From b01bc1a344c47ded15342f7872832daa1bf5cfce Mon Sep 17 00:00:00 2001 From: Runxi Yu Date: Sat, 21 Feb 2026 13:25:21 +0800 Subject: diff/lines: Line-based diffs via Myers --- diff/lines/diff.go | 223 +++++++++++++++++++++++++++++++++ diff/lines/diff_test.go | 326 ++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 549 insertions(+) create mode 100644 diff/lines/diff.go create mode 100644 diff/lines/diff_test.go diff --git a/diff/lines/diff.go b/diff/lines/diff.go new file mode 100644 index 00000000..fefec5de --- /dev/null +++ b/diff/lines/diff.go @@ -0,0 +1,223 @@ +// Package lines provides routines to perform line-based diffs. +package lines + +import "bytes" + +// Diff performs a line-based diff. +// Lines are bytes up to and including '\n' (final line may lack '\n'). +func Diff(oldB, newB []byte) ([]Chunk, error) { + type lineRef struct { + base []byte + start int + end int + } + + split := func(b []byte) []lineRef { + if len(b) == 0 { + return nil + } + var res []lineRef + start := 0 + for i := range b { + if b[i] == '\n' { + res = append(res, lineRef{base: b, start: start, end: i + 1}) + start = i + 1 + } + } + if start < len(b) { + res = append(res, lineRef{base: b, start: start, end: len(b)}) + } + return res + } + + oldLines := split(oldB) + newLines := split(newB) + + n := len(oldLines) + m := len(newLines) + if n == 0 && m == 0 { + return nil, nil + } + + idOf := make(map[string]int) + nextID := 0 + oldIDs := make([]int, n) + for i, ln := range oldLines { + key := string(ln.base[ln.start:ln.end]) + id, ok := idOf[key] + if !ok { + id = nextID + idOf[key] = id + nextID++ + } + oldIDs[i] = id + } + newIDs := make([]int, m) + for i, ln := range newLines { + key := string(ln.base[ln.start:ln.end]) + id, ok := idOf[key] + if !ok { + id = nextID + idOf[key] = id + nextID++ + } + newIDs[i] = id + } + + max := n + m + offset := max + trace := make([][]int, 0, max+1) + + Vprev := make([]int, 2*max+1) + for i := range Vprev { + Vprev[i] = -1 + } + + x0 := 0 + y0 := 0 + for x0 < n && y0 < m && oldIDs[x0] == newIDs[y0] { + x0++ + y0++ + } + Vprev[offset+0] = x0 + trace = append(trace, append([]int(nil), Vprev...)) + + found := x0 >= n && y0 >= m + + for D := 1; D <= max && !found; D++ { + V := make([]int, 2*max+1) + for i := range V { + V[i] = -1 + } + + for k := -D; k <= D; k += 2 { + var x int + if k == -D || (k != D && Vprev[offset+(k-1)] < Vprev[offset+(k+1)]) { + x = Vprev[offset+(k+1)] + } else { + x = Vprev[offset+(k-1)] + 1 + } + y := x - k + + for x < n && y < m && oldIDs[x] == newIDs[y] { + x++ + y++ + } + V[offset+k] = x + + if x >= n && y >= m { + trace = append(trace, V) + found = true + break + } + } + + if !found { + trace = append(trace, V) + Vprev = V + } + } + + type edit struct { + kind ChunkKind + lineref lineRef + } + revEdits := make([]edit, 0, n+m) + + x := n + y := m + for D := len(trace) - 1; D >= 0; D-- { + k := x - y + + var ( + prevK int + prevX int + prevY int + ) + if D > 0 { + prevV := trace[D-1] + if k == -D || (k != D && prevV[offset+(k-1)] < prevV[offset+(k+1)]) { + prevK = k + 1 + } else { + prevK = k - 1 + } + prevX = prevV[offset+prevK] + prevY = prevX - prevK + } + + for x > prevX && y > prevY { + x-- + y-- + revEdits = append(revEdits, edit{kind: ChunkKindUnchanged, lineref: oldLines[x]}) + } + + if D == 0 { + break + } + + if x == prevX { + y-- + revEdits = append(revEdits, edit{kind: ChunkKindAdded, lineref: newLines[y]}) + } else { + x-- + revEdits = append(revEdits, edit{kind: ChunkKindDeleted, lineref: oldLines[x]}) + } + } + + for i, j := 0, len(revEdits)-1; i < j; i, j = i+1, j-1 { + revEdits[i], revEdits[j] = revEdits[j], revEdits[i] + } + + var out []Chunk + type meta struct { + base []byte + start int + end int + } + var metas []meta + + for _, e := range revEdits { + curBase := e.lineref.base + curStart := e.lineref.start + curEnd := e.lineref.end + + if len(out) == 0 || out[len(out)-1].Kind != e.kind { + out = append(out, Chunk{Kind: e.kind, Data: curBase[curStart:curEnd]}) + metas = append(metas, meta{base: curBase, start: curStart, end: curEnd}) + continue + } + + lastIdx := len(out) - 1 + lastMeta := metas[lastIdx] + + if bytes.Equal(lastMeta.base, curBase) && lastMeta.end == curStart { + metas[lastIdx].end = curEnd + out[lastIdx].Data = curBase[metas[lastIdx].start:metas[lastIdx].end] + continue + } + + out[lastIdx].Data = append(out[lastIdx].Data, curBase[curStart:curEnd]...) + metas[lastIdx] = meta{base: nil, start: 0, end: 0} + } + + return out, nil +} + +// Chunk represents a contiguous region of lines categorized +// as unchanged, deleted, or added. +type Chunk struct { + Kind ChunkKind + Data []byte +} + +// ChunkKind enumerates the type of diff chunk. +type ChunkKind int + +const ( + // ChunkKindUnchanged represents an unchanged diff chunk. + ChunkKindUnchanged ChunkKind = iota + // ChunkKindDeleted represents a deleted diff chunk. + ChunkKindDeleted + // ChunkKindAdded represents an added diff chunk. + ChunkKindAdded +) diff --git a/diff/lines/diff_test.go b/diff/lines/diff_test.go new file mode 100644 index 00000000..9f0d1ab8 --- /dev/null +++ b/diff/lines/diff_test.go @@ -0,0 +1,326 @@ +package lines + +import ( + "bytes" + "strconv" + "strings" + "testing" +) + +func TestDiff(t *testing.T) { + t.Parallel() + + tests := []struct { + name string + oldInput string + newInput string + expected []Chunk + }{ + { + name: "empty inputs produce no chunks", + oldInput: "", + newInput: "", + expected: []Chunk{}, + }, + { + name: "only additions", + oldInput: "", + newInput: "alpha\nbeta\n", + expected: []Chunk{ + {Kind: ChunkKindAdded, Data: []byte("alpha\nbeta\n")}, + }, + }, + { + name: "only deletions", + oldInput: "alpha\nbeta\n", + newInput: "", + expected: []Chunk{ + {Kind: ChunkKindDeleted, Data: []byte("alpha\nbeta\n")}, + }, + }, + { + name: "unchanged content is grouped", + oldInput: "same\nlines\n", + newInput: "same\nlines\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("same\nlines\n")}, + }, + }, + { + name: "insertion in the middle", + oldInput: "a\nb\nc\n", + newInput: "a\nb\nX\nc\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("a\nb\n")}, + {Kind: ChunkKindAdded, Data: []byte("X\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("c\n")}, + }, + }, + { + name: "replacement without trailing newline", + oldInput: "first\nsecond", + newInput: "first\nsecond\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("first\n")}, + {Kind: ChunkKindDeleted, Data: []byte("second")}, + {Kind: ChunkKindAdded, Data: []byte("second\n")}, + }, + }, + { + name: "line replacement", + oldInput: "a\nb\nc\n", + newInput: "a\nB\nc\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("a\n")}, + {Kind: ChunkKindDeleted, Data: []byte("b\n")}, + {Kind: ChunkKindAdded, Data: []byte("B\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("c\n")}, + }, + }, + { + name: "swap adjacent lines", + oldInput: "A\nB\n", + newInput: "B\nA\n", + expected: []Chunk{ + {Kind: ChunkKindDeleted, Data: []byte("A\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("B\n")}, + {Kind: ChunkKindAdded, Data: []byte("A\n")}, + }, + }, + { + name: "indentation change is a full line replacement", + oldInput: "func main() {\n\treturn\n}\n", + newInput: "func main() {\n return\n}\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("func main() {\n")}, + {Kind: ChunkKindDeleted, Data: []byte("\treturn\n")}, + {Kind: ChunkKindAdded, Data: []byte(" return\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("}\n")}, + }, + }, + { + name: "commenting out lines", + oldInput: "code\n", + newInput: "// code\n", + expected: []Chunk{ + {Kind: ChunkKindDeleted, Data: []byte("code\n")}, + {Kind: ChunkKindAdded, Data: []byte("// code\n")}, + }, + }, + { + name: "reducing repeating lines", + oldInput: "log\nlog\nlog\n", + newInput: "log\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("log\n")}, + {Kind: ChunkKindDeleted, Data: []byte("log\nlog\n")}, + }, + }, + { + name: "expanding repeating lines", + oldInput: "tick\n", + newInput: "tick\ntick\ntick\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("tick\n")}, + {Kind: ChunkKindAdded, Data: []byte("tick\ntick\n")}, + }, + }, + { + name: "interleaved modifications", + oldInput: "keep\nchange\nkeep\nchange\n", + newInput: "keep\nfixed\nkeep\nfixed\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("keep\n")}, + {Kind: ChunkKindDeleted, Data: []byte("change\n")}, + {Kind: ChunkKindAdded, Data: []byte("fixed\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("keep\n")}, + {Kind: ChunkKindDeleted, Data: []byte("change\n")}, + {Kind: ChunkKindAdded, Data: []byte("fixed\n")}, + }, + }, + { + name: "large common header and footer", + oldInput: "header\nheader\nheader\nOLD\nfooter\nfooter\n", + newInput: "header\nheader\nheader\nNEW\nfooter\nfooter\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("header\nheader\nheader\n")}, + {Kind: ChunkKindDeleted, Data: []byte("OLD\n")}, + {Kind: ChunkKindAdded, Data: []byte("NEW\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("footer\nfooter\n")}, + }, + }, + { + name: "completely different content", + oldInput: "apple\nbanana\n", + newInput: "cherry\ndate\n", + expected: []Chunk{ + {Kind: ChunkKindDeleted, Data: []byte("apple\nbanana\n")}, + {Kind: ChunkKindAdded, Data: []byte("cherry\ndate\n")}, + }, + }, + { + name: "unicode and emoji changes", + oldInput: "Hello 🌍\nYay\n", + newInput: "Hello 🌎\nYay\n", + expected: []Chunk{ + {Kind: ChunkKindDeleted, Data: []byte("Hello 🌍\n")}, + {Kind: ChunkKindAdded, Data: []byte("Hello 🌎\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("Yay\n")}, + }, + }, + { + name: "binary data with embedded newlines", + oldInput: "\x00\x01\n\x02\x03\n", + newInput: "\x00\x01\n\x02\xFF\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("\x00\x01\n")}, + {Kind: ChunkKindDeleted, Data: []byte("\x02\x03\n")}, + {Kind: ChunkKindAdded, Data: []byte("\x02\xFF\n")}, + }, + }, + { + name: "adding trailing newline to last line", + oldInput: "Line 1\nLine 2", + newInput: "Line 1\nLine 2\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("Line 1\n")}, + {Kind: ChunkKindDeleted, Data: []byte("Line 2")}, + {Kind: ChunkKindAdded, Data: []byte("Line 2\n")}, + }, + }, + { + name: "removing trailing newline", + oldInput: "A\nB\n", + newInput: "A\nB", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("A\n")}, + {Kind: ChunkKindDeleted, Data: []byte("B\n")}, + {Kind: ChunkKindAdded, Data: []byte("B")}, + }, + }, + { + name: "inserting blank lines", + oldInput: "A\nB\n", + newInput: "A\n\n\nB\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("A\n")}, + {Kind: ChunkKindAdded, Data: []byte("\n\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("B\n")}, + }, + }, + { + name: "collapsing blank lines", + oldInput: "A\n\n\n\nB\n", + newInput: "A\nB\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("A\n")}, + {Kind: ChunkKindDeleted, Data: []byte("\n\n\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("B\n")}, + }, + }, + { + name: "case sensitivity check", + oldInput: "FOO\nbar\n", + newInput: "foo\nbar\n", + expected: []Chunk{ + {Kind: ChunkKindDeleted, Data: []byte("FOO\n")}, + {Kind: ChunkKindAdded, Data: []byte("foo\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("bar\n")}, + }, + }, + { + name: "partial line match is full mismatch", + oldInput: "The quick brown fox\n", + newInput: "The quick brown fox jumps\n", + expected: []Chunk{ + {Kind: ChunkKindDeleted, Data: []byte("The quick brown fox\n")}, + {Kind: ChunkKindAdded, Data: []byte("The quick brown fox jumps\n")}, + }, + }, + { + name: "inserting middle content", + oldInput: "Top\nBottom\n", + newInput: "Top\nMiddle\nBottom\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("Top\n")}, + {Kind: ChunkKindAdded, Data: []byte("Middle\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("Bottom\n")}, + }, + }, + { + name: "block move simulated", + oldInput: "BlockA\nBlockB\nBlockC\n", + newInput: "BlockA\nBlockC\nBlockB\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("BlockA\n")}, + {Kind: ChunkKindDeleted, Data: []byte("BlockB\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("BlockC\n")}, + {Kind: ChunkKindAdded, Data: []byte("BlockB\n")}, + }, + }, + { + name: "alternating additions", + oldInput: "A\nB\nC\n", + newInput: "A\n1\nB\n2\nC\n", + expected: []Chunk{ + {Kind: ChunkKindUnchanged, Data: []byte("A\n")}, + {Kind: ChunkKindAdded, Data: []byte("1\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("B\n")}, + {Kind: ChunkKindAdded, Data: []byte("2\n")}, + {Kind: ChunkKindUnchanged, Data: []byte("C\n")}, + }, + }, + } + + for _, tt := range tests { + t.Run(tt.name, func(t *testing.T) { + t.Parallel() + + chunks, err := Diff([]byte(tt.oldInput), []byte(tt.newInput)) + if err != nil { + t.Fatalf("Diff returned error: %v", err) + } + + if len(chunks) != len(tt.expected) { + t.Fatalf("expected %d chunks, got %d: %s", len(tt.expected), len(chunks), formatChunks(chunks)) + } + + for i := range tt.expected { + if chunks[i].Kind != tt.expected[i].Kind { + t.Fatalf("chunk %d kind mismatch: got %v, want %v; chunks: %s", i, chunks[i].Kind, tt.expected[i].Kind, formatChunks(chunks)) + } + if !bytes.Equal(chunks[i].Data, tt.expected[i].Data) { + t.Fatalf("chunk %d data mismatch: got %q, want %q; chunks: %s", i, string(chunks[i].Data), string(tt.expected[i].Data), formatChunks(chunks)) + } + } + }) + } +} + +func formatChunks(chunks []Chunk) string { + var b strings.Builder + b.WriteByte('[') + for i, chunk := range chunks { + if i > 0 { + b.WriteString(", ") + } + b.WriteString(chunkKindName(chunk.Kind)) + b.WriteByte(':') + b.WriteString(strconv.Quote(string(chunk.Data))) + } + b.WriteByte(']') + return b.String() +} + +func chunkKindName(kind ChunkKind) string { + switch kind { + case ChunkKindUnchanged: + return "U" + case ChunkKindDeleted: + return "D" + case ChunkKindAdded: + return "A" + default: + return "?" + } +} -- cgit v1.3.1-10-gc9f91