StringDist

This package provides the stringdist module, which includes functions for calculating raw and normalized versions of the following string distance measurements:

Levenshtein distance
Restricted Damerau-Levenshtein distance (a.k.a. optimal string alignment distance)

For optimal performance, the package compiles and uses a C extension module under the hood, but a Python implementation is included as well and will automatically be used if C extensions are not supported by the system (e.g. when the selected interpreter is PyPy).

Installation

To install this package, just use pip:

pip install StringDist

All Python versions >=3.3 should be supported.

Usage

To use the package, simply import the stringdist module and call the desired function, passing in two strings:

import stringdist
stringdist.levenshtein('test', 'testing')

The available functions are as follows:

levenshtein
levenshtein_norm
rdlevenshtein
rdlevenshtein_norm

Raw distances assume that every allowed operation has a cost of 1. Normalized distances are floats in the range [0.0, 1.0], where 0.0 always corresponds to a raw value of 0 and 1.0 always corresponds to the length of the longer string, i.e. the biggest possible raw value.

Note: The restricted Damerau-Levenshtein distance is not a true distance metric because it does not satisfy the triangle inequality. This makes it a poor choice for applications that involve evaluating the similarity of more than two strings, such as clustering.

Bugs and Requests

Please use GitHub Issues for bugs and feature requests, checking first to make sure you're not creating a duplicate issue.

Contributing

Pull requests are welcome. Please discuss your plans first by creating a GitHub issue and use good coding style. For Python, this means following the rules laid out in PEP 8 and other relevant PEPs. If in doubt, use a linter like Pylint.

To run unit tests:

git clone https://github.com/obulkin/string-dist.git {directory}
cd {directory}
python setup.py install
python -m unittest -v test_stringdist

You can run tests without installing the package, but this will always cause the Python implementation to be used as the C variant has to be compiled first. By the same token, any changes to the C code will require recompilation before showing up in the tests, which can be handled by running python setup.py install again.

Contributors