Shallow clone trees to save on memory usage

pkg/filetree/filetree.go

@@ -45,7 +45,14 @@ func New() *FileTree {
 	}
 }
 
-// Copy returns a Copy of the current FileTree.
+// Clone returns a shallow Copy of the current FileTree.
+func (t *FileTree) Clone() (ReadWriter, error) {
+	ct := New()
+	ct.tree = t.tree.Clone()
+	return ct, nil
+}
+
+// Copy returns a deep Copy of the current FileTree.
 func (t *FileTree) Copy() (ReadWriter, error) {
 	ct := New()
 	ct.tree = t.tree.Copy()
@@ -781,7 +788,8 @@ func (t *FileTree) Walk(fn func(path file.Path, f filenode.FileNode) error, cond
 
 // Merge takes the given Tree and combines it with the current Tree, preferring files in the other Tree if there
 // are path conflicts. This is the basis function for squashing (where the current Tree is the bottom Tree and the
-// given Tree is the top Tree).
+// given Tree is the top Tree). Note: existing nodes are not mutated during this operation, they are replaced with
+// a node from the other tree.
 //
 //nolint:gocognit,funlen
 func (t *FileTree) Merge(upper Reader) error {

pkg/filetree/interfaces.go

@@ -29,6 +29,7 @@ type PathReader interface {
 
 type Copier interface {
 	Copy() (ReadWriter, error)
+	Clone() (ReadWriter, error)
 }
 
 type Walker interface {

pkg/filetree/union_filetree.go

@@ -21,14 +21,18 @@ func (u *UnionFileTree) Squash() (ReadWriter, error) {
 	case 0:
 		return New(), nil
 	case 1:
-		return u.trees[0].Copy()
+		// important: use clone over copy to reduce memory footprint. If callers need a distinct tree they can call
+		// copy after the fact.
+		return u.trees[0].Clone()
 	}
 
 	var squashedTree ReadWriter
 	var err error
 	for layerIdx, refTree := range u.trees {
 		if layerIdx == 0 {
-			squashedTree, err = refTree.Copy()
+			// important: use clone over copy to reduce memory footprint. If callers need a distinct tree they can call
+			// copy after the fact.
+			squashedTree, err = refTree.Clone()
 			if err != nil {
 				return nil, err
 			}

pkg/tree/tree.go

@@ -2,6 +2,7 @@ package tree
 
 import (
 	"fmt"
+	"maps"
 
 	"github.com/anchore/stereoscope/pkg/tree/node"
 )
@@ -22,6 +23,16 @@ func NewTree() *Tree {
 	}
 }
 
+// Clone returns a shallow copy of the Tree.
+func (t *Tree) Clone() *Tree {
+	ct := NewTree()
+	ct.nodes = maps.Clone(t.nodes)
+	ct.parent = maps.Clone(t.parent)
+	ct.children = maps.Clone(t.children)
+	return ct
+}
+
+// Copy returns a deep copy of the Tree.
 func (t *Tree) Copy() *Tree {
 	ct := NewTree()
 	for k, v := range t.nodes {

test/integration/path_type_change_test.go

@@ -0,0 +1,62 @@
+package integration
+
+import (
+	"testing"
+
+	"github.com/anchore/stereoscope/pkg/file"
+	"github.com/anchore/stereoscope/pkg/filetree/filenode"
+	"github.com/anchore/stereoscope/pkg/imagetest"
+)
+
+func TestPathTypeChangeImage(t *testing.T) {
+	image := imagetest.GetFixtureImage(t, "docker-archive", "image-path-type-change")
+
+	layerAssertions := []map[string]file.Type{
+		make(map[string]file.Type),
+		{
+			"/chimera/a.txt": file.TypeRegular,
+			"/chimera/b.txt": file.TypeRegular,
+			"/chimera":       file.TypeDirectory,
+		},
+		{
+			"/chimera": file.TypeDirectory,
+		},
+		make(map[string]file.Type),
+		{
+			"/chimera": file.TypeRegular,
+		},
+		make(map[string]file.Type),
+		{
+			"/chimera": file.TypeSymLink,
+		},
+		make(map[string]file.Type),
+		{
+			"/chimera": file.TypeDirectory,
+		},
+	}
+
+	for idx, layer := range image.Layers {
+		assertions := layerAssertions[idx]
+		err := layer.SquashedTree.Walk(func(path file.Path, f filenode.FileNode) error {
+			expect, ok := assertions[string(path)]
+			if !ok {
+				return nil
+			}
+			if f.FileType != expect {
+				t.Errorf("at %v got %v want %v", path, f.FileType, expect)
+			}
+			delete(assertions, string(path))
+			return nil
+		}, nil)
+		if err != nil {
+			t.Error(err)
+		}
+	}
+	for i, a := range layerAssertions {
+		if len(a) > 0 {
+			for k, v := range a {
+				t.Errorf("for layer %v, never saw %v of type %v", i, k, v)
+			}
+		}
+	}
+}

test/integration/test-fixtures/image-path-type-change/Dockerfile

@@ -0,0 +1,20 @@
+FROM busybox:1.37.0@sha256:db142d433cdde11f10ae479dbf92f3b13d693fd1c91053da9979728cceb1dc68
+
+# makes /chimera/a.txt /chimera/b.txt
+COPY . /
+
+# make a layer where the dir is empty
+RUN rm /chimera/a.txt /chimera/b.txt
+
+RUN rmdir /chimera
+
+RUN touch /chimera
+
+RUN rm /chimera
+
+RUN ln -s /etc/hostname /chimera
+
+RUN unlink /chimera
+
+RUN mkdir /chimera
+