String Similarity Measures for Polars

This package provides python bindings to compute various string similarity measures directly on a polars dataframe. All string similarity measures are implemented in rust and computed in parallel.

The similarity measures that have been implemented are:

• Levenshtein
• Jaro
• Jaro-Winkler
• Jaccard
• Sørensen-Dice

Each similarity measure returns a value normalized between 0.0 and 1.0 (inclusive), where 0.0 indicates the inputs are maximally different and 1.0 means the strings are maximally similar.

Installing the Library

With pip

pip install polars-strsim

From Source

To build and install this library from source, first ensure you have cargo installed. You will also need maturin, which you can install via pip install 'maturin[patchelf]'

polars-strsim can then be installed in your current python environment by running maturin develop --release

Using the Library

Input:

import polars as pl
from polars_strsim import levenshtein, jaro, jaro_winkler, jaccard, sorensen_dice

df = pl.DataFrame(
    {
        "name_a": ["phillips", "phillips", ""        , "", None      , None],
        "name_b": ["phillips", "philips" , "phillips", "", "phillips", None],
    }
).with_columns(
    levenshtein=levenshtein("name_a", "name_b"),
    jaro=jaro("name_a", "name_b"),
    jaro_winkler=jaro_winkler("name_a", "name_b"),
    jaccard=jaccard("name_a", "name_b"),
    sorensen_dice=sorensen_dice("name_a", "name_b"),
)

with pl.Config(ascii_tables=True):
    print(df)

Output:

shape: (6, 7)
+----------+----------+-------------+----------+--------------+---------+---------------+
| name_a   | name_b   | levenshtein | jaro     | jaro_winkler | jaccard | sorensen_dice |
| ---      | ---      | ---         | ---      | ---          | ---     | ---           |
| str      | str      | f64         | f64      | f64          | f64     | f64           |
+=======================================================================================+
| phillips | phillips | 1.0         | 1.0      | 1.0          | 1.0     | 1.0           |
| phillips | philips  | 0.875       | 0.958333 | 0.975        | 0.875   | 0.933333      |
|          | phillips | 0.0         | 0.0      | 0.0          | 0.0     | 0.0           |
|          |          | 1.0         | 1.0      | 1.0          | 1.0     | 1.0           |
| null     | phillips | null        | null     | null         | null    | null          |
| null     | null     | null        | null     | null         | null    | null          |
+----------+----------+-------------+----------+--------------+---------+---------------+