A real diff does not print every unchanged line - it shows only the changed regions with a few lines of context around each. Today you group the edit script into hunks, one per run of changes, with a bounded band of context on each side.
Split an edit script into one hunk per change run, each carrying up to N lines of surrounding context.
Showing an entire file to point out a two-line change is wasteful, so diffs report hunks: each changed region plus a few unchanged context lines on either side (three by default), which help a reader - and a patch tool - locate the change. Today’s grouping is deliberately simple: find each maximal run of Delete/Insert operations, and wrap it with up to context kept lines before and after. Unchanged lines beyond that band are dropped, so the lonely line 8 between two far-apart changes appears in no hunk at all.
Record each hunk’s starting line numbers as you go - the 1-based position of its first line in the old and new files - because the header needs them later. When two changes are far apart, as here, their context bands are separate and this per-run grouping already gives correct, non-overlapping hunks. The interesting case is when two changes sit close together and their context bands overlap or touch; that needs a merge step, and it gets its own lesson shortly. For now, keep each change run in its own hunk and do not merge.
type Hunk struct {OldStart, NewStart intOps []Op}// for each maximal run of Delete/Insert ops, build a hunk from:// up to `context` Keep ops immediately before the run,// the run itself,// up to `context` Keep ops immediately after.// record OldStart/NewStart = 1-based line numbers of the hunk's first line.// (do NOT merge overlapping bands yet - that is the next-but-one lesson.)