ExStrsim

Fast string similarity metrics for Elixir via Rust NIF, wrapping the strsim crate.

Precompiled binaries are available — no Rust toolchain required.

Installation

def deps do
  [
    {:ex_strsim, "~> 0.1"}
  ]
end

Usage

ExStrsim.levenshtein("kitten", "sitting")
#=> 3

ExStrsim.normalized_levenshtein("kitten", "sitting")
#=> 0.5714285714285714

ExStrsim.jaro_winkler("cheeseburger", "cheese fries")
#=> 0.9

ExStrsim.damerau_levenshtein("ab", "ba")
#=> 1

# Word-level edit distance
ExStrsim.generic_levenshtein(["the", "quick", "fox"], ["the", "slow", "dog"])
#=> 2

Available Functions

String Distance

Function	Returns	Description
`levenshtein/2`	integer	Edit distance (insertions, deletions, substitutions)
`damerau_levenshtein/2`	integer	Like Levenshtein, but transpositions count as one edit
`hamming/2`	integer / `{:error, reason}`	Hamming distance (equal-length strings only)
`osa_distance/2`	integer	Optimal String Alignment (restricted edit distance)

String Similarity

Function	Returns	Description
`normalized_levenshtein/2`	float	Levenshtein normalized to 0.0–1.0 (1.0 = identical)
`normalized_damerau_levenshtein/2`	float	Damerau-Levenshtein normalized to 0.0–1.0
`jaro/2`	float	Jaro similarity (0.0–1.0)
`jaro_winkler/2`	float	Jaro-Winkler similarity with common prefix boost
`sorensen_dice/2`	float	Sorensen-Dice coefficient (0.0–1.0)

Generic (Word-Level)

Operate on lists of strings instead of individual characters.

Function	Returns	Description
`generic_levenshtein/2`	integer	Levenshtein on lists of strings
`generic_damerau_levenshtein/2`	integer	Damerau-Levenshtein on lists of strings
`generic_hamming/2`	integer / `{:error, reason}`	Hamming on lists of strings (equal length only)
`generic_jaro/2`	float	Jaro similarity on lists of strings
`generic_jaro_winkler/2`	float	Jaro-Winkler similarity on lists of strings

Building from Source

If you need to compile the NIF locally (requires Rust toolchain):

EX_STRSIM_BUILD=1 mix deps.compile ex_strsim

License

MIT