edit-distance-linear: Efficient implementation of the Levenshtein edit distance in linear memory.

[ algorithms, bsd3, library ] [ Propose Tags ]

Modules

[Index] [Quick Jump]

Flags

Manual Flags

NameDescriptionDefault
llvm

Use the LLVM code generator (strongly recommended)

Disabled
with-executable

Build the test executable (a fast and dirty way of benchmarking, for the package development only)

Disabled

Use -f <flag> to enable a flag, or -f -<flag> to disable that flag. More info

Downloads

Maintainer's Corner

Package maintainers

For package maintainers and hackage trustees

Candidates

  • No Candidates
Versions [RSS] 0.2.0.1, 0.2.0.2
Change log ChangeLog.md
Dependencies array, base (>=4.7 && <5), bytestring, edit-distance-linear [details]
License BSD-3-Clause
Copyright 2019 Georg Rudoy
Author Georg Rudoy
Maintainer 0xd34df00d@gmail.com
Category Algorithms
Home page https://github.com/0xd34df00d/edit-distance-linear#readme
Bug tracker https://github.com/0xd34df00d/edit-distance-linear/issues
Source repo head: git clone https://github.com/0xd34df00d/edit-distance-linear
Uploaded by 0xd34df00d at 2019-12-08T22:44:39Z
Distributions NixOS:0.2.0.2
Executables edit-distance-linear-exe
Downloads 805 total (6 in the last 30 days)
Rating (no votes yet) [estimated by Bayesian average]
Your Rating
  • λ
  • λ
  • λ
Status Docs available [build log]
Last success reported on 2019-12-08 [all 1 reports]

Readme for edit-distance-linear-0.2.0.2

[back to package description]

edit-distance-linear

Build Status Hackage

The pure Haskell implementation of the Levenshtein edit distance, with linear space complexity.

Comparison

There are already several other existing implementations, but the goals and design decisions vary. In particular, this package is intended to be used to:

  • compare long strings (think tens of thousands of characters), driving the implementation to live in the ST monad and aim at linear space complexity to lower GC pressure;
  • not care about Unicode, thus accepting ByteStrings and comparing them byte-by-byte rather than character-by-character (or glyph-by-glyph, or whatever is the right notion of an edit for Unicode).

Among the alternatives:

  • text-metrics — uses a similar algorithm, but cares about Unicode, making it 4-5 times slower.
  • edit-distance — uses a very different algorithm (which we might implement here one day with huge potential benefits), which tends to consume more memory (I'm not up for estimating its space asymptotics, though).