git: skip ignored files while walking worktree

Skip ignored files when walking through the worktree. This signigifantly improves the performance of `Status()`: In a repository with 3M ignored files `Status` now takes 5 s instead of 160 s.
go-git · silkeh · Jan 13, 2025 · Jan 13, 2025 · Jan 13, 2025 · 25c3930c981662c2e5a5d7302bf6e2e840d5caca
commit 25c3930c981662c2e5a5d7302bf6e2e840d5caca
diff --git a/plumbing/format/gitignore/noder.go b/plumbing/format/gitignore/noder.go
@@ -0,0 +1,137 @@
+package gitignore
+
+import (
+	"slices"
+
+	"github.com/go-git/go-git/v5/utils/merkletrie/noder"
+)
+
+var _ noder.Noder = (*MatchNoder)(nil)
+
+// MatchNoder is an implementation of [noder.Noder] that only includes nodes based on a pattern.
+type MatchNoder struct {
+	noder.Noder
+
+	matcher  Matcher
+	invert   bool
+	path     []string
+	children []noder.Noder
+}
+
+// IgnoreNoder returns a [MatchNoder] that filters out the given pattern.
+func IgnoreNoder(m Matcher, n noder.Noder) *MatchNoder {
+	var path []string
+	if name := n.Name(); name != "." {
+		path = []string{name}
+	}
+
+	return &MatchNoder{matcher: m, invert: true, Noder: n, path: path}
+}
+
+// Children returns matched children.
+// It implements [noder.Noder].
+func (n *MatchNoder) Children() ([]noder.Noder, error) {
+	if len(n.children) > 0 {
+		return n.children, nil
+	}
+
+	children, err := n.Noder.Children()
+	if err != nil {
+		return nil, err
+	}
+
+	n.children = n.ignoreChildren(children)
+
+	return n.children, nil
+}
+
+func (n *MatchNoder) ignoreChildren(children []noder.Noder) []noder.Noder {
+	found := make([]noder.Noder, 0, len(children))
+
+	for _, child := range children {
+		path := append(n.path, child.Name())
+		if n.match(path, child.IsDir()) {
+			continue
+		}
+
+		found = append(found, n.newChild(child, path))
+	}
+
+	return found
+}
+
+func (n *MatchNoder) match(path []string, isDir bool) bool {
+	if n.matcher != nil && n.matcher.Match(path, isDir) {
+		return n.invert
+	}
+
+	return !n.invert
+}
+
+func (n *MatchNoder) newChild(child noder.Noder, path []string) noder.Noder {
+	if !child.IsDir() {
+		return child
+	}
+
+	return &MatchNoder{
+		matcher: n.matcher,
+		invert:  n.invert,
+		Noder:   child,
+		path:    slices.Clone(path),
+	}
+}
+
+// NumChildren returns the number of children.
+// It implements [noder.Noder].
+func (n *MatchNoder) NumChildren() (int, error) {
+	children, err := n.Children()
+	if err != nil {
+		return 0, err
+	}
+
+	return len(children), nil
+}
+
+// PathIgnored returns true if the given [noder.Path] is ignored.
+func (n *MatchNoder) PathIgnored(path noder.Path) bool {
+	return n.match(n.noderPaths(path), path.IsDir())
+}
+
+// FindPath returns the corresponding [noder.Path] from the tree if there is one.
+// It does not apply patterns, allowing retrieval of ignored nodes.
+func (n *MatchNoder) FindPath(p noder.Path) (path noder.Path, found bool) {
+	node := n.Noder
+
+	for i := range p {
+		node, found = n.findChild(node, p[i].Name())
+		if !found {
+			return nil, false
+		}
+
+		path = append(path, node)
+	}
+
+	return
+}
+
+func (n *MatchNoder) findChild(node noder.Noder, name string) (noder.Noder, bool) {
+	children, _ := node.Children()
+
+	for _, child := range children {
+		if child.Name() == name {
+			return child, true
+		}
+	}
+
+	return nil, false
+}
+
+func (n *MatchNoder) noderPaths(path noder.Path) []string {
+	parts := make([]string, len(path))
+
+	for i, p := range path {
+		parts[i] = p.Name()
+	}
+
+	return parts
+}
diff --git a/plumbing/format/gitignore/noder_test.go b/plumbing/format/gitignore/noder_test.go
@@ -0,0 +1,144 @@
+package gitignore_test
+
+import (
+	"io/fs"
+	"testing"
+
+	"github.com/go-git/go-git/v5/plumbing/format/gitignore"
+	"github.com/go-git/go-git/v5/utils/merkletrie/noder"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+type mockNoder struct {
+	hash        []byte
+	string      string
+	name        string
+	isDir       bool
+	children    []noder.Noder
+	childrenErr error
+	skip        bool
+}
+
+func (m mockNoder) Hash() []byte                     { return m.hash }
+func (m mockNoder) String() string                   { return m.string }
+func (m mockNoder) Name() string                     { return m.name }
+func (m mockNoder) IsDir() bool                      { return m.isDir }
+func (m mockNoder) Children() ([]noder.Noder, error) { return m.children, m.childrenErr }
+func (m mockNoder) NumChildren() (int, error)        { return len(m.children), m.childrenErr }
+func (m mockNoder) Skip() bool                       { return m.skip }
+
+func TestMatchNoder_Children(t *testing.T) {
+	mock := mockNoder{
+		name: ".",
+		children: []noder.Noder{
+			mockNoder{name: "volcano"},
+			mockNoder{name: "caldera"},
+			mockNoder{name: "super", isDir: true, children: []noder.Noder{
+				mockNoder{name: "caldera", children: []noder.Noder{}},
+			}},
+		},
+	}
+	patterns := []gitignore.Pattern{
+		gitignore.ParsePattern("**/middle/v[uo]l?ano", nil),
+		gitignore.ParsePattern("volcano", nil),
+	}
+
+	tests := map[string]struct {
+		Matcher     gitignore.Matcher
+		Noder       mockNoder
+		ExpErr      error
+		ExpChildren []noder.Noder
+		Skip        bool
+	}{
+		"children": {
+			Matcher: gitignore.NewMatcher([]gitignore.Pattern{
+				gitignore.ParsePattern("**/middle/v[uo]l?ano", nil),
+				gitignore.ParsePattern("volcano", nil),
+			}),
+			Noder: mock,
+			ExpChildren: []noder.Noder{
+				mock.children[1],
+				gitignore.IgnoreNoder(gitignore.NewMatcher(patterns), mock.children[2]),
+			},
+		},
+		"error": {
+			Matcher: gitignore.NewMatcher([]gitignore.Pattern{
+				gitignore.ParsePattern("**/middle/v[uo]l?ano", nil),
+				gitignore.ParsePattern("volcano", nil),
+			}),
+			Noder:  mockNoder{name: ".", childrenErr: fs.ErrNotExist},
+			ExpErr: fs.ErrNotExist,
+		},
+	}
+
+	for name, tc := range tests {
+		t.Run(name, func(t *testing.T) {
+			ignoreNoder := gitignore.IgnoreNoder(tc.Matcher, tc.Noder)
+
+			children, err := ignoreNoder.Children()
+			require.ErrorIs(t, err, tc.ExpErr)
+			assert.Equal(t, tc.ExpChildren, children)
+
+			// Do it twice for the cached children
+			children, err = ignoreNoder.Children()
+			require.ErrorIs(t, err, tc.ExpErr)
+			assert.Equal(t, tc.ExpChildren, children)
+
+			num, err := ignoreNoder.NumChildren()
+			require.ErrorIs(t, err, tc.ExpErr)
+			assert.Equal(t, len(tc.ExpChildren), num)
+		})
+	}
+}
+
+func TestMatchNoder_PathIgnored(t *testing.T) {
+	matcher := gitignore.NewMatcher([]gitignore.Pattern{
+		gitignore.ParsePattern("**/middle/v[uo]l?ano", nil),
+		gitignore.ParsePattern("volcano", nil),
+	})
+
+	found := gitignore.IgnoreNoder(matcher, mockNoder{name: "."}).PathIgnored([]noder.Noder{
+		mockNoder{name: "head"},
+		mockNoder{name: "middle"},
+		mockNoder{name: "volcano"},
+	})
+	assert.True(t, found)
+
+	found = gitignore.IgnoreNoder(matcher, mockNoder{name: "."}).PathIgnored([]noder.Noder{
+		mockNoder{name: "head"},
+		mockNoder{name: "middle"},
+		mockNoder{name: "potato"},
+	})
+	assert.False(t, found)
+}
+
+func TestMatchNoder_FindPath(t *testing.T) {
+	mock := mockNoder{
+		name: ".",
+		children: []noder.Noder{
+			mockNoder{name: "volcano"},
+			mockNoder{name: "super", isDir: true, children: []noder.Noder{
+				mockNoder{name: "volcano", children: []noder.Noder{}},
+			}},
+		},
+	}
+	matcher := gitignore.NewMatcher([]gitignore.Pattern{
+		gitignore.ParsePattern("**/middle/v[uo]l?ano", nil),
+		gitignore.ParsePattern("volcano", nil),
+	})
+
+	node, found := gitignore.IgnoreNoder(matcher, mock).FindPath([]noder.Noder{
+		mockNoder{name: "super"},
+		mockNoder{name: "volcano"},
+	})
+	assert.True(t, found)
+	assert.Equal(t, noder.Path{mock.children[1], mock.children[1].(mockNoder).children[0]}, node)
+
+	node, found = gitignore.IgnoreNoder(matcher, mock).FindPath([]noder.Noder{
+		mockNoder{name: "super"},
+		mockNoder{name: "caldera"},
+	})
+	assert.False(t, found)
+	assert.Nil(t, node)
+}
diff --git a/utils/merkletrie/difftree.go b/utils/merkletrie/difftree.go
@@ -252,6 +252,7 @@ import (
 	"errors"
 	"fmt"
 
+	"github.com/go-git/go-git/v5/plumbing/format/gitignore"
 	"github.com/go-git/go-git/v5/utils/merkletrie/noder"
 )
 
@@ -297,8 +298,14 @@ func DiffTreeContext(ctx context.Context, fromTree, toTree noder.Noder,
 		case noMoreNoders:
 			return ret, nil
 		case onlyFromRemains:
-			if err = ret.AddRecursiveDelete(from); err != nil {
-				return nil, err
+			if node, ok := ignoredNode(toTree, from); ok {
+				if err = diffNodesSameName(&ret, ii, ii.from.current, node); err != nil {
+					return nil, err
+				}
+			} else {
+				if err = ret.AddRecursiveDelete(from); err != nil {
+					return nil, err
+				}
 			}
 			if err = ii.nextFrom(); err != nil {
 				return nil, err
@@ -353,8 +360,10 @@ func diffNodes(changes *Changes, ii *doubleIter) error {
 	// compare their full paths as strings
 	switch from.Compare(to) {
 	case -1:
-		if err = changes.AddRecursiveDelete(from); err != nil {
-			return err
+		if ok := isIgnoredNode(to[0], from); !ok {
+			if err = changes.AddRecursiveDelete(from); err != nil {
+				return err
+			}
 		}
 		if err = ii.nextFrom(); err != nil {
 			return err
@@ -367,19 +376,16 @@ func diffNodes(changes *Changes, ii *doubleIter) error {
 			return err
 		}
 	default:
-		if err := diffNodesSameName(changes, ii); err != nil {
+		if err := diffNodesSameName(changes, ii, ii.from.current, ii.to.current); err != nil {
 			return err
 		}
 	}
 
 	return nil
 }
 
-func diffNodesSameName(changes *Changes, ii *doubleIter) error {
-	from := ii.from.current
-	to := ii.to.current
-
-	status, err := ii.compare()
+func diffNodesSameName(changes *Changes, ii *doubleIter, from, to noder.Path) error {
+	status, err := ii.compareNoders(from, to)
 	if err != nil {
 		return err
 	}
@@ -451,3 +457,18 @@ func diffDirs(changes *Changes, ii *doubleIter) error {
 
 	return nil
 }
+
+func isIgnoredNode(tree noder.Noder, path noder.Path) bool {
+	in, ok := tree.(*gitignore.MatchNoder)
+
+	return ok && in.PathIgnored(path)
+}
+
+func ignoredNode(tree noder.Noder, path noder.Path) (noder.Path, bool) {
+	in, ok := tree.(*gitignore.MatchNoder)
+	if !ok || !in.PathIgnored(path) {
+		return nil, false
+	}
+
+	return in.FindPath(path)
+}
diff --git a/utils/merkletrie/doubleiter.go b/utils/merkletrie/doubleiter.go
@@ -137,21 +137,25 @@ const (
 // Compare returns the comparison between the current elements in the
 // merkletries.
 func (d *doubleIter) compare() (s comparison, err error) {
-	s.sameHash = d.hashEqual(d.from.current, d.to.current)
+	return d.compareNoders(d.from.current, d.to.current)
+}
+
+func (d *doubleIter) compareNoders(from, to noder.Noder) (s comparison, err error) {
+	s.sameHash = d.hashEqual(from, to)
 
-	fromIsDir := d.from.current.IsDir()
-	toIsDir := d.to.current.IsDir()
+	fromIsDir := from.IsDir()
+	toIsDir := to.IsDir()
 
 	s.bothAreDirs = fromIsDir && toIsDir
 	s.bothAreFiles = !fromIsDir && !toIsDir
 	s.fileAndDir = !s.bothAreDirs && !s.bothAreFiles
 
-	fromNumChildren, err := d.from.current.NumChildren()
+	fromNumChildren, err := from.NumChildren()
 	if err != nil {
 		return comparison{}, fmt.Errorf("from: %s", err)
 	}
 
-	toNumChildren, err := d.to.current.NumChildren()
+	toNumChildren, err := to.NumChildren()
 	if err != nil {
 		return comparison{}, fmt.Errorf("to: %s", err)
 	}