RapidFuzz/CHANGELOG

80 lines
2.3 KiB
Plaintext

## [0.6.0] - 2020 Apr 4
### Changed
- Remove vector of word based levenshtein since it was not much faster but added complexity
- use a bitmap based ratio to reduce the usage of levenshtein in extractOne for bad matches
- replace std::wstring_view with boost::wstring_view to add C++14 support
## [0.5.3] - 2020 Apr 4
### Fixed
- Process methods with embedded null characters in the returned string did only
return a string until the first null character
## [0.5.2] - 2020 Apr 3
### Changed
- the default processor replaces null characters in the string by whitespaces now
### Fixed
- Fixed a segfault that occured when extracting the boolean for unknown reason
replacing it by a short int appears to solve the problem
- Fixed a TypeError when strings have embedded null characters
## [0.5.1] - 2020 Apr 1
### Fixed
- Fixed a error in the Windows build
## [0.5.0] - 2020 Apr 1
### Changed
- levenshtein.weighted_distance does now support custom weighted
- PyBind11 is replaced by the Python C API, which makes the function calls a bit faster
### Fixed
- There is no longer a hidden module so this issue with pyinstaller can not occur anymore
## [0.4.0] - 2020 Mar 29
### Added
- levenshtein module that provides access to
- weighted distance
- normalized weighted distance
- distance
- normalized distance
### Fixed
- extract returned a list of tuples of empty string and 0 when not enough elements had a ratio
bigger than score_cutoff
- fix wrong sorting of results causing wrong results to be removed in extract
## [0.3.2] - 2020 Mar 29
### Added
- Readded QRatio as alias for ratio to stay conform with fuzzywuzzy
- add extractBests process as alias for extract since it is a better name for it
## [0.3.1] - 2020 Mar 27
### Changed
- Implement default_process in python since it is faster than the overhead to call into python
## [0.3.0] - 2020 Mar 26
### Changed
- When using score_cutoff there is now a guarantee that it returns 0 when result < score_cutoff
- each function has now a preprocess argument to specify whether strings should be preprocessed
- the default preprocessing does now lower case and trim
- extract and extractOne accept a custom processor method instead of utils.default_processor now
### Removed
- QRatio was removed since it does now exactly the same as normal ratio