Releases: rapidfuzz/rapidfuzz-cpp
Releases · rapidfuzz/rapidfuzz-cpp
Release 3.0.5
Fixed
- the editops implementation didn't properly account for some cells in the Levenshtein matrix.
This could lead both to incorrect results and crashes.
Release 3.0.4
Fixed
- fix tagged version
Release 3.0.3
Fixed
- fix potentially incorrect results of JaroWinkler when using high prefix weights
Release 3.0.2
Fixed
- fix assert leading to compilation failures
Release 3.0.1
Fixed
- fix doxygen warnings
Release 3.0.0
Performance
- add banded implementation of LCS / Indel. This improves the runtime from
O((|s1|/64) * |s2|)
toO((score_cutoff/64) * |s2|)
Changed
- changed many types in the interface from int64_t to size_t, since they can't be negative.
Fixed
- fix incorrect transposition calculation in simd implementation of Jaro similarity
- use posix_memalign on android
Release 2.2.3
Fixed
- use _mm_malloc/_mm_free on macOS if aligned_alloc is unsupported
Release 2.2.2
Fixed
- fix compilation failure on macOS
Release 2.2.1
Fixed
- fix wraparound issue in simd implementation of Jaro and Jaro Winkler
Release 2.2.0
Performance
- improve performance of simd implementation for LCS and Indel by up to 50%
- improve performance of simd implementation for Jaro and Jaro Winkler
- improve performance of Jaro and Jaro Winkler for long sequences