fuzzysearch 0.8.0

fuzzysearch is useful for finding approximate subsequence matches

Fuzzy search: Find parts of long text or data, allowing for some changes/typos.

Highly optimized, simple to use, does one thing well.

>>> find_near_matches('PATTERN', '---PATERN---', max_l_dist=1)
[Match(start=3, end=9, dist=1, matched="PATERN")]

• Two simple functions to use: one for in-memory data and one for files

  • Fastest search algorithm is chosen automatically

• Levenshtein Distance metric with configurable parameters

  • Separately configure the max. allowed distance, substitutions, deletions and/or insertions

• Advanced algorithms with optional C and Cython optimizations

• Properly handles Unicode; special optimizations for binary data

• Simple installation:

  • pip install fuzzysearch just works

  • pure-Python fallbacks for compiled modules

  • only one dependency (attrs)

• Extensively tested

• Free software: MIT license

For more info, see the documentation.

How is this different than FuzzyWuzzy or RapidFuzz?

The main difference is that fuzzysearch searches for fuzzy matches through long texts or data. FuzzyWuzzy and RapidFuzz, on the other hand, are intended for fuzzy comparison of pairs of strings, identifying how closely they match according to some metric such as the Levenshtein distance.

These are very different use-cases, and the solutions are very different as well.

How is this different than ElasticSearch and Lucene?

The main difference is that fuzzysearch does no indexing or other preparations; it directly searches through the given text or data for a given sub-string. Therefore, it is much simpler to use compared to systems based on text indexing.

Installation

fuzzysearch supports Python versions 3.8+, as well as PyPy 3.9 and 3.10.

$ pip install fuzzysearch

This will work even if installing the C and Cython extensions fails, using pure-Python fallbacks.

Usage

Just call find_near_matches() with the sub-sequence you’re looking for, the sequence to search, and the matching parameters:

>>> from fuzzysearch import find_near_matches
# search for 'PATTERN' with a maximum Levenshtein Distance of 1
>>> find_near_matches('PATTERN', '---PATERN---', max_l_dist=1)
[Match(start=3, end=9, dist=1, matched="PATERN")]

To search in a file, use find_near_matches_in_file():

>>> from fuzzysearch import find_near_matches_in_file
>>> with open('data_file', 'rb') as f:
...     find_near_matches_in_file(b'PATTERN', f, max_l_dist=1)
[Match(start=3, end=9, dist=1, matched="PATERN")]

Examples

fuzzysearch is great for ad-hoc searches of genetic data, such as DNA or protein sequences, before reaching for more complex tools:

>>> sequence = '''\
GACTAGCACTGTAGGGATAACAATTTCACACAGGTGGACAATTACATTGAAAATCACAGATTGGTCACACACACA
TTGGACATACATAGAAACACACACACATACATTAGATACGAACATAGAAACACACATTAGACGCGTACATAGACA
CAAACACATTGACAGGCAGTTCAGATGATGACGCCCGACTGATACTCGCGTAGTCGTGGGAGGCAAGGCACACAG
GGGATAGG'''
>>> subsequence = 'TGCACTGTAGGGATAACAAT' # distance = 1
>>> find_near_matches(subsequence, sequence, max_l_dist=2)
[Match(start=3, end=24, dist=1, matched="TAGCACTGTAGGGATAACAAT")]

BioPython sequences are also supported:

>>> from Bio.Seq import Seq
>>> from Bio.Alphabet import IUPAC
>>> sequence = Seq('''\
GACTAGCACTGTAGGGATAACAATTTCACACAGGTGGACAATTACATTGAAAATCACAGATTGGTCACACACACA
TTGGACATACATAGAAACACACACACATACATTAGATACGAACATAGAAACACACATTAGACGCGTACATAGACA
CAAACACATTGACAGGCAGTTCAGATGATGACGCCCGACTGATACTCGCGTAGTCGTGGGAGGCAAGGCACACAG
GGGATAGG''', IUPAC.unambiguous_dna)
>>> subsequence = Seq('TGCACTGTAGGGATAACAAT', IUPAC.unambiguous_dna)
>>> find_near_matches(subsequence, sequence, max_l_dist=2)
[Match(start=3, end=24, dist=1, matched="TAGCACTGTAGGGATAACAAT")]

Matching Criteria

The search function supports four possible match criteria, which may be supplied in any combination:

• maximum Levenshtein distance (max_l_dist)

• maximum # of subsitutions

• maximum # of deletions (“delete” = skip a character in the sub-sequence)

• maximum # of insertions (“insert” = skip a character in the sequence)

Not supplying a criterion means that there is no limit for it. For this reason, one must always supply max_l_dist and/or all other criteria.

>>> find_near_matches('PATTERN', '---PATERN---', max_l_dist=1)
[Match(start=3, end=9, dist=1, matched="PATERN")]

# this will not match since max-deletions is set to zero
>>> find_near_matches('PATTERN', '---PATERN---', max_l_dist=1, max_deletions=0)
[]

# note that a deletion + insertion may be combined to match a substution
>>> find_near_matches('PATTERN', '---PAT-ERN---', max_deletions=1, max_insertions=1, max_substitutions=0)
[Match(start=3, end=10, dist=1, matched="PAT-ERN")] # the Levenshtein distance is still 1

# ... but deletion + insertion may also match other, non-substitution differences
>>> find_near_matches('PATTERN', '---PATERRN---', max_deletions=1, max_insertions=1, max_substitutions=0)
[Match(start=3, end=10, dist=2, matched="PATERRN")]

History

0.8.0 (????-??-??)

• Dropped support for Python 2.7, 3.5, 3.6 and 3.7.

• Added support for Python 3.9, 3.10, 3.11 and 3.12.

0.7.3 (2020-06-27)

• Fixed segmentation faults due to wrong handling of inputs in bytes-like-only functions in C extensions.

0.7.2 (2020-05-07)

• Added PyPy support.

• Several minor bug fixes.

0.7.1 (2020-04-05)

• Dropped support for Python 3.4.

• Removed deprecation warning with Python 3.8.

• Fixed a couple of nasty bugs.

0.7.0 (2020-01-14)

• Added matched attribue to Match objects containing the matched part of the sequence.

• Added support for CPython 3.8. Now supporting CPython 2.7 and 3.4-3.8.

0.6.2 (2019-04-22)

• Fix calling search_exact() without passing end_index.

• Fix edge case: max. dist >= sub-sequence length.

0.6.1 (2018-12-08)

• Fixed some C compiler warnings for the C and Cython modules

0.6.0 (2018-12-07)

• Dropped support for Python versions 2.6, 3.2 and 3.3

• Added support and testing for Python 3.7

• Optimized the n-grams Levenshtein search for long sub-sequences

• Further optimized the n-grams Levenshtein search

• Cython versions of the optimized parts of the n-grams Levenshtein search

0.5.0 (2017-09-05)

• Fixed search_exact_byteslike() to support supplying start and end indexes

• Added support for lists, tuples and other Sequence types to search_exact()

• Fixed a bug where find_near_matches() could return a wrong Match.end with max_l_dist=0

• Added more tests and improved some existing ones.

0.4.0 (2017-07-06)

• Added support and testing for Python 3.5 and 3.6

• Many small improvements to README, setup.py and CI testing

0.3.0 (2015-02-12)

• Added C extensions for several search functions as well as internal functions

• Use C extensions if available, or pure-Python implementations otherwise

• setup.py attempts to build C extensions, but installs without if build fails

• Added --noexts setup.py option to avoid trying to build the C extensions

• Greatly improved testing and coverage

0.2.2 (2014-03-27)

• Added support for searching through BioPython Seq objects

• Added specialized search function allowing only subsitutions and insertions

• Fixed several bugs

0.2.1 (2014-03-14)

• Fixed major match grouping bug

0.2.0 (2013-03-13)

• New utility function find_near_matches() for easier use

• Additional documentation

0.1.0 (2013-11-12)

• Two working implementations

• Extensive test suite; all tests passing

• Full support for Python 2.6-2.7 and 3.1-3.3

• Bumped status from Pre-Alpha to Alpha

0.0.1 (2013-11-01)

• First release on PyPI.