fparkan/vendor/strsim/CHANGELOG.md

227 lines
5.8 KiB
Markdown
Raw Normal View History

# Change Log
This project attempts to adhere to [Semantic Versioning](http://semver.org).
## [Unreleased]
## [0.10.1] - (2024-01-07)
### Changed
- improve OSA implementation
- reduce runtime
- reduce binary size by more than `25%`
- reduce binary size of Levenshtein distance
- improve Damerau-Levenshtein implementation
- reduce memory usage from `O(N*M)` to `O(N+M)`
- reduce runtime in our own benchmark by more than `70%`
- reduce binary size by more than `25%`
- only boost similarity in Jaro-Winkler once the Jaro similarity exceeds 0.7
### Fixed
- Fix transposition counting in Jaro and Jaro-Winkler.
- Limit common prefix in Jaro-Winkler to 4 characters
## [0.10.0] - (2020-01-31)
### Added
- Sørensen-Dice implementation (thanks [@robjtede](https://github.com/robjtede))
## [0.9.3] - (2019-12-12)
### Fixed
- Fix Jaro and Jaro-Winkler when the arguments have lengths of 1 and are equal.
Previously, the functions would erroneously return 0 instead of 1. Thanks to
[@vvrably](https://github.com/vvrably) for pointing out the issue.
## [0.9.2] - (2019-05-09)
### Changed
- Revert back to the standard library hashmap because it will use hashbrown very
soon
- Remove ndarray in favor of using a single vector to represent the 2d grid in
Damerau-Levenshtein
## [0.9.1] - (2019-04-08)
### Changed
- Faster Damerau-Levenshtein implementation (thanks [@lovasoa](https://github.com/lovasoa))
## [0.9.0] - (2019-04-06)
### Added
- Generic distance functions (thanks [@lovasoa](https://github.com/lovasoa))
## [0.8.0] - (2018-08-19)
### Added
- Normalized versions of Levenshtein and Damerau-Levenshtein (thanks [@gentoid](https://github.com/gentoid))
## [0.7.0] - (2018-01-17)
### Changed
- Faster Levenshtein implementation (thanks [@wdv4758h](https://github.com/wdv4758h))
### Removed
- Remove the "against_vec" functions. They are one-liners now, so they don't
seem to add enough value to justify making the API larger. I didn't find
anybody using them when I skimmed through a GitHub search. If you do use them,
you can change the calls to something like:
```rust
let distances = strings.iter().map(|a| jaro(target, a)).collect();
```
## [0.6.0] - (2016-12-26)
### Added
- Add optimal string alignment distance
### Fixed
- Fix Damerau-Levenshtein implementation (previous implementation was actually
optimal string alignment; see this [Damerau-Levenshtein explanation])
## [0.5.2] - (2016-11-21)
### Changed
- Remove Cargo generated documentation in favor of a [docs.rs] link
## [0.5.1] - (2016-08-23)
### Added
- Add Cargo generated documentation
### Fixed
- Fix panic when Jaro or Jaro-Winkler are given strings both with a length of
one
## [0.5.0] - (2016-08-11)
### Changed
- Make Hamming faster (thanks @IBUzPE9) when the two strings have the same
length but slower when they have different lengths
## [0.4.1] - (2016-04-18)
### Added
- Add Vagrant setup for development
- Add AppVeyor configuration for Windows CI
### Fixed
- Fix metrics when given strings with multibyte characters (thanks @WanzenBug)
## [0.4.0] - (2015-06-10)
### Added
- For each metric, add a function that takes a vector of strings and returns a
vector of results (thanks @ovarene)
## [0.3.0] - (2015-04-30)
### Changed
- Remove usage of unstable Rust features
## [0.2.5] - (2015-04-24)
### Fixed
- Remove unnecessary `Float` import from doc tests
## [0.2.4] - (2015-04-15)
### Fixed
- Remove unused `core` feature flag
## [0.2.3] - (2015-04-01)
### Fixed
- Remove now unnecessary `Float` import
## [0.2.2] - (2015-03-29)
### Fixed
- Remove usage of `char_at` (marked as unstable)
## [0.2.1] - (2015-02-20)
### Fixed
- Update bit vector import to match Rust update
## [0.2.0] - (2015-02-19)
### Added
- Implement Damerau-Levenshtein
- Add tests in docs
## [0.1.1] - (2015-02-10)
### Added
- Configure Travis for CI
- Add rustdoc comments
### Fixed
- Limit Jaro-Winkler return value to a maximum of 1.0
- Fix float comparisons in tests
## [0.1.0] - (2015-02-09)
### Added
- Implement Hamming, Jaro, Jaro-Winkler, and Levenshtein
[Unreleased]: https://github.com/rapidfuzz/strsim-rs/compare/0.10.1...HEAD
[0.10.1]: https://github.com/rapidfuzz/strsim-rs/compare/0.10.0...0.10.1
[0.10.0]: https://github.com/rapidfuzz/strsim-rs/compare/0.9.3...0.10.0
[0.9.3]: https://github.com/rapidfuzz/strsim-rs/compare/0.9.2...0.9.3
[0.9.2]: https://github.com/rapidfuzz/strsim-rs/compare/0.9.1...0.9.2
[0.9.1]: https://github.com/rapidfuzz/strsim-rs/compare/0.9.0...0.9.1
[0.9.0]: https://github.com/rapidfuzz/strsim-rs/compare/0.8.0...0.9.0
[0.8.0]: https://github.com/rapidfuzz/strsim-rs/compare/0.7.0...0.8.0
[0.7.0]: https://github.com/rapidfuzz/strsim-rs/compare/0.6.0...0.7.0
[0.6.0]: https://github.com/rapidfuzz/strsim-rs/compare/0.5.2...0.6.0
[0.5.2]: https://github.com/rapidfuzz/strsim-rs/compare/0.5.1...0.5.2
[0.5.1]: https://github.com/rapidfuzz/strsim-rs/compare/0.5.0...0.5.1
[0.5.0]: https://github.com/rapidfuzz/strsim-rs/compare/0.4.1...0.5.0
[0.4.1]: https://github.com/rapidfuzz/strsim-rs/compare/0.4.0...0.4.1
[0.4.0]: https://github.com/rapidfuzz/strsim-rs/compare/0.3.0...0.4.0
[0.3.0]: https://github.com/rapidfuzz/strsim-rs/compare/0.2.5...0.3.0
[0.2.5]: https://github.com/rapidfuzz/strsim-rs/compare/0.2.4...0.2.5
[0.2.4]: https://github.com/rapidfuzz/strsim-rs/compare/0.2.3...0.2.4
[0.2.3]: https://github.com/rapidfuzz/strsim-rs/compare/0.2.2...0.2.3
[0.2.2]: https://github.com/rapidfuzz/strsim-rs/compare/0.2.1...0.2.2
[0.2.1]: https://github.com/rapidfuzz/strsim-rs/compare/0.2.0...0.2.1
[0.2.0]: https://github.com/rapidfuzz/strsim-rs/compare/0.1.1...0.2.0
[0.1.1]: https://github.com/rapidfuzz/strsim-rs/compare/0.1.0...0.1.1
[0.1.0]: https://github.com/rapidfuzz/strsim-rs/compare/fabad4...0.1.0
[docs.rs]: https://docs.rs/strsim/
[Damerau-Levenshtein explanation]:
http://scarcitycomputing.blogspot.com/2013/04/damerau-levenshtein-edit-distance.html