Eq + Hash + Sync, + S: BuildHasher + Sync, +{ + type Item = &'a T; + + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + self.a + .par_difference(self.b) + .chain(self.b.par_difference(self.a)) + .drive_unindexed(consumer) + } +} + +/// Parallel iterator over shared references to elements in the intersection of +/// sets. +/// +/// This iterator is created by the [`par_intersection`] method on [`HashSet`]. +/// See its documentation for more. +/// +/// [`par_intersection`]: /hashbrown/struct.HashSet.html#method.par_intersection +/// [`HashSet`]: /hashbrown/struct.HashSet.html +pub struct ParIntersection<'a, T, S> { + a: &'a HashSet<T, S>, + b: &'a HashSet<T, S>, +} + +impl<'a, T, S> ParallelIterator for ParIntersection<'a, T, S> +where + T: Eq + Hash + Sync, + S: BuildHasher + Sync, +{ + type Item = &'a T; + + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + self.a + .into_par_iter() + .filter(|&x| self.b.contains(x)) + .drive_unindexed(consumer) + } +} + +/// Parallel iterator over shared references to elements in the union of sets. +/// +/// This iterator is created by the [`par_union`] method on [`HashSet`]. +/// See its documentation for more. +/// +/// [`par_union`]: /hashbrown/struct.HashSet.html#method.par_union +/// [`HashSet`]: /hashbrown/struct.HashSet.html +pub struct ParUnion<'a, T, S> { + a: &'a HashSet<T, S>, + b: &'a HashSet<T, S>, +} + +impl<'a, T, S> ParallelIterator for ParUnion<'a, T, S> +where + T: Eq + Hash + Sync, + S: BuildHasher + Sync, +{ + type Item = &'a T; + + fn drive_unindexed<C>(self, consumer: C) -> C::Result + where + C: UnindexedConsumer<Self::Item>, + { + self.a + .into_par_iter() + .chain(self.b.par_difference(self.a)) + .drive_unindexed(consumer) + } +} + +impl<T, S> HashSet<T, S> +where + T: Eq + Hash + Sync, + S: BuildHasher + Sync, +{ + /// Visits (potentially in parallel) the values representing the difference, + /// i.e. the values that are in `self` but not in `other`. + #[cfg_attr(feature = "inline-more", inline)] + pub fn par_difference<'a>(&'a self, other: &'a Self) -> ParDifference<'a, T, S> { + ParDifference { a: self, b: other } + } + + /// Visits (potentially in parallel) the values representing the symmetric + /// difference, i.e. the values that are in `self` or in `other` but not in both. + #[cfg_attr(feature = "inline-more", inline)] + pub fn par_symmetric_difference<'a>( + &'a self, + other: &'a Self, + ) -> ParSymmetricDifference<'a, T, S> { + ParSymmetricDifference { a: self, b: other } + } + + /// Visits (potentially in parallel) the values representing the + /// intersection, i.e. the values that are both in `self` and `other`. + #[cfg_attr(feature = "inline-more", inline)] + pub fn par_intersection<'a>(&'a self, other: &'a Self) -> ParIntersection<'a, T, S> { + ParIntersection { a: self, b: other } + } + + /// Visits (potentially in parallel) the values representing the union, + /// i.e. all the values in `self` or `other`, without duplicates. + #[cfg_attr(feature = "inline-more", inline)] + pub fn par_union<'a>(&'a self, other: &'a Self) -> ParUnion<'a, T, S> { + ParUnion { a: self, b: other } + } + + /// Returns `true` if `self` has no elements in common with `other`. + /// This is equivalent to checking for an empty intersection. + /// + /// This method runs in a potentially parallel fashion. + pub fn par_is_disjoint(&self, other: &Self) -> bool { + self.into_par_iter().all(|x| !other.contains(x)) + } + + /// Returns `true` if the set is a subset of another, + /// i.e. `other` contains at least all the values in `self`. + /// + /// This method runs in a potentially parallel fashion. + pub fn par_is_subset(&self, other: &Self) -> bool { + if self.len() <= other.len() { + self.into_par_iter().all(|x| other.contains(x)) + } else { + false + } + } + + /// Returns `true` if the set is a superset of another, + /// i.e. `self` contains at least all the values in `other`. + /// + /// This method runs in a potentially parallel fashion. + pub fn par_is_superset(&self, other: &Self) -> bool { + other.par_is_subset(self) + } + + /// Returns `true` if the set is equal to another, + /// i.e. both sets contain the same values. + /// + /// This method runs in a potentially parallel fashion. + pub fn par_eq(&self, other: &Self) -> bool { + self.len() == other.len() && self.par_is_subset(other) + } +} + +impl<T, S> HashSet<T, S> +where + T: Eq + Hash + Send, + S: BuildHasher + Send, +{ + /// Consumes (potentially in parallel) all values in an arbitrary order, + /// while preserving the set's allocated memory for reuse. + #[cfg_attr(feature = "inline-more", inline)] + pub fn par_drain(&mut self) -> ParDrain<'_, T, S> { + ParDrain { set: self } + } +} + +impl<T: Send, S: Send> IntoParallelIterator for HashSet<T, S> { + type Item = T; + type Iter = IntoParIter<T, S>; + + #[cfg_attr(feature = "inline-more", inline)] + fn into_par_iter(self) -> Self::Iter { + IntoParIter { set: self } + } +} + +impl<'a, T: Sync, S: Sync> IntoParallelIterator for &'a HashSet<T, S> { + type Item = &'a T; + type Iter = ParIter<'a, T, S>; + + #[cfg_attr(feature = "inline-more", inline)] + fn into_par_iter(self) -> Self::Iter { + ParIter { set: self } + } +} + +/// Collect values from a parallel iterator into a hashset. +impl<T, S> FromParallelIterator<T> for HashSet<T, S> +where + T: Eq + Hash + Send, + S: BuildHasher + Default, +{ + fn from_par_iter<P>(par_iter: P) -> Self + where + P: IntoParallelIterator<Item = T>, + { + let mut set = HashSet::default(); + set.par_extend(par_iter); + set + } +} + +/// Extend a hash set with items from a parallel iterator. +impl<T, S> ParallelExtend<T> for HashSet<T, S> +where + T: Eq + Hash + Send, + S: BuildHasher, +{ + fn par_extend<I>(&mut self, par_iter: I) + where + I: IntoParallelIterator<Item = T>, + { + extend(self, par_iter); + } +} + +/// Extend a hash set with copied items from a parallel iterator. +impl<'a, T, S> ParallelExtend<&'a T> for HashSet<T, S> +where + T: 'a + Copy + Eq + Hash + Sync, + S: BuildHasher, +{ + fn par_extend<I>(&mut self, par_iter: I) + where + I: IntoParallelIterator<Item = &'a T>, + { + extend(self, par_iter); + } +} + +// This is equal to the normal `HashSet` -- no custom advantage. +fn extend<T, S, I>(set: &mut HashSet<T, S>, par_iter: I) +where + T: Eq + Hash, + S: BuildHasher, + I: IntoParallelIterator, + HashSet<T, S>: Extend<I::Item>, +{ + let (list, len) = super::helpers::collect(par_iter); + + // Values may be already present or show multiple times in the iterator. + // Reserve the entire length if the set is empty. + // Otherwise reserve half the length (rounded up), so the set + // will only resize twice in the worst case. + let reserve = if set.is_empty() { len } else { (len + 1) / 2 }; + set.reserve(reserve); + for vec in list { + set.extend(vec); + } +} + +#[cfg(test)] +mod test_par_set { + use alloc::vec::Vec; + use core::sync::atomic::{AtomicUsize, Ordering}; + + use rayon::prelude::*; + + use crate::hash_set::HashSet; + + #[test] + fn test_disjoint() { + let mut xs = HashSet::new(); + let mut ys = HashSet::new(); + assert!(xs.par_is_disjoint(&ys)); + assert!(ys.par_is_disjoint(&xs)); + assert!(xs.insert(5)); + assert!(ys.insert(11)); + assert!(xs.par_is_disjoint(&ys)); + assert!(ys.par_is_disjoint(&xs)); + assert!(xs.insert(7)); + assert!(xs.insert(19)); + assert!(xs.insert(4)); + assert!(ys.insert(2)); + assert!(ys.insert(-11)); + assert!(xs.par_is_disjoint(&ys)); + assert!(ys.par_is_disjoint(&xs)); + assert!(ys.insert(7)); + assert!(!xs.par_is_disjoint(&ys)); + assert!(!ys.par_is_disjoint(&xs)); + } + + #[test] + fn test_subset_and_superset() { + let mut a = HashSet::new(); + assert!(a.insert(0)); + assert!(a.insert(5)); + assert!(a.insert(11)); + assert!(a.insert(7)); + + let mut b = HashSet::new(); + assert!(b.insert(0)); + assert!(b.insert(7)); + assert!(b.insert(19)); + assert!(b.insert(250)); + assert!(b.insert(11)); + assert!(b.insert(200)); + + assert!(!a.par_is_subset(&b)); + assert!(!a.par_is_superset(&b)); + assert!(!b.par_is_subset(&a)); + assert!(!b.par_is_superset(&a)); + + assert!(b.insert(5)); + + assert!(a.par_is_subset(&b)); + assert!(!a.par_is_superset(&b)); + assert!(!b.par_is_subset(&a)); + assert!(b.par_is_superset(&a)); + } + + #[test] + fn test_iterate() { + let mut a = HashSet::new(); + for i in 0..32 { + assert!(a.insert(i)); + } + let observed = AtomicUsize::new(0); + a.par_iter().for_each(|k| { + observed.fetch_or(1 << *k, Ordering::Relaxed); + }); + assert_eq!(observed.into_inner(), 0xFFFF_FFFF); + } + + #[test] + fn test_intersection() { + let mut a = HashSet::new(); + let mut b = HashSet::new(); + + assert!(a.insert(11)); + assert!(a.insert(1)); + assert!(a.insert(3)); + assert!(a.insert(77)); + assert!(a.insert(103)); + assert!(a.insert(5)); + assert!(a.insert(-5)); + + assert!(b.insert(2)); + assert!(b.insert(11)); + assert!(b.insert(77)); + assert!(b.insert(-9)); + assert!(b.insert(-42)); + assert!(b.insert(5)); + assert!(b.insert(3)); + + let expected = [3, 5, 11, 77]; + let i = a + .par_intersection(&b) + .map(|x| { + assert!(expected.contains(x)); + 1 + }) + .sum::<usize>(); + assert_eq!(i, expected.len()); + } + + #[test] + fn test_difference() { + let mut a = HashSet::new(); + let mut b = HashSet::new(); + + assert!(a.insert(1)); + assert!(a.insert(3)); + assert!(a.insert(5)); + assert!(a.insert(9)); + assert!(a.insert(11)); + + assert!(b.insert(3)); + assert!(b.insert(9)); + + let expected = [1, 5, 11]; + let i = a + .par_difference(&b) + .map(|x| { + assert!(expected.contains(x)); + 1 + }) + .sum::<usize>(); + assert_eq!(i, expected.len()); + } + + #[test] + fn test_symmetric_difference() { + let mut a = HashSet::new(); + let mut b = HashSet::new(); + + assert!(a.insert(1)); + assert!(a.insert(3)); + assert!(a.insert(5)); + assert!(a.insert(9)); + assert!(a.insert(11)); + + assert!(b.insert(-2)); + assert!(b.insert(3)); + assert!(b.insert(9)); + assert!(b.insert(14)); + assert!(b.insert(22)); + + let expected = [-2, 1, 5, 11, 14, 22]; + let i = a + .par_symmetric_difference(&b) + .map(|x| { + assert!(expected.contains(x)); + 1 + }) + .sum::<usize>(); + assert_eq!(i, expected.len()); + } + + #[test] + fn test_union() { + let mut a = HashSet::new(); + let mut b = HashSet::new(); + + assert!(a.insert(1)); + assert!(a.insert(3)); + assert!(a.insert(5)); + assert!(a.insert(9)); + assert!(a.insert(11)); + assert!(a.insert(16)); + assert!(a.insert(19)); + assert!(a.insert(24)); + + assert!(b.insert(-2)); + assert!(b.insert(1)); + assert!(b.insert(5)); + assert!(b.insert(9)); + assert!(b.insert(13)); + assert!(b.insert(19)); + + let expected = [-2, 1, 3, 5, 9, 11, 13, 16, 19, 24]; + let i = a + .par_union(&b) + .map(|x| { + assert!(expected.contains(x)); + 1 + }) + .sum::<usize>(); + assert_eq!(i, expected.len()); + } + + #[test] + fn test_from_iter() { + let xs = [1, 2, 3, 4, 5, 6, 7, 8, 9]; + + let set: HashSet<_> = xs.par_iter().cloned().collect(); + + for x in &xs { + assert!(set.contains(x)); + } + } + + #[test] + fn test_move_iter() { + let hs = { + let mut hs = HashSet::new(); + + hs.insert('a'); + hs.insert('b'); + + hs + }; + + let v = hs.into_par_iter().collect::<Vec<char>>(); + assert!(v == ['a', 'b'] || v == ['b', 'a']); + } + + #[test] + fn test_eq() { + // These constants once happened to expose a bug in insert(). + // I'm keeping them around to prevent a regression. + let mut s1 = HashSet::new(); + + s1.insert(1); + s1.insert(2); + s1.insert(3); + + let mut s2 = HashSet::new(); + + s2.insert(1); + s2.insert(2); + + assert!(!s1.par_eq(&s2)); + + s2.insert(3); + + assert!(s1.par_eq(&s2)); + } + + #[test] + fn test_extend_ref() { + let mut a = HashSet::new(); + a.insert(1); + + a.par_extend(&[2, 3, 4][..]); + + assert_eq!(a.len(), 4); + assert!(a.contains(&1)); + assert!(a.contains(&2)); + assert!(a.contains(&3)); + assert!(a.contains(&4)); + + let mut b = HashSet::new(); + b.insert(5); + b.insert(6); + + a.par_extend(&b); + + assert_eq!(a.len(), 6); + assert!(a.contains(&1)); + assert!(a.contains(&2)); + assert!(a.contains(&3)); + assert!(a.contains(&4)); + assert!(a.contains(&5)); + assert!(a.contains(&6)); + } +} diff --git a/src/external_trait_impls/serde.rs b/src/external_trait_impls/serde.rs new file mode 100644 index 0000000..7816e78 --- /dev/null +++ b/src/external_trait_impls/serde.rs @@ -0,0 +1,200 @@ +mod size_hint { + use core::cmp; + + /// This presumably exists to prevent denial of service attacks. + /// + /// Original discussion: https://github.com/serde-rs/serde/issues/1114. + #[cfg_attr(feature = "inline-more", inline)] + pub(super) fn cautious(hint: Option<usize>) -> usize { + cmp::min(hint.unwrap_or(0), 4096) + } +} + +mod map { + use core::fmt; + use core::hash::{BuildHasher, Hash}; + use core::marker::PhantomData; + use serde::de::{Deserialize, Deserializer, MapAccess, Visitor}; + use serde::ser::{Serialize, Serializer}; + + use crate::hash_map::HashMap; + + use super::size_hint; + + impl<K, V, H> Serialize for HashMap<K, V, H> + where + K: Serialize + Eq + Hash, + V: Serialize, + H: BuildHasher, + { + #[cfg_attr(feature = "inline-more", inline)] + fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> + where + S: Serializer, + { + serializer.collect_map(self) + } + } + + impl<'de, K, V, S> Deserialize<'de> for HashMap<K, V, S> + where + K: Deserialize<'de> + Eq + Hash, + V: Deserialize<'de>, + S: BuildHasher + Default, + { + fn deserialize<D>(deserializer: D) -> Result<Self, D::Error> + where + D: Deserializer<'de>, + { + struct MapVisitor<K, V, S> { + marker: PhantomData<HashMap<K, V, S>>, + } + + impl<'de, K, V, S> Visitor<'de> for MapVisitor<K, V, S> + where + K: Deserialize<'de> + Eq + Hash, + V: Deserialize<'de>, + S: BuildHasher + Default, + { + type Value = HashMap<K, V, S>; + + fn expecting(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result { + formatter.write_str("a map") + } + + #[cfg_attr(feature = "inline-more", inline)] + fn visit_map<A>(self, mut map: A) -> Result<Self::Value, A::Error> + where + A: MapAccess<'de>, + { + let mut values = HashMap::with_capacity_and_hasher( + size_hint::cautious(map.size_hint()), + S::default(), + ); + + while let Some((key, value)) = map.next_entry()? { + values.insert(key, value); + } + + Ok(values) + } + } + + let visitor = MapVisitor { + marker: PhantomData, + }; + deserializer.deserialize_map(visitor) + } + } +} + +mod set { + use core::fmt; + use core::hash::{BuildHasher, Hash}; + use core::marker::PhantomData; + use serde::de::{Deserialize, Deserializer, SeqAccess, Visitor}; + use serde::ser::{Serialize, Serializer}; + + use crate::hash_set::HashSet; + + use super::size_hint; + + impl<T, H> Serialize for HashSet<T, H> + where + T: Serialize + Eq + Hash, + H: BuildHasher, + { + #[cfg_attr(feature = "inline-more", inline)] + fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> + where + S: Serializer, + { + serializer.collect_seq(self) + } + } + + impl<'de, T, S> Deserialize<'de> for HashSet<T, S> + where + T: Deserialize<'de> + Eq + Hash, + S: BuildHasher + Default, + { + fn deserialize<D>(deserializer: D) -> Result<Self, D::Error> + where + D: Deserializer<'de>, + { + struct SeqVisitor<T, S> { + marker: PhantomData<HashSet<T, S>>, + } + + impl<'de, T, S> Visitor<'de> for SeqVisitor<T, S> + where + T: Deserialize<'de> + Eq + Hash, + S: BuildHasher + Default, + { + type Value = HashSet<T, S>; + + fn expecting(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result { + formatter.write_str("a sequence") + } + + #[cfg_attr(feature = "inline-more", inline)] + fn visit_seq<A>(self, mut seq: A) -> Result<Self::Value, A::Error> + where + A: SeqAccess<'de>, + { + let mut values = HashSet::with_capacity_and_hasher( + size_hint::cautious(seq.size_hint()), + S::default(), + ); + + while let Some(value) = seq.next_element()? { + values.insert(value); + } + + Ok(values) + } + } + + let visitor = SeqVisitor { + marker: PhantomData, + }; + deserializer.deserialize_seq(visitor) + } + + fn deserialize_in_place<D>(deserializer: D, place: &mut Self) -> Result<(), D::Error> + where + D: Deserializer<'de>, + { + struct SeqInPlaceVisitor<'a, T, S>(&'a mut HashSet<T, S>); + + impl<'a, 'de, T, S> Visitor<'de> for SeqInPlaceVisitor<'a, T, S> + where + T: Deserialize<'de> + Eq + Hash, + S: BuildHasher + Default, + { + type Value = (); + + fn expecting(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result { + formatter.write_str("a sequence") + } + + #[cfg_attr(feature = "inline-more", inline)] + fn visit_seq<A>(self, mut seq: A) -> Result<Self::Value, A::Error> + where + A: SeqAccess<'de>, + { + self.0.clear(); + self.0.reserve(size_hint::cautious(seq.size_hint())); + + while let Some(value) = seq.next_element()? { + self.0.insert(value); + } + + Ok(()) + } + } + + deserializer.deserialize_seq(SeqInPlaceVisitor(place)) + } + } +} diff --git a/src/lib.rs b/src/lib.rs new file mode 100644 index 0000000..3aff40a --- /dev/null +++ b/src/lib.rs @@ -0,0 +1,112 @@ +//! This crate is a Rust port of Google's high-performance [SwissTable] hash +//! map, adapted to make it a drop-in replacement for Rust's standard `HashMap` +//! and `HashSet` types. +//! +//! The original C++ version of [SwissTable] can be found [here], and this +//! [CppCon talk] gives an overview of how the algorithm works. +//! +//! [SwissTable]: https://abseil.io/blog/20180927-swisstables +//! [here]: https://github.com/abseil/abseil-cpp/blob/master/absl/container/internal/raw_hash_set.h +//! [CppCon talk]: https://www.youtube.com/watch?v=ncHmEUmJZf4 + +#![no_std] +#![cfg_attr( + feature = "nightly", + feature(test, core_intrinsics, dropck_eyepatch, min_specialization, extend_one) +)] +#![allow( + clippy::doc_markdown, + clippy::module_name_repetitions, + clippy::must_use_candidate, + clippy::option_if_let_else +)] +#![warn(missing_docs)] +#![warn(rust_2018_idioms)] + +#[cfg(test)] +#[macro_use] +extern crate std; + +#[cfg_attr(test, macro_use)] +extern crate alloc; + +#[cfg(feature = "nightly")] +#[cfg(doctest)] +doc_comment::doctest!("../README.md"); + +#[macro_use] +mod macros; + +#[cfg(feature = "raw")] +/// Experimental and unsafe `RawTable` API. This module is only available if the +/// `raw` feature is enabled. +pub mod raw { + // The RawTable API is still experimental and is not properly documented yet. + #[allow(missing_docs)] + #[path = "mod.rs"] + mod inner; + pub use inner::*; + + #[cfg(feature = "rayon")] + pub mod rayon { + pub use crate::external_trait_impls::rayon::raw::*; + } +} +#[cfg(not(feature = "raw"))] +mod raw; + +mod external_trait_impls; +mod map; +#[cfg(feature = "rustc-internal-api")] +mod rustc_entry; +mod scopeguard; +mod set; + +pub mod hash_map { + //! A hash map implemented with quadratic probing and SIMD lookup. + pub use crate::map::*; + + #[cfg(feature = "rustc-internal-api")] + pub use crate::rustc_entry::*; + + #[cfg(feature = "rayon")] + /// [rayon]-based parallel iterator types for hash maps. + /// You will rarely need to interact with it directly unless you have need + /// to name one of the iterator types. + /// + /// [rayon]: https://docs.rs/rayon/1.0/rayon + pub mod rayon { + pub use crate::external_trait_impls::rayon::map::*; + } +} +pub mod hash_set { + //! A hash set implemented as a `HashMap` where the value is `()`. + pub use crate::set::*; + + #[cfg(feature = "rayon")] + /// [rayon]-based parallel iterator types for hash sets. + /// You will rarely need to interact with it directly unless you have need + /// to name one of the iterator types. + /// + /// [rayon]: https://docs.rs/rayon/1.0/rayon + pub mod rayon { + pub use crate::external_trait_impls::rayon::set::*; + } +} + +pub use crate::map::HashMap; +pub use crate::set::HashSet; + +/// The error type for `try_reserve` methods. +#[derive(Clone, PartialEq, Eq, Debug)] +pub enum TryReserveError { + /// Error due to the computed capacity exceeding the collection's maximum + /// (usually `isize::MAX` bytes). + CapacityOverflow, + + /// The memory allocator returned an error + AllocError { + /// The layout of the allocation request that failed. + layout: alloc::alloc::Layout, + }, +} diff --git a/src/macros.rs b/src/macros.rs new file mode 100644 index 0000000..0279597 --- /dev/null +++ b/src/macros.rs @@ -0,0 +1,69 @@ +// See the cfg-if crate. +macro_rules! cfg_if { + // match if/else chains with a final `else` + ($( + if #[cfg($($meta:meta),*)] { $($it:item)* } + ) else * else { + $($it2:item)* + }) => { + cfg_if! { + @__items + () ; + $( ( ($($meta),*) ($($it)*) ), )* + ( () ($($it2)*) ), + } + }; + + // match if/else chains lacking a final `else` + ( + if #[cfg($($i_met:meta),*)] { $($i_it:item)* } + $( + else if #[cfg($($e_met:meta),*)] { $($e_it:item)* } + )* + ) => { + cfg_if! { + @__items + () ; + ( ($($i_met),*) ($($i_it)*) ), + $( ( ($($e_met),*) ($($e_it)*) ), )* + ( () () ), + } + }; + + // Internal and recursive macro to emit all the items + // + // Collects all the negated cfgs in a list at the beginning and after the + // semicolon is all the remaining items + (@__items ($($not:meta,)*) ; ) => {}; + (@__items ($($not:meta,)*) ; ( ($($m:meta),*) ($($it:item)*) ), $($rest:tt)*) => { + // Emit all items within one block, applying an approprate #[cfg]. The + // #[cfg] will require all `$m` matchers specified and must also negate + // all previous matchers. + cfg_if! { @__apply cfg(all($($m,)* not(any($($not),*)))), $($it)* } + + // Recurse to emit all other items in `$rest`, and when we do so add all + // our `$m` matchers to the list of `$not` matchers as future emissions + // will have to negate everything we just matched as well. + cfg_if! { @__items ($($not,)* $($m,)*) ; $($rest)* } + }; + + // Internal macro to Apply a cfg attribute to a list of items + (@__apply $m:meta, $($it:item)*) => { + $(#[$m] $it)* + }; +} + +// Helper macro for specialization. This also helps avoid parse errors if the +// default fn syntax for specialization changes in the future. +#[cfg(feature = "nightly")] +macro_rules! default_fn { + ($($tt:tt)*) => { + default $($tt)* + } +} +#[cfg(not(feature = "nightly"))] +macro_rules! default_fn { + ($($tt:tt)*) => { + $($tt)* + } +} diff --git a/src/map.rs b/src/map.rs new file mode 100644 index 0000000..1ccba31 --- /dev/null +++ b/src/map.rs @@ -0,0 +1,4524 @@ +use crate::raw::{Bucket, RawDrain, RawIntoIter, RawIter, RawTable}; +use crate::TryReserveError; +use core::borrow::Borrow; +use core::fmt::{self, Debug}; +use core::hash::{BuildHasher, Hash, Hasher}; +use core::iter::{FromIterator, FusedIterator}; +use core::marker::PhantomData; +use core::mem; +use core::ops::Index; + +/// Default hasher for `HashMap`. +#[cfg(feature = "ahash")] +pub type DefaultHashBuilder = ahash::RandomState; + +/// Dummy default hasher for `HashMap`. +#[cfg(not(feature = "ahash"))] +pub enum DefaultHashBuilder {} + +/// A hash map implemented with quadratic probing and SIMD lookup. +/// +/// The default hashing algorithm is currently [`AHash`], though this is +/// subject to change at any point in the future. This hash function is very +/// fast for all types of keys, but this algorithm will typically *not* protect +/// against attacks such as HashDoS. +/// +/// The hashing algorithm can be replaced on a per-`HashMap` basis using the +/// [`default`], [`with_hasher`], and [`with_capacity_and_hasher`] methods. Many +/// alternative algorithms are available on crates.io, such as the [`fnv`] crate. +/// +/// It is required that the keys implement the [`Eq`] and [`Hash`] traits, although +/// this can frequently be achieved by using `#[derive(PartialEq, Eq, Hash)]`. +/// If you implement these yourself, it is important that the following +/// property holds: +/// +/// ```text +/// k1 == k2 -> hash(k1) == hash(k2) +/// ``` +/// +/// In other words, if two keys are equal, their hashes must be equal. +/// +/// It is a logic error for a key to be modified in such a way that the key's +/// hash, as determined by the [`Hash`] trait, or its equality, as determined by +/// the [`Eq`] trait, changes while it is in the map. This is normally only +/// possible through [`Cell`], [`RefCell`], global state, I/O, or unsafe code. +/// +/// It is also a logic error for the [`Hash`] implementation of a key to panic. +/// This is generally only possible if the trait is implemented manually. If a +/// panic does occur then the contents of the `HashMap` may become corrupted and +/// some items may be dropped from the table. +/// +/// # Examples +/// +/// ``` +/// use hashbrown::HashMap; +/// +/// // Type inference lets us omit an explicit type signature (which +/// // would be `HashMap<String, String>` in this example). +/// let mut book_reviews = HashMap::new(); +/// +/// // Review some books. +/// book_reviews.insert( +/// "Adventures of Huckleberry Finn".to_string(), +/// "My favorite book.".to_string(), +/// ); +/// book_reviews.insert( +/// "Grimms' Fairy Tales".to_string(), +/// "Masterpiece.".to_string(), +/// ); +/// book_reviews.insert( +/// "Pride and Prejudice".to_string(), +/// "Very enjoyable.".to_string(), +/// ); +/// book_reviews.insert( +/// "The Adventures of Sherlock Holmes".to_string(), +/// "Eye lyked it alot.".to_string(), +/// ); +/// +/// // Check for a specific one. +/// // When collections store owned values (String), they can still be +/// // queried using references (&str). +/// if !book_reviews.contains_key("Les Misérables") { +/// println!("We've got {} reviews, but Les Misérables ain't one.", +/// book_reviews.len()); +/// } +/// +/// // oops, this review has a lot of spelling mistakes, let's delete it. +/// book_reviews.remove("The Adventures of Sherlock Holmes"); +/// +/// // Look up the values associated with some keys. +/// let to_find = ["Pride and Prejudice", "Alice's Adventure in Wonderland"]; +/// for &book in &to_find { +/// match book_reviews.get(book) { +/// Some(review) => println!("{}: {}", book, review), +/// None => println!("{} is unreviewed.", book) +/// } +/// } +/// +/// // Look up the value for a key (will panic if the key is not found). +/// println!("Review for Jane: {}", book_reviews["Pride and Prejudice"]); +/// +/// // Iterate over everything. +/// for (book, review) in &book_reviews { +/// println!("{}: \"{}\"", book, review); +/// } +/// ``` +/// +/// `HashMap` also implements an [`Entry API`](#method.entry), which allows +/// for more complex methods of getting, setting, updating and removing keys and +/// their values: +/// +/// ``` +/// use hashbrown::HashMap; +/// +/// // type inference lets us omit an explicit type signature (which +/// // would be `HashMap<&str, u8>` in this example). +/// let mut player_stats = HashMap::new(); +/// +/// fn random_stat_buff() -> u8 { +/// // could actually return some random value here - let's just return +/// // some fixed value for now +/// 42 +/// } +/// +/// // insert a key only if it doesn't already exist +/// player_stats.entry("health").or_insert(100); +/// +/// // insert a key using a function that provides a new value only if it +/// // doesn't already exist +/// player_stats.entry("defence").or_insert_with(random_stat_buff); +/// +/// // update a key, guarding against the key possibly not being set +/// let stat = player_stats.entry("attack").or_insert(100); +/// *stat += random_stat_buff(); +/// ``` +/// +/// The easiest way to use `HashMap` with a custom key type is to derive [`Eq`] and [`Hash`]. +/// We must also derive [`PartialEq`]. +/// +/// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html +/// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html +/// [`PartialEq`]: https://doc.rust-lang.org/std/cmp/trait.PartialEq.html +/// [`RefCell`]: https://doc.rust-lang.org/std/cell/struct.RefCell.html +/// [`Cell`]: https://doc.rust-lang.org/std/cell/struct.Cell.html +/// [`default`]: #method.default +/// [`with_hasher`]: #method.with_hasher +/// [`with_capacity_and_hasher`]: #method.with_capacity_and_hasher +/// [`fnv`]: https://crates.io/crates/fnv +/// [`AHash`]: https://crates.io/crates/ahash +/// +/// ``` +/// use hashbrown::HashMap; +/// +/// #[derive(Hash, Eq, PartialEq, Debug)] +/// struct Viking { +/// name: String, +/// country: String, +/// } +/// +/// impl Viking { +/// /// Creates a new Viking. +/// fn new(name: &str, country: &str) -> Viking { +/// Viking { name: name.to_string(), country: country.to_string() } +/// } +/// } +/// +/// // Use a HashMap to store the vikings' health points. +/// let mut vikings = HashMap::new(); +/// +/// vikings.insert(Viking::new("Einar", "Norway"), 25); +/// vikings.insert(Viking::new("Olaf", "Denmark"), 24); +/// vikings.insert(Viking::new("Harald", "Iceland"), 12); +/// +/// // Use derived implementation to print the status of the vikings. +/// for (viking, health) in &vikings { +/// println!("{:?} has {} hp", viking, health); +/// } +/// ``` +/// +/// A `HashMap` with fixed list of elements can be initialized from an array: +/// +/// ``` +/// use hashbrown::HashMap; +/// +/// let timber_resources: HashMap<&str, i32> = [("Norway", 100), ("Denmark", 50), ("Iceland", 10)] +/// .iter().cloned().collect(); +/// // use the values stored in map +/// ``` +pub struct HashMap<K, V, S = DefaultHashBuilder> { + pub(crate) hash_builder: S, + pub(crate) table: RawTable<(K, V)>, +} + +impl<K: Clone, V: Clone, S: Clone> Clone for HashMap<K, V, S> { + fn clone(&self) -> Self { + HashMap { + hash_builder: self.hash_builder.clone(), + table: self.table.clone(), + } + } + + fn clone_from(&mut self, source: &Self) { + self.table.clone_from(&source.table); + + // Update hash_builder only if we successfully cloned all elements. + self.hash_builder.clone_from(&source.hash_builder); + } +} + +#[cfg_attr(feature = "inline-more", inline)] +pub(crate) fn make_hash<K: Hash + ?Sized>(hash_builder: &impl BuildHasher, val: &K) -> u64 { + let mut state = hash_builder.build_hasher(); + val.hash(&mut state); + state.finish() +} + +#[cfg(feature = "ahash")] +impl<K, V> HashMap<K, V, DefaultHashBuilder> { + /// Creates an empty `HashMap`. + /// + /// The hash map is initially created with a capacity of 0, so it will not allocate until it + /// is first inserted into. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// let mut map: HashMap<&str, i32> = HashMap::new(); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn new() -> Self { + Self::default() + } + + /// Creates an empty `HashMap` with the specified capacity. + /// + /// The hash map will be able to hold at least `capacity` elements without + /// reallocating. If `capacity` is 0, the hash map will not allocate. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// let mut map: HashMap<&str, i32> = HashMap::with_capacity(10); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn with_capacity(capacity: usize) -> Self { + Self::with_capacity_and_hasher(capacity, DefaultHashBuilder::default()) + } +} + +impl<K, V, S> HashMap<K, V, S> { + /// Creates an empty `HashMap` which will use the given hash builder to hash + /// keys. + /// + /// The created map has the default initial capacity. + /// + /// Warning: `hash_builder` is normally randomly generated, and + /// is designed to allow HashMaps to be resistant to attacks that + /// cause many collisions and very poor performance. Setting it + /// manually using this function can expose a DoS attack vector. + /// + /// The `hash_builder` passed should implement the [`BuildHasher`] trait for + /// the HashMap to be useful, see its documentation for details. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::DefaultHashBuilder; + /// + /// let s = DefaultHashBuilder::default(); + /// let mut map = HashMap::with_hasher(s); + /// map.insert(1, 2); + /// ``` + /// + /// [`BuildHasher`]: ../../std/hash/trait.BuildHasher.html + #[cfg_attr(feature = "inline-more", inline)] + pub const fn with_hasher(hash_builder: S) -> Self { + Self { + hash_builder, + table: RawTable::new(), + } + } + + /// Creates an empty `HashMap` with the specified capacity, using `hash_builder` + /// to hash the keys. + /// + /// The hash map will be able to hold at least `capacity` elements without + /// reallocating. If `capacity` is 0, the hash map will not allocate. + /// + /// Warning: `hash_builder` is normally randomly generated, and + /// is designed to allow HashMaps to be resistant to attacks that + /// cause many collisions and very poor performance. Setting it + /// manually using this function can expose a DoS attack vector. + /// + /// The `hash_builder` passed should implement the [`BuildHasher`] trait for + /// the HashMap to be useful, see its documentation for details. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::DefaultHashBuilder; + /// + /// let s = DefaultHashBuilder::default(); + /// let mut map = HashMap::with_capacity_and_hasher(10, s); + /// map.insert(1, 2); + /// ``` + /// + /// [`BuildHasher`]: ../../std/hash/trait.BuildHasher.html + #[cfg_attr(feature = "inline-more", inline)] + pub fn with_capacity_and_hasher(capacity: usize, hash_builder: S) -> Self { + Self { + hash_builder, + table: RawTable::with_capacity(capacity), + } + } + + /// Returns a reference to the map's [`BuildHasher`]. + /// + /// [`BuildHasher`]: https://doc.rust-lang.org/std/hash/trait.BuildHasher.html + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::DefaultHashBuilder; + /// + /// let hasher = DefaultHashBuilder::default(); + /// let map: HashMap<i32, i32> = HashMap::with_hasher(hasher); + /// let hasher: &DefaultHashBuilder = map.hasher(); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn hasher(&self) -> &S { + &self.hash_builder + } + + /// Returns the number of elements the map can hold without reallocating. + /// + /// This number is a lower bound; the `HashMap<K, V>` might be able to hold + /// more, but is guaranteed to be able to hold at least this many. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// let map: HashMap<i32, i32> = HashMap::with_capacity(100); + /// assert!(map.capacity() >= 100); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn capacity(&self) -> usize { + self.table.capacity() + } + + /// An iterator visiting all keys in arbitrary order. + /// The iterator element type is `&'a K`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert("a", 1); + /// map.insert("b", 2); + /// map.insert("c", 3); + /// + /// for key in map.keys() { + /// println!("{}", key); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn keys(&self) -> Keys<'_, K, V> { + Keys { inner: self.iter() } + } + + /// An iterator visiting all values in arbitrary order. + /// The iterator element type is `&'a V`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert("a", 1); + /// map.insert("b", 2); + /// map.insert("c", 3); + /// + /// for val in map.values() { + /// println!("{}", val); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn values(&self) -> Values<'_, K, V> { + Values { inner: self.iter() } + } + + /// An iterator visiting all values mutably in arbitrary order. + /// The iterator element type is `&'a mut V`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// + /// map.insert("a", 1); + /// map.insert("b", 2); + /// map.insert("c", 3); + /// + /// for val in map.values_mut() { + /// *val = *val + 10; + /// } + /// + /// for val in map.values() { + /// println!("{}", val); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn values_mut(&mut self) -> ValuesMut<'_, K, V> { + ValuesMut { + inner: self.iter_mut(), + } + } + + /// An iterator visiting all key-value pairs in arbitrary order. + /// The iterator element type is `(&'a K, &'a V)`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert("a", 1); + /// map.insert("b", 2); + /// map.insert("c", 3); + /// + /// for (key, val) in map.iter() { + /// println!("key: {} val: {}", key, val); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn iter(&self) -> Iter<'_, K, V> { + // Here we tie the lifetime of self to the iter. + unsafe { + Iter { + inner: self.table.iter(), + marker: PhantomData, + } + } + } + + /// An iterator visiting all key-value pairs in arbitrary order, + /// with mutable references to the values. + /// The iterator element type is `(&'a K, &'a mut V)`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert("a", 1); + /// map.insert("b", 2); + /// map.insert("c", 3); + /// + /// // Update all values + /// for (_, val) in map.iter_mut() { + /// *val *= 2; + /// } + /// + /// for (key, val) in &map { + /// println!("key: {} val: {}", key, val); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn iter_mut(&mut self) -> IterMut<'_, K, V> { + // Here we tie the lifetime of self to the iter. + unsafe { + IterMut { + inner: self.table.iter(), + marker: PhantomData, + } + } + } + + #[cfg(test)] + #[cfg_attr(feature = "inline-more", inline)] + fn raw_capacity(&self) -> usize { + self.table.buckets() + } + + /// Returns the number of elements in the map. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut a = HashMap::new(); + /// assert_eq!(a.len(), 0); + /// a.insert(1, "a"); + /// assert_eq!(a.len(), 1); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn len(&self) -> usize { + self.table.len() + } + + /// Returns `true` if the map contains no elements. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut a = HashMap::new(); + /// assert!(a.is_empty()); + /// a.insert(1, "a"); + /// assert!(!a.is_empty()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn is_empty(&self) -> bool { + self.len() == 0 + } + + /// Clears the map, returning all key-value pairs as an iterator. Keeps the + /// allocated memory for reuse. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut a = HashMap::new(); + /// a.insert(1, "a"); + /// a.insert(2, "b"); + /// + /// for (k, v) in a.drain().take(1) { + /// assert!(k == 1 || k == 2); + /// assert!(v == "a" || v == "b"); + /// } + /// + /// assert!(a.is_empty()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn drain(&mut self) -> Drain<'_, K, V> { + Drain { + inner: self.table.drain(), + } + } + + /// Retains only the elements specified by the predicate. + /// + /// In other words, remove all pairs `(k, v)` such that `f(&k,&mut v)` returns `false`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<i32, i32> = (0..8).map(|x|(x, x*10)).collect(); + /// map.retain(|&k, _| k % 2 == 0); + /// assert_eq!(map.len(), 4); + /// ``` + pub fn retain<F>(&mut self, mut f: F) + where + F: FnMut(&K, &mut V) -> bool, + { + // Here we only use `iter` as a temporary, preventing use-after-free + unsafe { + for item in self.table.iter() { + let &mut (ref key, ref mut value) = item.as_mut(); + if !f(key, value) { + self.table.erase(item); + } + } + } + } + + /// Drains elements which are true under the given predicate, + /// and returns an iterator over the removed items. + /// + /// In other words, move all pairs `(k, v)` such that `f(&k,&mut v)` returns `true` out + /// into another iterator. + /// + /// When the returned DrainedFilter is dropped, any remaining elements that satisfy + /// the predicate are dropped from the table. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<i32, i32> = (0..8).map(|x| (x, x)).collect(); + /// let drained: HashMap<i32, i32> = map.drain_filter(|k, _v| k % 2 == 0).collect(); + /// + /// let mut evens = drained.keys().cloned().collect::<Vec<_>>(); + /// let mut odds = map.keys().cloned().collect::<Vec<_>>(); + /// evens.sort(); + /// odds.sort(); + /// + /// assert_eq!(evens, vec![0, 2, 4, 6]); + /// assert_eq!(odds, vec![1, 3, 5, 7]); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn drain_filter<F>(&mut self, f: F) -> DrainFilter<'_, K, V, F> + where + F: FnMut(&K, &mut V) -> bool, + { + DrainFilter { + f, + inner: DrainFilterInner { + iter: unsafe { self.table.iter() }, + table: &mut self.table, + }, + } + } + + /// Clears the map, removing all key-value pairs. Keeps the allocated memory + /// for reuse. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut a = HashMap::new(); + /// a.insert(1, "a"); + /// a.clear(); + /// assert!(a.is_empty()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn clear(&mut self) { + self.table.clear(); + } +} + +impl<K, V, S> HashMap<K, V, S> +where + K: Eq + Hash, + S: BuildHasher, +{ + /// Reserves capacity for at least `additional` more elements to be inserted + /// in the `HashMap`. The collection may reserve more space to avoid + /// frequent reallocations. + /// + /// # Panics + /// + /// Panics if the new allocation size overflows [`usize`]. + /// + /// [`usize`]: https://doc.rust-lang.org/std/primitive.usize.html + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// let mut map: HashMap<&str, i32> = HashMap::new(); + /// map.reserve(10); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn reserve(&mut self, additional: usize) { + let hash_builder = &self.hash_builder; + self.table + .reserve(additional, |x| make_hash(hash_builder, &x.0)); + } + + /// Tries to reserve capacity for at least `additional` more elements to be inserted + /// in the given `HashMap<K,V>`. The collection may reserve more space to avoid + /// frequent reallocations. + /// + /// # Errors + /// + /// If the capacity overflows, or the allocator reports a failure, then an error + /// is returned. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// let mut map: HashMap<&str, isize> = HashMap::new(); + /// map.try_reserve(10).expect("why is the test harness OOMing on 10 bytes?"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn try_reserve(&mut self, additional: usize) -> Result<(), TryReserveError> { + let hash_builder = &self.hash_builder; + self.table + .try_reserve(additional, |x| make_hash(hash_builder, &x.0)) + } + + /// Shrinks the capacity of the map as much as possible. It will drop + /// down as much as possible while maintaining the internal rules + /// and possibly leaving some space in accordance with the resize policy. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<i32, i32> = HashMap::with_capacity(100); + /// map.insert(1, 2); + /// map.insert(3, 4); + /// assert!(map.capacity() >= 100); + /// map.shrink_to_fit(); + /// assert!(map.capacity() >= 2); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn shrink_to_fit(&mut self) { + let hash_builder = &self.hash_builder; + self.table.shrink_to(0, |x| make_hash(hash_builder, &x.0)); + } + + /// Shrinks the capacity of the map with a lower limit. It will drop + /// down no lower than the supplied limit while maintaining the internal rules + /// and possibly leaving some space in accordance with the resize policy. + /// + /// This function does nothing if the current capacity is smaller than the + /// supplied minimum capacity. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<i32, i32> = HashMap::with_capacity(100); + /// map.insert(1, 2); + /// map.insert(3, 4); + /// assert!(map.capacity() >= 100); + /// map.shrink_to(10); + /// assert!(map.capacity() >= 10); + /// map.shrink_to(0); + /// assert!(map.capacity() >= 2); + /// map.shrink_to(10); + /// assert!(map.capacity() >= 2); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn shrink_to(&mut self, min_capacity: usize) { + let hash_builder = &self.hash_builder; + self.table + .shrink_to(min_capacity, |x| make_hash(hash_builder, &x.0)); + } + + /// Gets the given key's corresponding entry in the map for in-place manipulation. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut letters = HashMap::new(); + /// + /// for ch in "a short treatise on fungi".chars() { + /// let counter = letters.entry(ch).or_insert(0); + /// *counter += 1; + /// } + /// + /// assert_eq!(letters[&'s'], 2); + /// assert_eq!(letters[&'t'], 3); + /// assert_eq!(letters[&'u'], 1); + /// assert_eq!(letters.get(&'y'), None); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn entry(&mut self, key: K) -> Entry<'_, K, V, S> { + let hash = make_hash(&self.hash_builder, &key); + if let Some(elem) = self.table.find(hash, |q| q.0.eq(&key)) { + Entry::Occupied(OccupiedEntry { + hash, + key: Some(key), + elem, + table: self, + }) + } else { + Entry::Vacant(VacantEntry { + hash, + key, + table: self, + }) + } + } + + /// Returns a reference to the value corresponding to the key. + /// + /// The key may be any borrowed form of the map's key type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the key type. + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert(1, "a"); + /// assert_eq!(map.get(&1), Some(&"a")); + /// assert_eq!(map.get(&2), None); + /// ``` + #[inline] + pub fn get<Q: ?Sized>(&self, k: &Q) -> Option<&V> + where + K: Borrow<Q>, + Q: Hash + Eq, + { + // Avoid `Option::map` because it bloats LLVM IR. + match self.get_inner(k) { + Some(&(_, ref v)) => Some(v), + None => None, + } + } + + /// Returns the key-value pair corresponding to the supplied key. + /// + /// The supplied key may be any borrowed form of the map's key type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the key type. + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert(1, "a"); + /// assert_eq!(map.get_key_value(&1), Some((&1, &"a"))); + /// assert_eq!(map.get_key_value(&2), None); + /// ``` + #[inline] + pub fn get_key_value<Q: ?Sized>(&self, k: &Q) -> Option<(&K, &V)> + where + K: Borrow<Q>, + Q: Hash + Eq, + { + // Avoid `Option::map` because it bloats LLVM IR. + match self.get_inner(k) { + Some(&(ref key, ref value)) => Some((key, value)), + None => None, + } + } + + #[inline] + fn get_inner<Q: ?Sized>(&self, k: &Q) -> Option<&(K, V)> + where + K: Borrow<Q>, + Q: Hash + Eq, + { + let hash = make_hash(&self.hash_builder, k); + self.table.get(hash, |x| k.eq(x.0.borrow())) + } + + /// Returns the key-value pair corresponding to the supplied key, with a mutable reference to value. + /// + /// The supplied key may be any borrowed form of the map's key type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the key type. + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert(1, "a"); + /// let (k, v) = map.get_key_value_mut(&1).unwrap(); + /// assert_eq!(k, &1); + /// assert_eq!(v, &mut "a"); + /// *v = "b"; + /// assert_eq!(map.get_key_value_mut(&1), Some((&1, &mut "b"))); + /// assert_eq!(map.get_key_value_mut(&2), None); + /// ``` + #[inline] + pub fn get_key_value_mut<Q: ?Sized>(&mut self, k: &Q) -> Option<(&K, &mut V)> + where + K: Borrow<Q>, + Q: Hash + Eq, + { + // Avoid `Option::map` because it bloats LLVM IR. + match self.get_inner_mut(k) { + Some(&mut (ref key, ref mut value)) => Some((key, value)), + None => None, + } + } + + /// Returns `true` if the map contains a value for the specified key. + /// + /// The key may be any borrowed form of the map's key type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the key type. + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert(1, "a"); + /// assert_eq!(map.contains_key(&1), true); + /// assert_eq!(map.contains_key(&2), false); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn contains_key<Q: ?Sized>(&self, k: &Q) -> bool + where + K: Borrow<Q>, + Q: Hash + Eq, + { + self.get_inner(k).is_some() + } + + /// Returns a mutable reference to the value corresponding to the key. + /// + /// The key may be any borrowed form of the map's key type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the key type. + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert(1, "a"); + /// if let Some(x) = map.get_mut(&1) { + /// *x = "b"; + /// } + /// assert_eq!(map[&1], "b"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn get_mut<Q: ?Sized>(&mut self, k: &Q) -> Option<&mut V> + where + K: Borrow<Q>, + Q: Hash + Eq, + { + // Avoid `Option::map` because it bloats LLVM IR. + match self.get_inner_mut(k) { + Some(&mut (_, ref mut v)) => Some(v), + None => None, + } + } + + #[inline] + fn get_inner_mut<Q: ?Sized>(&mut self, k: &Q) -> Option<&mut (K, V)> + where + K: Borrow<Q>, + Q: Hash + Eq, + { + let hash = make_hash(&self.hash_builder, k); + self.table.get_mut(hash, |x| k.eq(x.0.borrow())) + } + + /// Inserts a key-value pair into the map. + /// + /// If the map did not have this key present, [`None`] is returned. + /// + /// If the map did have this key present, the value is updated, and the old + /// value is returned. The key is not updated, though; this matters for + /// types that can be `==` without being identical. See the [module-level + /// documentation] for more. + /// + /// [`None`]: https://doc.rust-lang.org/std/option/enum.Option.html#variant.None + /// [module-level documentation]: index.html#insert-and-complex-keys + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// assert_eq!(map.insert(37, "a"), None); + /// assert_eq!(map.is_empty(), false); + /// + /// map.insert(37, "b"); + /// assert_eq!(map.insert(37, "c"), Some("b")); + /// assert_eq!(map[&37], "c"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(&mut self, k: K, v: V) -> Option<V> { + let hash = make_hash(&self.hash_builder, &k); + if let Some((_, item)) = self.table.get_mut(hash, |x| k.eq(&x.0)) { + Some(mem::replace(item, v)) + } else { + let hash_builder = &self.hash_builder; + self.table + .insert(hash, (k, v), |x| make_hash(hash_builder, &x.0)); + None + } + } + + /// Removes a key from the map, returning the value at the key if the key + /// was previously in the map. + /// + /// The key may be any borrowed form of the map's key type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the key type. + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert(1, "a"); + /// assert_eq!(map.remove(&1), Some("a")); + /// assert_eq!(map.remove(&1), None); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn remove<Q: ?Sized>(&mut self, k: &Q) -> Option<V> + where + K: Borrow<Q>, + Q: Hash + Eq, + { + // Avoid `Option::map` because it bloats LLVM IR. + match self.remove_entry(k) { + Some((_, v)) => Some(v), + None => None, + } + } + + /// Removes a key from the map, returning the stored key and value if the + /// key was previously in the map. + /// + /// The key may be any borrowed form of the map's key type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the key type. + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert(1, "a"); + /// assert_eq!(map.remove_entry(&1), Some((1, "a"))); + /// assert_eq!(map.remove(&1), None); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn remove_entry<Q: ?Sized>(&mut self, k: &Q) -> Option<(K, V)> + where + K: Borrow<Q>, + Q: Hash + Eq, + { + let hash = make_hash(&self.hash_builder, &k); + self.table.remove_entry(hash, |x| k.eq(x.0.borrow())) + } +} + +impl<K, V, S> HashMap<K, V, S> { + /// Creates a raw entry builder for the HashMap. + /// + /// Raw entries provide the lowest level of control for searching and + /// manipulating a map. They must be manually initialized with a hash and + /// then manually searched. After this, insertions into a vacant entry + /// still require an owned key to be provided. + /// + /// Raw entries are useful for such exotic situations as: + /// + /// * Hash memoization + /// * Deferring the creation of an owned key until it is known to be required + /// * Using a search key that doesn't work with the Borrow trait + /// * Using custom comparison logic without newtype wrappers + /// + /// Because raw entries provide much more low-level control, it's much easier + /// to put the HashMap into an inconsistent state which, while memory-safe, + /// will cause the map to produce seemingly random results. Higher-level and + /// more foolproof APIs like `entry` should be preferred when possible. + /// + /// In particular, the hash used to initialized the raw entry must still be + /// consistent with the hash of the key that is ultimately stored in the entry. + /// This is because implementations of HashMap may need to recompute hashes + /// when resizing, at which point only the keys are available. + /// + /// Raw entries give mutable access to the keys. This must not be used + /// to modify how the key would compare or hash, as the map will not re-evaluate + /// where the key should go, meaning the keys may become "lost" if their + /// location does not reflect their state. For instance, if you change a key + /// so that the map now contains keys which compare equal, search may start + /// acting erratically, with two keys randomly masking each other. Implementations + /// are free to assume this doesn't happen (within the limits of memory-safety). + #[cfg_attr(feature = "inline-more", inline)] + pub fn raw_entry_mut(&mut self) -> RawEntryBuilderMut<'_, K, V, S> { + RawEntryBuilderMut { map: self } + } + + /// Creates a raw immutable entry builder for the HashMap. + /// + /// Raw entries provide the lowest level of control for searching and + /// manipulating a map. They must be manually initialized with a hash and + /// then manually searched. + /// + /// This is useful for + /// * Hash memoization + /// * Using a search key that doesn't work with the Borrow trait + /// * Using custom comparison logic without newtype wrappers + /// + /// Unless you are in such a situation, higher-level and more foolproof APIs like + /// `get` should be preferred. + /// + /// Immutable raw entries have very limited use; you might instead want `raw_entry_mut`. + #[cfg_attr(feature = "inline-more", inline)] + pub fn raw_entry(&self) -> RawEntryBuilder<'_, K, V, S> { + RawEntryBuilder { map: self } + } +} + +impl<K, V, S> PartialEq for HashMap<K, V, S> +where + K: Eq + Hash, + V: PartialEq, + S: BuildHasher, +{ + fn eq(&self, other: &Self) -> bool { + if self.len() != other.len() { + return false; + } + + self.iter() + .all(|(key, value)| other.get(key).map_or(false, |v| *value == *v)) + } +} + +impl<K, V, S> Eq for HashMap<K, V, S> +where + K: Eq + Hash, + V: Eq, + S: BuildHasher, +{ +} + +impl<K, V, S> Debug for HashMap<K, V, S> +where + K: Debug, + V: Debug, +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_map().entries(self.iter()).finish() + } +} + +impl<K, V, S> Default for HashMap<K, V, S> +where + S: Default, +{ + /// Creates an empty `HashMap<K, V, S>`, with the `Default` value for the hasher. + #[cfg_attr(feature = "inline-more", inline)] + fn default() -> Self { + Self::with_hasher(Default::default()) + } +} + +impl<K, Q: ?Sized, V, S> Index<&Q> for HashMap<K, V, S> +where + K: Eq + Hash + Borrow<Q>, + Q: Eq + Hash, + S: BuildHasher, +{ + type Output = V; + + /// Returns a reference to the value corresponding to the supplied key. + /// + /// # Panics + /// + /// Panics if the key is not present in the `HashMap`. + #[cfg_attr(feature = "inline-more", inline)] + fn index(&self, key: &Q) -> &V { + self.get(key).expect("no entry found for key") + } +} + +/// An iterator over the entries of a `HashMap`. +/// +/// This `struct` is created by the [`iter`] method on [`HashMap`]. See its +/// documentation for more. +/// +/// [`iter`]: struct.HashMap.html#method.iter +/// [`HashMap`]: struct.HashMap.html +pub struct Iter<'a, K, V> { + inner: RawIter<(K, V)>, + marker: PhantomData<(&'a K, &'a V)>, +} + +// FIXME(#26925) Remove in favor of `#[derive(Clone)]` +impl<K, V> Clone for Iter<'_, K, V> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + Iter { + inner: self.inner.clone(), + marker: PhantomData, + } + } +} + +impl<K: Debug, V: Debug> fmt::Debug for Iter<'_, K, V> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.clone()).finish() + } +} + +/// A mutable iterator over the entries of a `HashMap`. +/// +/// This `struct` is created by the [`iter_mut`] method on [`HashMap`]. See its +/// documentation for more. +/// +/// [`iter_mut`]: struct.HashMap.html#method.iter_mut +/// [`HashMap`]: struct.HashMap.html +pub struct IterMut<'a, K, V> { + inner: RawIter<(K, V)>, + // To ensure invariance with respect to V + marker: PhantomData<(&'a K, &'a mut V)>, +} + +// We override the default Send impl which has K: Sync instead of K: Send. Both +// are correct, but this one is more general since it allows keys which +// implement Send but not Sync. +unsafe impl<K: Send, V: Send> Send for IterMut<'_, K, V> {} + +impl<K, V> IterMut<'_, K, V> { + /// Returns a iterator of references over the remaining items. + #[cfg_attr(feature = "inline-more", inline)] + pub(super) fn iter(&self) -> Iter<'_, K, V> { + Iter { + inner: self.inner.clone(), + marker: PhantomData, + } + } +} + +/// An owning iterator over the entries of a `HashMap`. +/// +/// This `struct` is created by the [`into_iter`] method on [`HashMap`] +/// (provided by the `IntoIterator` trait). See its documentation for more. +/// +/// [`into_iter`]: struct.HashMap.html#method.into_iter +/// [`HashMap`]: struct.HashMap.html +pub struct IntoIter<K, V> { + inner: RawIntoIter<(K, V)>, +} + +impl<K, V> IntoIter<K, V> { + /// Returns a iterator of references over the remaining items. + #[cfg_attr(feature = "inline-more", inline)] + pub(super) fn iter(&self) -> Iter<'_, K, V> { + Iter { + inner: self.inner.iter(), + marker: PhantomData, + } + } +} + +/// An iterator over the keys of a `HashMap`. +/// +/// This `struct` is created by the [`keys`] method on [`HashMap`]. See its +/// documentation for more. +/// +/// [`keys`]: struct.HashMap.html#method.keys +/// [`HashMap`]: struct.HashMap.html +pub struct Keys<'a, K, V> { + inner: Iter<'a, K, V>, +} + +// FIXME(#26925) Remove in favor of `#[derive(Clone)]` +impl<K, V> Clone for Keys<'_, K, V> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + Keys { + inner: self.inner.clone(), + } + } +} + +impl<K: Debug, V> fmt::Debug for Keys<'_, K, V> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.clone()).finish() + } +} + +/// An iterator over the values of a `HashMap`. +/// +/// This `struct` is created by the [`values`] method on [`HashMap`]. See its +/// documentation for more. +/// +/// [`values`]: struct.HashMap.html#method.values +/// [`HashMap`]: struct.HashMap.html +pub struct Values<'a, K, V> { + inner: Iter<'a, K, V>, +} + +// FIXME(#26925) Remove in favor of `#[derive(Clone)]` +impl<K, V> Clone for Values<'_, K, V> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + Values { + inner: self.inner.clone(), + } + } +} + +impl<K, V: Debug> fmt::Debug for Values<'_, K, V> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.clone()).finish() + } +} + +/// A draining iterator over the entries of a `HashMap`. +/// +/// This `struct` is created by the [`drain`] method on [`HashMap`]. See its +/// documentation for more. +/// +/// [`drain`]: struct.HashMap.html#method.drain +/// [`HashMap`]: struct.HashMap.html +pub struct Drain<'a, K, V> { + inner: RawDrain<'a, (K, V)>, +} + +impl<K, V> Drain<'_, K, V> { + /// Returns a iterator of references over the remaining items. + #[cfg_attr(feature = "inline-more", inline)] + pub(super) fn iter(&self) -> Iter<'_, K, V> { + Iter { + inner: self.inner.iter(), + marker: PhantomData, + } + } +} + +/// A draining iterator over entries of a `HashMap` which don't satisfy the predicate `f`. +/// +/// This `struct` is created by the [`drain_filter`] method on [`HashMap`]. See its +/// documentation for more. +/// +/// [`drain_filter`]: struct.HashMap.html#method.drain_filter +/// [`HashMap`]: struct.HashMap.html +pub struct DrainFilter<'a, K, V, F> +where + F: FnMut(&K, &mut V) -> bool, +{ + f: F, + inner: DrainFilterInner<'a, K, V>, +} + +impl<'a, K, V, F> Drop for DrainFilter<'a, K, V, F> +where + F: FnMut(&K, &mut V) -> bool, +{ + #[cfg_attr(feature = "inline-more", inline)] + fn drop(&mut self) { + while let Some(item) = self.next() { + let guard = ConsumeAllOnDrop(self); + drop(item); + mem::forget(guard); + } + } +} + +pub(super) struct ConsumeAllOnDrop<'a, T: Iterator>(pub &'a mut T); + +impl<T: Iterator> Drop for ConsumeAllOnDrop<'_, T> { + #[cfg_attr(feature = "inline-more", inline)] + fn drop(&mut self) { + self.0.for_each(drop) + } +} + +impl<K, V, F> Iterator for DrainFilter<'_, K, V, F> +where + F: FnMut(&K, &mut V) -> bool, +{ + type Item = (K, V); + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<Self::Item> { + self.inner.next(&mut self.f) + } + + #[inline] + fn size_hint(&self) -> (usize, Option<usize>) { + (0, self.inner.iter.size_hint().1) + } +} + +impl<K, V, F> FusedIterator for DrainFilter<'_, K, V, F> where F: FnMut(&K, &mut V) -> bool {} + +/// Portions of `DrainFilter` shared with `set::DrainFilter` +pub(super) struct DrainFilterInner<'a, K, V> { + pub iter: RawIter<(K, V)>, + pub table: &'a mut RawTable<(K, V)>, +} + +impl<K, V> DrainFilterInner<'_, K, V> { + #[cfg_attr(feature = "inline-more", inline)] + pub(super) fn next<F>(&mut self, f: &mut F) -> Option<(K, V)> + where + F: FnMut(&K, &mut V) -> bool, + { + unsafe { + while let Some(item) = self.iter.next() { + let &mut (ref key, ref mut value) = item.as_mut(); + if f(key, value) { + return Some(self.table.remove(item)); + } + } + } + None + } +} + +/// A mutable iterator over the values of a `HashMap`. +/// +/// This `struct` is created by the [`values_mut`] method on [`HashMap`]. See its +/// documentation for more. +/// +/// [`values_mut`]: struct.HashMap.html#method.values_mut +/// [`HashMap`]: struct.HashMap.html +pub struct ValuesMut<'a, K, V> { + inner: IterMut<'a, K, V>, +} + +/// A builder for computing where in a [`HashMap`] a key-value pair would be stored. +/// +/// See the [`HashMap::raw_entry_mut`] docs for usage examples. +/// +/// [`HashMap::raw_entry_mut`]: struct.HashMap.html#method.raw_entry_mut +pub struct RawEntryBuilderMut<'a, K, V, S> { + map: &'a mut HashMap<K, V, S>, +} + +/// A view into a single entry in a map, which may either be vacant or occupied. +/// +/// This is a lower-level version of [`Entry`]. +/// +/// This `enum` is constructed through the [`raw_entry_mut`] method on [`HashMap`], +/// then calling one of the methods of that [`RawEntryBuilderMut`]. +/// +/// [`HashMap`]: struct.HashMap.html +/// [`Entry`]: enum.Entry.html +/// [`raw_entry_mut`]: struct.HashMap.html#method.raw_entry_mut +/// [`RawEntryBuilderMut`]: struct.RawEntryBuilderMut.html +pub enum RawEntryMut<'a, K, V, S> { + /// An occupied entry. + Occupied(RawOccupiedEntryMut<'a, K, V, S>), + /// A vacant entry. + Vacant(RawVacantEntryMut<'a, K, V, S>), +} + +/// A view into an occupied entry in a `HashMap`. +/// It is part of the [`RawEntryMut`] enum. +/// +/// [`RawEntryMut`]: enum.RawEntryMut.html +pub struct RawOccupiedEntryMut<'a, K, V, S> { + elem: Bucket<(K, V)>, + table: &'a mut RawTable<(K, V)>, + hash_builder: &'a S, +} + +unsafe impl<K, V, S> Send for RawOccupiedEntryMut<'_, K, V, S> +where + K: Send, + V: Send, + S: Sync, +{ +} +unsafe impl<K, V, S> Sync for RawOccupiedEntryMut<'_, K, V, S> +where + K: Sync, + V: Sync, + S: Sync, +{ +} + +/// A view into a vacant entry in a `HashMap`. +/// It is part of the [`RawEntryMut`] enum. +/// +/// [`RawEntryMut`]: enum.RawEntryMut.html +pub struct RawVacantEntryMut<'a, K, V, S> { + table: &'a mut RawTable<(K, V)>, + hash_builder: &'a S, +} + +/// A builder for computing where in a [`HashMap`] a key-value pair would be stored. +/// +/// See the [`HashMap::raw_entry`] docs for usage examples. +/// +/// [`HashMap::raw_entry`]: struct.HashMap.html#method.raw_entry +pub struct RawEntryBuilder<'a, K, V, S> { + map: &'a HashMap<K, V, S>, +} + +impl<'a, K, V, S> RawEntryBuilderMut<'a, K, V, S> { + /// Creates a `RawEntryMut` from the given key. + #[cfg_attr(feature = "inline-more", inline)] + #[allow(clippy::wrong_self_convention)] + pub fn from_key<Q: ?Sized>(self, k: &Q) -> RawEntryMut<'a, K, V, S> + where + S: BuildHasher, + K: Borrow<Q>, + Q: Hash + Eq, + { + let mut hasher = self.map.hash_builder.build_hasher(); + k.hash(&mut hasher); + self.from_key_hashed_nocheck(hasher.finish(), k) + } + + /// Creates a `RawEntryMut` from the given key and its hash. + #[inline] + #[allow(clippy::wrong_self_convention)] + pub fn from_key_hashed_nocheck<Q: ?Sized>(self, hash: u64, k: &Q) -> RawEntryMut<'a, K, V, S> + where + K: Borrow<Q>, + Q: Eq, + { + self.from_hash(hash, |q| q.borrow().eq(k)) + } +} + +impl<'a, K, V, S> RawEntryBuilderMut<'a, K, V, S> { + /// Creates a `RawEntryMut` from the given hash. + #[cfg_attr(feature = "inline-more", inline)] + #[allow(clippy::wrong_self_convention)] + pub fn from_hash<F>(self, hash: u64, is_match: F) -> RawEntryMut<'a, K, V, S> + where + for<'b> F: FnMut(&'b K) -> bool, + { + self.search(hash, is_match) + } + + #[cfg_attr(feature = "inline-more", inline)] + fn search<F>(self, hash: u64, mut is_match: F) -> RawEntryMut<'a, K, V, S> + where + for<'b> F: FnMut(&'b K) -> bool, + { + match self.map.table.find(hash, |(k, _)| is_match(k)) { + Some(elem) => RawEntryMut::Occupied(RawOccupiedEntryMut { + elem, + table: &mut self.map.table, + hash_builder: &self.map.hash_builder, + }), + None => RawEntryMut::Vacant(RawVacantEntryMut { + table: &mut self.map.table, + hash_builder: &self.map.hash_builder, + }), + } + } +} + +impl<'a, K, V, S> RawEntryBuilder<'a, K, V, S> { + /// Access an entry by key. + #[cfg_attr(feature = "inline-more", inline)] + #[allow(clippy::wrong_self_convention)] + pub fn from_key<Q: ?Sized>(self, k: &Q) -> Option<(&'a K, &'a V)> + where + S: BuildHasher, + K: Borrow<Q>, + Q: Hash + Eq, + { + let mut hasher = self.map.hash_builder.build_hasher(); + k.hash(&mut hasher); + self.from_key_hashed_nocheck(hasher.finish(), k) + } + + /// Access an entry by a key and its hash. + #[cfg_attr(feature = "inline-more", inline)] + #[allow(clippy::wrong_self_convention)] + pub fn from_key_hashed_nocheck<Q: ?Sized>(self, hash: u64, k: &Q) -> Option<(&'a K, &'a V)> + where + K: Borrow<Q>, + Q: Eq, + { + self.from_hash(hash, |q| q.borrow().eq(k)) + } + + #[cfg_attr(feature = "inline-more", inline)] + fn search<F>(self, hash: u64, mut is_match: F) -> Option<(&'a K, &'a V)> + where + F: FnMut(&K) -> bool, + { + match self.map.table.get(hash, |(k, _)| is_match(k)) { + Some(&(ref key, ref value)) => Some((key, value)), + None => None, + } + } + + /// Access an entry by hash. + #[cfg_attr(feature = "inline-more", inline)] + #[allow(clippy::wrong_self_convention)] + pub fn from_hash<F>(self, hash: u64, is_match: F) -> Option<(&'a K, &'a V)> + where + F: FnMut(&K) -> bool, + { + self.search(hash, is_match) + } +} + +impl<'a, K, V, S> RawEntryMut<'a, K, V, S> { + /// Sets the value of the entry, and returns a RawOccupiedEntryMut. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// let entry = map.raw_entry_mut().from_key("horseyland").insert("horseyland", 37); + /// + /// assert_eq!(entry.remove_entry(), ("horseyland", 37)); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(self, key: K, value: V) -> RawOccupiedEntryMut<'a, K, V, S> + where + K: Hash, + S: BuildHasher, + { + match self { + RawEntryMut::Occupied(mut entry) => { + entry.insert(value); + entry + } + RawEntryMut::Vacant(entry) => entry.insert_entry(key, value), + } + } + + /// Ensures a value is in the entry by inserting the default if empty, and returns + /// mutable references to the key and value in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// map.raw_entry_mut().from_key("poneyland").or_insert("poneyland", 3); + /// assert_eq!(map["poneyland"], 3); + /// + /// *map.raw_entry_mut().from_key("poneyland").or_insert("poneyland", 10).1 *= 2; + /// assert_eq!(map["poneyland"], 6); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn or_insert(self, default_key: K, default_val: V) -> (&'a mut K, &'a mut V) + where + K: Hash, + S: BuildHasher, + { + match self { + RawEntryMut::Occupied(entry) => entry.into_key_value(), + RawEntryMut::Vacant(entry) => entry.insert(default_key, default_val), + } + } + + /// Ensures a value is in the entry by inserting the result of the default function if empty, + /// and returns mutable references to the key and value in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, String> = HashMap::new(); + /// + /// map.raw_entry_mut().from_key("poneyland").or_insert_with(|| { + /// ("poneyland", "hoho".to_string()) + /// }); + /// + /// assert_eq!(map["poneyland"], "hoho".to_string()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn or_insert_with<F>(self, default: F) -> (&'a mut K, &'a mut V) + where + F: FnOnce() -> (K, V), + K: Hash, + S: BuildHasher, + { + match self { + RawEntryMut::Occupied(entry) => entry.into_key_value(), + RawEntryMut::Vacant(entry) => { + let (k, v) = default(); + entry.insert(k, v) + } + } + } + + /// Provides in-place mutable access to an occupied entry before any + /// potential inserts into the map. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// map.raw_entry_mut() + /// .from_key("poneyland") + /// .and_modify(|_k, v| { *v += 1 }) + /// .or_insert("poneyland", 42); + /// assert_eq!(map["poneyland"], 42); + /// + /// map.raw_entry_mut() + /// .from_key("poneyland") + /// .and_modify(|_k, v| { *v += 1 }) + /// .or_insert("poneyland", 0); + /// assert_eq!(map["poneyland"], 43); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn and_modify<F>(self, f: F) -> Self + where + F: FnOnce(&mut K, &mut V), + { + match self { + RawEntryMut::Occupied(mut entry) => { + { + let (k, v) = entry.get_key_value_mut(); + f(k, v); + } + RawEntryMut::Occupied(entry) + } + RawEntryMut::Vacant(entry) => RawEntryMut::Vacant(entry), + } + } + + /// Provides shared access to the key and owned access to the value of + /// an occupied entry and allows to replace or remove it based on the + /// value of the returned option. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::RawEntryMut; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// let entry = map + /// .raw_entry_mut() + /// .from_key("poneyland") + /// .and_replace_entry_with(|_k, _v| panic!()); + /// + /// match entry { + /// RawEntryMut::Vacant(_) => {}, + /// RawEntryMut::Occupied(_) => panic!(), + /// } + /// + /// map.insert("poneyland", 42); + /// + /// let entry = map + /// .raw_entry_mut() + /// .from_key("poneyland") + /// .and_replace_entry_with(|k, v| { + /// assert_eq!(k, &"poneyland"); + /// assert_eq!(v, 42); + /// Some(v + 1) + /// }); + /// + /// match entry { + /// RawEntryMut::Occupied(e) => { + /// assert_eq!(e.key(), &"poneyland"); + /// assert_eq!(e.get(), &43); + /// }, + /// RawEntryMut::Vacant(_) => panic!(), + /// } + /// + /// assert_eq!(map["poneyland"], 43); + /// + /// let entry = map + /// .raw_entry_mut() + /// .from_key("poneyland") + /// .and_replace_entry_with(|_k, _v| None); + /// + /// match entry { + /// RawEntryMut::Vacant(_) => {}, + /// RawEntryMut::Occupied(_) => panic!(), + /// } + /// + /// assert!(!map.contains_key("poneyland")); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn and_replace_entry_with<F>(self, f: F) -> Self + where + F: FnOnce(&K, V) -> Option<V>, + { + match self { + RawEntryMut::Occupied(entry) => entry.replace_entry_with(f), + RawEntryMut::Vacant(_) => self, + } + } +} + +impl<'a, K, V, S> RawOccupiedEntryMut<'a, K, V, S> { + /// Gets a reference to the key in the entry. + #[cfg_attr(feature = "inline-more", inline)] + pub fn key(&self) -> &K { + unsafe { &self.elem.as_ref().0 } + } + + /// Gets a mutable reference to the key in the entry. + #[cfg_attr(feature = "inline-more", inline)] + pub fn key_mut(&mut self) -> &mut K { + unsafe { &mut self.elem.as_mut().0 } + } + + /// Converts the entry into a mutable reference to the key in the entry + /// with a lifetime bound to the map itself. + #[cfg_attr(feature = "inline-more", inline)] + pub fn into_key(self) -> &'a mut K { + unsafe { &mut self.elem.as_mut().0 } + } + + /// Gets a reference to the value in the entry. + #[cfg_attr(feature = "inline-more", inline)] + pub fn get(&self) -> &V { + unsafe { &self.elem.as_ref().1 } + } + + /// Converts the OccupiedEntry into a mutable reference to the value in the entry + /// with a lifetime bound to the map itself. + #[cfg_attr(feature = "inline-more", inline)] + pub fn into_mut(self) -> &'a mut V { + unsafe { &mut self.elem.as_mut().1 } + } + + /// Gets a mutable reference to the value in the entry. + #[cfg_attr(feature = "inline-more", inline)] + pub fn get_mut(&mut self) -> &mut V { + unsafe { &mut self.elem.as_mut().1 } + } + + /// Gets a reference to the key and value in the entry. + #[cfg_attr(feature = "inline-more", inline)] + pub fn get_key_value(&mut self) -> (&K, &V) { + unsafe { + let &(ref key, ref value) = self.elem.as_ref(); + (key, value) + } + } + + /// Gets a mutable reference to the key and value in the entry. + #[cfg_attr(feature = "inline-more", inline)] + pub fn get_key_value_mut(&mut self) -> (&mut K, &mut V) { + unsafe { + let &mut (ref mut key, ref mut value) = self.elem.as_mut(); + (key, value) + } + } + + /// Converts the OccupiedEntry into a mutable reference to the key and value in the entry + /// with a lifetime bound to the map itself. + #[cfg_attr(feature = "inline-more", inline)] + pub fn into_key_value(self) -> (&'a mut K, &'a mut V) { + unsafe { + let &mut (ref mut key, ref mut value) = self.elem.as_mut(); + (key, value) + } + } + + /// Sets the value of the entry, and returns the entry's old value. + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(&mut self, value: V) -> V { + mem::replace(self.get_mut(), value) + } + + /// Sets the value of the entry, and returns the entry's old value. + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert_key(&mut self, key: K) -> K { + mem::replace(self.key_mut(), key) + } + + /// Takes the value out of the entry, and returns it. + #[cfg_attr(feature = "inline-more", inline)] + pub fn remove(self) -> V { + self.remove_entry().1 + } + + /// Take the ownership of the key and value from the map. + #[cfg_attr(feature = "inline-more", inline)] + pub fn remove_entry(self) -> (K, V) { + unsafe { self.table.remove(self.elem) } + } + + /// Provides shared access to the key and owned access to the value of + /// the entry and allows to replace or remove it based on the + /// value of the returned option. + #[cfg_attr(feature = "inline-more", inline)] + pub fn replace_entry_with<F>(self, f: F) -> RawEntryMut<'a, K, V, S> + where + F: FnOnce(&K, V) -> Option<V>, + { + unsafe { + let still_occupied = self + .table + .replace_bucket_with(self.elem.clone(), |(key, value)| { + f(&key, value).map(|new_value| (key, new_value)) + }); + + if still_occupied { + RawEntryMut::Occupied(self) + } else { + RawEntryMut::Vacant(RawVacantEntryMut { + table: self.table, + hash_builder: self.hash_builder, + }) + } + } + } +} + +impl<'a, K, V, S> RawVacantEntryMut<'a, K, V, S> { + /// Sets the value of the entry with the VacantEntry's key, + /// and returns a mutable reference to it. + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(self, key: K, value: V) -> (&'a mut K, &'a mut V) + where + K: Hash, + S: BuildHasher, + { + let mut hasher = self.hash_builder.build_hasher(); + key.hash(&mut hasher); + self.insert_hashed_nocheck(hasher.finish(), key, value) + } + + /// Sets the value of the entry with the VacantEntry's key, + /// and returns a mutable reference to it. + #[cfg_attr(feature = "inline-more", inline)] + #[allow(clippy::shadow_unrelated)] + pub fn insert_hashed_nocheck(self, hash: u64, key: K, value: V) -> (&'a mut K, &'a mut V) + where + K: Hash, + S: BuildHasher, + { + let hash_builder = self.hash_builder; + self.insert_with_hasher(hash, key, value, |k| make_hash(hash_builder, k)) + } + + /// Set the value of an entry with a custom hasher function. + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert_with_hasher<H>( + self, + hash: u64, + key: K, + value: V, + hasher: H, + ) -> (&'a mut K, &'a mut V) + where + H: Fn(&K) -> u64, + { + let &mut (ref mut k, ref mut v) = self + .table + .insert_entry(hash, (key, value), |x| hasher(&x.0)); + (k, v) + } + + #[cfg_attr(feature = "inline-more", inline)] + fn insert_entry(self, key: K, value: V) -> RawOccupiedEntryMut<'a, K, V, S> + where + K: Hash, + S: BuildHasher, + { + let hash_builder = self.hash_builder; + let mut hasher = self.hash_builder.build_hasher(); + key.hash(&mut hasher); + + let elem = self.table.insert(hasher.finish(), (key, value), |k| { + make_hash(hash_builder, &k.0) + }); + RawOccupiedEntryMut { + elem, + table: self.table, + hash_builder: self.hash_builder, + } + } +} + +impl<K, V, S> Debug for RawEntryBuilderMut<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("RawEntryBuilder").finish() + } +} + +impl<K: Debug, V: Debug, S> Debug for RawEntryMut<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + match *self { + RawEntryMut::Vacant(ref v) => f.debug_tuple("RawEntry").field(v).finish(), + RawEntryMut::Occupied(ref o) => f.debug_tuple("RawEntry").field(o).finish(), + } + } +} + +impl<K: Debug, V: Debug, S> Debug for RawOccupiedEntryMut<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("RawOccupiedEntryMut") + .field("key", self.key()) + .field("value", self.get()) + .finish() + } +} + +impl<K, V, S> Debug for RawVacantEntryMut<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("RawVacantEntryMut").finish() + } +} + +impl<K, V, S> Debug for RawEntryBuilder<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("RawEntryBuilder").finish() + } +} + +/// A view into a single entry in a map, which may either be vacant or occupied. +/// +/// This `enum` is constructed from the [`entry`] method on [`HashMap`]. +/// +/// [`HashMap`]: struct.HashMap.html +/// [`entry`]: struct.HashMap.html#method.entry +pub enum Entry<'a, K, V, S> { + /// An occupied entry. + Occupied(OccupiedEntry<'a, K, V, S>), + + /// A vacant entry. + Vacant(VacantEntry<'a, K, V, S>), +} + +impl<K: Debug, V: Debug, S> Debug for Entry<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + match *self { + Entry::Vacant(ref v) => f.debug_tuple("Entry").field(v).finish(), + Entry::Occupied(ref o) => f.debug_tuple("Entry").field(o).finish(), + } + } +} + +/// A view into an occupied entry in a `HashMap`. +/// It is part of the [`Entry`] enum. +/// +/// [`Entry`]: enum.Entry.html +pub struct OccupiedEntry<'a, K, V, S> { + hash: u64, + key: Option<K>, + elem: Bucket<(K, V)>, + table: &'a mut HashMap<K, V, S>, +} + +unsafe impl<K, V, S> Send for OccupiedEntry<'_, K, V, S> +where + K: Send, + V: Send, + S: Send, +{ +} +unsafe impl<K, V, S> Sync for OccupiedEntry<'_, K, V, S> +where + K: Sync, + V: Sync, + S: Sync, +{ +} + +impl<K: Debug, V: Debug, S> Debug for OccupiedEntry<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("OccupiedEntry") + .field("key", self.key()) + .field("value", self.get()) + .finish() + } +} + +/// A view into a vacant entry in a `HashMap`. +/// It is part of the [`Entry`] enum. +/// +/// [`Entry`]: enum.Entry.html +pub struct VacantEntry<'a, K, V, S> { + hash: u64, + key: K, + table: &'a mut HashMap<K, V, S>, +} + +impl<K: Debug, V, S> Debug for VacantEntry<'_, K, V, S> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_tuple("VacantEntry").field(self.key()).finish() + } +} + +impl<'a, K, V, S> IntoIterator for &'a HashMap<K, V, S> { + type Item = (&'a K, &'a V); + type IntoIter = Iter<'a, K, V>; + + #[cfg_attr(feature = "inline-more", inline)] + fn into_iter(self) -> Iter<'a, K, V> { + self.iter() + } +} + +impl<'a, K, V, S> IntoIterator for &'a mut HashMap<K, V, S> { + type Item = (&'a K, &'a mut V); + type IntoIter = IterMut<'a, K, V>; + + #[cfg_attr(feature = "inline-more", inline)] + fn into_iter(self) -> IterMut<'a, K, V> { + self.iter_mut() + } +} + +impl<K, V, S> IntoIterator for HashMap<K, V, S> { + type Item = (K, V); + type IntoIter = IntoIter<K, V>; + + /// Creates a consuming iterator, that is, one that moves each key-value + /// pair out of the map in arbitrary order. The map cannot be used after + /// calling this. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map = HashMap::new(); + /// map.insert("a", 1); + /// map.insert("b", 2); + /// map.insert("c", 3); + /// + /// // Not possible with .iter() + /// let vec: Vec<(&str, i32)> = map.into_iter().collect(); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + fn into_iter(self) -> IntoIter<K, V> { + IntoIter { + inner: self.table.into_iter(), + } + } +} + +impl<'a, K, V> Iterator for Iter<'a, K, V> { + type Item = (&'a K, &'a V); + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<(&'a K, &'a V)> { + // Avoid `Option::map` because it bloats LLVM IR. + match self.inner.next() { + Some(x) => unsafe { + let r = x.as_ref(); + Some((&r.0, &r.1)) + }, + None => None, + } + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.inner.size_hint() + } +} +impl<K, V> ExactSizeIterator for Iter<'_, K, V> { + #[cfg_attr(feature = "inline-more", inline)] + fn len(&self) -> usize { + self.inner.len() + } +} + +impl<K, V> FusedIterator for Iter<'_, K, V> {} + +impl<'a, K, V> Iterator for IterMut<'a, K, V> { + type Item = (&'a K, &'a mut V); + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<(&'a K, &'a mut V)> { + // Avoid `Option::map` because it bloats LLVM IR. + match self.inner.next() { + Some(x) => unsafe { + let r = x.as_mut(); + Some((&r.0, &mut r.1)) + }, + None => None, + } + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.inner.size_hint() + } +} +impl<K, V> ExactSizeIterator for IterMut<'_, K, V> { + #[cfg_attr(feature = "inline-more", inline)] + fn len(&self) -> usize { + self.inner.len() + } +} +impl<K, V> FusedIterator for IterMut<'_, K, V> {} + +impl<K, V> fmt::Debug for IterMut<'_, K, V> +where + K: fmt::Debug, + V: fmt::Debug, +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.iter()).finish() + } +} + +impl<K, V> Iterator for IntoIter<K, V> { + type Item = (K, V); + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<(K, V)> { + self.inner.next() + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.inner.size_hint() + } +} +impl<K, V> ExactSizeIterator for IntoIter<K, V> { + #[cfg_attr(feature = "inline-more", inline)] + fn len(&self) -> usize { + self.inner.len() + } +} +impl<K, V> FusedIterator for IntoIter<K, V> {} + +impl<K: Debug, V: Debug> fmt::Debug for IntoIter<K, V> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.iter()).finish() + } +} + +impl<'a, K, V> Iterator for Keys<'a, K, V> { + type Item = &'a K; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<&'a K> { + // Avoid `Option::map` because it bloats LLVM IR. + match self.inner.next() { + Some((k, _)) => Some(k), + None => None, + } + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.inner.size_hint() + } +} +impl<K, V> ExactSizeIterator for Keys<'_, K, V> { + #[cfg_attr(feature = "inline-more", inline)] + fn len(&self) -> usize { + self.inner.len() + } +} +impl<K, V> FusedIterator for Keys<'_, K, V> {} + +impl<'a, K, V> Iterator for Values<'a, K, V> { + type Item = &'a V; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<&'a V> { + // Avoid `Option::map` because it bloats LLVM IR. + match self.inner.next() { + Some((_, v)) => Some(v), + None => None, + } + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.inner.size_hint() + } +} +impl<K, V> ExactSizeIterator for Values<'_, K, V> { + #[cfg_attr(feature = "inline-more", inline)] + fn len(&self) -> usize { + self.inner.len() + } +} +impl<K, V> FusedIterator for Values<'_, K, V> {} + +impl<'a, K, V> Iterator for ValuesMut<'a, K, V> { + type Item = &'a mut V; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<&'a mut V> { + // Avoid `Option::map` because it bloats LLVM IR. + match self.inner.next() { + Some((_, v)) => Some(v), + None => None, + } + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.inner.size_hint() + } +} +impl<K, V> ExactSizeIterator for ValuesMut<'_, K, V> { + #[cfg_attr(feature = "inline-more", inline)] + fn len(&self) -> usize { + self.inner.len() + } +} +impl<K, V> FusedIterator for ValuesMut<'_, K, V> {} + +impl<K, V> fmt::Debug for ValuesMut<'_, K, V> +where + K: fmt::Debug, + V: fmt::Debug, +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.inner.iter()).finish() + } +} + +impl<'a, K, V> Iterator for Drain<'a, K, V> { + type Item = (K, V); + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<(K, V)> { + self.inner.next() + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.inner.size_hint() + } +} +impl<K, V> ExactSizeIterator for Drain<'_, K, V> { + #[cfg_attr(feature = "inline-more", inline)] + fn len(&self) -> usize { + self.inner.len() + } +} +impl<K, V> FusedIterator for Drain<'_, K, V> {} + +impl<K, V> fmt::Debug for Drain<'_, K, V> +where + K: fmt::Debug, + V: fmt::Debug, +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.iter()).finish() + } +} + +impl<'a, K, V, S> Entry<'a, K, V, S> { + /// Sets the value of the entry, and returns an OccupiedEntry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// let entry = map.entry("horseyland").insert(37); + /// + /// assert_eq!(entry.key(), &"horseyland"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(self, value: V) -> OccupiedEntry<'a, K, V, S> + where + K: Hash, + S: BuildHasher, + { + match self { + Entry::Occupied(mut entry) => { + entry.insert(value); + entry + } + Entry::Vacant(entry) => entry.insert_entry(value), + } + } + + /// Ensures a value is in the entry by inserting the default if empty, and returns + /// a mutable reference to the value in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// map.entry("poneyland").or_insert(3); + /// assert_eq!(map["poneyland"], 3); + /// + /// *map.entry("poneyland").or_insert(10) *= 2; + /// assert_eq!(map["poneyland"], 6); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn or_insert(self, default: V) -> &'a mut V + where + K: Hash, + S: BuildHasher, + { + match self { + Entry::Occupied(entry) => entry.into_mut(), + Entry::Vacant(entry) => entry.insert(default), + } + } + + /// Ensures a value is in the entry by inserting the result of the default function if empty, + /// and returns a mutable reference to the value in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, String> = HashMap::new(); + /// let s = "hoho".to_string(); + /// + /// map.entry("poneyland").or_insert_with(|| s); + /// + /// assert_eq!(map["poneyland"], "hoho".to_string()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn or_insert_with<F: FnOnce() -> V>(self, default: F) -> &'a mut V + where + K: Hash, + S: BuildHasher, + { + match self { + Entry::Occupied(entry) => entry.into_mut(), + Entry::Vacant(entry) => entry.insert(default()), + } + } + + /// Ensures a value is in the entry by inserting, if empty, the result of the default function, + /// which takes the key as its argument, and returns a mutable reference to the value in the + /// entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, usize> = HashMap::new(); + /// + /// map.entry("poneyland").or_insert_with_key(|key| key.chars().count()); + /// + /// assert_eq!(map["poneyland"], 9); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn or_insert_with_key<F: FnOnce(&K) -> V>(self, default: F) -> &'a mut V + where + K: Hash, + S: BuildHasher, + { + match self { + Entry::Occupied(entry) => entry.into_mut(), + Entry::Vacant(entry) => { + let value = default(entry.key()); + entry.insert(value) + } + } + } + + /// Returns a reference to this entry's key. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// assert_eq!(map.entry("poneyland").key(), &"poneyland"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn key(&self) -> &K { + match *self { + Entry::Occupied(ref entry) => entry.key(), + Entry::Vacant(ref entry) => entry.key(), + } + } + + /// Provides in-place mutable access to an occupied entry before any + /// potential inserts into the map. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// map.entry("poneyland") + /// .and_modify(|e| { *e += 1 }) + /// .or_insert(42); + /// assert_eq!(map["poneyland"], 42); + /// + /// map.entry("poneyland") + /// .and_modify(|e| { *e += 1 }) + /// .or_insert(42); + /// assert_eq!(map["poneyland"], 43); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn and_modify<F>(self, f: F) -> Self + where + F: FnOnce(&mut V), + { + match self { + Entry::Occupied(mut entry) => { + f(entry.get_mut()); + Entry::Occupied(entry) + } + Entry::Vacant(entry) => Entry::Vacant(entry), + } + } + + /// Provides shared access to the key and owned access to the value of + /// an occupied entry and allows to replace or remove it based on the + /// value of the returned option. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::Entry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// let entry = map + /// .entry("poneyland") + /// .and_replace_entry_with(|_k, _v| panic!()); + /// + /// match entry { + /// Entry::Vacant(e) => { + /// assert_eq!(e.key(), &"poneyland"); + /// } + /// Entry::Occupied(_) => panic!(), + /// } + /// + /// map.insert("poneyland", 42); + /// + /// let entry = map + /// .entry("poneyland") + /// .and_replace_entry_with(|k, v| { + /// assert_eq!(k, &"poneyland"); + /// assert_eq!(v, 42); + /// Some(v + 1) + /// }); + /// + /// match entry { + /// Entry::Occupied(e) => { + /// assert_eq!(e.key(), &"poneyland"); + /// assert_eq!(e.get(), &43); + /// } + /// Entry::Vacant(_) => panic!(), + /// } + /// + /// assert_eq!(map["poneyland"], 43); + /// + /// let entry = map + /// .entry("poneyland") + /// .and_replace_entry_with(|_k, _v| None); + /// + /// match entry { + /// Entry::Vacant(e) => assert_eq!(e.key(), &"poneyland"), + /// Entry::Occupied(_) => panic!(), + /// } + /// + /// assert!(!map.contains_key("poneyland")); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn and_replace_entry_with<F>(self, f: F) -> Self + where + F: FnOnce(&K, V) -> Option<V>, + { + match self { + Entry::Occupied(entry) => entry.replace_entry_with(f), + Entry::Vacant(_) => self, + } + } +} + +impl<'a, K, V: Default, S> Entry<'a, K, V, S> { + /// Ensures a value is in the entry by inserting the default value if empty, + /// and returns a mutable reference to the value in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, Option<u32>> = HashMap::new(); + /// map.entry("poneyland").or_default(); + /// + /// assert_eq!(map["poneyland"], None); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn or_default(self) -> &'a mut V + where + K: Hash, + S: BuildHasher, + { + match self { + Entry::Occupied(entry) => entry.into_mut(), + Entry::Vacant(entry) => entry.insert(Default::default()), + } + } +} + +impl<'a, K, V, S> OccupiedEntry<'a, K, V, S> { + /// Gets a reference to the key in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.entry("poneyland").or_insert(12); + /// assert_eq!(map.entry("poneyland").key(), &"poneyland"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn key(&self) -> &K { + unsafe { &self.elem.as_ref().0 } + } + + /// Take the ownership of the key and value from the map. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::Entry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.entry("poneyland").or_insert(12); + /// + /// if let Entry::Occupied(o) = map.entry("poneyland") { + /// // We delete the entry from the map. + /// o.remove_entry(); + /// } + /// + /// assert_eq!(map.contains_key("poneyland"), false); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn remove_entry(self) -> (K, V) { + unsafe { self.table.table.remove(self.elem) } + } + + /// Gets a reference to the value in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::Entry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.entry("poneyland").or_insert(12); + /// + /// if let Entry::Occupied(o) = map.entry("poneyland") { + /// assert_eq!(o.get(), &12); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn get(&self) -> &V { + unsafe { &self.elem.as_ref().1 } + } + + /// Gets a mutable reference to the value in the entry. + /// + /// If you need a reference to the `OccupiedEntry` which may outlive the + /// destruction of the `Entry` value, see [`into_mut`]. + /// + /// [`into_mut`]: #method.into_mut + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::Entry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.entry("poneyland").or_insert(12); + /// + /// assert_eq!(map["poneyland"], 12); + /// if let Entry::Occupied(mut o) = map.entry("poneyland") { + /// *o.get_mut() += 10; + /// assert_eq!(*o.get(), 22); + /// + /// // We can use the same Entry multiple times. + /// *o.get_mut() += 2; + /// } + /// + /// assert_eq!(map["poneyland"], 24); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn get_mut(&mut self) -> &mut V { + unsafe { &mut self.elem.as_mut().1 } + } + + /// Converts the OccupiedEntry into a mutable reference to the value in the entry + /// with a lifetime bound to the map itself. + /// + /// If you need multiple references to the `OccupiedEntry`, see [`get_mut`]. + /// + /// [`get_mut`]: #method.get_mut + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::Entry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.entry("poneyland").or_insert(12); + /// + /// assert_eq!(map["poneyland"], 12); + /// if let Entry::Occupied(o) = map.entry("poneyland") { + /// *o.into_mut() += 10; + /// } + /// + /// assert_eq!(map["poneyland"], 22); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn into_mut(self) -> &'a mut V { + unsafe { &mut self.elem.as_mut().1 } + } + + /// Sets the value of the entry, and returns the entry's old value. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::Entry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.entry("poneyland").or_insert(12); + /// + /// if let Entry::Occupied(mut o) = map.entry("poneyland") { + /// assert_eq!(o.insert(15), 12); + /// } + /// + /// assert_eq!(map["poneyland"], 15); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(&mut self, mut value: V) -> V { + let old_value = self.get_mut(); + mem::swap(&mut value, old_value); + value + } + + /// Takes the value out of the entry, and returns it. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::Entry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.entry("poneyland").or_insert(12); + /// + /// if let Entry::Occupied(o) = map.entry("poneyland") { + /// assert_eq!(o.remove(), 12); + /// } + /// + /// assert_eq!(map.contains_key("poneyland"), false); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn remove(self) -> V { + self.remove_entry().1 + } + + /// Replaces the entry, returning the old key and value. The new key in the hash map will be + /// the key used to create this entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::hash_map::{Entry, HashMap}; + /// use std::rc::Rc; + /// + /// let mut map: HashMap<Rc<String>, u32> = HashMap::new(); + /// map.insert(Rc::new("Stringthing".to_string()), 15); + /// + /// let my_key = Rc::new("Stringthing".to_string()); + /// + /// if let Entry::Occupied(entry) = map.entry(my_key) { + /// // Also replace the key with a handle to our other key. + /// let (old_key, old_value): (Rc<String>, u32) = entry.replace_entry(16); + /// } + /// + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn replace_entry(self, value: V) -> (K, V) { + let entry = unsafe { self.elem.as_mut() }; + + let old_key = mem::replace(&mut entry.0, self.key.unwrap()); + let old_value = mem::replace(&mut entry.1, value); + + (old_key, old_value) + } + + /// Replaces the key in the hash map with the key used to create this entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::hash_map::{Entry, HashMap}; + /// use std::rc::Rc; + /// + /// let mut map: HashMap<Rc<String>, u32> = HashMap::new(); + /// let mut known_strings: Vec<Rc<String>> = Vec::new(); + /// + /// // Initialise known strings, run program, etc. + /// + /// reclaim_memory(&mut map, &known_strings); + /// + /// fn reclaim_memory(map: &mut HashMap<Rc<String>, u32>, known_strings: &[Rc<String>] ) { + /// for s in known_strings { + /// if let Entry::Occupied(entry) = map.entry(s.clone()) { + /// // Replaces the entry's key with our version of it in `known_strings`. + /// entry.replace_key(); + /// } + /// } + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn replace_key(self) -> K { + let entry = unsafe { self.elem.as_mut() }; + mem::replace(&mut entry.0, self.key.unwrap()) + } + + /// Provides shared access to the key and owned access to the value of + /// the entry and allows to replace or remove it based on the + /// value of the returned option. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::Entry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.insert("poneyland", 42); + /// + /// let entry = match map.entry("poneyland") { + /// Entry::Occupied(e) => { + /// e.replace_entry_with(|k, v| { + /// assert_eq!(k, &"poneyland"); + /// assert_eq!(v, 42); + /// Some(v + 1) + /// }) + /// } + /// Entry::Vacant(_) => panic!(), + /// }; + /// + /// match entry { + /// Entry::Occupied(e) => { + /// assert_eq!(e.key(), &"poneyland"); + /// assert_eq!(e.get(), &43); + /// } + /// Entry::Vacant(_) => panic!(), + /// } + /// + /// assert_eq!(map["poneyland"], 43); + /// + /// let entry = match map.entry("poneyland") { + /// Entry::Occupied(e) => e.replace_entry_with(|_k, _v| None), + /// Entry::Vacant(_) => panic!(), + /// }; + /// + /// match entry { + /// Entry::Vacant(e) => { + /// assert_eq!(e.key(), &"poneyland"); + /// } + /// Entry::Occupied(_) => panic!(), + /// } + /// + /// assert!(!map.contains_key("poneyland")); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn replace_entry_with<F>(self, f: F) -> Entry<'a, K, V, S> + where + F: FnOnce(&K, V) -> Option<V>, + { + unsafe { + let mut spare_key = None; + + self.table + .table + .replace_bucket_with(self.elem.clone(), |(key, value)| { + if let Some(new_value) = f(&key, value) { + Some((key, new_value)) + } else { + spare_key = Some(key); + None + } + }); + + if let Some(key) = spare_key { + Entry::Vacant(VacantEntry { + hash: self.hash, + key, + table: self.table, + }) + } else { + Entry::Occupied(self) + } + } + } +} + +impl<'a, K, V, S> VacantEntry<'a, K, V, S> { + /// Gets a reference to the key that would be used when inserting a value + /// through the `VacantEntry`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// assert_eq!(map.entry("poneyland").key(), &"poneyland"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn key(&self) -> &K { + &self.key + } + + /// Take ownership of the key. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::Entry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// if let Entry::Vacant(v) = map.entry("poneyland") { + /// v.into_key(); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn into_key(self) -> K { + self.key + } + + /// Sets the value of the entry with the VacantEntry's key, + /// and returns a mutable reference to it. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::Entry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// if let Entry::Vacant(o) = map.entry("poneyland") { + /// o.insert(37); + /// } + /// assert_eq!(map["poneyland"], 37); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(self, value: V) -> &'a mut V + where + K: Hash, + S: BuildHasher, + { + let hash_builder = &self.table.hash_builder; + let table = &mut self.table.table; + let entry = table.insert_entry(self.hash, (self.key, value), |x| { + make_hash(hash_builder, &x.0) + }); + &mut entry.1 + } + + #[cfg_attr(feature = "inline-more", inline)] + fn insert_entry(self, value: V) -> OccupiedEntry<'a, K, V, S> + where + K: Hash, + S: BuildHasher, + { + let hash_builder = &self.table.hash_builder; + let elem = self.table.table.insert(self.hash, (self.key, value), |x| { + make_hash(hash_builder, &x.0) + }); + OccupiedEntry { + hash: self.hash, + key: None, + elem, + table: self.table, + } + } +} + +impl<K, V, S> FromIterator<(K, V)> for HashMap<K, V, S> +where + K: Eq + Hash, + S: BuildHasher + Default, +{ + #[cfg_attr(feature = "inline-more", inline)] + fn from_iter<T: IntoIterator<Item = (K, V)>>(iter: T) -> Self { + let iter = iter.into_iter(); + let mut map = Self::with_capacity_and_hasher(iter.size_hint().0, S::default()); + iter.for_each(|(k, v)| { + map.insert(k, v); + }); + map + } +} + +/// Inserts all new key-values from the iterator and replaces values with existing +/// keys with new values returned from the iterator. +impl<K, V, S> Extend<(K, V)> for HashMap<K, V, S> +where + K: Eq + Hash, + S: BuildHasher, +{ + #[cfg_attr(feature = "inline-more", inline)] + fn extend<T: IntoIterator<Item = (K, V)>>(&mut self, iter: T) { + // Keys may be already present or show multiple times in the iterator. + // Reserve the entire hint lower bound if the map is empty. + // Otherwise reserve half the hint (rounded up), so the map + // will only resize twice in the worst case. + let iter = iter.into_iter(); + let reserve = if self.is_empty() { + iter.size_hint().0 + } else { + (iter.size_hint().0 + 1) / 2 + }; + self.reserve(reserve); + iter.for_each(move |(k, v)| { + self.insert(k, v); + }); + } + + #[inline] + #[cfg(feature = "nightly")] + fn extend_one(&mut self, (k, v): (K, V)) { + self.insert(k, v); + } + + #[inline] + #[cfg(feature = "nightly")] + fn extend_reserve(&mut self, additional: usize) { + // Keys may be already present or show multiple times in the iterator. + // Reserve the entire hint lower bound if the map is empty. + // Otherwise reserve half the hint (rounded up), so the map + // will only resize twice in the worst case. + let reserve = if self.is_empty() { + additional + } else { + (additional + 1) / 2 + }; + self.reserve(reserve); + } +} + +impl<'a, K, V, S> Extend<(&'a K, &'a V)> for HashMap<K, V, S> +where + K: Eq + Hash + Copy, + V: Copy, + S: BuildHasher, +{ + #[cfg_attr(feature = "inline-more", inline)] + fn extend<T: IntoIterator<Item = (&'a K, &'a V)>>(&mut self, iter: T) { + self.extend(iter.into_iter().map(|(&key, &value)| (key, value))); + } + + #[inline] + #[cfg(feature = "nightly")] + fn extend_one(&mut self, (k, v): (&'a K, &'a V)) { + self.insert(*k, *v); + } + + #[inline] + #[cfg(feature = "nightly")] + fn extend_reserve(&mut self, additional: usize) { + Extend::<(K, V)>::extend_reserve(self, additional); + } +} + +#[allow(dead_code)] +fn assert_covariance() { + fn map_key<'new>(v: HashMap<&'static str, u8>) -> HashMap<&'new str, u8> { + v + } + fn map_val<'new>(v: HashMap<u8, &'static str>) -> HashMap<u8, &'new str> { + v + } + fn iter_key<'a, 'new>(v: Iter<'a, &'static str, u8>) -> Iter<'a, &'new str, u8> { + v + } + fn iter_val<'a, 'new>(v: Iter<'a, u8, &'static str>) -> Iter<'a, u8, &'new str> { + v + } + fn into_iter_key<'new>(v: IntoIter<&'static str, u8>) -> IntoIter<&'new str, u8> { + v + } + fn into_iter_val<'new>(v: IntoIter<u8, &'static str>) -> IntoIter<u8, &'new str> { + v + } + fn keys_key<'a, 'new>(v: Keys<'a, &'static str, u8>) -> Keys<'a, &'new str, u8> { + v + } + fn keys_val<'a, 'new>(v: Keys<'a, u8, &'static str>) -> Keys<'a, u8, &'new str> { + v + } + fn values_key<'a, 'new>(v: Values<'a, &'static str, u8>) -> Values<'a, &'new str, u8> { + v + } + fn values_val<'a, 'new>(v: Values<'a, u8, &'static str>) -> Values<'a, u8, &'new str> { + v + } + fn drain<'new>( + d: Drain<'static, &'static str, &'static str>, + ) -> Drain<'new, &'new str, &'new str> { + d + } +} + +#[cfg(test)] +mod test_map { + use super::DefaultHashBuilder; + use super::Entry::{Occupied, Vacant}; + use super::{HashMap, RawEntryMut}; + use crate::TryReserveError::*; + use rand::{rngs::SmallRng, Rng, SeedableRng}; + use std::cell::RefCell; + use std::usize; + use std::vec::Vec; + + #[test] + fn test_zero_capacities() { + type HM = HashMap<i32, i32>; + + let m = HM::new(); + assert_eq!(m.capacity(), 0); + + let m = HM::default(); + assert_eq!(m.capacity(), 0); + + let m = HM::with_hasher(DefaultHashBuilder::default()); + assert_eq!(m.capacity(), 0); + + let m = HM::with_capacity(0); + assert_eq!(m.capacity(), 0); + + let m = HM::with_capacity_and_hasher(0, DefaultHashBuilder::default()); + assert_eq!(m.capacity(), 0); + + let mut m = HM::new(); + m.insert(1, 1); + m.insert(2, 2); + m.remove(&1); + m.remove(&2); + m.shrink_to_fit(); + assert_eq!(m.capacity(), 0); + + let mut m = HM::new(); + m.reserve(0); + assert_eq!(m.capacity(), 0); + } + + #[test] + fn test_create_capacity_zero() { + let mut m = HashMap::with_capacity(0); + + assert!(m.insert(1, 1).is_none()); + + assert!(m.contains_key(&1)); + assert!(!m.contains_key(&0)); + } + + #[test] + fn test_insert() { + let mut m = HashMap::new(); + assert_eq!(m.len(), 0); + assert!(m.insert(1, 2).is_none()); + assert_eq!(m.len(), 1); + assert!(m.insert(2, 4).is_none()); + assert_eq!(m.len(), 2); + assert_eq!(*m.get(&1).unwrap(), 2); + assert_eq!(*m.get(&2).unwrap(), 4); + } + + #[test] + fn test_clone() { + let mut m = HashMap::new(); + assert_eq!(m.len(), 0); + assert!(m.insert(1, 2).is_none()); + assert_eq!(m.len(), 1); + assert!(m.insert(2, 4).is_none()); + assert_eq!(m.len(), 2); + let m2 = m.clone(); + assert_eq!(*m2.get(&1).unwrap(), 2); + assert_eq!(*m2.get(&2).unwrap(), 4); + assert_eq!(m2.len(), 2); + } + + #[test] + fn test_clone_from() { + let mut m = HashMap::new(); + let mut m2 = HashMap::new(); + assert_eq!(m.len(), 0); + assert!(m.insert(1, 2).is_none()); + assert_eq!(m.len(), 1); + assert!(m.insert(2, 4).is_none()); + assert_eq!(m.len(), 2); + m2.clone_from(&m); + assert_eq!(*m2.get(&1).unwrap(), 2); + assert_eq!(*m2.get(&2).unwrap(), 4); + assert_eq!(m2.len(), 2); + } + + thread_local! { static DROP_VECTOR: RefCell<Vec<i32>> = RefCell::new(Vec::new()) } + + #[derive(Hash, PartialEq, Eq)] + struct Droppable { + k: usize, + } + + impl Droppable { + fn new(k: usize) -> Droppable { + DROP_VECTOR.with(|slot| { + slot.borrow_mut()[k] += 1; + }); + + Droppable { k } + } + } + + impl Drop for Droppable { + fn drop(&mut self) { + DROP_VECTOR.with(|slot| { + slot.borrow_mut()[self.k] -= 1; + }); + } + } + + impl Clone for Droppable { + fn clone(&self) -> Self { + Droppable::new(self.k) + } + } + + #[test] + fn test_drops() { + DROP_VECTOR.with(|slot| { + *slot.borrow_mut() = vec![0; 200]; + }); + + { + let mut m = HashMap::new(); + + DROP_VECTOR.with(|v| { + for i in 0..200 { + assert_eq!(v.borrow()[i], 0); + } + }); + + for i in 0..100 { + let d1 = Droppable::new(i); + let d2 = Droppable::new(i + 100); + m.insert(d1, d2); + } + + DROP_VECTOR.with(|v| { + for i in 0..200 { + assert_eq!(v.borrow()[i], 1); + } + }); + + for i in 0..50 { + let k = Droppable::new(i); + let v = m.remove(&k); + + assert!(v.is_some()); + + DROP_VECTOR.with(|v| { + assert_eq!(v.borrow()[i], 1); + assert_eq!(v.borrow()[i + 100], 1); + }); + } + + DROP_VECTOR.with(|v| { + for i in 0..50 { + assert_eq!(v.borrow()[i], 0); + assert_eq!(v.borrow()[i + 100], 0); + } + + for i in 50..100 { + assert_eq!(v.borrow()[i], 1); + assert_eq!(v.borrow()[i + 100], 1); + } + }); + } + + DROP_VECTOR.with(|v| { + for i in 0..200 { + assert_eq!(v.borrow()[i], 0); + } + }); + } + + #[test] + fn test_into_iter_drops() { + DROP_VECTOR.with(|v| { + *v.borrow_mut() = vec![0; 200]; + }); + + let hm = { + let mut hm = HashMap::new(); + + DROP_VECTOR.with(|v| { + for i in 0..200 { + assert_eq!(v.borrow()[i], 0); + } + }); + + for i in 0..100 { + let d1 = Droppable::new(i); + let d2 = Droppable::new(i + 100); + hm.insert(d1, d2); + } + + DROP_VECTOR.with(|v| { + for i in 0..200 { + assert_eq!(v.borrow()[i], 1); + } + }); + + hm + }; + + // By the way, ensure that cloning doesn't screw up the dropping. + drop(hm.clone()); + + { + let mut half = hm.into_iter().take(50); + + DROP_VECTOR.with(|v| { + for i in 0..200 { + assert_eq!(v.borrow()[i], 1); + } + }); + + for _ in half.by_ref() {} + + DROP_VECTOR.with(|v| { + let nk = (0..100).filter(|&i| v.borrow()[i] == 1).count(); + + let nv = (0..100).filter(|&i| v.borrow()[i + 100] == 1).count(); + + assert_eq!(nk, 50); + assert_eq!(nv, 50); + }); + }; + + DROP_VECTOR.with(|v| { + for i in 0..200 { + assert_eq!(v.borrow()[i], 0); + } + }); + } + + #[test] + fn test_empty_remove() { + let mut m: HashMap<i32, bool> = HashMap::new(); + assert_eq!(m.remove(&0), None); + } + + #[test] + fn test_empty_entry() { + let mut m: HashMap<i32, bool> = HashMap::new(); + match m.entry(0) { + Occupied(_) => panic!(), + Vacant(_) => {} + } + assert!(*m.entry(0).or_insert(true)); + assert_eq!(m.len(), 1); + } + + #[test] + fn test_empty_iter() { + let mut m: HashMap<i32, bool> = HashMap::new(); + assert_eq!(m.drain().next(), None); + assert_eq!(m.keys().next(), None); + assert_eq!(m.values().next(), None); + assert_eq!(m.values_mut().next(), None); + assert_eq!(m.iter().next(), None); + assert_eq!(m.iter_mut().next(), None); + assert_eq!(m.len(), 0); + assert!(m.is_empty()); + assert_eq!(m.into_iter().next(), None); + } + + #[test] + #[cfg_attr(miri, ignore)] // FIXME: takes too long + fn test_lots_of_insertions() { + let mut m = HashMap::new(); + + // Try this a few times to make sure we never screw up the hashmap's + // internal state. + for _ in 0..10 { + assert!(m.is_empty()); + + for i in 1..1001 { + assert!(m.insert(i, i).is_none()); + + for j in 1..=i { + let r = m.get(&j); + assert_eq!(r, Some(&j)); + } + + for j in i + 1..1001 { + let r = m.get(&j); + assert_eq!(r, None); + } + } + + for i in 1001..2001 { + assert!(!m.contains_key(&i)); + } + + // remove forwards + for i in 1..1001 { + assert!(m.remove(&i).is_some()); + + for j in 1..=i { + assert!(!m.contains_key(&j)); + } + + for j in i + 1..1001 { + assert!(m.contains_key(&j)); + } + } + + for i in 1..1001 { + assert!(!m.contains_key(&i)); + } + + for i in 1..1001 { + assert!(m.insert(i, i).is_none()); + } + + // remove backwards + for i in (1..1001).rev() { + assert!(m.remove(&i).is_some()); + + for j in i..1001 { + assert!(!m.contains_key(&j)); + } + + for j in 1..i { + assert!(m.contains_key(&j)); + } + } + } + } + + #[test] + fn test_find_mut() { + let mut m = HashMap::new(); + assert!(m.insert(1, 12).is_none()); + assert!(m.insert(2, 8).is_none()); + assert!(m.insert(5, 14).is_none()); + let new = 100; + match m.get_mut(&5) { + None => panic!(), + Some(x) => *x = new, + } + assert_eq!(m.get(&5), Some(&new)); + } + + #[test] + fn test_insert_overwrite() { + let mut m = HashMap::new(); + assert!(m.insert(1, 2).is_none()); + assert_eq!(*m.get(&1).unwrap(), 2); + assert!(!m.insert(1, 3).is_none()); + assert_eq!(*m.get(&1).unwrap(), 3); + } + + #[test] + fn test_insert_conflicts() { + let mut m = HashMap::with_capacity(4); + assert!(m.insert(1, 2).is_none()); + assert!(m.insert(5, 3).is_none()); + assert!(m.insert(9, 4).is_none()); + assert_eq!(*m.get(&9).unwrap(), 4); + assert_eq!(*m.get(&5).unwrap(), 3); + assert_eq!(*m.get(&1).unwrap(), 2); + } + + #[test] + fn test_conflict_remove() { + let mut m = HashMap::with_capacity(4); + assert!(m.insert(1, 2).is_none()); + assert_eq!(*m.get(&1).unwrap(), 2); + assert!(m.insert(5, 3).is_none()); + assert_eq!(*m.get(&1).unwrap(), 2); + assert_eq!(*m.get(&5).unwrap(), 3); + assert!(m.insert(9, 4).is_none()); + assert_eq!(*m.get(&1).unwrap(), 2); + assert_eq!(*m.get(&5).unwrap(), 3); + assert_eq!(*m.get(&9).unwrap(), 4); + assert!(m.remove(&1).is_some()); + assert_eq!(*m.get(&9).unwrap(), 4); + assert_eq!(*m.get(&5).unwrap(), 3); + } + + #[test] + fn test_is_empty() { + let mut m = HashMap::with_capacity(4); + assert!(m.insert(1, 2).is_none()); + assert!(!m.is_empty()); + assert!(m.remove(&1).is_some()); + assert!(m.is_empty()); + } + + #[test] + fn test_remove() { + let mut m = HashMap::new(); + m.insert(1, 2); + assert_eq!(m.remove(&1), Some(2)); + assert_eq!(m.remove(&1), None); + } + + #[test] + fn test_remove_entry() { + let mut m = HashMap::new(); + m.insert(1, 2); + assert_eq!(m.remove_entry(&1), Some((1, 2))); + assert_eq!(m.remove(&1), None); + } + + #[test] + fn test_iterate() { + let mut m = HashMap::with_capacity(4); + for i in 0..32 { + assert!(m.insert(i, i * 2).is_none()); + } + assert_eq!(m.len(), 32); + + let mut observed: u32 = 0; + + for (k, v) in &m { + assert_eq!(*v, *k * 2); + observed |= 1 << *k; + } + assert_eq!(observed, 0xFFFF_FFFF); + } + + #[test] + fn test_keys() { + let vec = vec![(1, 'a'), (2, 'b'), (3, 'c')]; + let map: HashMap<_, _> = vec.into_iter().collect(); + let keys: Vec<_> = map.keys().cloned().collect(); + assert_eq!(keys.len(), 3); + assert!(keys.contains(&1)); + assert!(keys.contains(&2)); + assert!(keys.contains(&3)); + } + + #[test] + fn test_values() { + let vec = vec![(1, 'a'), (2, 'b'), (3, 'c')]; + let map: HashMap<_, _> = vec.into_iter().collect(); + let values: Vec<_> = map.values().cloned().collect(); + assert_eq!(values.len(), 3); + assert!(values.contains(&'a')); + assert!(values.contains(&'b')); + assert!(values.contains(&'c')); + } + + #[test] + fn test_values_mut() { + let vec = vec![(1, 1), (2, 2), (3, 3)]; + let mut map: HashMap<_, _> = vec.into_iter().collect(); + for value in map.values_mut() { + *value = (*value) * 2 + } + let values: Vec<_> = map.values().cloned().collect(); + assert_eq!(values.len(), 3); + assert!(values.contains(&2)); + assert!(values.contains(&4)); + assert!(values.contains(&6)); + } + + #[test] + fn test_find() { + let mut m = HashMap::new(); + assert!(m.get(&1).is_none()); + m.insert(1, 2); + match m.get(&1) { + None => panic!(), + Some(v) => assert_eq!(*v, 2), + } + } + + #[test] + fn test_eq() { + let mut m1 = HashMap::new(); + m1.insert(1, 2); + m1.insert(2, 3); + m1.insert(3, 4); + + let mut m2 = HashMap::new(); + m2.insert(1, 2); + m2.insert(2, 3); + + assert!(m1 != m2); + + m2.insert(3, 4); + + assert_eq!(m1, m2); + } + + #[test] + fn test_show() { + let mut map = HashMap::new(); + let empty: HashMap<i32, i32> = HashMap::new(); + + map.insert(1, 2); + map.insert(3, 4); + + let map_str = format!("{:?}", map); + + assert!(map_str == "{1: 2, 3: 4}" || map_str == "{3: 4, 1: 2}"); + assert_eq!(format!("{:?}", empty), "{}"); + } + + #[test] + fn test_expand() { + let mut m = HashMap::new(); + + assert_eq!(m.len(), 0); + assert!(m.is_empty()); + + let mut i = 0; + let old_raw_cap = m.raw_capacity(); + while old_raw_cap == m.raw_capacity() { + m.insert(i, i); + i += 1; + } + + assert_eq!(m.len(), i); + assert!(!m.is_empty()); + } + + #[test] + fn test_behavior_resize_policy() { + let mut m = HashMap::new(); + + assert_eq!(m.len(), 0); + assert_eq!(m.raw_capacity(), 1); + assert!(m.is_empty()); + + m.insert(0, 0); + m.remove(&0); + assert!(m.is_empty()); + let initial_raw_cap = m.raw_capacity(); + m.reserve(initial_raw_cap); + let raw_cap = m.raw_capacity(); + + assert_eq!(raw_cap, initial_raw_cap * 2); + + let mut i = 0; + for _ in 0..raw_cap * 3 / 4 { + m.insert(i, i); + i += 1; + } + // three quarters full + + assert_eq!(m.len(), i); + assert_eq!(m.raw_capacity(), raw_cap); + + for _ in 0..raw_cap / 4 { + m.insert(i, i); + i += 1; + } + // half full + + let new_raw_cap = m.raw_capacity(); + assert_eq!(new_raw_cap, raw_cap * 2); + + for _ in 0..raw_cap / 2 - 1 { + i -= 1; + m.remove(&i); + assert_eq!(m.raw_capacity(), new_raw_cap); + } + // A little more than one quarter full. + m.shrink_to_fit(); + assert_eq!(m.raw_capacity(), raw_cap); + // again, a little more than half full + for _ in 0..raw_cap / 2 { + i -= 1; + m.remove(&i); + } + m.shrink_to_fit(); + + assert_eq!(m.len(), i); + assert!(!m.is_empty()); + assert_eq!(m.raw_capacity(), initial_raw_cap); + } + + #[test] + fn test_reserve_shrink_to_fit() { + let mut m = HashMap::new(); + m.insert(0, 0); + m.remove(&0); + assert!(m.capacity() >= m.len()); + for i in 0..128 { + m.insert(i, i); + } + m.reserve(256); + + let usable_cap = m.capacity(); + for i in 128..(128 + 256) { + m.insert(i, i); + assert_eq!(m.capacity(), usable_cap); + } + + for i in 100..(128 + 256) { + assert_eq!(m.remove(&i), Some(i)); + } + m.shrink_to_fit(); + + assert_eq!(m.len(), 100); + assert!(!m.is_empty()); + assert!(m.capacity() >= m.len()); + + for i in 0..100 { + assert_eq!(m.remove(&i), Some(i)); + } + m.shrink_to_fit(); + m.insert(0, 0); + + assert_eq!(m.len(), 1); + assert!(m.capacity() >= m.len()); + assert_eq!(m.remove(&0), Some(0)); + } + + #[test] + fn test_from_iter() { + let xs = [(1, 1), (2, 2), (2, 2), (3, 3), (4, 4), (5, 5), (6, 6)]; + + let map: HashMap<_, _> = xs.iter().cloned().collect(); + + for &(k, v) in &xs { + assert_eq!(map.get(&k), Some(&v)); + } + + assert_eq!(map.iter().len(), xs.len() - 1); + } + + #[test] + fn test_size_hint() { + let xs = [(1, 1), (2, 2), (3, 3), (4, 4), (5, 5), (6, 6)]; + + let map: HashMap<_, _> = xs.iter().cloned().collect(); + + let mut iter = map.iter(); + + for _ in iter.by_ref().take(3) {} + + assert_eq!(iter.size_hint(), (3, Some(3))); + } + + #[test] + fn test_iter_len() { + let xs = [(1, 1), (2, 2), (3, 3), (4, 4), (5, 5), (6, 6)]; + + let map: HashMap<_, _> = xs.iter().cloned().collect(); + + let mut iter = map.iter(); + + for _ in iter.by_ref().take(3) {} + + assert_eq!(iter.len(), 3); + } + + #[test] + fn test_mut_size_hint() { + let xs = [(1, 1), (2, 2), (3, 3), (4, 4), (5, 5), (6, 6)]; + + let mut map: HashMap<_, _> = xs.iter().cloned().collect(); + + let mut iter = map.iter_mut(); + + for _ in iter.by_ref().take(3) {} + + assert_eq!(iter.size_hint(), (3, Some(3))); + } + + #[test] + fn test_iter_mut_len() { + let xs = [(1, 1), (2, 2), (3, 3), (4, 4), (5, 5), (6, 6)]; + + let mut map: HashMap<_, _> = xs.iter().cloned().collect(); + + let mut iter = map.iter_mut(); + + for _ in iter.by_ref().take(3) {} + + assert_eq!(iter.len(), 3); + } + + #[test] + fn test_index() { + let mut map = HashMap::new(); + + map.insert(1, 2); + map.insert(2, 1); + map.insert(3, 4); + + assert_eq!(map[&2], 1); + } + + #[test] + #[should_panic] + fn test_index_nonexistent() { + let mut map = HashMap::new(); + + map.insert(1, 2); + map.insert(2, 1); + map.insert(3, 4); + + map[&4]; + } + + #[test] + fn test_entry() { + let xs = [(1, 10), (2, 20), (3, 30), (4, 40), (5, 50), (6, 60)]; + + let mut map: HashMap<_, _> = xs.iter().cloned().collect(); + + // Existing key (insert) + match map.entry(1) { + Vacant(_) => unreachable!(), + Occupied(mut view) => { + assert_eq!(view.get(), &10); + assert_eq!(view.insert(100), 10); + } + } + assert_eq!(map.get(&1).unwrap(), &100); + assert_eq!(map.len(), 6); + + // Existing key (update) + match map.entry(2) { + Vacant(_) => unreachable!(), + Occupied(mut view) => { + let v = view.get_mut(); + let new_v = (*v) * 10; + *v = new_v; + } + } + assert_eq!(map.get(&2).unwrap(), &200); + assert_eq!(map.len(), 6); + + // Existing key (take) + match map.entry(3) { + Vacant(_) => unreachable!(), + Occupied(view) => { + assert_eq!(view.remove(), 30); + } + } + assert_eq!(map.get(&3), None); + assert_eq!(map.len(), 5); + + // Inexistent key (insert) + match map.entry(10) { + Occupied(_) => unreachable!(), + Vacant(view) => { + assert_eq!(*view.insert(1000), 1000); + } + } + assert_eq!(map.get(&10).unwrap(), &1000); + assert_eq!(map.len(), 6); + } + + #[test] + fn test_entry_take_doesnt_corrupt() { + #![allow(deprecated)] //rand + // Test for #19292 + fn check(m: &HashMap<i32, ()>) { + for k in m.keys() { + assert!(m.contains_key(k), "{} is in keys() but not in the map?", k); + } + } + + let mut m = HashMap::new(); + + let mut rng = { + let seed = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]; + SmallRng::from_seed(seed) + }; + + // Populate the map with some items. + for _ in 0..50 { + let x = rng.gen_range(-10, 10); + m.insert(x, ()); + } + + for _ in 0..1000 { + let x = rng.gen_range(-10, 10); + match m.entry(x) { + Vacant(_) => {} + Occupied(e) => { + e.remove(); + } + } + + check(&m); + } + } + + #[test] + fn test_extend_ref() { + let mut a = HashMap::new(); + a.insert(1, "one"); + let mut b = HashMap::new(); + b.insert(2, "two"); + b.insert(3, "three"); + + a.extend(&b); + + assert_eq!(a.len(), 3); + assert_eq!(a[&1], "one"); + assert_eq!(a[&2], "two"); + assert_eq!(a[&3], "three"); + } + + #[test] + fn test_capacity_not_less_than_len() { + let mut a = HashMap::new(); + let mut item = 0; + + for _ in 0..116 { + a.insert(item, 0); + item += 1; + } + + assert!(a.capacity() > a.len()); + + let free = a.capacity() - a.len(); + for _ in 0..free { + a.insert(item, 0); + item += 1; + } + + assert_eq!(a.len(), a.capacity()); + + // Insert at capacity should cause allocation. + a.insert(item, 0); + assert!(a.capacity() > a.len()); + } + + #[test] + fn test_occupied_entry_key() { + let mut a = HashMap::new(); + let key = "hello there"; + let value = "value goes here"; + assert!(a.is_empty()); + a.insert(key.clone(), value.clone()); + assert_eq!(a.len(), 1); + assert_eq!(a[key], value); + + match a.entry(key.clone()) { + Vacant(_) => panic!(), + Occupied(e) => assert_eq!(key, *e.key()), + } + assert_eq!(a.len(), 1); + assert_eq!(a[key], value); + } + + #[test] + fn test_vacant_entry_key() { + let mut a = HashMap::new(); + let key = "hello there"; + let value = "value goes here"; + + assert!(a.is_empty()); + match a.entry(key.clone()) { + Occupied(_) => panic!(), + Vacant(e) => { + assert_eq!(key, *e.key()); + e.insert(value.clone()); + } + } + assert_eq!(a.len(), 1); + assert_eq!(a[key], value); + } + + #[test] + fn test_occupied_entry_replace_entry_with() { + let mut a = HashMap::new(); + + let key = "a key"; + let value = "an initial value"; + let new_value = "a new value"; + + let entry = a.entry(key).insert(value).replace_entry_with(|k, v| { + assert_eq!(k, &key); + assert_eq!(v, value); + Some(new_value) + }); + + match entry { + Occupied(e) => { + assert_eq!(e.key(), &key); + assert_eq!(e.get(), &new_value); + } + Vacant(_) => panic!(), + } + + assert_eq!(a[key], new_value); + assert_eq!(a.len(), 1); + + let entry = match a.entry(key) { + Occupied(e) => e.replace_entry_with(|k, v| { + assert_eq!(k, &key); + assert_eq!(v, new_value); + None + }), + Vacant(_) => panic!(), + }; + + match entry { + Vacant(e) => assert_eq!(e.key(), &key), + Occupied(_) => panic!(), + } + + assert!(!a.contains_key(key)); + assert_eq!(a.len(), 0); + } + + #[test] + fn test_entry_and_replace_entry_with() { + let mut a = HashMap::new(); + + let key = "a key"; + let value = "an initial value"; + let new_value = "a new value"; + + let entry = a.entry(key).and_replace_entry_with(|_, _| panic!()); + + match entry { + Vacant(e) => assert_eq!(e.key(), &key), + Occupied(_) => panic!(), + } + + a.insert(key, value); + + let entry = a.entry(key).and_replace_entry_with(|k, v| { + assert_eq!(k, &key); + assert_eq!(v, value); + Some(new_value) + }); + + match entry { + Occupied(e) => { + assert_eq!(e.key(), &key); + assert_eq!(e.get(), &new_value); + } + Vacant(_) => panic!(), + } + + assert_eq!(a[key], new_value); + assert_eq!(a.len(), 1); + + let entry = a.entry(key).and_replace_entry_with(|k, v| { + assert_eq!(k, &key); + assert_eq!(v, new_value); + None + }); + + match entry { + Vacant(e) => assert_eq!(e.key(), &key), + Occupied(_) => panic!(), + } + + assert!(!a.contains_key(key)); + assert_eq!(a.len(), 0); + } + + #[test] + fn test_raw_occupied_entry_replace_entry_with() { + let mut a = HashMap::new(); + + let key = "a key"; + let value = "an initial value"; + let new_value = "a new value"; + + let entry = a + .raw_entry_mut() + .from_key(&key) + .insert(key, value) + .replace_entry_with(|k, v| { + assert_eq!(k, &key); + assert_eq!(v, value); + Some(new_value) + }); + + match entry { + RawEntryMut::Occupied(e) => { + assert_eq!(e.key(), &key); + assert_eq!(e.get(), &new_value); + } + RawEntryMut::Vacant(_) => panic!(), + } + + assert_eq!(a[key], new_value); + assert_eq!(a.len(), 1); + + let entry = match a.raw_entry_mut().from_key(&key) { + RawEntryMut::Occupied(e) => e.replace_entry_with(|k, v| { + assert_eq!(k, &key); + assert_eq!(v, new_value); + None + }), + RawEntryMut::Vacant(_) => panic!(), + }; + + match entry { + RawEntryMut::Vacant(_) => {} + RawEntryMut::Occupied(_) => panic!(), + } + + assert!(!a.contains_key(key)); + assert_eq!(a.len(), 0); + } + + #[test] + fn test_raw_entry_and_replace_entry_with() { + let mut a = HashMap::new(); + + let key = "a key"; + let value = "an initial value"; + let new_value = "a new value"; + + let entry = a + .raw_entry_mut() + .from_key(&key) + .and_replace_entry_with(|_, _| panic!()); + + match entry { + RawEntryMut::Vacant(_) => {} + RawEntryMut::Occupied(_) => panic!(), + } + + a.insert(key, value); + + let entry = a + .raw_entry_mut() + .from_key(&key) + .and_replace_entry_with(|k, v| { + assert_eq!(k, &key); + assert_eq!(v, value); + Some(new_value) + }); + + match entry { + RawEntryMut::Occupied(e) => { + assert_eq!(e.key(), &key); + assert_eq!(e.get(), &new_value); + } + RawEntryMut::Vacant(_) => panic!(), + } + + assert_eq!(a[key], new_value); + assert_eq!(a.len(), 1); + + let entry = a + .raw_entry_mut() + .from_key(&key) + .and_replace_entry_with(|k, v| { + assert_eq!(k, &key); + assert_eq!(v, new_value); + None + }); + + match entry { + RawEntryMut::Vacant(_) => {} + RawEntryMut::Occupied(_) => panic!(), + } + + assert!(!a.contains_key(key)); + assert_eq!(a.len(), 0); + } + + #[test] + fn test_replace_entry_with_doesnt_corrupt() { + #![allow(deprecated)] //rand + // Test for #19292 + fn check(m: &HashMap<i32, ()>) { + for k in m.keys() { + assert!(m.contains_key(k), "{} is in keys() but not in the map?", k); + } + } + + let mut m = HashMap::new(); + + let mut rng = { + let seed = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]; + SmallRng::from_seed(seed) + }; + + // Populate the map with some items. + for _ in 0..50 { + let x = rng.gen_range(-10, 10); + m.insert(x, ()); + } + + for _ in 0..1000 { + let x = rng.gen_range(-10, 10); + m.entry(x).and_replace_entry_with(|_, _| None); + check(&m); + } + } + + #[test] + fn test_retain() { + let mut map: HashMap<i32, i32> = (0..100).map(|x| (x, x * 10)).collect(); + + map.retain(|&k, _| k % 2 == 0); + assert_eq!(map.len(), 50); + assert_eq!(map[&2], 20); + assert_eq!(map[&4], 40); + assert_eq!(map[&6], 60); + } + + #[test] + fn test_drain_filter() { + { + let mut map: HashMap<i32, i32> = (0..8).map(|x| (x, x * 10)).collect(); + let drained = map.drain_filter(|&k, _| k % 2 == 0); + let mut out = drained.collect::<Vec<_>>(); + out.sort_unstable(); + assert_eq!(vec![(0, 0), (2, 20), (4, 40), (6, 60)], out); + assert_eq!(map.len(), 4); + } + { + let mut map: HashMap<i32, i32> = (0..8).map(|x| (x, x * 10)).collect(); + drop(map.drain_filter(|&k, _| k % 2 == 0)); + assert_eq!(map.len(), 4); + } + } + + #[test] + #[cfg_attr(miri, ignore)] // FIXME: no OOM signalling (https://github.com/rust-lang/miri/issues/613) + fn test_try_reserve() { + let mut empty_bytes: HashMap<u8, u8> = HashMap::new(); + + const MAX_USIZE: usize = usize::MAX; + + if let Err(CapacityOverflow) = empty_bytes.try_reserve(MAX_USIZE) { + } else { + panic!("usize::MAX should trigger an overflow!"); + } + + if let Err(AllocError { .. }) = empty_bytes.try_reserve(MAX_USIZE / 8) { + } else { + // This may succeed if there is enough free memory. Attempt to + // allocate a second hashmap to ensure the allocation will fail. + let mut empty_bytes2: HashMap<u8, u8> = HashMap::new(); + if let Err(AllocError { .. }) = empty_bytes2.try_reserve(MAX_USIZE / 8) { + } else { + panic!("usize::MAX / 8 should trigger an OOM!"); + } + } + } + + #[test] + fn test_raw_entry() { + use super::RawEntryMut::{Occupied, Vacant}; + + let xs = [(1i32, 10i32), (2, 20), (3, 30), (4, 40), (5, 50), (6, 60)]; + + let mut map: HashMap<_, _> = xs.iter().cloned().collect(); + + let compute_hash = |map: &HashMap<i32, i32>, k: i32| -> u64 { + use core::hash::{BuildHasher, Hash, Hasher}; + + let mut hasher = map.hasher().build_hasher(); + k.hash(&mut hasher); + hasher.finish() + }; + + // Existing key (insert) + match map.raw_entry_mut().from_key(&1) { + Vacant(_) => unreachable!(), + Occupied(mut view) => { + assert_eq!(view.get(), &10); + assert_eq!(view.insert(100), 10); + } + } + let hash1 = compute_hash(&map, 1); + assert_eq!(map.raw_entry().from_key(&1).unwrap(), (&1, &100)); + assert_eq!( + map.raw_entry().from_hash(hash1, |k| *k == 1).unwrap(), + (&1, &100) + ); + assert_eq!( + map.raw_entry().from_key_hashed_nocheck(hash1, &1).unwrap(), + (&1, &100) + ); + assert_eq!(map.len(), 6); + + // Existing key (update) + match map.raw_entry_mut().from_key(&2) { + Vacant(_) => unreachable!(), + Occupied(mut view) => { + let v = view.get_mut(); + let new_v = (*v) * 10; + *v = new_v; + } + } + let hash2 = compute_hash(&map, 2); + assert_eq!(map.raw_entry().from_key(&2).unwrap(), (&2, &200)); + assert_eq!( + map.raw_entry().from_hash(hash2, |k| *k == 2).unwrap(), + (&2, &200) + ); + assert_eq!( + map.raw_entry().from_key_hashed_nocheck(hash2, &2).unwrap(), + (&2, &200) + ); + assert_eq!(map.len(), 6); + + // Existing key (take) + let hash3 = compute_hash(&map, 3); + match map.raw_entry_mut().from_key_hashed_nocheck(hash3, &3) { + Vacant(_) => unreachable!(), + Occupied(view) => { + assert_eq!(view.remove_entry(), (3, 30)); + } + } + assert_eq!(map.raw_entry().from_key(&3), None); + assert_eq!(map.raw_entry().from_hash(hash3, |k| *k == 3), None); + assert_eq!(map.raw_entry().from_key_hashed_nocheck(hash3, &3), None); + assert_eq!(map.len(), 5); + + // Nonexistent key (insert) + match map.raw_entry_mut().from_key(&10) { + Occupied(_) => unreachable!(), + Vacant(view) => { + assert_eq!(view.insert(10, 1000), (&mut 10, &mut 1000)); + } + } + assert_eq!(map.raw_entry().from_key(&10).unwrap(), (&10, &1000)); + assert_eq!(map.len(), 6); + + // Ensure all lookup methods produce equivalent results. + for k in 0..12 { + let hash = compute_hash(&map, k); + let v = map.get(&k).cloned(); + let kv = v.as_ref().map(|v| (&k, v)); + + assert_eq!(map.raw_entry().from_key(&k), kv); + assert_eq!(map.raw_entry().from_hash(hash, |q| *q == k), kv); + assert_eq!(map.raw_entry().from_key_hashed_nocheck(hash, &k), kv); + + match map.raw_entry_mut().from_key(&k) { + Occupied(mut o) => assert_eq!(Some(o.get_key_value()), kv), + Vacant(_) => assert_eq!(v, None), + } + match map.raw_entry_mut().from_key_hashed_nocheck(hash, &k) { + Occupied(mut o) => assert_eq!(Some(o.get_key_value()), kv), + Vacant(_) => assert_eq!(v, None), + } + match map.raw_entry_mut().from_hash(hash, |q| *q == k) { + Occupied(mut o) => assert_eq!(Some(o.get_key_value()), kv), + Vacant(_) => assert_eq!(v, None), + } + } + } + + #[test] + fn test_key_without_hash_impl() { + #[derive(Debug)] + struct IntWrapper(u64); + + let mut m: HashMap<IntWrapper, (), ()> = HashMap::default(); + { + assert!(m.raw_entry().from_hash(0, |k| k.0 == 0).is_none()); + } + { + let vacant_entry = match m.raw_entry_mut().from_hash(0, |k| k.0 == 0) { + RawEntryMut::Occupied(..) => panic!("Found entry for key 0"), + RawEntryMut::Vacant(e) => e, + }; + vacant_entry.insert_with_hasher(0, IntWrapper(0), (), |k| k.0); + } + { + assert!(m.raw_entry().from_hash(0, |k| k.0 == 0).is_some()); + assert!(m.raw_entry().from_hash(1, |k| k.0 == 1).is_none()); + assert!(m.raw_entry().from_hash(2, |k| k.0 == 2).is_none()); + } + { + let vacant_entry = match m.raw_entry_mut().from_hash(1, |k| k.0 == 1) { + RawEntryMut::Occupied(..) => panic!("Found entry for key 1"), + RawEntryMut::Vacant(e) => e, + }; + vacant_entry.insert_with_hasher(1, IntWrapper(1), (), |k| k.0); + } + { + assert!(m.raw_entry().from_hash(0, |k| k.0 == 0).is_some()); + assert!(m.raw_entry().from_hash(1, |k| k.0 == 1).is_some()); + assert!(m.raw_entry().from_hash(2, |k| k.0 == 2).is_none()); + } + { + let occupied_entry = match m.raw_entry_mut().from_hash(0, |k| k.0 == 0) { + RawEntryMut::Occupied(e) => e, + RawEntryMut::Vacant(..) => panic!("Couldn't find entry for key 0"), + }; + occupied_entry.remove(); + } + assert!(m.raw_entry().from_hash(0, |k| k.0 == 0).is_none()); + assert!(m.raw_entry().from_hash(1, |k| k.0 == 1).is_some()); + assert!(m.raw_entry().from_hash(2, |k| k.0 == 2).is_none()); + } + + #[test] + #[cfg(feature = "raw")] + fn test_into_iter_refresh() { + use core::hash::{BuildHasher, Hash, Hasher}; + + #[cfg(miri)] + const N: usize = 32; + #[cfg(not(miri))] + const N: usize = 128; + + let mut rng = rand::thread_rng(); + for n in 0..N { + let mut m = HashMap::new(); + for i in 0..n { + assert!(m.insert(i, 2 * i).is_none()); + } + let hasher = m.hasher().clone(); + + let mut it = unsafe { m.table.iter() }; + assert_eq!(it.len(), n); + + let mut i = 0; + let mut left = n; + let mut removed = Vec::new(); + loop { + // occasionally remove some elements + if i < n && rng.gen_bool(0.1) { + let mut hsh = hasher.build_hasher(); + i.hash(&mut hsh); + let hash = hsh.finish(); + + unsafe { + let e = m.table.find(hash, |q| q.0.eq(&i)); + if let Some(e) = e { + it.reflect_remove(&e); + let t = m.table.remove(e); + removed.push(t); + left -= 1; + } else { + assert!(removed.contains(&(i, 2 * i)), "{} not in {:?}", i, removed); + let e = m + .table + .insert(hash, (i, 2 * i), |x| super::make_hash(&hasher, &x.0)); + it.reflect_insert(&e); + if let Some(p) = removed.iter().position(|e| e == &(i, 2 * i)) { + removed.swap_remove(p); + } + left += 1; + } + } + } + + let e = it.next(); + if e.is_none() { + break; + } + assert!(i < n); + let t = unsafe { e.unwrap().as_ref() }; + assert!(!removed.contains(t)); + let (k, v) = t; + assert_eq!(*v, 2 * k); + i += 1; + } + assert!(i <= n); + + // just for safety: + assert_eq!(m.table.len(), left); + } + } + + #[test] + fn test_const_with_hasher() { + use core::hash::BuildHasher; + use std::borrow::ToOwned; + use std::collections::hash_map::DefaultHasher; + + #[derive(Clone)] + struct MyHasher; + impl BuildHasher for MyHasher { + type Hasher = DefaultHasher; + + fn build_hasher(&self) -> DefaultHasher { + DefaultHasher::new() + } + } + + const EMPTY_MAP: HashMap<u32, std::string::String, MyHasher> = + HashMap::with_hasher(MyHasher); + + let mut map = EMPTY_MAP.clone(); + map.insert(17, "seventeen".to_owned()); + assert_eq!("seventeen", map[&17]); + } +} diff --git a/src/raw/bitmask.rs b/src/raw/bitmask.rs new file mode 100644 index 0000000..99b2d53 --- /dev/null +++ b/src/raw/bitmask.rs @@ -0,0 +1,122 @@ +use super::imp::{BitMaskWord, BITMASK_MASK, BITMASK_STRIDE}; +#[cfg(feature = "nightly")] +use core::intrinsics; + +/// A bit mask which contains the result of a `Match` operation on a `Group` and +/// allows iterating through them. +/// +/// The bit mask is arranged so that low-order bits represent lower memory +/// addresses for group match results. +/// +/// For implementation reasons, the bits in the set may be sparsely packed, so +/// that there is only one bit-per-byte used (the high bit, 7). If this is the +/// case, `BITMASK_STRIDE` will be 8 to indicate a divide-by-8 should be +/// performed on counts/indices to normalize this difference. `BITMASK_MASK` is +/// similarly a mask of all the actually-used bits. +#[derive(Copy, Clone)] +pub struct BitMask(pub BitMaskWord); + +#[allow(clippy::use_self)] +impl BitMask { + /// Returns a new `BitMask` with all bits inverted. + #[inline] + #[must_use] + pub fn invert(self) -> Self { + BitMask(self.0 ^ BITMASK_MASK) + } + + /// Flip the bit in the mask for the entry at the given index. + /// + /// Returns the bit's previous state. + #[inline] + #[allow(clippy::cast_ptr_alignment)] + #[cfg(feature = "raw")] + pub unsafe fn flip(&mut self, index: usize) -> bool { + // NOTE: The + BITMASK_STRIDE - 1 is to set the high bit. + let mask = 1 << (index * BITMASK_STRIDE + BITMASK_STRIDE - 1); + self.0 ^= mask; + // The bit was set if the bit is now 0. + self.0 & mask == 0 + } + + /// Returns a new `BitMask` with the lowest bit removed. + #[inline] + #[must_use] + pub fn remove_lowest_bit(self) -> Self { + BitMask(self.0 & (self.0 - 1)) + } + /// Returns whether the `BitMask` has at least one set bit. + #[inline] + pub fn any_bit_set(self) -> bool { + self.0 != 0 + } + + /// Returns the first set bit in the `BitMask`, if there is one. + #[inline] + pub fn lowest_set_bit(self) -> Option<usize> { + if self.0 == 0 { + None + } else { + Some(unsafe { self.lowest_set_bit_nonzero() }) + } + } + + /// Returns the first set bit in the `BitMask`, if there is one. The + /// bitmask must not be empty. + #[inline] + #[cfg(feature = "nightly")] + pub unsafe fn lowest_set_bit_nonzero(self) -> usize { + intrinsics::cttz_nonzero(self.0) as usize / BITMASK_STRIDE + } + #[inline] + #[cfg(not(feature = "nightly"))] + pub unsafe fn lowest_set_bit_nonzero(self) -> usize { + self.trailing_zeros() + } + + /// Returns the number of trailing zeroes in the `BitMask`. + #[inline] + pub fn trailing_zeros(self) -> usize { + // ARM doesn't have a trailing_zeroes instruction, and instead uses + // reverse_bits (RBIT) + leading_zeroes (CLZ). However older ARM + // versions (pre-ARMv7) don't have RBIT and need to emulate it + // instead. Since we only have 1 bit set in each byte on ARM, we can + // use swap_bytes (REV) + leading_zeroes instead. + if cfg!(target_arch = "arm") && BITMASK_STRIDE % 8 == 0 { + self.0.swap_bytes().leading_zeros() as usize / BITMASK_STRIDE + } else { + self.0.trailing_zeros() as usize / BITMASK_STRIDE + } + } + + /// Returns the number of leading zeroes in the `BitMask`. + #[inline] + pub fn leading_zeros(self) -> usize { + self.0.leading_zeros() as usize / BITMASK_STRIDE + } +} + +impl IntoIterator for BitMask { + type Item = usize; + type IntoIter = BitMaskIter; + + #[inline] + fn into_iter(self) -> BitMaskIter { + BitMaskIter(self) + } +} + +/// Iterator over the contents of a `BitMask`, returning the indicies of set +/// bits. +pub struct BitMaskIter(BitMask); + +impl Iterator for BitMaskIter { + type Item = usize; + + #[inline] + fn next(&mut self) -> Option<usize> { + let bit = self.0.lowest_set_bit()?; + self.0 = self.0.remove_lowest_bit(); + Some(bit) + } +} diff --git a/src/raw/generic.rs b/src/raw/generic.rs new file mode 100644 index 0000000..26f8c58 --- /dev/null +++ b/src/raw/generic.rs @@ -0,0 +1,151 @@ +use super::bitmask::BitMask; +use super::EMPTY; +use core::{mem, ptr}; + +// Use the native word size as the group size. Using a 64-bit group size on +// a 32-bit architecture will just end up being more expensive because +// shifts and multiplies will need to be emulated. +#[cfg(any( + target_pointer_width = "64", + target_arch = "aarch64", + target_arch = "x86_64", +))] +type GroupWord = u64; +#[cfg(all( + target_pointer_width = "32", + not(target_arch = "aarch64"), + not(target_arch = "x86_64"), +))] +type GroupWord = u32; + +pub type BitMaskWord = GroupWord; +pub const BITMASK_STRIDE: usize = 8; +// We only care about the highest bit of each byte for the mask. +#[allow(clippy::cast_possible_truncation, clippy::unnecessary_cast)] +pub const BITMASK_MASK: BitMaskWord = 0x8080_8080_8080_8080_u64 as GroupWord; + +/// Helper function to replicate a byte across a `GroupWord`. +#[inline] +fn repeat(byte: u8) -> GroupWord { + GroupWord::from_ne_bytes([byte; Group::WIDTH]) +} + +/// Abstraction over a group of control bytes which can be scanned in +/// parallel. +/// +/// This implementation uses a word-sized integer. +#[derive(Copy, Clone)] +pub struct Group(GroupWord); + +// We perform all operations in the native endianess, and convert to +// little-endian just before creating a BitMask. The can potentially +// enable the compiler to eliminate unnecessary byte swaps if we are +// only checking whether a BitMask is empty. +#[allow(clippy::use_self)] +impl Group { + /// Number of bytes in the group. + pub const WIDTH: usize = mem::size_of::<Self>(); + + /// Returns a full group of empty bytes, suitable for use as the initial + /// value for an empty hash table. + /// + /// This is guaranteed to be aligned to the group size. + pub const fn static_empty() -> &'static [u8; Group::WIDTH] { + #[repr(C)] + struct AlignedBytes { + _align: [Group; 0], + bytes: [u8; Group::WIDTH], + }; + const ALIGNED_BYTES: AlignedBytes = AlignedBytes { + _align: [], + bytes: [EMPTY; Group::WIDTH], + }; + &ALIGNED_BYTES.bytes + } + + /// Loads a group of bytes starting at the given address. + #[inline] + #[allow(clippy::cast_ptr_alignment)] // unaligned load + pub unsafe fn load(ptr: *const u8) -> Self { + Group(ptr::read_unaligned(ptr as *const _)) + } + + /// Loads a group of bytes starting at the given address, which must be + /// aligned to `mem::align_of::<Group>()`. + #[inline] + #[allow(clippy::cast_ptr_alignment)] + pub unsafe fn load_aligned(ptr: *const u8) -> Self { + // FIXME: use align_offset once it stabilizes + debug_assert_eq!(ptr as usize & (mem::align_of::<Self>() - 1), 0); + Group(ptr::read(ptr as *const _)) + } + + /// Stores the group of bytes to the given address, which must be + /// aligned to `mem::align_of::<Group>()`. + #[inline] + #[allow(clippy::cast_ptr_alignment)] + pub unsafe fn store_aligned(self, ptr: *mut u8) { + // FIXME: use align_offset once it stabilizes + debug_assert_eq!(ptr as usize & (mem::align_of::<Self>() - 1), 0); + ptr::write(ptr as *mut _, self.0); + } + + /// Returns a `BitMask` indicating all bytes in the group which *may* + /// have the given value. + /// + /// This function may return a false positive in certain cases where + /// the byte in the group differs from the searched value only in its + /// lowest bit. This is fine because: + /// - This never happens for `EMPTY` and `DELETED`, only full entries. + /// - The check for key equality will catch these. + /// - This only happens if there is at least 1 true match. + /// - The chance of this happening is very low (< 1% chance per byte). + #[inline] + pub fn match_byte(self, byte: u8) -> BitMask { + // This algorithm is derived from + // http://graphics.stanford.edu/~seander/bithacks.html##ValueInWord + let cmp = self.0 ^ repeat(byte); + BitMask((cmp.wrapping_sub(repeat(0x01)) & !cmp & repeat(0x80)).to_le()) + } + + /// Returns a `BitMask` indicating all bytes in the group which are + /// `EMPTY`. + #[inline] + pub fn match_empty(self) -> BitMask { + // If the high bit is set, then the byte must be either: + // 1111_1111 (EMPTY) or 1000_0000 (DELETED). + // So we can just check if the top two bits are 1 by ANDing them. + BitMask((self.0 & (self.0 << 1) & repeat(0x80)).to_le()) + } + + /// Returns a `BitMask` indicating all bytes in the group which are + /// `EMPTY` or `DELETED`. + #[inline] + pub fn match_empty_or_deleted(self) -> BitMask { + // A byte is EMPTY or DELETED iff the high bit is set + BitMask((self.0 & repeat(0x80)).to_le()) + } + + /// Returns a `BitMask` indicating all bytes in the group which are full. + #[inline] + pub fn match_full(self) -> BitMask { + self.match_empty_or_deleted().invert() + } + + /// Performs the following transformation on all bytes in the group: + /// - `EMPTY => EMPTY` + /// - `DELETED => EMPTY` + /// - `FULL => DELETED` + #[inline] + pub fn convert_special_to_empty_and_full_to_deleted(self) -> Self { + // Map high_bit = 1 (EMPTY or DELETED) to 1111_1111 + // and high_bit = 0 (FULL) to 1000_0000 + // + // Here's this logic expanded to concrete values: + // let full = 1000_0000 (true) or 0000_0000 (false) + // !1000_0000 + 1 = 0111_1111 + 1 = 1000_0000 (no carry) + // !0000_0000 + 0 = 1111_1111 + 0 = 1111_1111 (no carry) + let full = !self.0 & repeat(0x80); + Group(!full + (full >> 7)) + } +} diff --git a/src/raw/mod.rs b/src/raw/mod.rs new file mode 100644 index 0000000..32fec98 --- /dev/null +++ b/src/raw/mod.rs @@ -0,0 +1,1924 @@ +use crate::alloc::alloc::{alloc, dealloc, handle_alloc_error}; +use crate::scopeguard::guard; +use crate::TryReserveError; +use core::alloc::Layout; +use core::hint; +use core::iter::FusedIterator; +use core::marker::PhantomData; +use core::mem; +use core::mem::ManuallyDrop; +use core::ptr::NonNull; + +cfg_if! { + // Use the SSE2 implementation if possible: it allows us to scan 16 buckets + // at once instead of 8. We don't bother with AVX since it would require + // runtime dispatch and wouldn't gain us much anyways: the probability of + // finding a match drops off drastically after the first few buckets. + // + // I attempted an implementation on ARM using NEON instructions, but it + // turns out that most NEON instructions have multi-cycle latency, which in + // the end outweighs any gains over the generic implementation. + if #[cfg(all( + target_feature = "sse2", + any(target_arch = "x86", target_arch = "x86_64"), + not(miri) + ))] { + mod sse2; + use sse2 as imp; + } else { + #[path = "generic.rs"] + mod generic; + use generic as imp; + } +} + +mod bitmask; + +use self::bitmask::{BitMask, BitMaskIter}; +use self::imp::Group; + +// Branch prediction hint. This is currently only available on nightly but it +// consistently improves performance by 10-15%. +#[cfg(feature = "nightly")] +use core::intrinsics::{likely, unlikely}; +#[cfg(not(feature = "nightly"))] +#[inline] +fn likely(b: bool) -> bool { + b +} +#[cfg(not(feature = "nightly"))] +#[inline] +fn unlikely(b: bool) -> bool { + b +} + +#[cfg(feature = "nightly")] +#[cfg_attr(feature = "inline-more", inline)] +unsafe fn offset_from<T>(to: *const T, from: *const T) -> usize { + to.offset_from(from) as usize +} +#[cfg(not(feature = "nightly"))] +#[cfg_attr(feature = "inline-more", inline)] +unsafe fn offset_from<T>(to: *const T, from: *const T) -> usize { + (to as usize - from as usize) / mem::size_of::<T>() +} + +/// Whether memory allocation errors should return an error or abort. +#[derive(Copy, Clone)] +enum Fallibility { + Fallible, + Infallible, +} + +impl Fallibility { + /// Error to return on capacity overflow. + #[cfg_attr(feature = "inline-more", inline)] + fn capacity_overflow(self) -> TryReserveError { + match self { + Fallibility::Fallible => TryReserveError::CapacityOverflow, + Fallibility::Infallible => panic!("Hash table capacity overflow"), + } + } + + /// Error to return on allocation error. + #[cfg_attr(feature = "inline-more", inline)] + fn alloc_err(self, layout: Layout) -> TryReserveError { + match self { + Fallibility::Fallible => TryReserveError::AllocError { layout }, + Fallibility::Infallible => handle_alloc_error(layout), + } + } +} + +/// Control byte value for an empty bucket. +const EMPTY: u8 = 0b1111_1111; + +/// Control byte value for a deleted bucket. +const DELETED: u8 = 0b1000_0000; + +/// Checks whether a control byte represents a full bucket (top bit is clear). +#[inline] +fn is_full(ctrl: u8) -> bool { + ctrl & 0x80 == 0 +} + +/// Checks whether a control byte represents a special value (top bit is set). +#[inline] +fn is_special(ctrl: u8) -> bool { + ctrl & 0x80 != 0 +} + +/// Checks whether a special control value is EMPTY (just check 1 bit). +#[inline] +fn special_is_empty(ctrl: u8) -> bool { + debug_assert!(is_special(ctrl)); + ctrl & 0x01 != 0 +} + +/// Primary hash function, used to select the initial bucket to probe from. +#[inline] +#[allow(clippy::cast_possible_truncation)] +fn h1(hash: u64) -> usize { + // On 32-bit platforms we simply ignore the higher hash bits. + hash as usize +} + +/// Secondary hash function, saved in the low 7 bits of the control byte. +#[inline] +#[allow(clippy::cast_possible_truncation)] +fn h2(hash: u64) -> u8 { + // Grab the top 7 bits of the hash. While the hash is normally a full 64-bit + // value, some hash functions (such as FxHash) produce a usize result + // instead, which means that the top 32 bits are 0 on 32-bit platforms. + let hash_len = usize::min(mem::size_of::<usize>(), mem::size_of::<u64>()); + let top7 = hash >> (hash_len * 8 - 7); + (top7 & 0x7f) as u8 // truncation +} + +/// Probe sequence based on triangular numbers, which is guaranteed (since our +/// table size is a power of two) to visit every group of elements exactly once. +/// +/// A triangular probe has us jump by 1 more group every time. So first we +/// jump by 1 group (meaning we just continue our linear scan), then 2 groups +/// (skipping over 1 group), then 3 groups (skipping over 2 groups), and so on. +/// +/// Proof that the probe will visit every group in the table: +/// <https://fgiesen.wordpress.com/2015/02/22/triangular-numbers-mod-2n/> +struct ProbeSeq { + bucket_mask: usize, + pos: usize, + stride: usize, +} + +impl Iterator for ProbeSeq { + type Item = usize; + + #[inline] + fn next(&mut self) -> Option<usize> { + // We should have found an empty bucket by now and ended the probe. + debug_assert!( + self.stride <= self.bucket_mask, + "Went past end of probe sequence" + ); + + let result = self.pos; + self.stride += Group::WIDTH; + self.pos += self.stride; + self.pos &= self.bucket_mask; + Some(result) + } +} + +/// Returns the number of buckets needed to hold the given number of items, +/// taking the maximum load factor into account. +/// +/// Returns `None` if an overflow occurs. +// Workaround for emscripten bug emscripten-core/emscripten-fastcomp#258 +#[cfg_attr(target_os = "emscripten", inline(never))] +#[cfg_attr(not(target_os = "emscripten"), inline)] +fn capacity_to_buckets(cap: usize) -> Option<usize> { + debug_assert_ne!(cap, 0); + + // For small tables we require at least 1 empty bucket so that lookups are + // guaranteed to terminate if an element doesn't exist in the table. + if cap < 8 { + // We don't bother with a table size of 2 buckets since that can only + // hold a single element. Instead we skip directly to a 4 bucket table + // which can hold 3 elements. + return Some(if cap < 4 { 4 } else { 8 }); + } + + // Otherwise require 1/8 buckets to be empty (87.5% load) + // + // Be careful when modifying this, calculate_layout relies on the + // overflow check here. + let adjusted_cap = cap.checked_mul(8)? / 7; + + // Any overflows will have been caught by the checked_mul. Also, any + // rounding errors from the division above will be cleaned up by + // next_power_of_two (which can't overflow because of the previous divison). + Some(adjusted_cap.next_power_of_two()) +} + +/// Returns the maximum effective capacity for the given bucket mask, taking +/// the maximum load factor into account. +#[inline] +fn bucket_mask_to_capacity(bucket_mask: usize) -> usize { + if bucket_mask < 8 { + // For tables with 1/2/4/8 buckets, we always reserve one empty slot. + // Keep in mind that the bucket mask is one less than the bucket count. + bucket_mask + } else { + // For larger tables we reserve 12.5% of the slots as empty. + ((bucket_mask + 1) / 8) * 7 + } +} + +/// Returns a Layout which describes the allocation required for a hash table, +/// and the offset of the control bytes in the allocation. +/// (the offset is also one past last element of buckets) +/// +/// Returns `None` if an overflow occurs. +#[cfg_attr(feature = "inline-more", inline)] +#[cfg(feature = "nightly")] +fn calculate_layout<T>(buckets: usize) -> Option<(Layout, usize)> { + debug_assert!(buckets.is_power_of_two()); + + // Array of buckets + let data = Layout::array::<T>(buckets).ok()?; + + // Array of control bytes. This must be aligned to the group size. + // + // We add `Group::WIDTH` control bytes at the end of the array which + // replicate the bytes at the start of the array and thus avoids the need to + // perform bounds-checking while probing. + // + // There is no possible overflow here since buckets is a power of two and + // Group::WIDTH is a small number. + let ctrl = unsafe { Layout::from_size_align_unchecked(buckets + Group::WIDTH, Group::WIDTH) }; + + data.extend(ctrl).ok() +} + +/// Returns a Layout which describes the allocation required for a hash table, +/// and the offset of the control bytes in the allocation. +/// (the offset is also one past last element of buckets) +/// +/// Returns `None` if an overflow occurs. +#[cfg_attr(feature = "inline-more", inline)] +#[cfg(not(feature = "nightly"))] +fn calculate_layout<T>(buckets: usize) -> Option<(Layout, usize)> { + debug_assert!(buckets.is_power_of_two()); + + // Manual layout calculation since Layout methods are not yet stable. + let ctrl_align = usize::max(mem::align_of::<T>(), Group::WIDTH); + let ctrl_offset = mem::size_of::<T>() + .checked_mul(buckets)? + .checked_add(ctrl_align - 1)? + & !(ctrl_align - 1); + let len = ctrl_offset.checked_add(buckets + Group::WIDTH)?; + + Some(( + unsafe { Layout::from_size_align_unchecked(len, ctrl_align) }, + ctrl_offset, + )) +} + +/// A reference to a hash table bucket containing a `T`. +/// +/// This is usually just a pointer to the element itself. However if the element +/// is a ZST, then we instead track the index of the element in the table so +/// that `erase` works properly. +pub struct Bucket<T> { + // Actually it is pointer to next element than element itself + // this is needed to maintain pointer arithmetic invariants + // keeping direct pointer to element introduces difficulty. + // Using `NonNull` for variance and niche layout + ptr: NonNull<T>, +} + +// This Send impl is needed for rayon support. This is safe since Bucket is +// never exposed in a public API. +unsafe impl<T> Send for Bucket<T> {} + +impl<T> Clone for Bucket<T> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + Self { ptr: self.ptr } + } +} + +impl<T> Bucket<T> { + #[cfg_attr(feature = "inline-more", inline)] + unsafe fn from_base_index(base: NonNull<T>, index: usize) -> Self { + let ptr = if mem::size_of::<T>() == 0 { + // won't overflow because index must be less than length + (index + 1) as *mut T + } else { + base.as_ptr().sub(index) + }; + Self { + ptr: NonNull::new_unchecked(ptr), + } + } + #[cfg_attr(feature = "inline-more", inline)] + unsafe fn to_base_index(&self, base: NonNull<T>) -> usize { + if mem::size_of::<T>() == 0 { + self.ptr.as_ptr() as usize - 1 + } else { + offset_from(base.as_ptr(), self.ptr.as_ptr()) + } + } + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn as_ptr(&self) -> *mut T { + if mem::size_of::<T>() == 0 { + // Just return an arbitrary ZST pointer which is properly aligned + mem::align_of::<T>() as *mut T + } else { + self.ptr.as_ptr().sub(1) + } + } + #[cfg_attr(feature = "inline-more", inline)] + unsafe fn next_n(&self, offset: usize) -> Self { + let ptr = if mem::size_of::<T>() == 0 { + (self.ptr.as_ptr() as usize + offset) as *mut T + } else { + self.ptr.as_ptr().sub(offset) + }; + Self { + ptr: NonNull::new_unchecked(ptr), + } + } + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn drop(&self) { + self.as_ptr().drop_in_place(); + } + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn read(&self) -> T { + self.as_ptr().read() + } + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn write(&self, val: T) { + self.as_ptr().write(val); + } + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn as_ref<'a>(&self) -> &'a T { + &*self.as_ptr() + } + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn as_mut<'a>(&self) -> &'a mut T { + &mut *self.as_ptr() + } + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn copy_from_nonoverlapping(&self, other: &Self) { + self.as_ptr().copy_from_nonoverlapping(other.as_ptr(), 1); + } +} + +/// A raw hash table with an unsafe API. +pub struct RawTable<T> { + // Mask to get an index from a hash value. The value is one less than the + // number of buckets in the table. + bucket_mask: usize, + + // [Padding], T1, T2, ..., Tlast, C1, C2, ... + // ^ points here + ctrl: NonNull<u8>, + + // Number of elements that can be inserted before we need to grow the table + growth_left: usize, + + // Number of elements in the table, only really used by len() + items: usize, + + // Tell dropck that we own instances of T. + marker: PhantomData<T>, +} + +impl<T> RawTable<T> { + /// Creates a new empty hash table without allocating any memory. + /// + /// In effect this returns a table with exactly 1 bucket. However we can + /// leave the data pointer dangling since that bucket is never written to + /// due to our load factor forcing us to always have at least 1 free bucket. + #[cfg_attr(feature = "inline-more", inline)] + pub const fn new() -> Self { + Self { + // Be careful to cast the entire slice to a raw pointer. + ctrl: unsafe { NonNull::new_unchecked(Group::static_empty() as *const _ as *mut u8) }, + bucket_mask: 0, + items: 0, + growth_left: 0, + marker: PhantomData, + } + } + + /// Allocates a new hash table with the given number of buckets. + /// + /// The control bytes are left uninitialized. + #[cfg_attr(feature = "inline-more", inline)] + unsafe fn new_uninitialized( + buckets: usize, + fallability: Fallibility, + ) -> Result<Self, TryReserveError> { + debug_assert!(buckets.is_power_of_two()); + + // Avoid `Option::ok_or_else` because it bloats LLVM IR. + let (layout, ctrl_offset) = match calculate_layout::<T>(buckets) { + Some(lco) => lco, + None => return Err(fallability.capacity_overflow()), + }; + let ptr = match NonNull::new(alloc(layout)) { + Some(ptr) => ptr, + None => return Err(fallability.alloc_err(layout)), + }; + let ctrl = NonNull::new_unchecked(ptr.as_ptr().add(ctrl_offset)); + Ok(Self { + ctrl, + bucket_mask: buckets - 1, + items: 0, + growth_left: bucket_mask_to_capacity(buckets - 1), + marker: PhantomData, + }) + } + + /// Attempts to allocate a new hash table with at least enough capacity + /// for inserting the given number of elements without reallocating. + fn fallible_with_capacity( + capacity: usize, + fallability: Fallibility, + ) -> Result<Self, TryReserveError> { + if capacity == 0 { + Ok(Self::new()) + } else { + unsafe { + // Avoid `Option::ok_or_else` because it bloats LLVM IR. + let buckets = match capacity_to_buckets(capacity) { + Some(buckets) => buckets, + None => return Err(fallability.capacity_overflow()), + }; + let result = Self::new_uninitialized(buckets, fallability)?; + result.ctrl(0).write_bytes(EMPTY, result.num_ctrl_bytes()); + + Ok(result) + } + } + } + + /// Attempts to allocate a new hash table with at least enough capacity + /// for inserting the given number of elements without reallocating. + #[cfg(feature = "raw")] + pub fn try_with_capacity(capacity: usize) -> Result<Self, TryReserveError> { + Self::fallible_with_capacity(capacity, Fallibility::Fallible) + } + + /// Allocates a new hash table with at least enough capacity for inserting + /// the given number of elements without reallocating. + pub fn with_capacity(capacity: usize) -> Self { + // Avoid `Result::unwrap_or_else` because it bloats LLVM IR. + match Self::fallible_with_capacity(capacity, Fallibility::Infallible) { + Ok(capacity) => capacity, + Err(_) => unsafe { hint::unreachable_unchecked() }, + } + } + + /// Deallocates the table without dropping any entries. + #[cfg_attr(feature = "inline-more", inline)] + unsafe fn free_buckets(&mut self) { + // Avoid `Option::unwrap_or_else` because it bloats LLVM IR. + let (layout, ctrl_offset) = match calculate_layout::<T>(self.buckets()) { + Some(lco) => lco, + None => hint::unreachable_unchecked(), + }; + dealloc(self.ctrl.as_ptr().sub(ctrl_offset), layout); + } + + /// Returns pointer to one past last element of data table. + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn data_end(&self) -> NonNull<T> { + NonNull::new_unchecked(self.ctrl.as_ptr() as *mut T) + } + + /// Returns pointer to start of data table. + #[cfg_attr(feature = "inline-more", inline)] + #[cfg(feature = "nightly")] + pub unsafe fn data_start(&self) -> *mut T { + self.data_end().as_ptr().wrapping_sub(self.buckets()) + } + + /// Returns the index of a bucket from a `Bucket`. + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn bucket_index(&self, bucket: &Bucket<T>) -> usize { + bucket.to_base_index(self.data_end()) + } + + /// Returns a pointer to a control byte. + #[cfg_attr(feature = "inline-more", inline)] + unsafe fn ctrl(&self, index: usize) -> *mut u8 { + debug_assert!(index < self.num_ctrl_bytes()); + self.ctrl.as_ptr().add(index) + } + + /// Returns a pointer to an element in the table. + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn bucket(&self, index: usize) -> Bucket<T> { + debug_assert_ne!(self.bucket_mask, 0); + debug_assert!(index < self.buckets()); + Bucket::from_base_index(self.data_end(), index) + } + + /// Erases an element from the table without dropping it. + #[cfg_attr(feature = "inline-more", inline)] + #[deprecated(since = "0.8.1", note = "use erase or remove instead")] + pub unsafe fn erase_no_drop(&mut self, item: &Bucket<T>) { + let index = self.bucket_index(item); + debug_assert!(is_full(*self.ctrl(index))); + let index_before = index.wrapping_sub(Group::WIDTH) & self.bucket_mask; + let empty_before = Group::load(self.ctrl(index_before)).match_empty(); + let empty_after = Group::load(self.ctrl(index)).match_empty(); + + // If we are inside a continuous block of Group::WIDTH full or deleted + // cells then a probe window may have seen a full block when trying to + // insert. We therefore need to keep that block non-empty so that + // lookups will continue searching to the next probe window. + // + // Note that in this context `leading_zeros` refers to the bytes at the + // end of a group, while `trailing_zeros` refers to the bytes at the + // begining of a group. + let ctrl = if empty_before.leading_zeros() + empty_after.trailing_zeros() >= Group::WIDTH { + DELETED + } else { + self.growth_left += 1; + EMPTY + }; + self.set_ctrl(index, ctrl); + self.items -= 1; + } + + /// Erases an element from the table, dropping it in place. + #[cfg_attr(feature = "inline-more", inline)] + #[allow(clippy::needless_pass_by_value)] + #[allow(deprecated)] + pub unsafe fn erase(&mut self, item: Bucket<T>) { + // Erase the element from the table first since drop might panic. + self.erase_no_drop(&item); + item.drop(); + } + + /// Finds and erases an element from the table, dropping it in place. + /// Returns true if an element was found. + #[cfg(feature = "raw")] + #[cfg_attr(feature = "inline-more", inline)] + pub fn erase_entry(&mut self, hash: u64, eq: impl FnMut(&T) -> bool) -> bool { + // Avoid `Option::map` because it bloats LLVM IR. + if let Some(bucket) = self.find(hash, eq) { + unsafe { self.erase(bucket) }; + true + } else { + false + } + } + + /// Removes an element from the table, returning it. + #[cfg_attr(feature = "inline-more", inline)] + #[allow(clippy::needless_pass_by_value)] + #[allow(deprecated)] + pub unsafe fn remove(&mut self, item: Bucket<T>) -> T { + self.erase_no_drop(&item); + item.read() + } + + /// Finds and removes an element from the table, returning it. + #[cfg_attr(feature = "inline-more", inline)] + pub fn remove_entry(&mut self, hash: u64, eq: impl FnMut(&T) -> bool) -> Option<T> { + // Avoid `Option::map` because it bloats LLVM IR. + match self.find(hash, eq) { + Some(bucket) => Some(unsafe { self.remove(bucket) }), + None => None, + } + } + + /// Returns an iterator for a probe sequence on the table. + /// + /// This iterator never terminates, but is guaranteed to visit each bucket + /// group exactly once. The loop using `probe_seq` must terminate upon + /// reaching a group containing an empty bucket. + #[cfg_attr(feature = "inline-more", inline)] + fn probe_seq(&self, hash: u64) -> ProbeSeq { + ProbeSeq { + bucket_mask: self.bucket_mask, + pos: h1(hash) & self.bucket_mask, + stride: 0, + } + } + + /// Sets a control byte, and possibly also the replicated control byte at + /// the end of the array. + #[cfg_attr(feature = "inline-more", inline)] + unsafe fn set_ctrl(&self, index: usize, ctrl: u8) { + // Replicate the first Group::WIDTH control bytes at the end of + // the array without using a branch: + // - If index >= Group::WIDTH then index == index2. + // - Otherwise index2 == self.bucket_mask + 1 + index. + // + // The very last replicated control byte is never actually read because + // we mask the initial index for unaligned loads, but we write it + // anyways because it makes the set_ctrl implementation simpler. + // + // If there are fewer buckets than Group::WIDTH then this code will + // replicate the buckets at the end of the trailing group. For example + // with 2 buckets and a group size of 4, the control bytes will look + // like this: + // + // Real | Replicated + // --------------------------------------------- + // | [A] | [B] | [EMPTY] | [EMPTY] | [A] | [B] | + // --------------------------------------------- + let index2 = ((index.wrapping_sub(Group::WIDTH)) & self.bucket_mask) + Group::WIDTH; + + *self.ctrl(index) = ctrl; + *self.ctrl(index2) = ctrl; + } + + /// Searches for an empty or deleted bucket which is suitable for inserting + /// a new element. + /// + /// There must be at least 1 empty bucket in the table. + #[cfg_attr(feature = "inline-more", inline)] + fn find_insert_slot(&self, hash: u64) -> usize { + for pos in self.probe_seq(hash) { + unsafe { + let group = Group::load(self.ctrl(pos)); + if let Some(bit) = group.match_empty_or_deleted().lowest_set_bit() { + let result = (pos + bit) & self.bucket_mask; + + // In tables smaller than the group width, trailing control + // bytes outside the range of the table are filled with + // EMPTY entries. These will unfortunately trigger a + // match, but once masked may point to a full bucket that + // is already occupied. We detect this situation here and + // perform a second scan starting at the begining of the + // table. This second scan is guaranteed to find an empty + // slot (due to the load factor) before hitting the trailing + // control bytes (containing EMPTY). + if unlikely(is_full(*self.ctrl(result))) { + debug_assert!(self.bucket_mask < Group::WIDTH); + debug_assert_ne!(pos, 0); + return Group::load_aligned(self.ctrl(0)) + .match_empty_or_deleted() + .lowest_set_bit_nonzero(); + } else { + return result; + } + } + } + } + + // probe_seq never returns. + unreachable!(); + } + + /// Marks all table buckets as empty without dropping their contents. + #[cfg_attr(feature = "inline-more", inline)] + pub fn clear_no_drop(&mut self) { + if !self.is_empty_singleton() { + unsafe { + self.ctrl(0).write_bytes(EMPTY, self.num_ctrl_bytes()); + } + } + self.items = 0; + self.growth_left = bucket_mask_to_capacity(self.bucket_mask); + } + + /// Removes all elements from the table without freeing the backing memory. + #[cfg_attr(feature = "inline-more", inline)] + pub fn clear(&mut self) { + // Ensure that the table is reset even if one of the drops panic + let self_ = guard(self, |self_| self_.clear_no_drop()); + + if mem::needs_drop::<T>() && self_.len() != 0 { + unsafe { + for item in self_.iter() { + item.drop(); + } + } + } + } + + /// Shrinks the table to fit `max(self.len(), min_size)` elements. + #[cfg_attr(feature = "inline-more", inline)] + pub fn shrink_to(&mut self, min_size: usize, hasher: impl Fn(&T) -> u64) { + // Calculate the minimal number of elements that we need to reserve + // space for. + let min_size = usize::max(self.items, min_size); + if min_size == 0 { + *self = Self::new(); + return; + } + + // Calculate the number of buckets that we need for this number of + // elements. If the calculation overflows then the requested bucket + // count must be larger than what we have right and nothing needs to be + // done. + let min_buckets = match capacity_to_buckets(min_size) { + Some(buckets) => buckets, + None => return, + }; + + // If we have more buckets than we need, shrink the table. + if min_buckets < self.buckets() { + // Fast path if the table is empty + if self.items == 0 { + *self = Self::with_capacity(min_size) + } else { + // Avoid `Result::unwrap_or_else` because it bloats LLVM IR. + if self + .resize(min_size, hasher, Fallibility::Infallible) + .is_err() + { + unsafe { hint::unreachable_unchecked() } + } + } + } + } + + /// Ensures that at least `additional` items can be inserted into the table + /// without reallocation. + #[cfg_attr(feature = "inline-more", inline)] + pub fn reserve(&mut self, additional: usize, hasher: impl Fn(&T) -> u64) { + if additional > self.growth_left { + // Avoid `Result::unwrap_or_else` because it bloats LLVM IR. + if self + .reserve_rehash(additional, hasher, Fallibility::Infallible) + .is_err() + { + unsafe { hint::unreachable_unchecked() } + } + } + } + + /// Tries to ensure that at least `additional` items can be inserted into + /// the table without reallocation. + #[cfg_attr(feature = "inline-more", inline)] + pub fn try_reserve( + &mut self, + additional: usize, + hasher: impl Fn(&T) -> u64, + ) -> Result<(), TryReserveError> { + if additional > self.growth_left { + self.reserve_rehash(additional, hasher, Fallibility::Fallible) + } else { + Ok(()) + } + } + + /// Out-of-line slow path for `reserve` and `try_reserve`. + #[cold] + #[inline(never)] + fn reserve_rehash( + &mut self, + additional: usize, + hasher: impl Fn(&T) -> u64, + fallability: Fallibility, + ) -> Result<(), TryReserveError> { + // Avoid `Option::ok_or_else` because it bloats LLVM IR. + let new_items = match self.items.checked_add(additional) { + Some(new_items) => new_items, + None => return Err(fallability.capacity_overflow()), + }; + let full_capacity = bucket_mask_to_capacity(self.bucket_mask); + if new_items <= full_capacity / 2 { + // Rehash in-place without re-allocating if we have plenty of spare + // capacity that is locked up due to DELETED entries. + self.rehash_in_place(hasher); + Ok(()) + } else { + // Otherwise, conservatively resize to at least the next size up + // to avoid churning deletes into frequent rehashes. + self.resize( + usize::max(new_items, full_capacity + 1), + hasher, + fallability, + ) + } + } + + /// Rehashes the contents of the table in place (i.e. without changing the + /// allocation). + /// + /// If `hasher` panics then some the table's contents may be lost. + fn rehash_in_place(&mut self, hasher: impl Fn(&T) -> u64) { + unsafe { + // Bulk convert all full control bytes to DELETED, and all DELETED + // control bytes to EMPTY. This effectively frees up all buckets + // containing a DELETED entry. + for i in (0..self.buckets()).step_by(Group::WIDTH) { + let group = Group::load_aligned(self.ctrl(i)); + let group = group.convert_special_to_empty_and_full_to_deleted(); + group.store_aligned(self.ctrl(i)); + } + + // Fix up the trailing control bytes. See the comments in set_ctrl + // for the handling of tables smaller than the group width. + if self.buckets() < Group::WIDTH { + self.ctrl(0) + .copy_to(self.ctrl(Group::WIDTH), self.buckets()); + } else { + self.ctrl(0) + .copy_to(self.ctrl(self.buckets()), Group::WIDTH); + } + + // If the hash function panics then properly clean up any elements + // that we haven't rehashed yet. We unfortunately can't preserve the + // element since we lost their hash and have no way of recovering it + // without risking another panic. + let mut guard = guard(self, |self_| { + if mem::needs_drop::<T>() { + for i in 0..self_.buckets() { + if *self_.ctrl(i) == DELETED { + self_.set_ctrl(i, EMPTY); + self_.bucket(i).drop(); + self_.items -= 1; + } + } + } + self_.growth_left = bucket_mask_to_capacity(self_.bucket_mask) - self_.items; + }); + + // At this point, DELETED elements are elements that we haven't + // rehashed yet. Find them and re-insert them at their ideal + // position. + 'outer: for i in 0..guard.buckets() { + if *guard.ctrl(i) != DELETED { + continue; + } + 'inner: loop { + // Hash the current item + let item = guard.bucket(i); + let hash = hasher(item.as_ref()); + + // Search for a suitable place to put it + let new_i = guard.find_insert_slot(hash); + + // Probing works by scanning through all of the control + // bytes in groups, which may not be aligned to the group + // size. If both the new and old position fall within the + // same unaligned group, then there is no benefit in moving + // it and we can just continue to the next item. + let probe_index = |pos: usize| { + (pos.wrapping_sub(guard.probe_seq(hash).pos) & guard.bucket_mask) + / Group::WIDTH + }; + if likely(probe_index(i) == probe_index(new_i)) { + guard.set_ctrl(i, h2(hash)); + continue 'outer; + } + + // We are moving the current item to a new position. Write + // our H2 to the control byte of the new position. + let prev_ctrl = *guard.ctrl(new_i); + guard.set_ctrl(new_i, h2(hash)); + + if prev_ctrl == EMPTY { + // If the target slot is empty, simply move the current + // element into the new slot and clear the old control + // byte. + guard.set_ctrl(i, EMPTY); + guard.bucket(new_i).copy_from_nonoverlapping(&item); + continue 'outer; + } else { + // If the target slot is occupied, swap the two elements + // and then continue processing the element that we just + // swapped into the old slot. + debug_assert_eq!(prev_ctrl, DELETED); + mem::swap(guard.bucket(new_i).as_mut(), item.as_mut()); + continue 'inner; + } + } + } + + guard.growth_left = bucket_mask_to_capacity(guard.bucket_mask) - guard.items; + mem::forget(guard); + } + } + + /// Allocates a new table of a different size and moves the contents of the + /// current table into it. + fn resize( + &mut self, + capacity: usize, + hasher: impl Fn(&T) -> u64, + fallability: Fallibility, + ) -> Result<(), TryReserveError> { + unsafe { + debug_assert!(self.items <= capacity); + + // Allocate and initialize the new table. + let mut new_table = Self::fallible_with_capacity(capacity, fallability)?; + new_table.growth_left -= self.items; + new_table.items = self.items; + + // The hash function may panic, in which case we simply free the new + // table without dropping any elements that may have been copied into + // it. + // + // This guard is also used to free the old table on success, see + // the comment at the bottom of this function. + let mut new_table = guard(ManuallyDrop::new(new_table), |new_table| { + if !new_table.is_empty_singleton() { + new_table.free_buckets(); + } + }); + + // Copy all elements to the new table. + for item in self.iter() { + // This may panic. + let hash = hasher(item.as_ref()); + + // We can use a simpler version of insert() here since: + // - there are no DELETED entries. + // - we know there is enough space in the table. + // - all elements are unique. + let index = new_table.find_insert_slot(hash); + new_table.set_ctrl(index, h2(hash)); + new_table.bucket(index).copy_from_nonoverlapping(&item); + } + + // We successfully copied all elements without panicking. Now replace + // self with the new table. The old table will have its memory freed but + // the items will not be dropped (since they have been moved into the + // new table). + mem::swap(self, &mut new_table); + + Ok(()) + } + } + + /// Inserts a new element into the table, and returns its raw bucket. + /// + /// This does not check if the given element already exists in the table. + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(&mut self, hash: u64, value: T, hasher: impl Fn(&T) -> u64) -> Bucket<T> { + unsafe { + let mut index = self.find_insert_slot(hash); + + // We can avoid growing the table once we have reached our load + // factor if we are replacing a tombstone. This works since the + // number of EMPTY slots does not change in this case. + let old_ctrl = *self.ctrl(index); + if unlikely(self.growth_left == 0 && special_is_empty(old_ctrl)) { + self.reserve(1, hasher); + index = self.find_insert_slot(hash); + } + + let bucket = self.bucket(index); + self.growth_left -= special_is_empty(old_ctrl) as usize; + self.set_ctrl(index, h2(hash)); + bucket.write(value); + self.items += 1; + bucket + } + } + + /// Inserts a new element into the table, and returns a mutable reference to it. + /// + /// This does not check if the given element already exists in the table. + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert_entry(&mut self, hash: u64, value: T, hasher: impl Fn(&T) -> u64) -> &mut T { + unsafe { self.insert(hash, value, hasher).as_mut() } + } + + /// Inserts a new element into the table, without growing the table. + /// + /// There must be enough space in the table to insert the new element. + /// + /// This does not check if the given element already exists in the table. + #[cfg_attr(feature = "inline-more", inline)] + #[cfg(any(feature = "raw", feature = "rustc-internal-api"))] + pub fn insert_no_grow(&mut self, hash: u64, value: T) -> Bucket<T> { + unsafe { + let index = self.find_insert_slot(hash); + let bucket = self.bucket(index); + + // If we are replacing a DELETED entry then we don't need to update + // the load counter. + let old_ctrl = *self.ctrl(index); + self.growth_left -= special_is_empty(old_ctrl) as usize; + + self.set_ctrl(index, h2(hash)); + bucket.write(value); + self.items += 1; + bucket + } + } + + /// Temporary removes a bucket, applying the given function to the removed + /// element and optionally put back the returned value in the same bucket. + /// + /// Returns `true` if the bucket still contains an element + /// + /// This does not check if the given bucket is actually occupied. + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn replace_bucket_with<F>(&mut self, bucket: Bucket<T>, f: F) -> bool + where + F: FnOnce(T) -> Option<T>, + { + let index = self.bucket_index(&bucket); + let old_ctrl = *self.ctrl(index); + debug_assert!(is_full(old_ctrl)); + let old_growth_left = self.growth_left; + let item = self.remove(bucket); + if let Some(new_item) = f(item) { + self.growth_left = old_growth_left; + self.set_ctrl(index, old_ctrl); + self.items += 1; + self.bucket(index).write(new_item); + true + } else { + false + } + } + + /// Searches for an element in the table. + #[inline] + pub fn find(&self, hash: u64, mut eq: impl FnMut(&T) -> bool) -> Option<Bucket<T>> { + unsafe { + for bucket in self.iter_hash(hash) { + let elm = bucket.as_ref(); + if likely(eq(elm)) { + return Some(bucket); + } + } + None + } + } + + /// Gets a reference to an element in the table. + #[inline] + pub fn get(&self, hash: u64, eq: impl FnMut(&T) -> bool) -> Option<&T> { + // Avoid `Option::map` because it bloats LLVM IR. + match self.find(hash, eq) { + Some(bucket) => Some(unsafe { bucket.as_ref() }), + None => None, + } + } + + /// Gets a mutable reference to an element in the table. + #[inline] + pub fn get_mut(&mut self, hash: u64, eq: impl FnMut(&T) -> bool) -> Option<&mut T> { + // Avoid `Option::map` because it bloats LLVM IR. + match self.find(hash, eq) { + Some(bucket) => Some(unsafe { bucket.as_mut() }), + None => None, + } + } + + /// Returns the number of elements the map can hold without reallocating. + /// + /// This number is a lower bound; the table might be able to hold + /// more, but is guaranteed to be able to hold at least this many. + #[cfg_attr(feature = "inline-more", inline)] + pub fn capacity(&self) -> usize { + self.items + self.growth_left + } + + /// Returns the number of elements in the table. + #[cfg_attr(feature = "inline-more", inline)] + pub fn len(&self) -> usize { + self.items + } + + /// Returns the number of buckets in the table. + #[cfg_attr(feature = "inline-more", inline)] + pub fn buckets(&self) -> usize { + self.bucket_mask + 1 + } + + /// Returns the number of control bytes in the table. + #[cfg_attr(feature = "inline-more", inline)] + fn num_ctrl_bytes(&self) -> usize { + self.bucket_mask + 1 + Group::WIDTH + } + + /// Returns whether this table points to the empty singleton with a capacity + /// of 0. + #[cfg_attr(feature = "inline-more", inline)] + fn is_empty_singleton(&self) -> bool { + self.bucket_mask == 0 + } + + /// Returns an iterator over every element in the table. It is up to + /// the caller to ensure that the `RawTable` outlives the `RawIter`. + /// Because we cannot make the `next` method unsafe on the `RawIter` + /// struct, we have to make the `iter` method unsafe. + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn iter(&self) -> RawIter<T> { + let data = Bucket::from_base_index(self.data_end(), 0); + RawIter { + iter: RawIterRange::new(self.ctrl.as_ptr(), data, self.buckets()), + items: self.items, + } + } + + /// Returns an iterator over occupied buckets that could match a given hash. + /// + /// In rare cases, the iterator may return a bucket with a different hash. + /// + /// It is up to the caller to ensure that the `RawTable` outlives the + /// `RawIterHash`. Because we cannot make the `next` method unsafe on the + /// `RawIterHash` struct, we have to make the `iter_hash` method unsafe. + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn iter_hash(&self, hash: u64) -> RawIterHash<'_, T> { + RawIterHash::new(self, hash) + } + + /// Returns an iterator which removes all elements from the table without + /// freeing the memory. + #[cfg_attr(feature = "inline-more", inline)] + pub fn drain(&mut self) -> RawDrain<'_, T> { + unsafe { + let iter = self.iter(); + self.drain_iter_from(iter) + } + } + + /// Returns an iterator which removes all elements from the table without + /// freeing the memory. + /// + /// Iteration starts at the provided iterator's current location. + /// + /// It is up to the caller to ensure that the iterator is valid for this + /// `RawTable` and covers all items that remain in the table. + #[cfg_attr(feature = "inline-more", inline)] + pub unsafe fn drain_iter_from(&mut self, iter: RawIter<T>) -> RawDrain<'_, T> { + debug_assert_eq!(iter.len(), self.len()); + RawDrain { + iter, + table: ManuallyDrop::new(mem::replace(self, Self::new())), + orig_table: NonNull::from(self), + marker: PhantomData, + } + } + + /// Returns an iterator which consumes all elements from the table. + /// + /// Iteration starts at the provided iterator's current location. + /// + /// It is up to the caller to ensure that the iterator is valid for this + /// `RawTable` and covers all items that remain in the table. + pub unsafe fn into_iter_from(self, iter: RawIter<T>) -> RawIntoIter<T> { + debug_assert_eq!(iter.len(), self.len()); + + let alloc = self.into_alloc(); + RawIntoIter { + iter, + alloc, + marker: PhantomData, + } + } + + /// Converts the table into a raw allocation. The contents of the table + /// should be dropped using a `RawIter` before freeing the allocation. + #[cfg_attr(feature = "inline-more", inline)] + pub(crate) fn into_alloc(self) -> Option<(NonNull<u8>, Layout)> { + let alloc = if self.is_empty_singleton() { + None + } else { + // Avoid `Option::unwrap_or_else` because it bloats LLVM IR. + let (layout, ctrl_offset) = match calculate_layout::<T>(self.buckets()) { + Some(lco) => lco, + None => unsafe { hint::unreachable_unchecked() }, + }; + Some(( + unsafe { NonNull::new_unchecked(self.ctrl.as_ptr().sub(ctrl_offset)) }, + layout, + )) + }; + mem::forget(self); + alloc + } +} + +unsafe impl<T> Send for RawTable<T> where T: Send {} +unsafe impl<T> Sync for RawTable<T> where T: Sync {} + +impl<T: Clone> Clone for RawTable<T> { + fn clone(&self) -> Self { + if self.is_empty_singleton() { + Self::new() + } else { + unsafe { + let mut new_table = ManuallyDrop::new( + // Avoid `Result::ok_or_else` because it bloats LLVM IR. + match Self::new_uninitialized(self.buckets(), Fallibility::Infallible) { + Ok(table) => table, + Err(_) => hint::unreachable_unchecked(), + }, + ); + + new_table.clone_from_spec(self, |new_table| { + // We need to free the memory allocated for the new table. + new_table.free_buckets(); + }); + + // Return the newly created table. + ManuallyDrop::into_inner(new_table) + } + } + } + + fn clone_from(&mut self, source: &Self) { + if source.is_empty_singleton() { + *self = Self::new(); + } else { + unsafe { + // First, drop all our elements without clearing the control bytes. + if mem::needs_drop::<T>() && self.len() != 0 { + for item in self.iter() { + item.drop(); + } + } + + // If necessary, resize our table to match the source. + if self.buckets() != source.buckets() { + // Skip our drop by using ptr::write. + if !self.is_empty_singleton() { + self.free_buckets(); + } + (self as *mut Self).write( + // Avoid `Result::unwrap_or_else` because it bloats LLVM IR. + match Self::new_uninitialized(source.buckets(), Fallibility::Infallible) { + Ok(table) => table, + Err(_) => hint::unreachable_unchecked(), + }, + ); + } + + self.clone_from_spec(source, |self_| { + // We need to leave the table in an empty state. + self_.clear_no_drop() + }); + } + } + } +} + +/// Specialization of `clone_from` for `Copy` types +trait RawTableClone { + unsafe fn clone_from_spec(&mut self, source: &Self, on_panic: impl FnMut(&mut Self)); +} +impl<T: Clone> RawTableClone for RawTable<T> { + #[cfg_attr(feature = "inline-more", inline)] + default_fn! { + unsafe fn clone_from_spec(&mut self, source: &Self, on_panic: impl FnMut(&mut Self)) { + self.clone_from_impl(source, on_panic); + } + } +} +#[cfg(feature = "nightly")] +impl<T: Copy> RawTableClone for RawTable<T> { + #[cfg_attr(feature = "inline-more", inline)] + unsafe fn clone_from_spec(&mut self, source: &Self, _on_panic: impl FnMut(&mut Self)) { + source + .ctrl(0) + .copy_to_nonoverlapping(self.ctrl(0), self.num_ctrl_bytes()); + source + .data_start() + .copy_to_nonoverlapping(self.data_start(), self.buckets()); + + self.items = source.items; + self.growth_left = source.growth_left; + } +} + +impl<T: Clone> RawTable<T> { + /// Common code for clone and clone_from. Assumes `self.buckets() == source.buckets()`. + #[cfg_attr(feature = "inline-more", inline)] + unsafe fn clone_from_impl(&mut self, source: &Self, mut on_panic: impl FnMut(&mut Self)) { + // Copy the control bytes unchanged. We do this in a single pass + source + .ctrl(0) + .copy_to_nonoverlapping(self.ctrl(0), self.num_ctrl_bytes()); + + // The cloning of elements may panic, in which case we need + // to make sure we drop only the elements that have been + // cloned so far. + let mut guard = guard((0, &mut *self), |(index, self_)| { + if mem::needs_drop::<T>() && self_.len() != 0 { + for i in 0..=*index { + if is_full(*self_.ctrl(i)) { + self_.bucket(i).drop(); + } + } + } + + // Depending on whether we were called from clone or clone_from, we + // either need to free the memory for the destination table or just + // clear the control bytes. + on_panic(self_); + }); + + for from in source.iter() { + let index = source.bucket_index(&from); + let to = guard.1.bucket(index); + to.write(from.as_ref().clone()); + + // Update the index in case we need to unwind. + guard.0 = index; + } + + // Successfully cloned all items, no need to clean up. + mem::forget(guard); + + self.items = source.items; + self.growth_left = source.growth_left; + } + + /// Variant of `clone_from` to use when a hasher is available. + #[cfg(feature = "raw")] + pub fn clone_from_with_hasher(&mut self, source: &Self, hasher: impl Fn(&T) -> u64) { + // If we have enough capacity in the table, just clear it and insert + // elements one by one. We don't do this if we have the same number of + // buckets as the source since we can just copy the contents directly + // in that case. + if self.buckets() != source.buckets() + && bucket_mask_to_capacity(self.bucket_mask) >= source.len() + { + self.clear(); + + let guard_self = guard(&mut *self, |self_| { + // Clear the partially copied table if a panic occurs, otherwise + // items and growth_left will be out of sync with the contents + // of the table. + self_.clear(); + }); + + unsafe { + for item in source.iter() { + // This may panic. + let item = item.as_ref().clone(); + let hash = hasher(&item); + + // We can use a simpler version of insert() here since: + // - there are no DELETED entries. + // - we know there is enough space in the table. + // - all elements are unique. + let index = guard_self.find_insert_slot(hash); + guard_self.set_ctrl(index, h2(hash)); + guard_self.bucket(index).write(item); + } + } + + // Successfully cloned all items, no need to clean up. + mem::forget(guard_self); + + self.items = source.items; + self.growth_left -= source.items; + } else { + self.clone_from(source); + } + } +} + +#[cfg(feature = "nightly")] +unsafe impl<#[may_dangle] T> Drop for RawTable<T> { + #[cfg_attr(feature = "inline-more", inline)] + fn drop(&mut self) { + if !self.is_empty_singleton() { + unsafe { + if mem::needs_drop::<T>() && self.len() != 0 { + for item in self.iter() { + item.drop(); + } + } + self.free_buckets(); + } + } + } +} +#[cfg(not(feature = "nightly"))] +impl<T> Drop for RawTable<T> { + #[cfg_attr(feature = "inline-more", inline)] + fn drop(&mut self) { + if !self.is_empty_singleton() { + unsafe { + if mem::needs_drop::<T>() && self.len() != 0 { + for item in self.iter() { + item.drop(); + } + } + self.free_buckets(); + } + } + } +} + +impl<T> IntoIterator for RawTable<T> { + type Item = T; + type IntoIter = RawIntoIter<T>; + + #[cfg_attr(feature = "inline-more", inline)] + fn into_iter(self) -> RawIntoIter<T> { + unsafe { + let iter = self.iter(); + self.into_iter_from(iter) + } + } +} + +/// Iterator over a sub-range of a table. Unlike `RawIter` this iterator does +/// not track an item count. +pub(crate) struct RawIterRange<T> { + // Mask of full buckets in the current group. Bits are cleared from this + // mask as each element is processed. + current_group: BitMask, + + // Pointer to the buckets for the current group. + data: Bucket<T>, + + // Pointer to the next group of control bytes, + // Must be aligned to the group size. + next_ctrl: *const u8, + + // Pointer one past the last control byte of this range. + end: *const u8, +} + +impl<T> RawIterRange<T> { + /// Returns a `RawIterRange` covering a subset of a table. + /// + /// The control byte address must be aligned to the group size. + #[cfg_attr(feature = "inline-more", inline)] + unsafe fn new(ctrl: *const u8, data: Bucket<T>, len: usize) -> Self { + debug_assert_ne!(len, 0); + debug_assert_eq!(ctrl as usize % Group::WIDTH, 0); + let end = ctrl.add(len); + + // Load the first group and advance ctrl to point to the next group + let current_group = Group::load_aligned(ctrl).match_full(); + let next_ctrl = ctrl.add(Group::WIDTH); + + Self { + current_group, + data, + next_ctrl, + end, + } + } + + /// Splits a `RawIterRange` into two halves. + /// + /// Returns `None` if the remaining range is smaller than or equal to the + /// group width. + #[cfg_attr(feature = "inline-more", inline)] + #[cfg(feature = "rayon")] + pub(crate) fn split(mut self) -> (Self, Option<RawIterRange<T>>) { + unsafe { + if self.end <= self.next_ctrl { + // Nothing to split if the group that we are current processing + // is the last one. + (self, None) + } else { + // len is the remaining number of elements after the group that + // we are currently processing. It must be a multiple of the + // group size (small tables are caught by the check above). + let len = offset_from(self.end, self.next_ctrl); + debug_assert_eq!(len % Group::WIDTH, 0); + + // Split the remaining elements into two halves, but round the + // midpoint down in case there is an odd number of groups + // remaining. This ensures that: + // - The tail is at least 1 group long. + // - The split is roughly even considering we still have the + // current group to process. + let mid = (len / 2) & !(Group::WIDTH - 1); + + let tail = Self::new( + self.next_ctrl.add(mid), + self.data.next_n(Group::WIDTH).next_n(mid), + len - mid, + ); + debug_assert_eq!( + self.data.next_n(Group::WIDTH).next_n(mid).ptr, + tail.data.ptr + ); + debug_assert_eq!(self.end, tail.end); + self.end = self.next_ctrl.add(mid); + debug_assert_eq!(self.end.add(Group::WIDTH), tail.next_ctrl); + (self, Some(tail)) + } + } + } +} + +// We make raw iterators unconditionally Send and Sync, and let the PhantomData +// in the actual iterator implementations determine the real Send/Sync bounds. +unsafe impl<T> Send for RawIterRange<T> {} +unsafe impl<T> Sync for RawIterRange<T> {} + +impl<T> Clone for RawIterRange<T> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + Self { + data: self.data.clone(), + next_ctrl: self.next_ctrl, + current_group: self.current_group, + end: self.end, + } + } +} + +impl<T> Iterator for RawIterRange<T> { + type Item = Bucket<T>; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<Bucket<T>> { + unsafe { + loop { + if let Some(index) = self.current_group.lowest_set_bit() { + self.current_group = self.current_group.remove_lowest_bit(); + return Some(self.data.next_n(index)); + } + + if self.next_ctrl >= self.end { + return None; + } + + // We might read past self.end up to the next group boundary, + // but this is fine because it only occurs on tables smaller + // than the group size where the trailing control bytes are all + // EMPTY. On larger tables self.end is guaranteed to be aligned + // to the group size (since tables are power-of-two sized). + self.current_group = Group::load_aligned(self.next_ctrl).match_full(); + self.data = self.data.next_n(Group::WIDTH); + self.next_ctrl = self.next_ctrl.add(Group::WIDTH); + } + } + } + + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + // We don't have an item count, so just guess based on the range size. + ( + 0, + Some(unsafe { offset_from(self.end, self.next_ctrl) + Group::WIDTH }), + ) + } +} + +impl<T> FusedIterator for RawIterRange<T> {} + +/// Iterator which returns a raw pointer to every full bucket in the table. +/// +/// For maximum flexibility this iterator is not bound by a lifetime, but you +/// must observe several rules when using it: +/// - You must not free the hash table while iterating (including via growing/shrinking). +/// - It is fine to erase a bucket that has been yielded by the iterator. +/// - Erasing a bucket that has not yet been yielded by the iterator may still +/// result in the iterator yielding that bucket (unless `reflect_remove` is called). +/// - It is unspecified whether an element inserted after the iterator was +/// created will be yielded by that iterator (unless `reflect_insert` is called). +/// - The order in which the iterator yields bucket is unspecified and may +/// change in the future. +pub struct RawIter<T> { + pub(crate) iter: RawIterRange<T>, + items: usize, +} + +impl<T> RawIter<T> { + /// Refresh the iterator so that it reflects a removal from the given bucket. + /// + /// For the iterator to remain valid, this method must be called once + /// for each removed bucket before `next` is called again. + /// + /// This method should be called _before_ the removal is made. It is not necessary to call this + /// method if you are removing an item that this iterator yielded in the past. + #[cfg(feature = "raw")] + pub fn reflect_remove(&mut self, b: &Bucket<T>) { + self.reflect_toggle_full(b, false); + } + + /// Refresh the iterator so that it reflects an insertion into the given bucket. + /// + /// For the iterator to remain valid, this method must be called once + /// for each insert before `next` is called again. + /// + /// This method does not guarantee that an insertion of a bucket witha greater + /// index than the last one yielded will be reflected in the iterator. + /// + /// This method should be called _after_ the given insert is made. + #[cfg(feature = "raw")] + pub fn reflect_insert(&mut self, b: &Bucket<T>) { + self.reflect_toggle_full(b, true); + } + + /// Refresh the iterator so that it reflects a change to the state of the given bucket. + #[cfg(feature = "raw")] + fn reflect_toggle_full(&mut self, b: &Bucket<T>, is_insert: bool) { + unsafe { + if b.as_ptr() > self.iter.data.as_ptr() { + // The iterator has already passed the bucket's group. + // So the toggle isn't relevant to this iterator. + return; + } + + if self.iter.next_ctrl < self.iter.end + && b.as_ptr() <= self.iter.data.next_n(Group::WIDTH).as_ptr() + { + // The iterator has not yet reached the bucket's group. + // We don't need to reload anything, but we do need to adjust the item count. + + if cfg!(debug_assertions) { + // Double-check that the user isn't lying to us by checking the bucket state. + // To do that, we need to find its control byte. We know that self.iter.data is + // at self.iter.next_ctrl - Group::WIDTH, so we work from there: + let offset = offset_from(self.iter.data.as_ptr(), b.as_ptr()); + let ctrl = self.iter.next_ctrl.sub(Group::WIDTH).add(offset); + // This method should be called _before_ a removal, or _after_ an insert, + // so in both cases the ctrl byte should indicate that the bucket is full. + assert!(is_full(*ctrl)); + } + + if is_insert { + self.items += 1; + } else { + self.items -= 1; + } + + return; + } + + // The iterator is at the bucket group that the toggled bucket is in. + // We need to do two things: + // + // - Determine if the iterator already yielded the toggled bucket. + // If it did, we're done. + // - Otherwise, update the iterator cached group so that it won't + // yield a to-be-removed bucket, or _will_ yield a to-be-added bucket. + // We'll also need ot update the item count accordingly. + if let Some(index) = self.iter.current_group.lowest_set_bit() { + let next_bucket = self.iter.data.next_n(index); + if b.as_ptr() > next_bucket.as_ptr() { + // The toggled bucket is "before" the bucket the iterator would yield next. We + // therefore don't need to do anything --- the iterator has already passed the + // bucket in question. + // + // The item count must already be correct, since a removal or insert "prior" to + // the iterator's position wouldn't affect the item count. + } else { + // The removed bucket is an upcoming bucket. We need to make sure it does _not_ + // get yielded, and also that it's no longer included in the item count. + // + // NOTE: We can't just reload the group here, both since that might reflect + // inserts we've already passed, and because that might inadvertently unset the + // bits for _other_ removals. If we do that, we'd have to also decrement the + // item count for those other bits that we unset. But the presumably subsequent + // call to reflect for those buckets might _also_ decrement the item count. + // Instead, we _just_ flip the bit for the particular bucket the caller asked + // us to reflect. + let our_bit = offset_from(self.iter.data.as_ptr(), b.as_ptr()); + let was_full = self.iter.current_group.flip(our_bit); + debug_assert_ne!(was_full, is_insert); + + if is_insert { + self.items += 1; + } else { + self.items -= 1; + } + + if cfg!(debug_assertions) { + if b.as_ptr() == next_bucket.as_ptr() { + // The removed bucket should no longer be next + debug_assert_ne!(self.iter.current_group.lowest_set_bit(), Some(index)); + } else { + // We should not have changed what bucket comes next. + debug_assert_eq!(self.iter.current_group.lowest_set_bit(), Some(index)); + } + } + } + } else { + // We must have already iterated past the removed item. + } + } + } +} + +impl<T> Clone for RawIter<T> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + Self { + iter: self.iter.clone(), + items: self.items, + } + } +} + +impl<T> Iterator for RawIter<T> { + type Item = Bucket<T>; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<Bucket<T>> { + if let Some(b) = self.iter.next() { + self.items -= 1; + Some(b) + } else { + // We don't check against items == 0 here to allow the + // compiler to optimize away the item count entirely if the + // iterator length is never queried. + debug_assert_eq!(self.items, 0); + None + } + } + + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + (self.items, Some(self.items)) + } +} + +impl<T> ExactSizeIterator for RawIter<T> {} +impl<T> FusedIterator for RawIter<T> {} + +/// Iterator which consumes a table and returns elements. +pub struct RawIntoIter<T> { + iter: RawIter<T>, + alloc: Option<(NonNull<u8>, Layout)>, + marker: PhantomData<T>, +} + +impl<T> RawIntoIter<T> { + #[cfg_attr(feature = "inline-more", inline)] + pub fn iter(&self) -> RawIter<T> { + self.iter.clone() + } +} + +unsafe impl<T> Send for RawIntoIter<T> where T: Send {} +unsafe impl<T> Sync for RawIntoIter<T> where T: Sync {} + +#[cfg(feature = "nightly")] +unsafe impl<#[may_dangle] T> Drop for RawIntoIter<T> { + #[cfg_attr(feature = "inline-more", inline)] + fn drop(&mut self) { + unsafe { + // Drop all remaining elements + if mem::needs_drop::<T>() && self.iter.len() != 0 { + while let Some(item) = self.iter.next() { + item.drop(); + } + } + + // Free the table + if let Some((ptr, layout)) = self.alloc { + dealloc(ptr.as_ptr(), layout); + } + } + } +} +#[cfg(not(feature = "nightly"))] +impl<T> Drop for RawIntoIter<T> { + #[cfg_attr(feature = "inline-more", inline)] + fn drop(&mut self) { + unsafe { + // Drop all remaining elements + if mem::needs_drop::<T>() && self.iter.len() != 0 { + while let Some(item) = self.iter.next() { + item.drop(); + } + } + + // Free the table + if let Some((ptr, layout)) = self.alloc { + dealloc(ptr.as_ptr(), layout); + } + } + } +} + +impl<T> Iterator for RawIntoIter<T> { + type Item = T; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<T> { + unsafe { Some(self.iter.next()?.read()) } + } + + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.iter.size_hint() + } +} + +impl<T> ExactSizeIterator for RawIntoIter<T> {} +impl<T> FusedIterator for RawIntoIter<T> {} + +/// Iterator which consumes elements without freeing the table storage. +pub struct RawDrain<'a, T> { + iter: RawIter<T>, + + // The table is moved into the iterator for the duration of the drain. This + // ensures that an empty table is left if the drain iterator is leaked + // without dropping. + table: ManuallyDrop<RawTable<T>>, + orig_table: NonNull<RawTable<T>>, + + // We don't use a &'a mut RawTable<T> because we want RawDrain to be + // covariant over T. + marker: PhantomData<&'a RawTable<T>>, +} + +impl<T> RawDrain<'_, T> { + #[cfg_attr(feature = "inline-more", inline)] + pub fn iter(&self) -> RawIter<T> { + self.iter.clone() + } +} + +unsafe impl<T> Send for RawDrain<'_, T> where T: Send {} +unsafe impl<T> Sync for RawDrain<'_, T> where T: Sync {} + +impl<T> Drop for RawDrain<'_, T> { + #[cfg_attr(feature = "inline-more", inline)] + fn drop(&mut self) { + unsafe { + // Drop all remaining elements. Note that this may panic. + if mem::needs_drop::<T>() && self.iter.len() != 0 { + while let Some(item) = self.iter.next() { + item.drop(); + } + } + + // Reset the contents of the table now that all elements have been + // dropped. + self.table.clear_no_drop(); + + // Move the now empty table back to its original location. + self.orig_table + .as_ptr() + .copy_from_nonoverlapping(&*self.table, 1); + } + } +} + +impl<T> Iterator for RawDrain<'_, T> { + type Item = T; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<T> { + unsafe { + let item = self.iter.next()?; + Some(item.read()) + } + } + + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.iter.size_hint() + } +} + +impl<T> ExactSizeIterator for RawDrain<'_, T> {} +impl<T> FusedIterator for RawDrain<'_, T> {} + +/// Iterator over occupied buckets that could match a given hash. +/// +/// In rare cases, the iterator may return a bucket with a different hash. +pub struct RawIterHash<'a, T> { + table: &'a RawTable<T>, + + // The top 7 bits of the hash. + h2_hash: u8, + + // The sequence of groups to probe in the search. + probe_seq: ProbeSeq, + + // The current group and its position. + pos: usize, + group: Group, + + // The elements within the group with a matching h2-hash. + bitmask: BitMaskIter, +} + +impl<'a, T> RawIterHash<'a, T> { + fn new(table: &'a RawTable<T>, hash: u64) -> Self { + unsafe { + let h2_hash = h2(hash); + let mut probe_seq = table.probe_seq(hash); + let pos = probe_seq.next().unwrap(); + let group = Group::load(table.ctrl(pos)); + let bitmask = group.match_byte(h2_hash).into_iter(); + + RawIterHash { + table, + h2_hash, + probe_seq, + pos, + group, + bitmask, + } + } + } +} + +impl<'a, T> Iterator for RawIterHash<'a, T> { + type Item = Bucket<T>; + + fn next(&mut self) -> Option<Bucket<T>> { + unsafe { + loop { + if let Some(bit) = self.bitmask.next() { + let index = (self.pos + bit) & self.table.bucket_mask; + let bucket = self.table.bucket(index); + return Some(bucket); + } + if likely(self.group.match_empty().any_bit_set()) { + return None; + } + self.pos = self.probe_seq.next().unwrap(); + self.group = Group::load(self.table.ctrl(self.pos)); + self.bitmask = self.group.match_byte(self.h2_hash).into_iter(); + } + } + } +} diff --git a/src/raw/sse2.rs b/src/raw/sse2.rs new file mode 100644 index 0000000..a27bc09 --- /dev/null +++ b/src/raw/sse2.rs @@ -0,0 +1,144 @@ +use super::bitmask::BitMask; +use super::EMPTY; +use core::mem; + +#[cfg(target_arch = "x86")] +use core::arch::x86; +#[cfg(target_arch = "x86_64")] +use core::arch::x86_64 as x86; + +pub type BitMaskWord = u16; +pub const BITMASK_STRIDE: usize = 1; +pub const BITMASK_MASK: BitMaskWord = 0xffff; + +/// Abstraction over a group of control bytes which can be scanned in +/// parallel. +/// +/// This implementation uses a 128-bit SSE value. +#[derive(Copy, Clone)] +pub struct Group(x86::__m128i); + +// FIXME: https://github.com/rust-lang/rust-clippy/issues/3859 +#[allow(clippy::use_self)] +impl Group { + /// Number of bytes in the group. + pub const WIDTH: usize = mem::size_of::<Self>(); + + /// Returns a full group of empty bytes, suitable for use as the initial + /// value for an empty hash table. + /// + /// This is guaranteed to be aligned to the group size. + pub const fn static_empty() -> &'static [u8; Group::WIDTH] { + #[repr(C)] + struct AlignedBytes { + _align: [Group; 0], + bytes: [u8; Group::WIDTH], + }; + const ALIGNED_BYTES: AlignedBytes = AlignedBytes { + _align: [], + bytes: [EMPTY; Group::WIDTH], + }; + &ALIGNED_BYTES.bytes + } + + /// Loads a group of bytes starting at the given address. + #[inline] + #[allow(clippy::cast_ptr_alignment)] // unaligned load + pub unsafe fn load(ptr: *const u8) -> Self { + Group(x86::_mm_loadu_si128(ptr as *const _)) + } + + /// Loads a group of bytes starting at the given address, which must be + /// aligned to `mem::align_of::<Group>()`. + #[inline] + #[allow(clippy::cast_ptr_alignment)] + pub unsafe fn load_aligned(ptr: *const u8) -> Self { + // FIXME: use align_offset once it stabilizes + debug_assert_eq!(ptr as usize & (mem::align_of::<Self>() - 1), 0); + Group(x86::_mm_load_si128(ptr as *const _)) + } + + /// Stores the group of bytes to the given address, which must be + /// aligned to `mem::align_of::<Group>()`. + #[inline] + #[allow(clippy::cast_ptr_alignment)] + pub unsafe fn store_aligned(self, ptr: *mut u8) { + // FIXME: use align_offset once it stabilizes + debug_assert_eq!(ptr as usize & (mem::align_of::<Self>() - 1), 0); + x86::_mm_store_si128(ptr as *mut _, self.0); + } + + /// Returns a `BitMask` indicating all bytes in the group which have + /// the given value. + #[inline] + pub fn match_byte(self, byte: u8) -> BitMask { + #[allow( + clippy::cast_possible_wrap, // byte: u8 as i8 + // byte: i32 as u16 + // note: _mm_movemask_epi8 returns a 16-bit mask in a i32, the + // upper 16-bits of the i32 are zeroed: + clippy::cast_sign_loss, + clippy::cast_possible_truncation + )] + unsafe { + let cmp = x86::_mm_cmpeq_epi8(self.0, x86::_mm_set1_epi8(byte as i8)); + BitMask(x86::_mm_movemask_epi8(cmp) as u16) + } + } + + /// Returns a `BitMask` indicating all bytes in the group which are + /// `EMPTY`. + #[inline] + pub fn match_empty(self) -> BitMask { + self.match_byte(EMPTY) + } + + /// Returns a `BitMask` indicating all bytes in the group which are + /// `EMPTY` or `DELETED`. + #[inline] + pub fn match_empty_or_deleted(self) -> BitMask { + #[allow( + // byte: i32 as u16 + // note: _mm_movemask_epi8 returns a 16-bit mask in a i32, the + // upper 16-bits of the i32 are zeroed: + clippy::cast_sign_loss, + clippy::cast_possible_truncation + )] + unsafe { + // A byte is EMPTY or DELETED iff the high bit is set + BitMask(x86::_mm_movemask_epi8(self.0) as u16) + } + } + + /// Returns a `BitMask` indicating all bytes in the group which are full. + #[inline] + pub fn match_full(&self) -> BitMask { + self.match_empty_or_deleted().invert() + } + + /// Performs the following transformation on all bytes in the group: + /// - `EMPTY => EMPTY` + /// - `DELETED => EMPTY` + /// - `FULL => DELETED` + #[inline] + pub fn convert_special_to_empty_and_full_to_deleted(self) -> Self { + // Map high_bit = 1 (EMPTY or DELETED) to 1111_1111 + // and high_bit = 0 (FULL) to 1000_0000 + // + // Here's this logic expanded to concrete values: + // let special = 0 > byte = 1111_1111 (true) or 0000_0000 (false) + // 1111_1111 | 1000_0000 = 1111_1111 + // 0000_0000 | 1000_0000 = 1000_0000 + #[allow( + clippy::cast_possible_wrap, // byte: 0x80_u8 as i8 + )] + unsafe { + let zero = x86::_mm_setzero_si128(); + let special = x86::_mm_cmpgt_epi8(zero, self.0); + Group(x86::_mm_or_si128( + special, + x86::_mm_set1_epi8(0x80_u8 as i8), + )) + } + } +} diff --git a/src/rustc_entry.rs b/src/rustc_entry.rs new file mode 100644 index 0000000..b6ea7bc --- /dev/null +++ b/src/rustc_entry.rs @@ -0,0 +1,618 @@ +use self::RustcEntry::*; +use crate::map::{make_hash, Drain, HashMap, IntoIter, Iter, IterMut}; +use crate::raw::{Bucket, RawTable}; +use core::fmt::{self, Debug}; +use core::hash::{BuildHasher, Hash}; +use core::mem; + +impl<K, V, S> HashMap<K, V, S> +where + K: Eq + Hash, + S: BuildHasher, +{ + /// Gets the given key's corresponding entry in the map for in-place manipulation. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut letters = HashMap::new(); + /// + /// for ch in "a short treatise on fungi".chars() { + /// let counter = letters.rustc_entry(ch).or_insert(0); + /// *counter += 1; + /// } + /// + /// assert_eq!(letters[&'s'], 2); + /// assert_eq!(letters[&'t'], 3); + /// assert_eq!(letters[&'u'], 1); + /// assert_eq!(letters.get(&'y'), None); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn rustc_entry(&mut self, key: K) -> RustcEntry<'_, K, V> { + let hash = make_hash(&self.hash_builder, &key); + if let Some(elem) = self.table.find(hash, |q| q.0.eq(&key)) { + RustcEntry::Occupied(RustcOccupiedEntry { + key: Some(key), + elem, + table: &mut self.table, + }) + } else { + // Ideally we would put this in VacantEntry::insert, but Entry is not + // generic over the BuildHasher and adding a generic parameter would be + // a breaking change. + self.reserve(1); + + RustcEntry::Vacant(RustcVacantEntry { + hash, + key, + table: &mut self.table, + }) + } + } +} + +/// A view into a single entry in a map, which may either be vacant or occupied. +/// +/// This `enum` is constructed from the [`entry`] method on [`HashMap`]. +/// +/// [`HashMap`]: struct.HashMap.html +/// [`entry`]: struct.HashMap.html#method.rustc_entry +pub enum RustcEntry<'a, K, V> { + /// An occupied entry. + Occupied(RustcOccupiedEntry<'a, K, V>), + + /// A vacant entry. + Vacant(RustcVacantEntry<'a, K, V>), +} + +impl<K: Debug, V: Debug> Debug for RustcEntry<'_, K, V> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + match *self { + Vacant(ref v) => f.debug_tuple("Entry").field(v).finish(), + Occupied(ref o) => f.debug_tuple("Entry").field(o).finish(), + } + } +} + +/// A view into an occupied entry in a `HashMap`. +/// It is part of the [`RustcEntry`] enum. +/// +/// [`RustcEntry`]: enum.RustcEntry.html +pub struct RustcOccupiedEntry<'a, K, V> { + key: Option<K>, + elem: Bucket<(K, V)>, + table: &'a mut RawTable<(K, V)>, +} + +unsafe impl<K, V> Send for RustcOccupiedEntry<'_, K, V> +where + K: Send, + V: Send, +{ +} +unsafe impl<K, V> Sync for RustcOccupiedEntry<'_, K, V> +where + K: Sync, + V: Sync, +{ +} + +impl<K: Debug, V: Debug> Debug for RustcOccupiedEntry<'_, K, V> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("OccupiedEntry") + .field("key", self.key()) + .field("value", self.get()) + .finish() + } +} + +/// A view into a vacant entry in a `HashMap`. +/// It is part of the [`RustcEntry`] enum. +/// +/// [`RustcEntry`]: enum.RustcEntry.html +pub struct RustcVacantEntry<'a, K, V> { + hash: u64, + key: K, + table: &'a mut RawTable<(K, V)>, +} + +impl<K: Debug, V> Debug for RustcVacantEntry<'_, K, V> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_tuple("VacantEntry").field(self.key()).finish() + } +} + +impl<'a, K, V> RustcEntry<'a, K, V> { + /// Sets the value of the entry, and returns a RustcOccupiedEntry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// let entry = map.entry("horseyland").insert(37); + /// + /// assert_eq!(entry.key(), &"horseyland"); + /// ``` + pub fn insert(self, value: V) -> RustcOccupiedEntry<'a, K, V> { + match self { + Vacant(entry) => entry.insert_entry(value), + Occupied(mut entry) => { + entry.insert(value); + entry + } + } + } + + /// Ensures a value is in the entry by inserting the default if empty, and returns + /// a mutable reference to the value in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// map.rustc_entry("poneyland").or_insert(3); + /// assert_eq!(map["poneyland"], 3); + /// + /// *map.rustc_entry("poneyland").or_insert(10) *= 2; + /// assert_eq!(map["poneyland"], 6); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn or_insert(self, default: V) -> &'a mut V + where + K: Hash, + { + match self { + Occupied(entry) => entry.into_mut(), + Vacant(entry) => entry.insert(default), + } + } + + /// Ensures a value is in the entry by inserting the result of the default function if empty, + /// and returns a mutable reference to the value in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, String> = HashMap::new(); + /// let s = "hoho".to_string(); + /// + /// map.rustc_entry("poneyland").or_insert_with(|| s); + /// + /// assert_eq!(map["poneyland"], "hoho".to_string()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn or_insert_with<F: FnOnce() -> V>(self, default: F) -> &'a mut V + where + K: Hash, + { + match self { + Occupied(entry) => entry.into_mut(), + Vacant(entry) => entry.insert(default()), + } + } + + /// Returns a reference to this entry's key. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// assert_eq!(map.rustc_entry("poneyland").key(), &"poneyland"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn key(&self) -> &K { + match *self { + Occupied(ref entry) => entry.key(), + Vacant(ref entry) => entry.key(), + } + } + + /// Provides in-place mutable access to an occupied entry before any + /// potential inserts into the map. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// map.rustc_entry("poneyland") + /// .and_modify(|e| { *e += 1 }) + /// .or_insert(42); + /// assert_eq!(map["poneyland"], 42); + /// + /// map.rustc_entry("poneyland") + /// .and_modify(|e| { *e += 1 }) + /// .or_insert(42); + /// assert_eq!(map["poneyland"], 43); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn and_modify<F>(self, f: F) -> Self + where + F: FnOnce(&mut V), + { + match self { + Occupied(mut entry) => { + f(entry.get_mut()); + Occupied(entry) + } + Vacant(entry) => Vacant(entry), + } + } +} + +impl<'a, K, V: Default> RustcEntry<'a, K, V> { + /// Ensures a value is in the entry by inserting the default value if empty, + /// and returns a mutable reference to the value in the entry. + /// + /// # Examples + /// + /// ``` + /// # fn main() { + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, Option<u32>> = HashMap::new(); + /// map.rustc_entry("poneyland").or_default(); + /// + /// assert_eq!(map["poneyland"], None); + /// # } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn or_default(self) -> &'a mut V + where + K: Hash, + { + match self { + Occupied(entry) => entry.into_mut(), + Vacant(entry) => entry.insert(Default::default()), + } + } +} + +impl<'a, K, V> RustcOccupiedEntry<'a, K, V> { + /// Gets a reference to the key in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.rustc_entry("poneyland").or_insert(12); + /// assert_eq!(map.rustc_entry("poneyland").key(), &"poneyland"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn key(&self) -> &K { + unsafe { &self.elem.as_ref().0 } + } + + /// Take the ownership of the key and value from the map. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::RustcEntry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.rustc_entry("poneyland").or_insert(12); + /// + /// if let RustcEntry::Occupied(o) = map.rustc_entry("poneyland") { + /// // We delete the entry from the map. + /// o.remove_entry(); + /// } + /// + /// assert_eq!(map.contains_key("poneyland"), false); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn remove_entry(self) -> (K, V) { + unsafe { self.table.remove(self.elem) } + } + + /// Gets a reference to the value in the entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::RustcEntry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.rustc_entry("poneyland").or_insert(12); + /// + /// if let RustcEntry::Occupied(o) = map.rustc_entry("poneyland") { + /// assert_eq!(o.get(), &12); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn get(&self) -> &V { + unsafe { &self.elem.as_ref().1 } + } + + /// Gets a mutable reference to the value in the entry. + /// + /// If you need a reference to the `RustcOccupiedEntry` which may outlive the + /// destruction of the `RustcEntry` value, see [`into_mut`]. + /// + /// [`into_mut`]: #method.into_mut + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::RustcEntry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.rustc_entry("poneyland").or_insert(12); + /// + /// assert_eq!(map["poneyland"], 12); + /// if let RustcEntry::Occupied(mut o) = map.rustc_entry("poneyland") { + /// *o.get_mut() += 10; + /// assert_eq!(*o.get(), 22); + /// + /// // We can use the same RustcEntry multiple times. + /// *o.get_mut() += 2; + /// } + /// + /// assert_eq!(map["poneyland"], 24); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn get_mut(&mut self) -> &mut V { + unsafe { &mut self.elem.as_mut().1 } + } + + /// Converts the RustcOccupiedEntry into a mutable reference to the value in the entry + /// with a lifetime bound to the map itself. + /// + /// If you need multiple references to the `RustcOccupiedEntry`, see [`get_mut`]. + /// + /// [`get_mut`]: #method.get_mut + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::RustcEntry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.rustc_entry("poneyland").or_insert(12); + /// + /// assert_eq!(map["poneyland"], 12); + /// if let RustcEntry::Occupied(o) = map.rustc_entry("poneyland") { + /// *o.into_mut() += 10; + /// } + /// + /// assert_eq!(map["poneyland"], 22); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn into_mut(self) -> &'a mut V { + unsafe { &mut self.elem.as_mut().1 } + } + + /// Sets the value of the entry, and returns the entry's old value. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::RustcEntry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.rustc_entry("poneyland").or_insert(12); + /// + /// if let RustcEntry::Occupied(mut o) = map.rustc_entry("poneyland") { + /// assert_eq!(o.insert(15), 12); + /// } + /// + /// assert_eq!(map["poneyland"], 15); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(&mut self, mut value: V) -> V { + let old_value = self.get_mut(); + mem::swap(&mut value, old_value); + value + } + + /// Takes the value out of the entry, and returns it. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::RustcEntry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// map.rustc_entry("poneyland").or_insert(12); + /// + /// if let RustcEntry::Occupied(o) = map.rustc_entry("poneyland") { + /// assert_eq!(o.remove(), 12); + /// } + /// + /// assert_eq!(map.contains_key("poneyland"), false); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn remove(self) -> V { + self.remove_entry().1 + } + + /// Replaces the entry, returning the old key and value. The new key in the hash map will be + /// the key used to create this entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::hash_map::{RustcEntry, HashMap}; + /// use std::rc::Rc; + /// + /// let mut map: HashMap<Rc<String>, u32> = HashMap::new(); + /// map.insert(Rc::new("Stringthing".to_string()), 15); + /// + /// let my_key = Rc::new("Stringthing".to_string()); + /// + /// if let RustcEntry::Occupied(entry) = map.rustc_entry(my_key) { + /// // Also replace the key with a handle to our other key. + /// let (old_key, old_value): (Rc<String>, u32) = entry.replace_entry(16); + /// } + /// + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn replace_entry(self, value: V) -> (K, V) { + let entry = unsafe { self.elem.as_mut() }; + + let old_key = mem::replace(&mut entry.0, self.key.unwrap()); + let old_value = mem::replace(&mut entry.1, value); + + (old_key, old_value) + } + + /// Replaces the key in the hash map with the key used to create this entry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::hash_map::{RustcEntry, HashMap}; + /// use std::rc::Rc; + /// + /// let mut map: HashMap<Rc<String>, u32> = HashMap::new(); + /// let mut known_strings: Vec<Rc<String>> = Vec::new(); + /// + /// // Initialise known strings, run program, etc. + /// + /// reclaim_memory(&mut map, &known_strings); + /// + /// fn reclaim_memory(map: &mut HashMap<Rc<String>, u32>, known_strings: &[Rc<String>] ) { + /// for s in known_strings { + /// if let RustcEntry::Occupied(entry) = map.rustc_entry(s.clone()) { + /// // Replaces the entry's key with our version of it in `known_strings`. + /// entry.replace_key(); + /// } + /// } + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn replace_key(self) -> K { + let entry = unsafe { self.elem.as_mut() }; + mem::replace(&mut entry.0, self.key.unwrap()) + } +} + +impl<'a, K, V> RustcVacantEntry<'a, K, V> { + /// Gets a reference to the key that would be used when inserting a value + /// through the `RustcVacantEntry`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// assert_eq!(map.rustc_entry("poneyland").key(), &"poneyland"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn key(&self) -> &K { + &self.key + } + + /// Take ownership of the key. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::RustcEntry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// if let RustcEntry::Vacant(v) = map.rustc_entry("poneyland") { + /// v.into_key(); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn into_key(self) -> K { + self.key + } + + /// Sets the value of the entry with the RustcVacantEntry's key, + /// and returns a mutable reference to it. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::RustcEntry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// if let RustcEntry::Vacant(o) = map.rustc_entry("poneyland") { + /// o.insert(37); + /// } + /// assert_eq!(map["poneyland"], 37); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(self, value: V) -> &'a mut V { + let bucket = self.table.insert_no_grow(self.hash, (self.key, value)); + unsafe { &mut bucket.as_mut().1 } + } + + /// Sets the value of the entry with the RustcVacantEntry's key, + /// and returns a RustcOccupiedEntry. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashMap; + /// use hashbrown::hash_map::RustcEntry; + /// + /// let mut map: HashMap<&str, u32> = HashMap::new(); + /// + /// if let RustcEntry::Vacant(v) = map.rustc_entry("poneyland") { + /// let o = v.insert_entry(37); + /// assert_eq!(o.get(), &37); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert_entry(self, value: V) -> RustcOccupiedEntry<'a, K, V> { + let bucket = self.table.insert_no_grow(self.hash, (self.key, value)); + RustcOccupiedEntry { + key: None, + elem: bucket, + table: self.table, + } + } +} + +impl<K, V> IterMut<'_, K, V> { + /// Returns a iterator of references over the remaining items. + #[cfg_attr(feature = "inline-more", inline)] + pub fn rustc_iter(&self) -> Iter<'_, K, V> { + self.iter() + } +} + +impl<K, V> IntoIter<K, V> { + /// Returns a iterator of references over the remaining items. + #[cfg_attr(feature = "inline-more", inline)] + pub fn rustc_iter(&self) -> Iter<'_, K, V> { + self.iter() + } +} + +impl<K, V> Drain<'_, K, V> { + /// Returns a iterator of references over the remaining items. + #[cfg_attr(feature = "inline-more", inline)] + pub fn rustc_iter(&self) -> Iter<'_, K, V> { + self.iter() + } +} diff --git a/src/scopeguard.rs b/src/scopeguard.rs new file mode 100644 index 0000000..32c9694 --- /dev/null +++ b/src/scopeguard.rs @@ -0,0 +1,49 @@ +// Extracted from the scopeguard crate +use core::ops::{Deref, DerefMut}; + +pub struct ScopeGuard<T, F> +where + F: FnMut(&mut T), +{ + dropfn: F, + value: T, +} + +#[cfg_attr(feature = "inline-more", inline)] +pub fn guard<T, F>(value: T, dropfn: F) -> ScopeGuard<T, F> +where + F: FnMut(&mut T), +{ + ScopeGuard { dropfn, value } +} + +impl<T, F> Deref for ScopeGuard<T, F> +where + F: FnMut(&mut T), +{ + type Target = T; + #[cfg_attr(feature = "inline-more", inline)] + fn deref(&self) -> &T { + &self.value + } +} + +impl<T, F> DerefMut for ScopeGuard<T, F> +where + F: FnMut(&mut T), +{ + #[cfg_attr(feature = "inline-more", inline)] + fn deref_mut(&mut self) -> &mut T { + &mut self.value + } +} + +impl<T, F> Drop for ScopeGuard<T, F> +where + F: FnMut(&mut T), +{ + #[cfg_attr(feature = "inline-more", inline)] + fn drop(&mut self) { + (self.dropfn)(&mut self.value) + } +} diff --git a/src/set.rs b/src/set.rs new file mode 100644 index 0000000..b8460fd --- /dev/null +++ b/src/set.rs @@ -0,0 +1,2119 @@ +use crate::TryReserveError; +use alloc::borrow::ToOwned; +use core::borrow::Borrow; +use core::fmt; +use core::hash::{BuildHasher, Hash}; +use core::iter::{Chain, FromIterator, FusedIterator}; +use core::mem; +use core::ops::{BitAnd, BitOr, BitXor, Sub}; + +use super::map::{self, ConsumeAllOnDrop, DefaultHashBuilder, DrainFilterInner, HashMap, Keys}; + +// Future Optimization (FIXME!) +// ============================= +// +// Iteration over zero sized values is a noop. There is no need +// for `bucket.val` in the case of HashSet. I suppose we would need HKT +// to get rid of it properly. + +/// A hash set implemented as a `HashMap` where the value is `()`. +/// +/// As with the [`HashMap`] type, a `HashSet` requires that the elements +/// implement the [`Eq`] and [`Hash`] traits. This can frequently be achieved by +/// using `#[derive(PartialEq, Eq, Hash)]`. If you implement these yourself, +/// it is important that the following property holds: +/// +/// ```text +/// k1 == k2 -> hash(k1) == hash(k2) +/// ``` +/// +/// In other words, if two keys are equal, their hashes must be equal. +/// +/// +/// It is a logic error for an item to be modified in such a way that the +/// item's hash, as determined by the [`Hash`] trait, or its equality, as +/// determined by the [`Eq`] trait, changes while it is in the set. This is +/// normally only possible through [`Cell`], [`RefCell`], global state, I/O, or +/// unsafe code. +/// +/// It is also a logic error for the [`Hash`] implementation of a key to panic. +/// This is generally only possible if the trait is implemented manually. If a +/// panic does occur then the contents of the `HashSet` may become corrupted and +/// some items may be dropped from the table. +/// +/// # Examples +/// +/// ``` +/// use hashbrown::HashSet; +/// // Type inference lets us omit an explicit type signature (which +/// // would be `HashSet<String>` in this example). +/// let mut books = HashSet::new(); +/// +/// // Add some books. +/// books.insert("A Dance With Dragons".to_string()); +/// books.insert("To Kill a Mockingbird".to_string()); +/// books.insert("The Odyssey".to_string()); +/// books.insert("The Great Gatsby".to_string()); +/// +/// // Check for a specific one. +/// if !books.contains("The Winds of Winter") { +/// println!("We have {} books, but The Winds of Winter ain't one.", +/// books.len()); +/// } +/// +/// // Remove a book. +/// books.remove("The Odyssey"); +/// +/// // Iterate over everything. +/// for book in &books { +/// println!("{}", book); +/// } +/// ``` +/// +/// The easiest way to use `HashSet` with a custom type is to derive +/// [`Eq`] and [`Hash`]. We must also derive [`PartialEq`], this will in the +/// future be implied by [`Eq`]. +/// +/// ``` +/// use hashbrown::HashSet; +/// #[derive(Hash, Eq, PartialEq, Debug)] +/// struct Viking { +/// name: String, +/// power: usize, +/// } +/// +/// let mut vikings = HashSet::new(); +/// +/// vikings.insert(Viking { name: "Einar".to_string(), power: 9 }); +/// vikings.insert(Viking { name: "Einar".to_string(), power: 9 }); +/// vikings.insert(Viking { name: "Olaf".to_string(), power: 4 }); +/// vikings.insert(Viking { name: "Harald".to_string(), power: 8 }); +/// +/// // Use derived implementation to print the vikings. +/// for x in &vikings { +/// println!("{:?}", x); +/// } +/// ``` +/// +/// A `HashSet` with fixed list of elements can be initialized from an array: +/// +/// ``` +/// use hashbrown::HashSet; +/// +/// let viking_names: HashSet<&'static str> = +/// [ "Einar", "Olaf", "Harald" ].iter().cloned().collect(); +/// // use the values stored in the set +/// ``` +/// +/// [`Cell`]: https://doc.rust-lang.org/std/cell/struct.Cell.html +/// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html +/// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html +/// [`HashMap`]: struct.HashMap.html +/// [`PartialEq`]: https://doc.rust-lang.org/std/cmp/trait.PartialEq.html +/// [`RefCell`]: https://doc.rust-lang.org/std/cell/struct.RefCell.html +pub struct HashSet<T, S = DefaultHashBuilder> { + pub(crate) map: HashMap<T, (), S>, +} + +impl<T: Clone, S: Clone> Clone for HashSet<T, S> { + fn clone(&self) -> Self { + HashSet { + map: self.map.clone(), + } + } + + fn clone_from(&mut self, source: &Self) { + self.map.clone_from(&source.map); + } +} + +#[cfg(feature = "ahash")] +impl<T> HashSet<T, DefaultHashBuilder> { + /// Creates an empty `HashSet`. + /// + /// The hash set is initially created with a capacity of 0, so it will not allocate until it + /// is first inserted into. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let set: HashSet<i32> = HashSet::new(); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn new() -> Self { + Self { + map: HashMap::new(), + } + } + + /// Creates an empty `HashSet` with the specified capacity. + /// + /// The hash set will be able to hold at least `capacity` elements without + /// reallocating. If `capacity` is 0, the hash set will not allocate. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let set: HashSet<i32> = HashSet::with_capacity(10); + /// assert!(set.capacity() >= 10); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn with_capacity(capacity: usize) -> Self { + Self { + map: HashMap::with_capacity(capacity), + } + } +} + +impl<T, S> HashSet<T, S> { + /// Creates a new empty hash set which will use the given hasher to hash + /// keys. + /// + /// The hash set is also created with the default initial capacity. + /// + /// Warning: `hasher` is normally randomly generated, and + /// is designed to allow `HashSet`s to be resistant to attacks that + /// cause many collisions and very poor performance. Setting it + /// manually using this function can expose a DoS attack vector. + /// + /// The `hash_builder` passed should implement the [`BuildHasher`] trait for + /// the HashMap to be useful, see its documentation for details. + /// + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// use hashbrown::hash_map::DefaultHashBuilder; + /// + /// let s = DefaultHashBuilder::default(); + /// let mut set = HashSet::with_hasher(s); + /// set.insert(2); + /// ``` + /// + /// [`BuildHasher`]: ../../std/hash/trait.BuildHasher.html + #[cfg_attr(feature = "inline-more", inline)] + pub const fn with_hasher(hasher: S) -> Self { + Self { + map: HashMap::with_hasher(hasher), + } + } + + /// Creates an empty `HashSet` with the specified capacity, using + /// `hasher` to hash the keys. + /// + /// The hash set will be able to hold at least `capacity` elements without + /// reallocating. If `capacity` is 0, the hash set will not allocate. + /// + /// Warning: `hasher` is normally randomly generated, and + /// is designed to allow `HashSet`s to be resistant to attacks that + /// cause many collisions and very poor performance. Setting it + /// manually using this function can expose a DoS attack vector. + /// + /// The `hash_builder` passed should implement the [`BuildHasher`] trait for + /// the HashMap to be useful, see its documentation for details. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// use hashbrown::hash_map::DefaultHashBuilder; + /// + /// let s = DefaultHashBuilder::default(); + /// let mut set = HashSet::with_capacity_and_hasher(10, s); + /// set.insert(1); + /// ``` + /// + /// [`BuildHasher`]: ../../std/hash/trait.BuildHasher.html + #[cfg_attr(feature = "inline-more", inline)] + pub fn with_capacity_and_hasher(capacity: usize, hasher: S) -> Self { + Self { + map: HashMap::with_capacity_and_hasher(capacity, hasher), + } + } + + /// Returns the number of elements the set can hold without reallocating. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let set: HashSet<i32> = HashSet::with_capacity(100); + /// assert!(set.capacity() >= 100); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn capacity(&self) -> usize { + self.map.capacity() + } + + /// An iterator visiting all elements in arbitrary order. + /// The iterator element type is `&'a T`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let mut set = HashSet::new(); + /// set.insert("a"); + /// set.insert("b"); + /// + /// // Will print in an arbitrary order. + /// for x in set.iter() { + /// println!("{}", x); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn iter(&self) -> Iter<'_, T> { + Iter { + iter: self.map.keys(), + } + } + + /// Returns the number of elements in the set. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut v = HashSet::new(); + /// assert_eq!(v.len(), 0); + /// v.insert(1); + /// assert_eq!(v.len(), 1); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn len(&self) -> usize { + self.map.len() + } + + /// Returns `true` if the set contains no elements. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut v = HashSet::new(); + /// assert!(v.is_empty()); + /// v.insert(1); + /// assert!(!v.is_empty()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn is_empty(&self) -> bool { + self.map.is_empty() + } + + /// Clears the set, returning all elements in an iterator. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// assert!(!set.is_empty()); + /// + /// // print 1, 2, 3 in an arbitrary order + /// for i in set.drain() { + /// println!("{}", i); + /// } + /// + /// assert!(set.is_empty()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn drain(&mut self) -> Drain<'_, T> { + Drain { + iter: self.map.drain(), + } + } + + /// Retains only the elements specified by the predicate. + /// + /// In other words, remove all elements `e` such that `f(&e)` returns `false`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let xs = [1,2,3,4,5,6]; + /// let mut set: HashSet<i32> = xs.iter().cloned().collect(); + /// set.retain(|&k| k % 2 == 0); + /// assert_eq!(set.len(), 3); + /// ``` + pub fn retain<F>(&mut self, mut f: F) + where + F: FnMut(&T) -> bool, + { + self.map.retain(|k, _| f(k)); + } + + /// Drains elements which are true under the given predicate, + /// and returns an iterator over the removed items. + /// + /// In other words, move all elements `e` such that `f(&e)` returns `true` out + /// into another iterator. + /// + /// When the returned DrainedFilter is dropped, any remaining elements that satisfy + /// the predicate are dropped from the set. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set: HashSet<i32> = (0..8).collect(); + /// let drained: HashSet<i32> = set.drain_filter(|v| v % 2 == 0).collect(); + /// + /// let mut evens = drained.into_iter().collect::<Vec<_>>(); + /// let mut odds = set.into_iter().collect::<Vec<_>>(); + /// evens.sort(); + /// odds.sort(); + /// + /// assert_eq!(evens, vec![0, 2, 4, 6]); + /// assert_eq!(odds, vec![1, 3, 5, 7]); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn drain_filter<F>(&mut self, f: F) -> DrainFilter<'_, T, F> + where + F: FnMut(&T) -> bool, + { + DrainFilter { + f, + inner: DrainFilterInner { + iter: unsafe { self.map.table.iter() }, + table: &mut self.map.table, + }, + } + } + + /// Clears the set, removing all values. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut v = HashSet::new(); + /// v.insert(1); + /// v.clear(); + /// assert!(v.is_empty()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn clear(&mut self) { + self.map.clear() + } + + /// Returns a reference to the set's [`BuildHasher`]. + /// + /// [`BuildHasher`]: https://doc.rust-lang.org/std/hash/trait.BuildHasher.html + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// use hashbrown::hash_map::DefaultHashBuilder; + /// + /// let hasher = DefaultHashBuilder::default(); + /// let set: HashSet<i32> = HashSet::with_hasher(hasher); + /// let hasher: &DefaultHashBuilder = set.hasher(); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn hasher(&self) -> &S { + self.map.hasher() + } +} + +impl<T, S> HashSet<T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ + /// Reserves capacity for at least `additional` more elements to be inserted + /// in the `HashSet`. The collection may reserve more space to avoid + /// frequent reallocations. + /// + /// # Panics + /// + /// Panics if the new allocation size overflows `usize`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let mut set: HashSet<i32> = HashSet::new(); + /// set.reserve(10); + /// assert!(set.capacity() >= 10); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn reserve(&mut self, additional: usize) { + self.map.reserve(additional) + } + + /// Tries to reserve capacity for at least `additional` more elements to be inserted + /// in the given `HashSet<K,V>`. The collection may reserve more space to avoid + /// frequent reallocations. + /// + /// # Errors + /// + /// If the capacity overflows, or the allocator reports a failure, then an error + /// is returned. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let mut set: HashSet<i32> = HashSet::new(); + /// set.try_reserve(10).expect("why is the test harness OOMing on 10 bytes?"); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn try_reserve(&mut self, additional: usize) -> Result<(), TryReserveError> { + self.map.try_reserve(additional) + } + + /// Shrinks the capacity of the set as much as possible. It will drop + /// down as much as possible while maintaining the internal rules + /// and possibly leaving some space in accordance with the resize policy. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set = HashSet::with_capacity(100); + /// set.insert(1); + /// set.insert(2); + /// assert!(set.capacity() >= 100); + /// set.shrink_to_fit(); + /// assert!(set.capacity() >= 2); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn shrink_to_fit(&mut self) { + self.map.shrink_to_fit() + } + + /// Shrinks the capacity of the set with a lower limit. It will drop + /// down no lower than the supplied limit while maintaining the internal rules + /// and possibly leaving some space in accordance with the resize policy. + /// + /// Panics if the current capacity is smaller than the supplied + /// minimum capacity. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set = HashSet::with_capacity(100); + /// set.insert(1); + /// set.insert(2); + /// assert!(set.capacity() >= 100); + /// set.shrink_to(10); + /// assert!(set.capacity() >= 10); + /// set.shrink_to(0); + /// assert!(set.capacity() >= 2); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn shrink_to(&mut self, min_capacity: usize) { + self.map.shrink_to(min_capacity) + } + + /// Visits the values representing the difference, + /// i.e., the values that are in `self` but not in `other`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let a: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// let b: HashSet<_> = [4, 2, 3, 4].iter().cloned().collect(); + /// + /// // Can be seen as `a - b`. + /// for x in a.difference(&b) { + /// println!("{}", x); // Print 1 + /// } + /// + /// let diff: HashSet<_> = a.difference(&b).collect(); + /// assert_eq!(diff, [1].iter().collect()); + /// + /// // Note that difference is not symmetric, + /// // and `b - a` means something else: + /// let diff: HashSet<_> = b.difference(&a).collect(); + /// assert_eq!(diff, [4].iter().collect()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn difference<'a>(&'a self, other: &'a Self) -> Difference<'a, T, S> { + Difference { + iter: self.iter(), + other, + } + } + + /// Visits the values representing the symmetric difference, + /// i.e., the values that are in `self` or in `other` but not in both. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let a: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// let b: HashSet<_> = [4, 2, 3, 4].iter().cloned().collect(); + /// + /// // Print 1, 4 in arbitrary order. + /// for x in a.symmetric_difference(&b) { + /// println!("{}", x); + /// } + /// + /// let diff1: HashSet<_> = a.symmetric_difference(&b).collect(); + /// let diff2: HashSet<_> = b.symmetric_difference(&a).collect(); + /// + /// assert_eq!(diff1, diff2); + /// assert_eq!(diff1, [1, 4].iter().collect()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn symmetric_difference<'a>(&'a self, other: &'a Self) -> SymmetricDifference<'a, T, S> { + SymmetricDifference { + iter: self.difference(other).chain(other.difference(self)), + } + } + + /// Visits the values representing the intersection, + /// i.e., the values that are both in `self` and `other`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let a: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// let b: HashSet<_> = [4, 2, 3, 4].iter().cloned().collect(); + /// + /// // Print 2, 3 in arbitrary order. + /// for x in a.intersection(&b) { + /// println!("{}", x); + /// } + /// + /// let intersection: HashSet<_> = a.intersection(&b).collect(); + /// assert_eq!(intersection, [2, 3].iter().collect()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn intersection<'a>(&'a self, other: &'a Self) -> Intersection<'a, T, S> { + let (smaller, larger) = if self.len() <= other.len() { + (self, other) + } else { + (other, self) + }; + Intersection { + iter: smaller.iter(), + other: larger, + } + } + + /// Visits the values representing the union, + /// i.e., all the values in `self` or `other`, without duplicates. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let a: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// let b: HashSet<_> = [4, 2, 3, 4].iter().cloned().collect(); + /// + /// // Print 1, 2, 3, 4 in arbitrary order. + /// for x in a.union(&b) { + /// println!("{}", x); + /// } + /// + /// let union: HashSet<_> = a.union(&b).collect(); + /// assert_eq!(union, [1, 2, 3, 4].iter().collect()); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn union<'a>(&'a self, other: &'a Self) -> Union<'a, T, S> { + let (smaller, larger) = if self.len() >= other.len() { + (self, other) + } else { + (other, self) + }; + Union { + iter: larger.iter().chain(smaller.difference(larger)), + } + } + + /// Returns `true` if the set contains a value. + /// + /// The value may be any borrowed form of the set's value type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the value type. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let set: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// assert_eq!(set.contains(&1), true); + /// assert_eq!(set.contains(&4), false); + /// ``` + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + #[cfg_attr(feature = "inline-more", inline)] + pub fn contains<Q: ?Sized>(&self, value: &Q) -> bool + where + T: Borrow<Q>, + Q: Hash + Eq, + { + self.map.contains_key(value) + } + + /// Returns a reference to the value in the set, if any, that is equal to the given value. + /// + /// The value may be any borrowed form of the set's value type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the value type. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let set: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// assert_eq!(set.get(&2), Some(&2)); + /// assert_eq!(set.get(&4), None); + /// ``` + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + #[cfg_attr(feature = "inline-more", inline)] + pub fn get<Q: ?Sized>(&self, value: &Q) -> Option<&T> + where + T: Borrow<Q>, + Q: Hash + Eq, + { + // Avoid `Option::map` because it bloats LLVM IR. + match self.map.get_key_value(value) { + Some((k, _)) => Some(k), + None => None, + } + } + + /// Inserts the given `value` into the set if it is not present, then + /// returns a reference to the value in the set. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// assert_eq!(set.len(), 3); + /// assert_eq!(set.get_or_insert(2), &2); + /// assert_eq!(set.get_or_insert(100), &100); + /// assert_eq!(set.len(), 4); // 100 was inserted + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn get_or_insert(&mut self, value: T) -> &T { + // Although the raw entry gives us `&mut T`, we only return `&T` to be consistent with + // `get`. Key mutation is "raw" because you're not supposed to affect `Eq` or `Hash`. + self.map + .raw_entry_mut() + .from_key(&value) + .or_insert(value, ()) + .0 + } + + /// Inserts an owned copy of the given `value` into the set if it is not + /// present, then returns a reference to the value in the set. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set: HashSet<String> = ["cat", "dog", "horse"] + /// .iter().map(|&pet| pet.to_owned()).collect(); + /// + /// assert_eq!(set.len(), 3); + /// for &pet in &["cat", "dog", "fish"] { + /// let value = set.get_or_insert_owned(pet); + /// assert_eq!(value, pet); + /// } + /// assert_eq!(set.len(), 4); // a new "fish" was inserted + /// ``` + #[inline] + pub fn get_or_insert_owned<Q: ?Sized>(&mut self, value: &Q) -> &T + where + T: Borrow<Q>, + Q: Hash + Eq + ToOwned<Owned = T>, + { + // Although the raw entry gives us `&mut T`, we only return `&T` to be consistent with + // `get`. Key mutation is "raw" because you're not supposed to affect `Eq` or `Hash`. + self.map + .raw_entry_mut() + .from_key(value) + .or_insert_with(|| (value.to_owned(), ())) + .0 + } + + /// Inserts a value computed from `f` into the set if the given `value` is + /// not present, then returns a reference to the value in the set. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set: HashSet<String> = ["cat", "dog", "horse"] + /// .iter().map(|&pet| pet.to_owned()).collect(); + /// + /// assert_eq!(set.len(), 3); + /// for &pet in &["cat", "dog", "fish"] { + /// let value = set.get_or_insert_with(pet, str::to_owned); + /// assert_eq!(value, pet); + /// } + /// assert_eq!(set.len(), 4); // a new "fish" was inserted + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn get_or_insert_with<Q: ?Sized, F>(&mut self, value: &Q, f: F) -> &T + where + T: Borrow<Q>, + Q: Hash + Eq, + F: FnOnce(&Q) -> T, + { + // Although the raw entry gives us `&mut T`, we only return `&T` to be consistent with + // `get`. Key mutation is "raw" because you're not supposed to affect `Eq` or `Hash`. + self.map + .raw_entry_mut() + .from_key(value) + .or_insert_with(|| (f(value), ())) + .0 + } + + /// Returns `true` if `self` has no elements in common with `other`. + /// This is equivalent to checking for an empty intersection. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let a: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// let mut b = HashSet::new(); + /// + /// assert_eq!(a.is_disjoint(&b), true); + /// b.insert(4); + /// assert_eq!(a.is_disjoint(&b), true); + /// b.insert(1); + /// assert_eq!(a.is_disjoint(&b), false); + /// ``` + pub fn is_disjoint(&self, other: &Self) -> bool { + self.iter().all(|v| !other.contains(v)) + } + + /// Returns `true` if the set is a subset of another, + /// i.e., `other` contains at least all the values in `self`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let sup: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// let mut set = HashSet::new(); + /// + /// assert_eq!(set.is_subset(&sup), true); + /// set.insert(2); + /// assert_eq!(set.is_subset(&sup), true); + /// set.insert(4); + /// assert_eq!(set.is_subset(&sup), false); + /// ``` + pub fn is_subset(&self, other: &Self) -> bool { + self.len() <= other.len() && self.iter().all(|v| other.contains(v)) + } + + /// Returns `true` if the set is a superset of another, + /// i.e., `self` contains at least all the values in `other`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let sub: HashSet<_> = [1, 2].iter().cloned().collect(); + /// let mut set = HashSet::new(); + /// + /// assert_eq!(set.is_superset(&sub), false); + /// + /// set.insert(0); + /// set.insert(1); + /// assert_eq!(set.is_superset(&sub), false); + /// + /// set.insert(2); + /// assert_eq!(set.is_superset(&sub), true); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn is_superset(&self, other: &Self) -> bool { + other.is_subset(self) + } + + /// Adds a value to the set. + /// + /// If the set did not have this value present, `true` is returned. + /// + /// If the set did have this value present, `false` is returned. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set = HashSet::new(); + /// + /// assert_eq!(set.insert(2), true); + /// assert_eq!(set.insert(2), false); + /// assert_eq!(set.len(), 1); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn insert(&mut self, value: T) -> bool { + self.map.insert(value, ()).is_none() + } + + /// Adds a value to the set, replacing the existing value, if any, that is equal to the given + /// one. Returns the replaced value. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set = HashSet::new(); + /// set.insert(Vec::<i32>::new()); + /// + /// assert_eq!(set.get(&[][..]).unwrap().capacity(), 0); + /// set.replace(Vec::with_capacity(10)); + /// assert_eq!(set.get(&[][..]).unwrap().capacity(), 10); + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + pub fn replace(&mut self, value: T) -> Option<T> { + match self.map.entry(value) { + map::Entry::Occupied(occupied) => Some(occupied.replace_key()), + map::Entry::Vacant(vacant) => { + vacant.insert(()); + None + } + } + } + + /// Removes a value from the set. Returns whether the value was + /// present in the set. + /// + /// The value may be any borrowed form of the set's value type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the value type. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set = HashSet::new(); + /// + /// set.insert(2); + /// assert_eq!(set.remove(&2), true); + /// assert_eq!(set.remove(&2), false); + /// ``` + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + #[cfg_attr(feature = "inline-more", inline)] + pub fn remove<Q: ?Sized>(&mut self, value: &Q) -> bool + where + T: Borrow<Q>, + Q: Hash + Eq, + { + self.map.remove(value).is_some() + } + + /// Removes and returns the value in the set, if any, that is equal to the given one. + /// + /// The value may be any borrowed form of the set's value type, but + /// [`Hash`] and [`Eq`] on the borrowed form *must* match those for + /// the value type. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let mut set: HashSet<_> = [1, 2, 3].iter().cloned().collect(); + /// assert_eq!(set.take(&2), Some(2)); + /// assert_eq!(set.take(&2), None); + /// ``` + /// + /// [`Eq`]: https://doc.rust-lang.org/std/cmp/trait.Eq.html + /// [`Hash`]: https://doc.rust-lang.org/std/hash/trait.Hash.html + #[cfg_attr(feature = "inline-more", inline)] + pub fn take<Q: ?Sized>(&mut self, value: &Q) -> Option<T> + where + T: Borrow<Q>, + Q: Hash + Eq, + { + // Avoid `Option::map` because it bloats LLVM IR. + match self.map.remove_entry(value) { + Some((k, _)) => Some(k), + None => None, + } + } +} + +impl<T, S> PartialEq for HashSet<T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ + fn eq(&self, other: &Self) -> bool { + if self.len() != other.len() { + return false; + } + + self.iter().all(|key| other.contains(key)) + } +} + +impl<T, S> Eq for HashSet<T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ +} + +impl<T, S> fmt::Debug for HashSet<T, S> +where + T: Eq + Hash + fmt::Debug, + S: BuildHasher, +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_set().entries(self.iter()).finish() + } +} + +impl<T, S> FromIterator<T> for HashSet<T, S> +where + T: Eq + Hash, + S: BuildHasher + Default, +{ + #[cfg_attr(feature = "inline-more", inline)] + fn from_iter<I: IntoIterator<Item = T>>(iter: I) -> Self { + let mut set = Self::with_hasher(Default::default()); + set.extend(iter); + set + } +} + +impl<T, S> Extend<T> for HashSet<T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ + #[cfg_attr(feature = "inline-more", inline)] + fn extend<I: IntoIterator<Item = T>>(&mut self, iter: I) { + self.map.extend(iter.into_iter().map(|k| (k, ()))); + } + + #[inline] + #[cfg(feature = "nightly")] + fn extend_one(&mut self, k: T) { + self.map.insert(k, ()); + } + + #[inline] + #[cfg(feature = "nightly")] + fn extend_reserve(&mut self, additional: usize) { + Extend::<(T, ())>::extend_reserve(&mut self.map, additional); + } +} + +impl<'a, T, S> Extend<&'a T> for HashSet<T, S> +where + T: 'a + Eq + Hash + Copy, + S: BuildHasher, +{ + #[cfg_attr(feature = "inline-more", inline)] + fn extend<I: IntoIterator<Item = &'a T>>(&mut self, iter: I) { + self.extend(iter.into_iter().cloned()); + } + + #[inline] + #[cfg(feature = "nightly")] + fn extend_one(&mut self, k: &'a T) { + self.map.insert(*k, ()); + } + + #[inline] + #[cfg(feature = "nightly")] + fn extend_reserve(&mut self, additional: usize) { + Extend::<(T, ())>::extend_reserve(&mut self.map, additional); + } +} + +impl<T, S> Default for HashSet<T, S> +where + S: Default, +{ + /// Creates an empty `HashSet<T, S>` with the `Default` value for the hasher. + #[cfg_attr(feature = "inline-more", inline)] + fn default() -> Self { + Self { + map: HashMap::default(), + } + } +} + +impl<T, S> BitOr<&HashSet<T, S>> for &HashSet<T, S> +where + T: Eq + Hash + Clone, + S: BuildHasher + Default, +{ + type Output = HashSet<T, S>; + + /// Returns the union of `self` and `rhs` as a new `HashSet<T, S>`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let a: HashSet<_> = vec![1, 2, 3].into_iter().collect(); + /// let b: HashSet<_> = vec![3, 4, 5].into_iter().collect(); + /// + /// let set = &a | &b; + /// + /// let mut i = 0; + /// let expected = [1, 2, 3, 4, 5]; + /// for x in &set { + /// assert!(expected.contains(x)); + /// i += 1; + /// } + /// assert_eq!(i, expected.len()); + /// ``` + fn bitor(self, rhs: &HashSet<T, S>) -> HashSet<T, S> { + self.union(rhs).cloned().collect() + } +} + +impl<T, S> BitAnd<&HashSet<T, S>> for &HashSet<T, S> +where + T: Eq + Hash + Clone, + S: BuildHasher + Default, +{ + type Output = HashSet<T, S>; + + /// Returns the intersection of `self` and `rhs` as a new `HashSet<T, S>`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let a: HashSet<_> = vec![1, 2, 3].into_iter().collect(); + /// let b: HashSet<_> = vec![2, 3, 4].into_iter().collect(); + /// + /// let set = &a & &b; + /// + /// let mut i = 0; + /// let expected = [2, 3]; + /// for x in &set { + /// assert!(expected.contains(x)); + /// i += 1; + /// } + /// assert_eq!(i, expected.len()); + /// ``` + fn bitand(self, rhs: &HashSet<T, S>) -> HashSet<T, S> { + self.intersection(rhs).cloned().collect() + } +} + +impl<T, S> BitXor<&HashSet<T, S>> for &HashSet<T, S> +where + T: Eq + Hash + Clone, + S: BuildHasher + Default, +{ + type Output = HashSet<T, S>; + + /// Returns the symmetric difference of `self` and `rhs` as a new `HashSet<T, S>`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let a: HashSet<_> = vec![1, 2, 3].into_iter().collect(); + /// let b: HashSet<_> = vec![3, 4, 5].into_iter().collect(); + /// + /// let set = &a ^ &b; + /// + /// let mut i = 0; + /// let expected = [1, 2, 4, 5]; + /// for x in &set { + /// assert!(expected.contains(x)); + /// i += 1; + /// } + /// assert_eq!(i, expected.len()); + /// ``` + fn bitxor(self, rhs: &HashSet<T, S>) -> HashSet<T, S> { + self.symmetric_difference(rhs).cloned().collect() + } +} + +impl<T, S> Sub<&HashSet<T, S>> for &HashSet<T, S> +where + T: Eq + Hash + Clone, + S: BuildHasher + Default, +{ + type Output = HashSet<T, S>; + + /// Returns the difference of `self` and `rhs` as a new `HashSet<T, S>`. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// + /// let a: HashSet<_> = vec![1, 2, 3].into_iter().collect(); + /// let b: HashSet<_> = vec![3, 4, 5].into_iter().collect(); + /// + /// let set = &a - &b; + /// + /// let mut i = 0; + /// let expected = [1, 2]; + /// for x in &set { + /// assert!(expected.contains(x)); + /// i += 1; + /// } + /// assert_eq!(i, expected.len()); + /// ``` + fn sub(self, rhs: &HashSet<T, S>) -> HashSet<T, S> { + self.difference(rhs).cloned().collect() + } +} + +/// An iterator over the items of a `HashSet`. +/// +/// This `struct` is created by the [`iter`] method on [`HashSet`]. +/// See its documentation for more. +/// +/// [`HashSet`]: struct.HashSet.html +/// [`iter`]: struct.HashSet.html#method.iter +pub struct Iter<'a, K> { + iter: Keys<'a, K, ()>, +} + +/// An owning iterator over the items of a `HashSet`. +/// +/// This `struct` is created by the [`into_iter`] method on [`HashSet`] +/// (provided by the `IntoIterator` trait). See its documentation for more. +/// +/// [`HashSet`]: struct.HashSet.html +/// [`into_iter`]: struct.HashSet.html#method.into_iter +pub struct IntoIter<K> { + iter: map::IntoIter<K, ()>, +} + +/// A draining iterator over the items of a `HashSet`. +/// +/// This `struct` is created by the [`drain`] method on [`HashSet`]. +/// See its documentation for more. +/// +/// [`HashSet`]: struct.HashSet.html +/// [`drain`]: struct.HashSet.html#method.drain +pub struct Drain<'a, K> { + iter: map::Drain<'a, K, ()>, +} + +/// A draining iterator over entries of a `HashSet` which don't satisfy the predicate `f`. +/// +/// This `struct` is created by the [`drain_filter`] method on [`HashSet`]. See its +/// documentation for more. +/// +/// [`drain_filter`]: struct.HashSet.html#method.drain_filter +/// [`HashSet`]: struct.HashSet.html +pub struct DrainFilter<'a, K, F> +where + F: FnMut(&K) -> bool, +{ + f: F, + inner: DrainFilterInner<'a, K, ()>, +} + +/// A lazy iterator producing elements in the intersection of `HashSet`s. +/// +/// This `struct` is created by the [`intersection`] method on [`HashSet`]. +/// See its documentation for more. +/// +/// [`HashSet`]: struct.HashSet.html +/// [`intersection`]: struct.HashSet.html#method.intersection +pub struct Intersection<'a, T, S> { + // iterator of the first set + iter: Iter<'a, T>, + // the second set + other: &'a HashSet<T, S>, +} + +/// A lazy iterator producing elements in the difference of `HashSet`s. +/// +/// This `struct` is created by the [`difference`] method on [`HashSet`]. +/// See its documentation for more. +/// +/// [`HashSet`]: struct.HashSet.html +/// [`difference`]: struct.HashSet.html#method.difference +pub struct Difference<'a, T, S> { + // iterator of the first set + iter: Iter<'a, T>, + // the second set + other: &'a HashSet<T, S>, +} + +/// A lazy iterator producing elements in the symmetric difference of `HashSet`s. +/// +/// This `struct` is created by the [`symmetric_difference`] method on +/// [`HashSet`]. See its documentation for more. +/// +/// [`HashSet`]: struct.HashSet.html +/// [`symmetric_difference`]: struct.HashSet.html#method.symmetric_difference +pub struct SymmetricDifference<'a, T, S> { + iter: Chain<Difference<'a, T, S>, Difference<'a, T, S>>, +} + +/// A lazy iterator producing elements in the union of `HashSet`s. +/// +/// This `struct` is created by the [`union`] method on [`HashSet`]. +/// See its documentation for more. +/// +/// [`HashSet`]: struct.HashSet.html +/// [`union`]: struct.HashSet.html#method.union +pub struct Union<'a, T, S> { + iter: Chain<Iter<'a, T>, Difference<'a, T, S>>, +} + +impl<'a, T, S> IntoIterator for &'a HashSet<T, S> { + type Item = &'a T; + type IntoIter = Iter<'a, T>; + + #[cfg_attr(feature = "inline-more", inline)] + fn into_iter(self) -> Iter<'a, T> { + self.iter() + } +} + +impl<T, S> IntoIterator for HashSet<T, S> { + type Item = T; + type IntoIter = IntoIter<T>; + + /// Creates a consuming iterator, that is, one that moves each value out + /// of the set in arbitrary order. The set cannot be used after calling + /// this. + /// + /// # Examples + /// + /// ``` + /// use hashbrown::HashSet; + /// let mut set = HashSet::new(); + /// set.insert("a".to_string()); + /// set.insert("b".to_string()); + /// + /// // Not possible to collect to a Vec<String> with a regular `.iter()`. + /// let v: Vec<String> = set.into_iter().collect(); + /// + /// // Will print in an arbitrary order. + /// for x in &v { + /// println!("{}", x); + /// } + /// ``` + #[cfg_attr(feature = "inline-more", inline)] + fn into_iter(self) -> IntoIter<T> { + IntoIter { + iter: self.map.into_iter(), + } + } +} + +impl<K> Clone for Iter<'_, K> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + Iter { + iter: self.iter.clone(), + } + } +} +impl<'a, K> Iterator for Iter<'a, K> { + type Item = &'a K; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<&'a K> { + self.iter.next() + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.iter.size_hint() + } +} +impl<'a, K> ExactSizeIterator for Iter<'a, K> { + #[cfg_attr(feature = "inline-more", inline)] + fn len(&self) -> usize { + self.iter.len() + } +} +impl<K> FusedIterator for Iter<'_, K> {} + +impl<K: fmt::Debug> fmt::Debug for Iter<'_, K> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.clone()).finish() + } +} + +impl<K> Iterator for IntoIter<K> { + type Item = K; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<K> { + // Avoid `Option::map` because it bloats LLVM IR. + match self.iter.next() { + Some((k, _)) => Some(k), + None => None, + } + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.iter.size_hint() + } +} +impl<K> ExactSizeIterator for IntoIter<K> { + #[cfg_attr(feature = "inline-more", inline)] + fn len(&self) -> usize { + self.iter.len() + } +} +impl<K> FusedIterator for IntoIter<K> {} + +impl<K: fmt::Debug> fmt::Debug for IntoIter<K> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + let entries_iter = self.iter.iter().map(|(k, _)| k); + f.debug_list().entries(entries_iter).finish() + } +} + +impl<K> Iterator for Drain<'_, K> { + type Item = K; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<K> { + // Avoid `Option::map` because it bloats LLVM IR. + match self.iter.next() { + Some((k, _)) => Some(k), + None => None, + } + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.iter.size_hint() + } +} +impl<K> ExactSizeIterator for Drain<'_, K> { + #[cfg_attr(feature = "inline-more", inline)] + fn len(&self) -> usize { + self.iter.len() + } +} +impl<K> FusedIterator for Drain<'_, K> {} + +impl<K: fmt::Debug> fmt::Debug for Drain<'_, K> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + let entries_iter = self.iter.iter().map(|(k, _)| k); + f.debug_list().entries(entries_iter).finish() + } +} + +impl<'a, K, F> Drop for DrainFilter<'a, K, F> +where + F: FnMut(&K) -> bool, +{ + #[cfg_attr(feature = "inline-more", inline)] + fn drop(&mut self) { + while let Some(item) = self.next() { + let guard = ConsumeAllOnDrop(self); + drop(item); + mem::forget(guard); + } + } +} + +impl<K, F> Iterator for DrainFilter<'_, K, F> +where + F: FnMut(&K) -> bool, +{ + type Item = K; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<Self::Item> { + let f = &mut self.f; + let (k, _) = self.inner.next(&mut |k, _| f(k))?; + Some(k) + } + + #[inline] + fn size_hint(&self) -> (usize, Option<usize>) { + (0, self.inner.iter.size_hint().1) + } +} + +impl<K, F> FusedIterator for DrainFilter<'_, K, F> where F: FnMut(&K) -> bool {} + +impl<T, S> Clone for Intersection<'_, T, S> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + Intersection { + iter: self.iter.clone(), + ..*self + } + } +} + +impl<'a, T, S> Iterator for Intersection<'a, T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ + type Item = &'a T; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<&'a T> { + loop { + let elt = self.iter.next()?; + if self.other.contains(elt) { + return Some(elt); + } + } + } + + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + let (_, upper) = self.iter.size_hint(); + (0, upper) + } +} + +impl<T, S> fmt::Debug for Intersection<'_, T, S> +where + T: fmt::Debug + Eq + Hash, + S: BuildHasher, +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.clone()).finish() + } +} + +impl<T, S> FusedIterator for Intersection<'_, T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ +} + +impl<T, S> Clone for Difference<'_, T, S> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + Difference { + iter: self.iter.clone(), + ..*self + } + } +} + +impl<'a, T, S> Iterator for Difference<'a, T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ + type Item = &'a T; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<&'a T> { + loop { + let elt = self.iter.next()?; + if !self.other.contains(elt) { + return Some(elt); + } + } + } + + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + let (_, upper) = self.iter.size_hint(); + (0, upper) + } +} + +impl<T, S> FusedIterator for Difference<'_, T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ +} + +impl<T, S> fmt::Debug for Difference<'_, T, S> +where + T: fmt::Debug + Eq + Hash, + S: BuildHasher, +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.clone()).finish() + } +} + +impl<T, S> Clone for SymmetricDifference<'_, T, S> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + SymmetricDifference { + iter: self.iter.clone(), + } + } +} + +impl<'a, T, S> Iterator for SymmetricDifference<'a, T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ + type Item = &'a T; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<&'a T> { + self.iter.next() + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.iter.size_hint() + } +} + +impl<T, S> FusedIterator for SymmetricDifference<'_, T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ +} + +impl<T, S> fmt::Debug for SymmetricDifference<'_, T, S> +where + T: fmt::Debug + Eq + Hash, + S: BuildHasher, +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.clone()).finish() + } +} + +impl<T, S> Clone for Union<'_, T, S> { + #[cfg_attr(feature = "inline-more", inline)] + fn clone(&self) -> Self { + Union { + iter: self.iter.clone(), + } + } +} + +impl<T, S> FusedIterator for Union<'_, T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ +} + +impl<T, S> fmt::Debug for Union<'_, T, S> +where + T: fmt::Debug + Eq + Hash, + S: BuildHasher, +{ + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.clone()).finish() + } +} + +impl<'a, T, S> Iterator for Union<'a, T, S> +where + T: Eq + Hash, + S: BuildHasher, +{ + type Item = &'a T; + + #[cfg_attr(feature = "inline-more", inline)] + fn next(&mut self) -> Option<&'a T> { + self.iter.next() + } + #[cfg_attr(feature = "inline-more", inline)] + fn size_hint(&self) -> (usize, Option<usize>) { + self.iter.size_hint() + } +} + +#[allow(dead_code)] +fn assert_covariance() { + fn set<'new>(v: HashSet<&'static str>) -> HashSet<&'new str> { + v + } + fn iter<'a, 'new>(v: Iter<'a, &'static str>) -> Iter<'a, &'new str> { + v + } + fn into_iter<'new>(v: IntoIter<&'static str>) -> IntoIter<&'new str> { + v + } + fn difference<'a, 'new>( + v: Difference<'a, &'static str, DefaultHashBuilder>, + ) -> Difference<'a, &'new str, DefaultHashBuilder> { + v + } + fn symmetric_difference<'a, 'new>( + v: SymmetricDifference<'a, &'static str, DefaultHashBuilder>, + ) -> SymmetricDifference<'a, &'new str, DefaultHashBuilder> { + v + } + fn intersection<'a, 'new>( + v: Intersection<'a, &'static str, DefaultHashBuilder>, + ) -> Intersection<'a, &'new str, DefaultHashBuilder> { + v + } + fn union<'a, 'new>( + v: Union<'a, &'static str, DefaultHashBuilder>, + ) -> Union<'a, &'new str, DefaultHashBuilder> { + v + } + fn drain<'new>(d: Drain<'static, &'static str>) -> Drain<'new, &'new str> { + d + } +} + +#[cfg(test)] +mod test_set { + use super::super::map::DefaultHashBuilder; + use super::HashSet; + use std::vec::Vec; + + #[test] + fn test_zero_capacities() { + type HS = HashSet<i32>; + + let s = HS::new(); + assert_eq!(s.capacity(), 0); + + let s = HS::default(); + assert_eq!(s.capacity(), 0); + + let s = HS::with_hasher(DefaultHashBuilder::default()); + assert_eq!(s.capacity(), 0); + + let s = HS::with_capacity(0); + assert_eq!(s.capacity(), 0); + + let s = HS::with_capacity_and_hasher(0, DefaultHashBuilder::default()); + assert_eq!(s.capacity(), 0); + + let mut s = HS::new(); + s.insert(1); + s.insert(2); + s.remove(&1); + s.remove(&2); + s.shrink_to_fit(); + assert_eq!(s.capacity(), 0); + + let mut s = HS::new(); + s.reserve(0); + assert_eq!(s.capacity(), 0); + } + + #[test] + fn test_disjoint() { + let mut xs = HashSet::new(); + let mut ys = HashSet::new(); + assert!(xs.is_disjoint(&ys)); + assert!(ys.is_disjoint(&xs)); + assert!(xs.insert(5)); + assert!(ys.insert(11)); + assert!(xs.is_disjoint(&ys)); + assert!(ys.is_disjoint(&xs)); + assert!(xs.insert(7)); + assert!(xs.insert(19)); + assert!(xs.insert(4)); + assert!(ys.insert(2)); + assert!(ys.insert(-11)); + assert!(xs.is_disjoint(&ys)); + assert!(ys.is_disjoint(&xs)); + assert!(ys.insert(7)); + assert!(!xs.is_disjoint(&ys)); + assert!(!ys.is_disjoint(&xs)); + } + + #[test] + fn test_subset_and_superset() { + let mut a = HashSet::new(); + assert!(a.insert(0)); + assert!(a.insert(5)); + assert!(a.insert(11)); + assert!(a.insert(7)); + + let mut b = HashSet::new(); + assert!(b.insert(0)); + assert!(b.insert(7)); + assert!(b.insert(19)); + assert!(b.insert(250)); + assert!(b.insert(11)); + assert!(b.insert(200)); + + assert!(!a.is_subset(&b)); + assert!(!a.is_superset(&b)); + assert!(!b.is_subset(&a)); + assert!(!b.is_superset(&a)); + + assert!(b.insert(5)); + + assert!(a.is_subset(&b)); + assert!(!a.is_superset(&b)); + assert!(!b.is_subset(&a)); + assert!(b.is_superset(&a)); + } + + #[test] + fn test_iterate() { + let mut a = HashSet::new(); + for i in 0..32 { + assert!(a.insert(i)); + } + let mut observed: u32 = 0; + for k in &a { + observed |= 1 << *k; + } + assert_eq!(observed, 0xFFFF_FFFF); + } + + #[test] + fn test_intersection() { + let mut a = HashSet::new(); + let mut b = HashSet::new(); + + assert!(a.insert(11)); + assert!(a.insert(1)); + assert!(a.insert(3)); + assert!(a.insert(77)); + assert!(a.insert(103)); + assert!(a.insert(5)); + assert!(a.insert(-5)); + + assert!(b.insert(2)); + assert!(b.insert(11)); + assert!(b.insert(77)); + assert!(b.insert(-9)); + assert!(b.insert(-42)); + assert!(b.insert(5)); + assert!(b.insert(3)); + + let mut i = 0; + let expected = [3, 5, 11, 77]; + for x in a.intersection(&b) { + assert!(expected.contains(x)); + i += 1 + } + assert_eq!(i, expected.len()); + } + + #[test] + fn test_difference() { + let mut a = HashSet::new(); + let mut b = HashSet::new(); + + assert!(a.insert(1)); + assert!(a.insert(3)); + assert!(a.insert(5)); + assert!(a.insert(9)); + assert!(a.insert(11)); + + assert!(b.insert(3)); + assert!(b.insert(9)); + + let mut i = 0; + let expected = [1, 5, 11]; + for x in a.difference(&b) { + assert!(expected.contains(x)); + i += 1 + } + assert_eq!(i, expected.len()); + } + + #[test] + fn test_symmetric_difference() { + let mut a = HashSet::new(); + let mut b = HashSet::new(); + + assert!(a.insert(1)); + assert!(a.insert(3)); + assert!(a.insert(5)); + assert!(a.insert(9)); + assert!(a.insert(11)); + + assert!(b.insert(-2)); + assert!(b.insert(3)); + assert!(b.insert(9)); + assert!(b.insert(14)); + assert!(b.insert(22)); + + let mut i = 0; + let expected = [-2, 1, 5, 11, 14, 22]; + for x in a.symmetric_difference(&b) { + assert!(expected.contains(x)); + i += 1 + } + assert_eq!(i, expected.len()); + } + + #[test] + fn test_union() { + let mut a = HashSet::new(); + let mut b = HashSet::new(); + + assert!(a.insert(1)); + assert!(a.insert(3)); + assert!(a.insert(5)); + assert!(a.insert(9)); + assert!(a.insert(11)); + assert!(a.insert(16)); + assert!(a.insert(19)); + assert!(a.insert(24)); + + assert!(b.insert(-2)); + assert!(b.insert(1)); + assert!(b.insert(5)); + assert!(b.insert(9)); + assert!(b.insert(13)); + assert!(b.insert(19)); + + let mut i = 0; + let expected = [-2, 1, 3, 5, 9, 11, 13, 16, 19, 24]; + for x in a.union(&b) { + assert!(expected.contains(x)); + i += 1 + } + assert_eq!(i, expected.len()); + } + + #[test] + fn test_from_iter() { + let xs = [1, 2, 2, 3, 4, 5, 6, 7, 8, 9]; + + let set: HashSet<_> = xs.iter().cloned().collect(); + + for x in &xs { + assert!(set.contains(x)); + } + + assert_eq!(set.iter().len(), xs.len() - 1); + } + + #[test] + fn test_move_iter() { + let hs = { + let mut hs = HashSet::new(); + + hs.insert('a'); + hs.insert('b'); + + hs + }; + + let v = hs.into_iter().collect::<Vec<char>>(); + assert!(v == ['a', 'b'] || v == ['b', 'a']); + } + + #[test] + fn test_eq() { + // These constants once happened to expose a bug in insert(). + // I'm keeping them around to prevent a regression. + let mut s1 = HashSet::new(); + + s1.insert(1); + s1.insert(2); + s1.insert(3); + + let mut s2 = HashSet::new(); + + s2.insert(1); + s2.insert(2); + + assert!(s1 != s2); + + s2.insert(3); + + assert_eq!(s1, s2); + } + + #[test] + fn test_show() { + let mut set = HashSet::new(); + let empty = HashSet::<i32>::new(); + + set.insert(1); + set.insert(2); + + let set_str = format!("{:?}", set); + + assert!(set_str == "{1, 2}" || set_str == "{2, 1}"); + assert_eq!(format!("{:?}", empty), "{}"); + } + + #[test] + fn test_trivial_drain() { + let mut s = HashSet::<i32>::new(); + for _ in s.drain() {} + assert!(s.is_empty()); + drop(s); + + let mut s = HashSet::<i32>::new(); + drop(s.drain()); + assert!(s.is_empty()); + } + + #[test] + fn test_drain() { + let mut s: HashSet<_> = (1..100).collect(); + + // try this a bunch of times to make sure we don't screw up internal state. + for _ in 0..20 { + assert_eq!(s.len(), 99); + + { + let mut last_i = 0; + let mut d = s.drain(); + for (i, x) in d.by_ref().take(50).enumerate() { + last_i = i; + assert!(x != 0); + } + assert_eq!(last_i, 49); + } + + for _ in &s { + panic!("s should be empty!"); + } + + // reset to try again. + s.extend(1..100); + } + } + + #[test] + fn test_replace() { + use core::hash; + + #[derive(Debug)] + struct Foo(&'static str, i32); + + impl PartialEq for Foo { + fn eq(&self, other: &Self) -> bool { + self.0 == other.0 + } + } + + impl Eq for Foo {} + + impl hash::Hash for Foo { + fn hash<H: hash::Hasher>(&self, h: &mut H) { + self.0.hash(h); + } + } + + let mut s = HashSet::new(); + assert_eq!(s.replace(Foo("a", 1)), None); + assert_eq!(s.len(), 1); + assert_eq!(s.replace(Foo("a", 2)), Some(Foo("a", 1))); + assert_eq!(s.len(), 1); + + let mut it = s.iter(); + assert_eq!(it.next(), Some(&Foo("a", 2))); + assert_eq!(it.next(), None); + } + + #[test] + fn test_extend_ref() { + let mut a = HashSet::new(); + a.insert(1); + + a.extend(&[2, 3, 4]); + + assert_eq!(a.len(), 4); + assert!(a.contains(&1)); + assert!(a.contains(&2)); + assert!(a.contains(&3)); + assert!(a.contains(&4)); + + let mut b = HashSet::new(); + b.insert(5); + b.insert(6); + + a.extend(&b); + + assert_eq!(a.len(), 6); + assert!(a.contains(&1)); + assert!(a.contains(&2)); + assert!(a.contains(&3)); + assert!(a.contains(&4)); + assert!(a.contains(&5)); + assert!(a.contains(&6)); + } + + #[test] + fn test_retain() { + let xs = [1, 2, 3, 4, 5, 6]; + let mut set: HashSet<i32> = xs.iter().cloned().collect(); + set.retain(|&k| k % 2 == 0); + assert_eq!(set.len(), 3); + assert!(set.contains(&2)); + assert!(set.contains(&4)); + assert!(set.contains(&6)); + } + + #[test] + fn test_drain_filter() { + { + let mut set: HashSet<i32> = (0..8).collect(); + let drained = set.drain_filter(|&k| k % 2 == 0); + let mut out = drained.collect::<Vec<_>>(); + out.sort_unstable(); + assert_eq!(vec![0, 2, 4, 6], out); + assert_eq!(set.len(), 4); + } + { + let mut set: HashSet<i32> = (0..8).collect(); + drop(set.drain_filter(|&k| k % 2 == 0)); + assert_eq!(set.len(), 4, "Removes non-matching items on drop"); + } + } + + #[test] + fn test_const_with_hasher() { + use core::hash::BuildHasher; + use std::collections::hash_map::DefaultHasher; + + #[derive(Clone)] + struct MyHasher; + impl BuildHasher for MyHasher { + type Hasher = DefaultHasher; + + fn build_hasher(&self) -> DefaultHasher { + DefaultHasher::new() + } + } + + const EMPTY_SET: HashSet<u32, MyHasher> = HashSet::with_hasher(MyHasher); + + let mut set = EMPTY_SET.clone(); + set.insert(19); + assert!(set.contains(&19)); + } +} diff --git a/tests/hasher.rs b/tests/hasher.rs new file mode 100644 index 0000000..e455e3d --- /dev/null +++ b/tests/hasher.rs @@ -0,0 +1,65 @@ +//! Sanity check that alternate hashers work correctly. + +#![cfg(not(miri))] // FIXME: takes too long + +use hashbrown::HashSet; +use std::hash::{BuildHasher, BuildHasherDefault, Hasher}; + +fn check<S: BuildHasher + Default>() { + let range = 0..1_000; + + let mut set = HashSet::<i32, S>::default(); + set.extend(range.clone()); + + assert!(!set.contains(&i32::min_value())); + assert!(!set.contains(&(range.start - 1))); + for i in range.clone() { + assert!(set.contains(&i)); + } + assert!(!set.contains(&range.end)); + assert!(!set.contains(&i32::max_value())); +} + +/// Use hashbrown's default hasher. +#[test] +fn default() { + check::<hashbrown::hash_map::DefaultHashBuilder>(); +} + +/// Use std's default hasher. +#[test] +fn random_state() { + check::<std::collections::hash_map::RandomState>(); +} + +/// Use a constant 0 hash. +#[test] +fn zero() { + #[derive(Default)] + struct ZeroHasher; + + impl Hasher for ZeroHasher { + fn finish(&self) -> u64 { + 0 + } + fn write(&mut self, _: &[u8]) {} + } + + check::<BuildHasherDefault<ZeroHasher>>(); +} + +/// Use a constant maximum hash. +#[test] +fn max() { + #[derive(Default)] + struct MaxHasher; + + impl Hasher for MaxHasher { + fn finish(&self) -> u64 { + u64::max_value() + } + fn write(&mut self, _: &[u8]) {} + } + + check::<BuildHasherDefault<MaxHasher>>(); +} diff --git a/tests/rayon.rs b/tests/rayon.rs new file mode 100644 index 0000000..39b4770 --- /dev/null +++ b/tests/rayon.rs @@ -0,0 +1,533 @@ +#![cfg(feature = "rayon")] + +#[macro_use] +extern crate lazy_static; + +use hashbrown::{HashMap, HashSet}; +use rayon::iter::{ + IntoParallelIterator, IntoParallelRefIterator, IntoParallelRefMutIterator, ParallelExtend, + ParallelIterator, +}; + +macro_rules! assert_eq3 { + ($e1:expr, $e2:expr, $e3:expr) => {{ + assert_eq!($e1, $e2); + assert_eq!($e1, $e3); + assert_eq!($e2, $e3); + }}; +} + +lazy_static! { + static ref MAP_EMPTY: HashMap<char, u32> = HashMap::new(); + static ref MAP: HashMap<char, u32> = { + let mut m = HashMap::new(); + m.insert('b', 20); + m.insert('a', 10); + m.insert('c', 30); + m.insert('e', 50); + m.insert('f', 60); + m.insert('d', 40); + m + }; +} + +#[test] +fn map_seq_par_equivalence_iter_empty() { + let vec_seq = MAP_EMPTY.iter().collect::<Vec<_>>(); + let vec_par = MAP_EMPTY.par_iter().collect::<Vec<_>>(); + + assert_eq3!(vec_seq, vec_par, []); +} + +#[test] +fn map_seq_par_equivalence_iter() { + let mut vec_seq = MAP.iter().collect::<Vec<_>>(); + let mut vec_par = MAP.par_iter().collect::<Vec<_>>(); + + assert_eq!(vec_seq, vec_par); + + // Do not depend on the exact order of values + let expected_sorted = [ + (&'a', &10), + (&'b', &20), + (&'c', &30), + (&'d', &40), + (&'e', &50), + (&'f', &60), + ]; + + vec_seq.sort_unstable(); + vec_par.sort_unstable(); + + assert_eq3!(vec_seq, vec_par, expected_sorted); +} + +#[test] +fn map_seq_par_equivalence_keys_empty() { + let vec_seq = MAP_EMPTY.keys().collect::<Vec<&char>>(); + let vec_par = MAP_EMPTY.par_keys().collect::<Vec<&char>>(); + + let expected: [&char; 0] = []; + + assert_eq3!(vec_seq, vec_par, expected); +} + +#[test] +fn map_seq_par_equivalence_keys() { + let mut vec_seq = MAP.keys().collect::<Vec<_>>(); + let mut vec_par = MAP.par_keys().collect::<Vec<_>>(); + + assert_eq!(vec_seq, vec_par); + + // Do not depend on the exact order of values + let expected_sorted = [&'a', &'b', &'c', &'d', &'e', &'f']; + + vec_seq.sort_unstable(); + vec_par.sort_unstable(); + + assert_eq3!(vec_seq, vec_par, expected_sorted); +} + +#[test] +fn map_seq_par_equivalence_values_empty() { + let vec_seq = MAP_EMPTY.values().collect::<Vec<_>>(); + let vec_par = MAP_EMPTY.par_values().collect::<Vec<_>>(); + + let expected: [&u32; 0] = []; + + assert_eq3!(vec_seq, vec_par, expected); +} + +#[test] +fn map_seq_par_equivalence_values() { + let mut vec_seq = MAP.values().collect::<Vec<_>>(); + let mut vec_par = MAP.par_values().collect::<Vec<_>>(); + + assert_eq!(vec_seq, vec_par); + + // Do not depend on the exact order of values + let expected_sorted = [&10, &20, &30, &40, &50, &60]; + + vec_seq.sort_unstable(); + vec_par.sort_unstable(); + + assert_eq3!(vec_seq, vec_par, expected_sorted); +} + +#[test] +fn map_seq_par_equivalence_iter_mut_empty() { + let mut map1 = MAP_EMPTY.clone(); + let mut map2 = MAP_EMPTY.clone(); + + let vec_seq = map1.iter_mut().collect::<Vec<_>>(); + let vec_par = map2.par_iter_mut().collect::<Vec<_>>(); + + assert_eq3!(vec_seq, vec_par, []); +} + +#[test] +fn map_seq_par_equivalence_iter_mut() { + let mut map1 = MAP.clone(); + let mut map2 = MAP.clone(); + + let mut vec_seq = map1.iter_mut().collect::<Vec<_>>(); + let mut vec_par = map2.par_iter_mut().collect::<Vec<_>>(); + + assert_eq!(vec_seq, vec_par); + + // Do not depend on the exact order of values + let expected_sorted = [ + (&'a', &mut 10), + (&'b', &mut 20), + (&'c', &mut 30), + (&'d', &mut 40), + (&'e', &mut 50), + (&'f', &mut 60), + ]; + + vec_seq.sort_unstable(); + vec_par.sort_unstable(); + + assert_eq3!(vec_seq, vec_par, expected_sorted); +} + +#[test] +fn map_seq_par_equivalence_values_mut_empty() { + let mut map1 = MAP_EMPTY.clone(); + let mut map2 = MAP_EMPTY.clone(); + + let vec_seq = map1.values_mut().collect::<Vec<_>>(); + let vec_par = map2.par_values_mut().collect::<Vec<_>>(); + + let expected: [&u32; 0] = []; + + assert_eq3!(vec_seq, vec_par, expected); +} + +#[test] +fn map_seq_par_equivalence_values_mut() { + let mut map1 = MAP.clone(); + let mut map2 = MAP.clone(); + + let mut vec_seq = map1.values_mut().collect::<Vec<_>>(); + let mut vec_par = map2.par_values_mut().collect::<Vec<_>>(); + + assert_eq!(vec_seq, vec_par); + + // Do not depend on the exact order of values + let expected_sorted = [&mut 10, &mut 20, &mut 30, &mut 40, &mut 50, &mut 60]; + + vec_seq.sort_unstable(); + vec_par.sort_unstable(); + + assert_eq3!(vec_seq, vec_par, expected_sorted); +} + +#[test] +fn map_seq_par_equivalence_into_iter_empty() { + let vec_seq = MAP_EMPTY.clone().into_iter().collect::<Vec<_>>(); + let vec_par = MAP_EMPTY.clone().into_par_iter().collect::<Vec<_>>(); + + assert_eq3!(vec_seq, vec_par, []); +} + +#[test] +fn map_seq_par_equivalence_into_iter() { + let mut vec_seq = MAP.clone().into_iter().collect::<Vec<_>>(); + let mut vec_par = MAP.clone().into_par_iter().collect::<Vec<_>>(); + + assert_eq!(vec_seq, vec_par); + + // Do not depend on the exact order of values + let expected_sorted = [ + ('a', 10), + ('b', 20), + ('c', 30), + ('d', 40), + ('e', 50), + ('f', 60), + ]; + + vec_seq.sort_unstable(); + vec_par.sort_unstable(); + + assert_eq3!(vec_seq, vec_par, expected_sorted); +} + +lazy_static! { + static ref MAP_VEC_EMPTY: Vec<(char, u32)> = vec![]; + static ref MAP_VEC: Vec<(char, u32)> = vec![ + ('b', 20), + ('a', 10), + ('c', 30), + ('e', 50), + ('f', 60), + ('d', 40), + ]; +} + +#[test] +fn map_seq_par_equivalence_collect_empty() { + let map_expected = MAP_EMPTY.clone(); + let map_seq = MAP_VEC_EMPTY.clone().into_iter().collect::<HashMap<_, _>>(); + let map_par = MAP_VEC_EMPTY + .clone() + .into_par_iter() + .collect::<HashMap<_, _>>(); + + assert_eq!(map_seq, map_par); + assert_eq!(map_seq, map_expected); + assert_eq!(map_par, map_expected); +} + +#[test] +fn map_seq_par_equivalence_collect() { + let map_expected = MAP.clone(); + let map_seq = MAP_VEC.clone().into_iter().collect::<HashMap<_, _>>(); + let map_par = MAP_VEC.clone().into_par_iter().collect::<HashMap<_, _>>(); + + assert_eq!(map_seq, map_par); + assert_eq!(map_seq, map_expected); + assert_eq!(map_par, map_expected); +} + +lazy_static! { + static ref MAP_EXISTING_EMPTY: HashMap<char, u32> = HashMap::new(); + static ref MAP_EXISTING: HashMap<char, u32> = { + let mut m = HashMap::new(); + m.insert('b', 20); + m.insert('a', 10); + m + }; + static ref MAP_EXTENSION_EMPTY: Vec<(char, u32)> = vec![]; + static ref MAP_EXTENSION: Vec<(char, u32)> = vec![('c', 30), ('e', 50), ('f', 60), ('d', 40),]; +} + +#[test] +fn map_seq_par_equivalence_existing_empty_extend_empty() { + let expected = HashMap::new(); + let mut map_seq = MAP_EXISTING_EMPTY.clone(); + let mut map_par = MAP_EXISTING_EMPTY.clone(); + + map_seq.extend(MAP_EXTENSION_EMPTY.iter().cloned()); + map_par.par_extend(MAP_EXTENSION_EMPTY.par_iter().cloned()); + + assert_eq3!(map_seq, map_par, expected); +} + +#[test] +fn map_seq_par_equivalence_existing_empty_extend() { + let expected = MAP_EXTENSION.iter().cloned().collect::<HashMap<_, _>>(); + let mut map_seq = MAP_EXISTING_EMPTY.clone(); + let mut map_par = MAP_EXISTING_EMPTY.clone(); + + map_seq.extend(MAP_EXTENSION.iter().cloned()); + map_par.par_extend(MAP_EXTENSION.par_iter().cloned()); + + assert_eq3!(map_seq, map_par, expected); +} + +#[test] +fn map_seq_par_equivalence_existing_extend_empty() { + let expected = MAP_EXISTING.clone(); + let mut map_seq = MAP_EXISTING.clone(); + let mut map_par = MAP_EXISTING.clone(); + + map_seq.extend(MAP_EXTENSION_EMPTY.iter().cloned()); + map_par.par_extend(MAP_EXTENSION_EMPTY.par_iter().cloned()); + + assert_eq3!(map_seq, map_par, expected); +} + +#[test] +fn map_seq_par_equivalence_existing_extend() { + let expected = MAP.clone(); + let mut map_seq = MAP_EXISTING.clone(); + let mut map_par = MAP_EXISTING.clone(); + + map_seq.extend(MAP_EXTENSION.iter().cloned()); + map_par.par_extend(MAP_EXTENSION.par_iter().cloned()); + + assert_eq3!(map_seq, map_par, expected); +} + +lazy_static! { + static ref SET_EMPTY: HashSet<char> = HashSet::new(); + static ref SET: HashSet<char> = { + let mut s = HashSet::new(); + s.insert('b'); + s.insert('a'); + s.insert('c'); + s.insert('e'); + s.insert('f'); + s.insert('d'); + s + }; +} + +#[test] +fn set_seq_par_equivalence_iter_empty() { + let vec_seq = SET_EMPTY.iter().collect::<Vec<_>>(); + let vec_par = SET_EMPTY.par_iter().collect::<Vec<_>>(); + + let expected: [&char; 0] = []; + + assert_eq3!(vec_seq, vec_par, expected); +} + +#[test] +fn set_seq_par_equivalence_iter() { + let mut vec_seq = SET.iter().collect::<Vec<_>>(); + let mut vec_par = SET.par_iter().collect::<Vec<_>>(); + + assert_eq!(vec_seq, vec_par); + + // Do not depend on the exact order of values + let expected_sorted = [&'a', &'b', &'c', &'d', &'e', &'f']; + + vec_seq.sort_unstable(); + vec_par.sort_unstable(); + + assert_eq3!(vec_seq, vec_par, expected_sorted); +} + +#[test] +fn set_seq_par_equivalence_into_iter_empty() { + let vec_seq = SET_EMPTY.clone().into_iter().collect::<Vec<_>>(); + let vec_par = SET_EMPTY.clone().into_par_iter().collect::<Vec<_>>(); + + assert_eq3!(vec_seq, vec_par, []); +} + +#[test] +fn set_seq_par_equivalence_into_iter() { + let mut vec_seq = SET.clone().into_iter().collect::<Vec<_>>(); + let mut vec_par = SET.clone().into_par_iter().collect::<Vec<_>>(); + + assert_eq!(vec_seq, vec_par); + + // Do not depend on the exact order of values + let expected_sorted = ['a', 'b', 'c', 'd', 'e', 'f']; + + vec_seq.sort_unstable(); + vec_par.sort_unstable(); + + assert_eq3!(vec_seq, vec_par, expected_sorted); +} + +lazy_static! { + static ref SET_VEC_EMPTY: Vec<char> = vec![]; + static ref SET_VEC: Vec<char> = vec!['b', 'a', 'c', 'e', 'f', 'd',]; +} + +#[test] +fn set_seq_par_equivalence_collect_empty() { + let set_expected = SET_EMPTY.clone(); + let set_seq = SET_VEC_EMPTY.clone().into_iter().collect::<HashSet<_>>(); + let set_par = SET_VEC_EMPTY + .clone() + .into_par_iter() + .collect::<HashSet<_>>(); + + assert_eq!(set_seq, set_par); + assert_eq!(set_seq, set_expected); + assert_eq!(set_par, set_expected); +} + +#[test] +fn set_seq_par_equivalence_collect() { + let set_expected = SET.clone(); + let set_seq = SET_VEC.clone().into_iter().collect::<HashSet<_>>(); + let set_par = SET_VEC.clone().into_par_iter().collect::<HashSet<_>>(); + + assert_eq!(set_seq, set_par); + assert_eq!(set_seq, set_expected); + assert_eq!(set_par, set_expected); +} + +lazy_static! { + static ref SET_EXISTING_EMPTY: HashSet<char> = HashSet::new(); + static ref SET_EXISTING: HashSet<char> = { + let mut s = HashSet::new(); + s.insert('b'); + s.insert('a'); + s + }; + static ref SET_EXTENSION_EMPTY: Vec<char> = vec![]; + static ref SET_EXTENSION: Vec<char> = vec!['c', 'e', 'f', 'd',]; +} + +#[test] +fn set_seq_par_equivalence_existing_empty_extend_empty() { + let expected = HashSet::new(); + let mut set_seq = SET_EXISTING_EMPTY.clone(); + let mut set_par = SET_EXISTING_EMPTY.clone(); + + set_seq.extend(SET_EXTENSION_EMPTY.iter().cloned()); + set_par.par_extend(SET_EXTENSION_EMPTY.par_iter().cloned()); + + assert_eq3!(set_seq, set_par, expected); +} + +#[test] +fn set_seq_par_equivalence_existing_empty_extend() { + let expected = SET_EXTENSION.iter().cloned().collect::<HashSet<_>>(); + let mut set_seq = SET_EXISTING_EMPTY.clone(); + let mut set_par = SET_EXISTING_EMPTY.clone(); + + set_seq.extend(SET_EXTENSION.iter().cloned()); + set_par.par_extend(SET_EXTENSION.par_iter().cloned()); + + assert_eq3!(set_seq, set_par, expected); +} + +#[test] +fn set_seq_par_equivalence_existing_extend_empty() { + let expected = SET_EXISTING.clone(); + let mut set_seq = SET_EXISTING.clone(); + let mut set_par = SET_EXISTING.clone(); + + set_seq.extend(SET_EXTENSION_EMPTY.iter().cloned()); + set_par.par_extend(SET_EXTENSION_EMPTY.par_iter().cloned()); + + assert_eq3!(set_seq, set_par, expected); +} + +#[test] +fn set_seq_par_equivalence_existing_extend() { + let expected = SET.clone(); + let mut set_seq = SET_EXISTING.clone(); + let mut set_par = SET_EXISTING.clone(); + + set_seq.extend(SET_EXTENSION.iter().cloned()); + set_par.par_extend(SET_EXTENSION.par_iter().cloned()); + + assert_eq3!(set_seq, set_par, expected); +} + +lazy_static! { + static ref SET_A: HashSet<char> = ['a', 'b', 'c', 'd'].iter().cloned().collect(); + static ref SET_B: HashSet<char> = ['a', 'b', 'e', 'f'].iter().cloned().collect(); + static ref SET_DIFF_AB: HashSet<char> = ['c', 'd'].iter().cloned().collect(); + static ref SET_DIFF_BA: HashSet<char> = ['e', 'f'].iter().cloned().collect(); + static ref SET_SYMM_DIFF_AB: HashSet<char> = ['c', 'd', 'e', 'f'].iter().cloned().collect(); + static ref SET_INTERSECTION_AB: HashSet<char> = ['a', 'b'].iter().cloned().collect(); + static ref SET_UNION_AB: HashSet<char> = + ['a', 'b', 'c', 'd', 'e', 'f'].iter().cloned().collect(); +} + +#[test] +fn set_seq_par_equivalence_difference() { + let diff_ab_seq = SET_A.difference(&*SET_B).cloned().collect::<HashSet<_>>(); + let diff_ab_par = SET_A + .par_difference(&*SET_B) + .cloned() + .collect::<HashSet<_>>(); + + assert_eq3!(diff_ab_seq, diff_ab_par, *SET_DIFF_AB); + + let diff_ba_seq = SET_B.difference(&*SET_A).cloned().collect::<HashSet<_>>(); + let diff_ba_par = SET_B + .par_difference(&*SET_A) + .cloned() + .collect::<HashSet<_>>(); + + assert_eq3!(diff_ba_seq, diff_ba_par, *SET_DIFF_BA); +} + +#[test] +fn set_seq_par_equivalence_symmetric_difference() { + let symm_diff_ab_seq = SET_A + .symmetric_difference(&*SET_B) + .cloned() + .collect::<HashSet<_>>(); + let symm_diff_ab_par = SET_A + .par_symmetric_difference(&*SET_B) + .cloned() + .collect::<HashSet<_>>(); + + assert_eq3!(symm_diff_ab_seq, symm_diff_ab_par, *SET_SYMM_DIFF_AB); +} + +#[test] +fn set_seq_par_equivalence_intersection() { + let intersection_ab_seq = SET_A.intersection(&*SET_B).cloned().collect::<HashSet<_>>(); + let intersection_ab_par = SET_A + .par_intersection(&*SET_B) + .cloned() + .collect::<HashSet<_>>(); + + assert_eq3!( + intersection_ab_seq, + intersection_ab_par, + *SET_INTERSECTION_AB + ); +} + +#[test] +fn set_seq_par_equivalence_union() { + let union_ab_seq = SET_A.union(&*SET_B).cloned().collect::<HashSet<_>>(); + let union_ab_par = SET_A.par_union(&*SET_B).cloned().collect::<HashSet<_>>(); + + assert_eq3!(union_ab_seq, union_ab_par, *SET_UNION_AB); +} diff --git a/tests/serde.rs b/tests/serde.rs new file mode 100644 index 0000000..570bf70 --- /dev/null +++ b/tests/serde.rs @@ -0,0 +1,65 @@ +#![cfg(feature = "serde")] + +use core::hash::BuildHasherDefault; +use hashbrown::{HashMap, HashSet}; +use rustc_hash::FxHasher; +use serde_test::{assert_tokens, Token}; + +// We use FxHash for this test because we rely on the ordering +type FxHashMap<K, V> = HashMap<K, V, BuildHasherDefault<FxHasher>>; +type FxHashSet<T> = HashSet<T, BuildHasherDefault<FxHasher>>; + +#[test] +fn map_serde_tokens_empty() { + let map = FxHashMap::<char, u32>::default(); + + assert_tokens(&map, &[Token::Map { len: Some(0) }, Token::MapEnd]); +} + +#[test] +fn map_serde_tokens() { + let mut map = FxHashMap::default(); + map.insert('b', 20); + map.insert('a', 10); + map.insert('c', 30); + + assert_tokens( + &map, + &[ + Token::Map { len: Some(3) }, + Token::Char('a'), + Token::I32(10), + Token::Char('b'), + Token::I32(20), + Token::Char('c'), + Token::I32(30), + Token::MapEnd, + ], + ); +} + +#[test] +fn set_serde_tokens_empty() { + let set = FxHashSet::<u32>::default(); + + assert_tokens(&set, &[Token::Seq { len: Some(0) }, Token::SeqEnd]); +} + +#[test] +fn set_serde_tokens() { + let mut set = FxHashSet::default(); + set.insert(20); + set.insert(10); + set.insert(30); + + assert_tokens( + &set, + &[ + Token::Seq { len: Some(3) }, + Token::I32(20), + Token::I32(10), + Token::I32(30), + Token::SeqEnd, + ], + ); +} diff --git a/tests/set.rs b/tests/set.rs new file mode 100644 index 0000000..3fc0717 --- /dev/null +++ b/tests/set.rs @@ -0,0 +1,30 @@ +#![cfg(not(miri))] // FIXME: takes too long + +use hashbrown::HashSet; +use rand::{distributions::Alphanumeric, rngs::SmallRng, Rng, SeedableRng}; + +#[test] +fn test_hashset_insert_remove() { + let mut m: HashSet<Vec<char>> = HashSet::new(); + //let num: u32 = 4096; + //let tx: Vec<Vec<u8>> = (0..num).map(|i| (i..(16 + i)).collect()).collect(); + let seed: [u8; 16] = [ + 130, 220, 246, 217, 111, 124, 221, 189, 190, 234, 121, 93, 67, 95, 100, 43, + ]; + + let rng = &mut SmallRng::from_seed(seed); + let tx: Vec<Vec<char>> = (0..4096) + .map(|_| (rng.sample_iter(&Alphanumeric).take(32).collect())) + .collect(); + + for _ in 0..32 { + for i in 0..4096 { + assert_eq!(m.contains(&tx[i].clone()), false); + assert_eq!(m.insert(tx[i].clone()), true); + } + for i in 0..4096 { + println!("removing {} {:?}", i, tx[i]); + assert_eq!(m.remove(&tx[i]), true); + } + } +} |