Module engine 

The diff engine: Myers’ O(ND) algorithm in linear space, hunk grouping, token-level diffs for intra-line emphasis and the diff -u text format.

Elements are interned to integers first, a common prefix and suffix are trimmed, and regions with no element in common are resolved without a search, so large inputs with few changes stay fast. Past an edit cost of COST_LIMIT the search splits at its furthest-reaching point instead of the exact middle snake (git’s heuristic): the script stays valid, it just may not be minimal.

Structs

Hunk

A group of changes with surrounding context, as diff -u prints them.

TextDiff

A line-level diff of two texts.

Enums

Op

One run of an edit script. Ranges index the compared sequences: lines for diff_lines, bytes of the input strings for diff_words and diff_chars. The side a run does not touch has an empty range at the position where the run applies.

Constants

COST_LIMIT

The edit cost past which the middle-snake search takes a heuristic split.

Functions

diff_chars

Diff two strings character by character; ranges are byte offsets.

diff_lines

Diff two sequences of lines. Include line terminators in the elements when a missing final newline should count as a change, as diff does.

diff_slices

Diff two sequences of any hashable elements.

diff_words

Diff two strings by tokenized words; ranges are byte offsets.

group_hunks

Group an edit script into hunks with context unchanged elements around every change; changes closer than twice the context share a hunk.

hunk_header

@@ -a,b +c,d @@ for 0-based line ranges, with diff -u’s conventions: a single line omits ,1, and an empty range names the line before it.

tokenize

Split text into word, whitespace and punctuation tokens, as byte ranges. A word is a run of alphanumerics and _; whitespace runs are one token; every other character is its own token.