Skip to content

Latest commit

 

History

History
185 lines (124 loc) · 6.24 KB

README.rst

File metadata and controls

185 lines (124 loc) · 6.24 KB
Raw

fuzzysearch

Latest Version Test Coverage Wheels Supported Python versions Supported Python implementations License

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")]