feat: Allow recursing into submodules when using git walker

cmd/format/format.go

@@ -159,7 +159,7 @@ func Run(v *viper.Viper, statz *stats.Stats, cmd *cobra.Command, paths []string)
 	}
 
 	// create a new walker for traversing the paths
-	walker, err := walk.NewCompositeReader(walkType, cfg.TreeRoot, paths, db, statz)
+	walker, err := walk.NewCompositeReader(walkType, cfg.TreeRoot, paths, cfg.IncludeGitSubmodules, db, statz)
 	if err != nil {
 		return fmt.Errorf("failed to create walker: %w", err)
 	}

config/config.go

@@ -26,6 +26,7 @@ type Config struct {
 	TreeRootFile          string   `mapstructure:"tree-root-file"          toml:"tree-root-file,omitempty"`
 	Verbose               uint8    `mapstructure:"verbose"                 toml:"verbose,omitempty"`
 	Walk                  string   `mapstructure:"walk"                    toml:"walk,omitempty"`
+	IncludeGitSubmodules  bool     `mapstructure:"include-git-submodules   toml:"include-git-subsmodules,omitempty"`
 	WorkingDirectory      string   `mapstructure:"working-dir"             toml:"-"`
 	Stdin                 bool     `mapstructure:"stdin"                   toml:"-"` // not allowed in config
 
@@ -115,6 +116,10 @@ func SetFlags(fs *pflag.FlagSet) {
 		"The method used to traverse the files within the tree root. Currently supports "+
 			"<auto|git|filesystem>. (env $TREEFMT_WALK)",
 	)
+	fs.Bool(
+		"include-git-submodules", false,
+		"Also format files in submodules",
+	)
 	fs.StringP(
 		"working-dir", "C", ".",
 		"Run as if treefmt was started in the specified working directory instead of the current working "+

walk/git.go

@@ -19,6 +19,7 @@ import (
 type GitReader struct {
 	root string
 	path string
+	includeSubmodules bool
 
 	log   *log.Logger
 	stats *stats.Stats
@@ -38,7 +39,13 @@ func (g *GitReader) Read(ctx context.Context, files []*File) (n int, err error)
 		r, w := io.Pipe()
 
 		// create a command which will execute from the specified sub path within root
-		cmd := exec.Command("git", "ls-files")
+		var cmd *exec.Cmd
+		if g.includeSubmodules {
+			cmd = exec.Command("git", "ls-files")
+	  	} else {
+			cmd = exec.Command("git", "ls-files", "--recurse-submodules")
+		}
+
 		cmd.Dir = filepath.Join(g.root, g.path)
 		cmd.Stdout = w
 
@@ -133,6 +140,7 @@ func (g *GitReader) Close() error {
 func NewGitReader(
 	root string,
 	path string,
+	includeSubmodules bool,
 	statz *stats.Stats,
 ) (*GitReader, error) {
 	// check if the root is a git repository
@@ -148,6 +156,7 @@ func NewGitReader(
 	return &GitReader{
 		root:  root,
 		path:  path,
+		includeSubmodules: includeSubmodules,
 		stats: statz,
 		eg:    &errgroup.Group{},
 		log:   log.WithPrefix("walk | git"),

walk/git_test.go

@@ -26,7 +26,7 @@ func TestGitReader(t *testing.T) {
 
 	// read empty worktree
 	statz := stats.New()
-	reader, err := walk.NewGitReader(tempDir, "", &statz)
+	reader, err := walk.NewGitReader(tempDir, "", false, &statz)
 	as.NoError(err)
 
 	files := make([]*walk.File, 8)
@@ -42,7 +42,7 @@ func TestGitReader(t *testing.T) {
 	cmd.Dir = tempDir
 	as.NoError(cmd.Run(), "failed to add everything to the index")
 
-	reader, err = walk.NewGitReader(tempDir, "", &statz)
+	reader, err = walk.NewGitReader(tempDir, "", false, &statz)
 	as.NoError(err)
 
 	count := 0

walk/walk.go

@@ -192,6 +192,7 @@ func NewReader(
 	walkType Type,
 	root string,
 	path string,
+	includeGitSubmodules bool,
 	db *bolt.DB,
 	statz *stats.Stats,
 ) (Reader, error) {
@@ -203,9 +204,9 @@ func NewReader(
 	switch walkType {
 	case Auto:
 		// for now, we keep it simple and try git first, filesystem second
-		reader, err = NewReader(Git, root, path, db, statz)
+		reader, err = NewReader(Git, root, path, includeGitSubmodules, db, statz)
 		if err != nil {
-			reader, err = NewReader(Filesystem, root, path, db, statz)
+			reader, err = NewReader(Filesystem, root, path, includeGitSubmodules, db, statz)
 		}
 
 		return reader, err
@@ -214,7 +215,7 @@ func NewReader(
 	case Filesystem:
 		reader = NewFilesystemReader(root, path, statz, BatchSize)
 	case Git:
-		reader, err = NewGitReader(root, path, statz)
+		reader, err = NewGitReader(root, path, includeGitSubmodules, statz)
 
 	default:
 		return nil, fmt.Errorf("unknown walk type: %v", walkType)
@@ -238,12 +239,13 @@ func NewCompositeReader(
 	walkType Type,
 	root string,
 	paths []string,
+	includeGitSubmodules bool,
 	db *bolt.DB,
 	statz *stats.Stats,
 ) (Reader, error) {
 	// if not paths are provided we default to processing the tree root
 	if len(paths) == 0 {
-		return NewReader(walkType, root, "", db, statz)
+		return NewReader(walkType, root, "", includeGitSubmodules, db, statz)
 	}
 
 	readers := make([]Reader, len(paths))
@@ -275,10 +277,10 @@ func NewCompositeReader(
 
 		if info.IsDir() {
 			// for directories, we honour the walk type as we traverse them
-			readers[idx], err = NewReader(walkType, root, relPath, db, statz)
+			readers[idx], err = NewReader(walkType, root, relPath, includeGitSubmodules, db, statz)
 		} else {
 			// for files, we enforce a simple filesystem read
-			readers[idx], err = NewReader(Filesystem, root, relPath, db, statz)
+			readers[idx], err = NewReader(Filesystem, root, relPath, includeGitSubmodules, db, statz)
 		}
 
 		if err != nil {