GitoxideLabs/gitoxide gix-merge-v0.15.0

gix-merge v0.15.0

on GitHub

Chore

• make fuzz test myers algorithm resilient against pathological inputs.
  Achieved by removing Myers as it's way of operation can't really be fixed.
• add package.include directives to control which files are packaged.

New Features

• add blob::text::PreparedMerge to allow re-running with different Conflict resolutions.
  This is useful to quickly check different resolution results.

Bug Fixes

• enable diff-postprocessing in when merging blobs
  This can lead to cleaner, more Git-like merges.

• coalesce split Myers hunks to prevent false merge conflicts
  When imara-diff's Myers algorithm diffs two files, it sometimes splits
  what is logically one change into a non-empty deletion hunk and a
  separate empty insertion hunk, with one unchanged base line between
  them. This is a valid minimal edit script, but it differs from the
  alignment that git's xdiff (also Myers-based) would choose.

  When the empty insertion lands at a base position that the other side
  of a 3-way merge also touches, take_intersecting reports an overlap
  and the merge produces a conflict — even though git merge-file
  resolves the same inputs cleanly.

  Fix this by adding a pre-processing step after sorting hunks: for each
  empty-range insertion hunk, look backwards for the nearest same-side
  non-empty hunk within a gap of at most one unchanged base line. If
  found, extend that hunk to cover the gap and the insertion point. This
  re-joins the split hunk, making the merge robust to different diff
  algorithm alignment choices.

  The coalescing is conservative: it only applies when (a) the insertion
  hunk has an empty before-range, (b) there is a same-side non-empty
  hunk nearby (gap ≤ 1), and (c) that hunk is the nearest same-side
  hunk. This avoids affecting cases like zdiff3-interesting where empty
  insertions are standalone and represent genuinely different changes.

New Features (BREAKING)

• Use imara-diff-v2 with git sliders processing
  The slider post-processing imrpoves the diff quality for about 8% slower diffs.
  Line-counts, however, will be 50% faster to compute.

Commit Statistics

• 20 commits contributed to the release over the course of 32 calendar days.
• 32 days passed between releases.
• 6 commits were understood as conventional.
• 0 issues like '(#ID)' were seen in commit messages

Thanks Clippy

Clippy helped 1 time to make code idiomatic.

Commit Details