toastdriven / pylev

A pure Python Levenshtein implementation that's not freaking GPL'd.

pylev

A pure Python Levenshtein implementation that's not freaking GPL'd.

Based off the Wikipedia code samples at http://en.wikipedia.org/wiki/Levenshtein_distance.

Requirements

• Python 2.7.X, Python 3.3+ or PyPy 1.6.0+

Usage

Usage is fairly straightforward:

import pylev
distance = pylev.levenshtein('kitten', 'sitting')
assert distance == 3

License

New BSD.

Tests

Setup:

$ git clone https://github.com/toastdriven/pylev.git
$ cd pylev

Running:

$ python -m unittest tests

Alternatives

• https://pypi.org/project/levenshtein/ - GPL
• https://pypi.org/project/python-Levenshtein/ - GPL
• https://pypi.org/project/fuzzywuzzy/ - GPL
• https://pypi.org/project/pylevenshtein/ - GPL
• https://pypi.org/project/leven/ - Unknown license & requires Cython

Version History

• v1.5.0

  • Added an alternatives section. Thanks to @graingert for the start!

• v1.4.0

  • Updated for current versions of Python
  • Integrated a better Travis matrix. Thanks to @graingert!
  • Fixed mistaken docs about the assert. Thanks to @adamchainz!
  • Reorganized the package.
  • Blacked all the source code.

• v1.3.0

  • Implemented a considerably faster variants (orders of magnitude).
  • Tested & working on Python 2.7.4, Python 3.3.1 & PyPy 1.9.0.

• v1.2.0

  • Fixed all incorrect spellings of "Levenshtein" (there's no "c" in it).
  • Old methods are aliased for backward-compatibility.

• v1.1.0

  • Implemented a much faster variant (several orders of magnitude).
  • The older variant was renamed to classic_levenschtein.
  • Tested & working on Python 3.3 & PyPy 1.6.0 as well.

• v1.0.2

  • Python packaging is REALLY hard. Including the README this time.

• v1.0.1

  • Python packaging is hard. Including the README this time.

• v1.0.0

  • Initial release, just the naive implementation of Levenshtein.

pylev's Issues

Clarify that the damerau_levenshtein function calculates the restricted Damerau-Levenshtein distance

There are two different distances called Damerau-Levenshtein (see the Wikipedia page, for example), one is the restricted distance and the other is the original one, the unrestricted distance. The damerau_levenshtein function in pylev calculates the restricted distance (so the distance between "CA" and "ABC" is 3, not 2). It would be good for this to be made clear in the documentation.

Thanks!

Use pylev.levenshtein

Hi,

I have downloaded the package and it works great.
Now I want to use the pylev.levenshtein package as the metric for the scipy linkage function, something like:
Z = linkage(my_array, method='average', metric=pylev.levenshtein).

What exactly do I need to use?
Thanks

This project is active?

Hi, I note that this project is used on clikit and is a depedency of poetry.

I'm trying to package poetry on Debian and I need package pylev too. I was
wondering to know if this project is still maintained.

Cheers

prev_cost variable never used

prev_cost variable is never used

pylev/pylev/damerau.py

Line 44 in 700700e

prev_cost = 0

LICENSE not included in pypi source

I'm working on packaging this project for fedora, and during the review, it was pointed out that the 1.2.0 sources that are available from PyPI don't include the LICENSE file - which they apparently should.

Test case for pylev.damerau_levenshtein() function is wrong

Link:

pylev/tests.py

Line 27 in 700700e

assert pylev.damerau_levenshtein("foobar", "foobra") == 1

What it is:

assert pylev.damerau_levenshtein("foobar", "foobra") == 1

What it should be:

assert pylev.damerau_levenshtein("foobar", "foobra") == 2

This project is active?

Hi, I note that this project is used on clikit and is a depedency of poetry.

I'm trying to package poetry on Debian and I need package pylev too. I was
wondering to know if this project is still maintained.

Cheers

Add v1.3.0 tag/release

I notice that the version on Pypi is 1.3.0 and the version in the README says 1.3.0 as of commit 72e3d49, but there isn't a Github release or a tag for v1.3.0. Can you add one? I'm packaging (NixOS/nixpkgs#53599) this for nixpkgs and it would be helpful.

Enable Travis CI

There's a .travis.yml but Travis had no builds yet so probably the hook was never added?

Convenience function: search

Happy to see this project active again.

I've copy pasted variations of this function between projects, perhaps it's worth considering inclusion.

def search(
    needle       : str,
    haystack     : Iterable[str],
    max_distance : int = 2,
    distance_func: Callable[[str, str], int] = pylev.damerau_levenshtein,
) -> List[str]:
    closest: List[Tuple[int, str]] = []
    for maybe_needle in haystack:
        dist = distance_func(needle, maybe_needle)
        if dist <= max_distance:
            closest.append((dist, maybe_needle))
    return [val for _dist, val in sorted(closest)]

Example usage.

if key not in VALID_KEYS:
    alternatives = pylev.search(key, VALID_KEYS)
    if alternatives: 
        raise KeyError(key, f"Maybe {alternatives} ?")
    else:
        raise KeyError(key)