fix: record cache entries for files that don't match formatters

Signed-off-by: Brian McGee <brian@bmcgee.ie>

cache/cache.go

@@ -5,9 +5,7 @@ import (
 	"crypto/sha1"
 	"encoding/hex"
 	"fmt"
-	"io/fs"
 	"os"
-	"path/filepath"
 	"runtime"
 	"time"
 
@@ -180,7 +178,7 @@ func putEntry(bucket *bolt.Bucket, path string, entry *Entry) error {
 
 // ChangeSet is used to walk a filesystem, starting at root, and outputting any new or changed paths using pathsCh.
 // It determines if a path is new or has changed by comparing against cache entries.
-func ChangeSet(ctx context.Context, walker walk.Walker, pathsCh chan<- string) error {
+func ChangeSet(ctx context.Context, walker walk.Walker, filesCh chan<- *walk.File) error {
 	start := time.Now()
 
 	defer func() {
@@ -198,24 +196,21 @@ func ChangeSet(ctx context.Context, walker walk.Walker, pathsCh chan<- string) e
 		}
 	}()
 
-	// for quick removal of tree root from paths
+	return walker.Walk(ctx, func(file *walk.File, err error) error {
-	relPathOffset := len(walker.Root()) + 1
-
-	return walker.Walk(ctx, func(path string, info fs.FileInfo, err error) error {
 		select {
 		case <-ctx.Done():
 			return ctx.Err()
 		default:
 			if err != nil {
 				return fmt.Errorf("%w: failed to walk path", err)
-			} else if info.IsDir() {
+			} else if file.Info.IsDir() {
 				// ignore directories
 				return nil
 			}
 		}
 
 		// ignore symlinks
-		if info.Mode()&os.ModeSymlink == os.ModeSymlink {
+		if file.Info.Mode()&os.ModeSymlink == os.ModeSymlink {
 			return nil
 		}
 
@@ -229,13 +224,12 @@ func ChangeSet(ctx context.Context, walker walk.Walker, pathsCh chan<- string) e
 			bucket = tx.Bucket([]byte(pathsBucket))
 		}
 
-		relPath := path[relPathOffset:]
+		cached, err := getEntry(bucket, file.RelPath)
-		cached, err := getEntry(bucket, relPath)
 		if err != nil {
 			return err
 		}
 
-		changedOrNew := cached == nil || !(cached.Modified == info.ModTime() && cached.Size == info.Size())
+		changedOrNew := cached == nil || !(cached.Modified == file.Info.ModTime() && cached.Size == file.Info.Size())
 
 		stats.Add(stats.Traversed, 1)
 		if !changedOrNew {
@@ -250,7 +244,7 @@ func ChangeSet(ctx context.Context, walker walk.Walker, pathsCh chan<- string) e
 		case <-ctx.Done():
 			return ctx.Err()
 		default:
-			pathsCh <- relPath
+			filesCh <- file
 		}
 
 		// close the current tx if we have reached the batch size
@@ -266,47 +260,35 @@ func ChangeSet(ctx context.Context, walker walk.Walker, pathsCh chan<- string) e
 }
 
 // Update is used to record updated cache information for the specified list of paths.
-func Update(treeRoot string, paths []string) (int, error) {
+func Update(files []*walk.File) error {
 	start := time.Now()
 	defer func() {
-		logger.Infof("finished updating %v paths in %v", len(paths), time.Since(start))
+		logger.Infof("finished processing %v paths in %v", len(files), time.Since(start))
 	}()
 
-	if len(paths) == 0 {
+	if len(files) == 0 {
-		return 0, nil
+		return nil
 	}
 
-	var changes int
+	return db.Update(func(tx *bolt.Tx) error {
-
-	return changes, db.Update(func(tx *bolt.Tx) error {
 		bucket := tx.Bucket([]byte(pathsBucket))
 
-		for _, path := range paths {
+		for _, f := range files {
-			cached, err := getEntry(bucket, path)
+			currentInfo, err := os.Stat(f.Path)
 			if err != nil {
 				return err
 			}
 
-			pathInfo, err := os.Stat(filepath.Join(treeRoot, path))
+			if !(f.Info.ModTime() == currentInfo.ModTime() && f.Info.Size() == currentInfo.Size()) {
-			if err != nil {
+				stats.Add(stats.Formatted, 1)
-				return err
 			}
 
-			if cached == nil || !(cached.Modified == pathInfo.ModTime() && cached.Size == pathInfo.Size()) {
-				changes += 1
-			} else {
-				// no change to write
-				continue
-			}
-
-			stats.Add(stats.Formatted, 1)
-
 			entry := Entry{
-				Size:     pathInfo.Size(),
+				Size:     currentInfo.Size(),
-				Modified: pathInfo.ModTime(),
+				Modified: currentInfo.ModTime(),
 			}
 
-			if err = putEntry(bucket, path, &entry); err != nil {
+			if err = putEntry(bucket, f.RelPath, &entry); err != nil {
 				return err
 			}
 		}

cli/format.go

@@ -5,7 +5,6 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"io/fs"
 	"os"
 	"os/signal"
 	"path/filepath"
@@ -35,8 +34,8 @@ var (
 	globalExcludes []glob.Glob
 	formatters     map[string]*format.Formatter
 	pipelines      map[string]*format.Pipeline
-	pathsCh        chan string
+	filesCh        chan *walk.File
-	processedCh    chan string
+	processedCh    chan *walk.File
 
 	ErrFailOnChange = errors.New("unexpected changes detected, --fail-on-change is enabled")
 )
@@ -142,10 +141,10 @@ func (f *Format) Run() (err error) {
 
 	// create a channel for paths to be processed
 	// we use a multiple of batch size here to allow for greater concurrency
-	pathsCh = make(chan string, BatchSize*runtime.NumCPU())
+	filesCh = make(chan *walk.File, BatchSize*runtime.NumCPU())
 
 	// create a channel for tracking paths that have been processed
-	processedCh = make(chan string, cap(pathsCh))
+	processedCh = make(chan *walk.File, cap(filesCh))
 
 	// start concurrent processing tasks
 	eg.Go(updateCache(ctx))
@@ -185,26 +184,26 @@ func walkFilesystem(ctx context.Context) func() error {
 			return fmt.Errorf("failed to create walker: %w", err)
 		}
 
-		defer close(pathsCh)
+		defer close(filesCh)
 
 		if Cli.NoCache {
-			return walker.Walk(ctx, func(path string, info fs.FileInfo, err error) error {
+			return walker.Walk(ctx, func(file *walk.File, err error) error {
 				select {
 				case <-ctx.Done():
 					return ctx.Err()
 				default:
 					// ignore symlinks and directories
-					if !(info.IsDir() || info.Mode()&os.ModeSymlink == os.ModeSymlink) {
+					if !(file.Info.IsDir() || file.Info.Mode()&os.ModeSymlink == os.ModeSymlink) {
 						stats.Add(stats.Traversed, 1)
 						stats.Add(stats.Emitted, 1)
-						pathsCh <- path
+						filesCh <- file
 					}
 					return nil
 				}
 			})
 		}
 
-		if err = cache.ChangeSet(ctx, walker, pathsCh); err != nil {
+		if err = cache.ChangeSet(ctx, walker, filesCh); err != nil {
 			return fmt.Errorf("failed to generate change set: %w", err)
 		}
 		return nil
@@ -213,19 +212,11 @@ func walkFilesystem(ctx context.Context) func() error {
 
 func updateCache(ctx context.Context) func() error {
 	return func() error {
-		batch := make([]string, 0, BatchSize)
+		batch := make([]*walk.File, 0, BatchSize)
-
-		var changes int
 
 		processBatch := func() error {
-			if Cli.NoCache {
+			if err := cache.Update(batch); err != nil {
-				changes += len(batch)
+				return err
-			} else {
-				count, err := cache.Update(Cli.TreeRoot, batch)
-				if err != nil {
-					return err
-				}
-				changes += count
 			}
 			batch = batch[:0]
 			return nil
@@ -254,7 +245,7 @@ func updateCache(ctx context.Context) func() error {
 			return err
 		}
 
-		if Cli.FailOnChange && changes != 0 {
+		if Cli.FailOnChange && stats.Value(stats.Formatted) != 0 {
 			return ErrFailOnChange
 		}
 
@@ -265,28 +256,28 @@ func updateCache(ctx context.Context) func() error {
 
 func applyFormatters(ctx context.Context) func() error {
 	fg, ctx := errgroup.WithContext(ctx)
-	batches := make(map[string][]string)
+	batches := make(map[string][]*walk.File)
 
-	tryApply := func(key string, path string) {
+	tryApply := func(key string, file *walk.File) {
 		batch, ok := batches[key]
 		if !ok {
-			batch = make([]string, 0, BatchSize)
+			batch = make([]*walk.File, 0, BatchSize)
 		}
-		batch = append(batch, path)
+		batch = append(batch, file)
 		batches[key] = batch
 
 		if len(batch) == BatchSize {
 			pipeline := pipelines[key]
 
 			// copy the batch
-			paths := make([]string, len(batch))
+			files := make([]*walk.File, len(batch))
-			copy(paths, batch)
+			copy(files, batch)
 
 			fg.Go(func() error {
-				if err := pipeline.Apply(ctx, paths); err != nil {
+				if err := pipeline.Apply(ctx, files); err != nil {
 					return err
 				}
-				for _, path := range paths {
+				for _, path := range files {
 					processedCh <- path
 				}
 				return nil
@@ -322,17 +313,19 @@ func applyFormatters(ctx context.Context) func() error {
 			close(processedCh)
 		}()
 
-		for path := range pathsCh {
+		for file := range filesCh {
 			var matched bool
 			for key, pipeline := range pipelines {
-				if !pipeline.Wants(path) {
+				if !pipeline.Wants(file) {
 					continue
 				}
 				matched = true
-				tryApply(key, path)
+				tryApply(key, file)
 			}
 			if matched {
 				stats.Add(stats.Matched, 1)
+			} else {
+				processedCh <- file
 			}
 		}
 

cli/format_test.go

@@ -51,42 +51,42 @@ func TestSpecifyingFormatters(t *testing.T) {
 	test.WriteConfig(t, configPath, config2.Config{
 		Formatters: map[string]*config2.Formatter{
 			"elm": {
-				Command:  "echo",
+				Command:  "touch",
 				Includes: []string{"*.elm"},
 			},
 			"nix": {
-				Command:  "echo",
+				Command:  "touch",
 				Includes: []string{"*.nix"},
 			},
 			"ruby": {
-				Command:  "echo",
+				Command:  "touch",
 				Includes: []string{"*.rb"},
 			},
 		},
 	})
 
-	out, err := cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
+	_, err := cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 3)
+	assertStats(t, as, 31, 31, 3, 3)
 
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--formatters", "elm,nix")
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--formatters", "elm,nix")
 	as.NoError(err)
-	assertFormatted(t, as, out, 2)
+	assertStats(t, as, 31, 31, 2, 2)
 
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--formatters", "ruby,nix")
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--formatters", "ruby,nix")
 	as.NoError(err)
-	assertFormatted(t, as, out, 2)
+	assertStats(t, as, 31, 31, 2, 2)
 
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--formatters", "nix")
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--formatters", "nix")
 	as.NoError(err)
-	assertFormatted(t, as, out, 1)
+	assertStats(t, as, 31, 31, 1, 1)
 
 	// test bad names
 
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--formatters", "foo")
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--formatters", "foo")
 	as.Errorf(err, "formatter not found in config: foo")
 
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--formatters", "bar,foo")
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--formatters", "bar,foo")
 	as.Errorf(err, "formatter not found in config: bar")
 }
 
@@ -94,7 +94,7 @@ func TestIncludesAndExcludes(t *testing.T) {
 	as := require.New(t)
 
 	tempDir := test.TempExamples(t)
-	configPath := tempDir + "/echo.toml"
+	configPath := tempDir + "/touch.toml"
 
 	// test without any excludes
 	cfg := config2.Config{
@@ -107,25 +107,25 @@ func TestIncludesAndExcludes(t *testing.T) {
 	}
 
 	test.WriteConfig(t, configPath, cfg)
-	out, err := cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
+	_, err := cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 31)
+	assertStats(t, as, 31, 31, 31, 0)
 
 	// globally exclude nix files
 	cfg.Global.Excludes = []string{"*.nix"}
 
 	test.WriteConfig(t, configPath, cfg)
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 30)
+	assertStats(t, as, 31, 31, 30, 0)
 
 	// add haskell files to the global exclude
 	cfg.Global.Excludes = []string{"*.nix", "*.hs"}
 
 	test.WriteConfig(t, configPath, cfg)
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 24)
+	assertStats(t, as, 31, 31, 24, 0)
 
 	echo := cfg.Formatters["echo"]
 
@@ -133,40 +133,40 @@ func TestIncludesAndExcludes(t *testing.T) {
 	echo.Excludes = []string{"*.py"}
 
 	test.WriteConfig(t, configPath, cfg)
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 22)
+	assertStats(t, as, 31, 31, 22, 0)
 
 	// remove go files from the echo formatter
 	echo.Excludes = []string{"*.py", "*.go"}
 
 	test.WriteConfig(t, configPath, cfg)
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 21)
+	assertStats(t, as, 31, 31, 21, 0)
 
 	// adjust the includes for echo to only include elm files
 	echo.Includes = []string{"*.elm"}
 
 	test.WriteConfig(t, configPath, cfg)
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 1)
+	assertStats(t, as, 31, 31, 1, 0)
 
 	// add js files to echo formatter
 	echo.Includes = []string{"*.elm", "*.js"}
 
 	test.WriteConfig(t, configPath, cfg)
-	out, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 2)
+	assertStats(t, as, 31, 31, 2, 0)
 }
 
 func TestCache(t *testing.T) {
 	as := require.New(t)
 
 	tempDir := test.TempExamples(t)
-	configPath := tempDir + "/echo.toml"
+	configPath := tempDir + "/touch.toml"
 
 	// test without any excludes
 	cfg := config2.Config{
@@ -181,7 +181,7 @@ func TestCache(t *testing.T) {
 	test.WriteConfig(t, configPath, cfg)
 	out, err := cmd(t, "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 31)
+	assertStats(t, as, 31, 31, 31, 0)
 
 	out, err = cmd(t, "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
@@ -190,7 +190,7 @@ func TestCache(t *testing.T) {
 	// clear cache
 	out, err = cmd(t, "--config-file", configPath, "--tree-root", tempDir, "-c")
 	as.NoError(err)
-	assertFormatted(t, as, out, 31)
+	assertStats(t, as, 31, 31, 31, 0)
 
 	out, err = cmd(t, "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
@@ -199,7 +199,7 @@ func TestCache(t *testing.T) {
 	// clear cache
 	out, err = cmd(t, "--config-file", configPath, "--tree-root", tempDir, "-c")
 	as.NoError(err)
-	assertFormatted(t, as, out, 31)
+	assertStats(t, as, 31, 31, 31, 0)
 
 	out, err = cmd(t, "--config-file", configPath, "--tree-root", tempDir)
 	as.NoError(err)
@@ -208,7 +208,7 @@ func TestCache(t *testing.T) {
 	// no cache
 	out, err = cmd(t, "--config-file", configPath, "--tree-root", tempDir, "--no-cache")
 	as.NoError(err)
-	assertStats(t, as, out, 31, 31, 31, 0)
+	assertStats(t, as, 31, 31, 31, 0)
 }
 
 func TestChangeWorkingDirectory(t *testing.T) {
@@ -240,22 +240,22 @@ func TestChangeWorkingDirectory(t *testing.T) {
 
 	// by default, we look for ./treefmt.toml and use the cwd for the tree root
 	// this should fail if the working directory hasn't been changed first
-	out, err := cmd(t, "-C", tempDir)
+	_, err = cmd(t, "-C", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 31)
+	assertStats(t, as, 31, 31, 31, 0)
 }
 
 func TestFailOnChange(t *testing.T) {
 	as := require.New(t)
 
 	tempDir := test.TempExamples(t)
-	configPath := tempDir + "/echo.toml"
+	configPath := tempDir + "/touch.toml"
 
 	// test without any excludes
 	cfg := config2.Config{
 		Formatters: map[string]*config2.Formatter{
-			"echo": {
+			"touch": {
-				Command:  "echo",
+				Command:  "touch",
 				Includes: []string{"*"},
 			},
 		},
@@ -270,7 +270,7 @@ func TestBustCacheOnFormatterChange(t *testing.T) {
 	as := require.New(t)
 
 	tempDir := test.TempExamples(t)
-	configPath := tempDir + "/echo.toml"
+	configPath := tempDir + "/touch.toml"
 
 	// symlink some formatters into temp dir, so we can mess with their mod times
 	binPath := tempDir + "/bin"
@@ -304,33 +304,33 @@ func TestBustCacheOnFormatterChange(t *testing.T) {
 
 	test.WriteConfig(t, configPath, cfg)
 	args := []string{"--config-file", configPath, "--tree-root", tempDir}
-	out, err := cmd(t, args...)
+	_, err := cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 3)
+	assertStats(t, as, 31, 31, 3, 0)
 
 	// tweak mod time of elm formatter
 	as.NoError(test.RecreateSymlink(t, binPath+"/"+"elm-format"))
 
-	out, err = cmd(t, args...)
+	_, err = cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 3)
+	assertStats(t, as, 31, 31, 3, 0)
 
 	// check cache is working
-	out, err = cmd(t, args...)
+	_, err = cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 0)
+	assertStats(t, as, 31, 0, 0, 0)
 
 	// tweak mod time of python formatter
 	as.NoError(test.RecreateSymlink(t, binPath+"/"+"black"))
 
-	out, err = cmd(t, args...)
+	_, err = cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 3)
+	assertStats(t, as, 31, 31, 3, 0)
 
 	// check cache is working
-	out, err = cmd(t, args...)
+	_, err = cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 0)
+	assertStats(t, as, 31, 0, 0, 0)
 
 	// add go formatter
 	cfg.Formatters["go"] = &config2.Formatter{
@@ -340,40 +340,40 @@ func TestBustCacheOnFormatterChange(t *testing.T) {
 	}
 	test.WriteConfig(t, configPath, cfg)
 
-	out, err = cmd(t, args...)
+	_, err = cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 4)
+	assertStats(t, as, 31, 31, 4, 0)
 
 	// check cache is working
-	out, err = cmd(t, args...)
+	_, err = cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 0)
+	assertStats(t, as, 31, 0, 0, 0)
 
 	// remove python formatter
 	delete(cfg.Formatters, "python")
 	test.WriteConfig(t, configPath, cfg)
 
-	out, err = cmd(t, args...)
+	_, err = cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 2)
+	assertStats(t, as, 31, 31, 2, 0)
 
 	// check cache is working
-	out, err = cmd(t, args...)
+	_, err = cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 0)
+	assertStats(t, as, 31, 0, 0, 0)
 
 	// remove elm formatter
 	delete(cfg.Formatters, "elm")
 	test.WriteConfig(t, configPath, cfg)
 
-	out, err = cmd(t, args...)
+	_, err = cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 1)
+	assertStats(t, as, 31, 31, 1, 0)
 
 	// check cache is working
-	out, err = cmd(t, args...)
+	_, err = cmd(t, args...)
 	as.NoError(err)
-	assertFormatted(t, as, out, 0)
+	assertStats(t, as, 31, 0, 0, 0)
 }
 
 func TestGitWorktree(t *testing.T) {
@@ -407,28 +407,28 @@ func TestGitWorktree(t *testing.T) {
 	wt, err := repo.Worktree()
 	as.NoError(err, "failed to get git worktree")
 
-	run := func(formatted int) {
+	run := func(traversed int, emitted int, matched int, formatted int) {
 		out, err := cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir)
 		as.NoError(err)
 		assertFormatted(t, as, out, formatted)
 	}
 
 	// run before adding anything to the worktree
-	run(0)
+	run(0, 0, 0, 0)
 
 	// add everything to the worktree
 	as.NoError(wt.AddGlob("."))
 	as.NoError(err)
-	run(31)
+	run(31, 31, 31, 0)
 
 	// remove python directory
 	as.NoError(wt.RemoveGlob("python/*"))
-	run(28)
+	run(28, 28, 28, 0)
 
 	// walk with filesystem instead of git
-	out, err := cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--walk", "filesystem")
+	_, err = cmd(t, "-c", "--config-file", configPath, "--tree-root", tempDir, "--walk", "filesystem")
 	as.NoError(err)
-	assertFormatted(t, as, out, 59)
+	assertStats(t, as, 59, 59, 59, 0)
 }
 
 func TestPathsArg(t *testing.T) {
@@ -461,17 +461,17 @@ func TestPathsArg(t *testing.T) {
 	test.WriteConfig(t, configPath, cfg)
 
 	// without any path args
-	out, err := cmd(t, "-C", tempDir)
+	_, err = cmd(t, "-C", tempDir)
 	as.NoError(err)
-	assertFormatted(t, as, out, 31)
+	assertStats(t, as, 31, 31, 31, 0)
 
 	// specify some explicit paths
-	out, err = cmd(t, "-C", tempDir, "-c", "elm/elm.json", "haskell/Nested/Foo.hs")
+	_, err = cmd(t, "-C", tempDir, "-c", "elm/elm.json", "haskell/Nested/Foo.hs")
 	as.NoError(err)
-	assertFormatted(t, as, out, 2)
+	assertStats(t, as, 4, 4, 4, 0)
 
 	// specify a bad path
-	out, err = cmd(t, "-C", tempDir, "-c", "elm/elm.json", "haskell/Nested/Bar.hs")
+	_, err = cmd(t, "-C", tempDir, "-c", "elm/elm.json", "haskell/Nested/Bar.hs")
 	as.ErrorContains(err, "no such file or directory")
 }
 
@@ -526,9 +526,9 @@ go/main.go
 		_, _ = stdin.Seek(0, 0)
 	}()
 
-	out, err := cmd(t, "-C", tempDir, "--stdin")
+	_, err = cmd(t, "-C", tempDir, "--stdin")
 	as.NoError(err)
-	assertFormatted(t, as, out, 3)
+	assertStats(t, as, 6, 6, 6, 0)
 }
 
 func TestDeterministicOrderingInPipeline(t *testing.T) {

cli/helpers_test.go

@@ -2,6 +2,7 @@ package cli
 
 import (
 	"fmt"
+	"git.numtide.com/numtide/treefmt/stats"
 	"io"
 	"os"
 	"path/filepath"
@@ -70,12 +71,12 @@ func cmd(t *testing.T, args ...string) ([]byte, error) {
 	return out, nil
 }
 
-func assertStats(t *testing.T, as *require.Assertions, output []byte, traversed int32, emitted int32, matched int32, formatted int32) {
+func assertStats(t *testing.T, as *require.Assertions, traversed int32, emitted int32, matched int32, formatted int32) {
 	t.Helper()
-	as.Contains(string(output), fmt.Sprintf("traversed %d files", traversed))
+	as.Equal(traversed, stats.Value(stats.Traversed))
-	as.Contains(string(output), fmt.Sprintf("emitted %d files", emitted))
+	as.Equal(emitted, stats.Value(stats.Emitted))
-	as.Contains(string(output), fmt.Sprintf("matched %d files", matched))
+	as.Equal(matched, stats.Value(stats.Matched))
-	as.Contains(string(output), fmt.Sprintf("formatted %d files", formatted))
+	as.Equal(formatted, stats.Value(stats.Formatted))
 }
 
 func assertFormatted(t *testing.T, as *require.Assertions, output []byte, count int) {

format/formatter.go

@@ -8,6 +8,8 @@ import (
 	"os/exec"
 	"time"
 
+	"git.numtide.com/numtide/treefmt/walk"
+
 	"git.numtide.com/numtide/treefmt/config"
 
 	"github.com/charmbracelet/log"
@@ -38,7 +40,7 @@ func (f *Formatter) Executable() string {
 	return f.executable
 }
 
-func (f *Formatter) Apply(ctx context.Context, paths []string, filter bool) error {
+func (f *Formatter) Apply(ctx context.Context, files []*walk.File, filter bool) error {
 	start := time.Now()
 
 	// construct args, starting with config
@@ -52,9 +54,9 @@ func (f *Formatter) Apply(ctx context.Context, paths []string, filter bool) erro
 		f.batch = f.batch[:0]
 
 		// filter paths
-		for _, path := range paths {
+		for _, file := range files {
-			if f.Wants(path) {
+			if f.Wants(file) {
-				f.batch = append(f.batch, path)
+				f.batch = append(f.batch, file.RelPath)
 			}
 		}
 
@@ -67,12 +69,14 @@ func (f *Formatter) Apply(ctx context.Context, paths []string, filter bool) erro
 		args = append(args, f.batch...)
 	} else {
 		// exit early if nothing to process
-		if len(paths) == 0 {
+		if len(files) == 0 {
 			return nil
 		}
 
 		// append paths to the args
-		args = append(args, paths...)
+		for _, file := range files {
+			args = append(args, file.RelPath)
+		}
 	}
 
 	// execute the command
@@ -88,17 +92,17 @@ func (f *Formatter) Apply(ctx context.Context, paths []string, filter bool) erro
 
 	//
 
-	f.log.Infof("%v files processed in %v", len(paths), time.Now().Sub(start))
+	f.log.Infof("%v files processed in %v", len(files), time.Now().Sub(start))
 
 	return nil
 }
 
 // Wants is used to test if a Formatter wants a path based on it's configured Includes and Excludes patterns.
 // Returns true if the Formatter should be applied to path, false otherwise.
-func (f *Formatter) Wants(path string) bool {
+func (f *Formatter) Wants(file *walk.File) bool {
-	match := !PathMatches(path, f.excludes) && PathMatches(path, f.includes)
+	match := !PathMatches(file.RelPath, f.excludes) && PathMatches(file.RelPath, f.includes)
 	if match {
-		f.log.Debugf("match: %v", path)
+		f.log.Debugf("match: %v", file)
 	}
 	return match
 }

format/pipeline.go

@@ -3,6 +3,8 @@ package format
 import (
 	"context"
 	"slices"
+
+	"git.numtide.com/numtide/treefmt/walk"
 )
 
 type Pipeline struct {
@@ -17,7 +19,7 @@ func (p *Pipeline) Add(f *Formatter) {
 	})
 }
 
-func (p *Pipeline) Wants(path string) bool {
+func (p *Pipeline) Wants(path *walk.File) bool {
 	var match bool
 	for _, f := range p.sequence {
 		match = f.Wants(path)
@@ -28,7 +30,7 @@ func (p *Pipeline) Wants(path string) bool {
 	return match
 }
 
-func (p *Pipeline) Apply(ctx context.Context, paths []string) error {
+func (p *Pipeline) Apply(ctx context.Context, paths []*walk.File) error {
 	for _, f := range p.sequence {
 		if err := f.Apply(ctx, paths, len(p.sequence) > 1); err != nil {
 			return err

test/examples/nixpkgs.toml

@@ -2,6 +2,7 @@
 
 [formatter.deadnix]
 command = "deadnix"
+options = ["--edit"]
 includes = ["*.nix"]
 pipeline = "nix"
 priority = 1

test/examples/touch.toml

@@ -1,3 +1,3 @@
 [formatter.echo]
-command = "echo"
+command = "touch"
 includes = [ "*.*" ]

walk/filesystem.go

@@ -2,6 +2,7 @@ package walk
 
 import (
 	"context"
+	"io/fs"
 	"os"
 	"path/filepath"
 )
@@ -15,18 +16,42 @@ func (f filesystemWalker) Root() string {
 	return f.root
 }
 
-func (f filesystemWalker) Walk(_ context.Context, fn filepath.WalkFunc) error {
+func (f filesystemWalker) Walk(_ context.Context, fn WalkFunc) error {
+	relPathOffset := len(f.root) + 1
+
+	relPathFn := func(path string) (relPath string) {
+		if len(path) >= relPathOffset {
+			relPath = path[relPathOffset:]
+		}
+		return
+	}
+
+	walkFn := func(path string, info fs.FileInfo, err error) error {
+		file := File{
+			Path:    path,
+			RelPath: relPathFn(path),
+			Info:    info,
+		}
+		return fn(&file, err)
+	}
+
 	if len(f.paths) == 0 {
-		return filepath.Walk(f.root, fn)
+		return filepath.Walk(f.root, walkFn)
 	}
 
 	for _, path := range f.paths {
 		info, err := os.Stat(path)
-		if err = filepath.Walk(path, fn); err != nil {
+		if err = filepath.Walk(path, walkFn); err != nil {
 			return err
 		}
 
-		if err = fn(path, info, err); err != nil {
+		file := File{
+			Path:    path,
+			RelPath: relPathFn(path),
+			Info:    info,
+		}
+
+		if err = fn(&file, err); err != nil {
 			return err
 		}
 	}

walk/git.go

@@ -24,7 +24,17 @@ func (g *gitWalker) Root() string {
 	return g.root
 }
 
-func (g *gitWalker) Walk(ctx context.Context, fn filepath.WalkFunc) error {
+func (g *gitWalker) Walk(ctx context.Context, fn WalkFunc) error {
+	// for quick relative paths
+	relPathOffset := len(g.root) + 1
+
+	relPathFn := func(path string) (relPath string) {
+		if len(path) >= relPathOffset {
+			relPath = path[relPathOffset:]
+		}
+		return
+	}
+
 	idx, err := g.repo.Storer.Index()
 	if err != nil {
 		return fmt.Errorf("%w: failed to open index", err)
@@ -49,7 +59,13 @@ func (g *gitWalker) Walk(ctx context.Context, fn filepath.WalkFunc) error {
 					return nil
 				}
 
-				return fn(path, info, err)
+				file := File{
+					Path:    path,
+					RelPath: relPathFn(path),
+					Info:    info,
+				}
+
+				return fn(&file, err)
 			})
 			if err != nil {
 				return err
@@ -66,7 +82,14 @@ func (g *gitWalker) Walk(ctx context.Context, fn filepath.WalkFunc) error {
 
 				// stat the file
 				info, err := os.Lstat(path)
-				if err = fn(path, info, err); err != nil {
+
+				file := File{
+					Path:    path,
+					RelPath: relPathFn(path),
+					Info:    info,
+				}
+
+				if err = fn(&file, err); err != nil {
 					return err
 				}
 			}

walk/walker.go

@@ -3,7 +3,7 @@ package walk
 import (
 	"context"
 	"fmt"
-	"path/filepath"
+	"io/fs"
 )
 
 type Type string
@@ -14,9 +14,21 @@ const (
 	Filesystem Type = "filesystem"
 )
 
+type File struct {
+	Path    string
+	RelPath string
+	Info    fs.FileInfo
+}
+
+func (f File) String() string {
+	return f.Path
+}
+
+type WalkFunc func(file *File, err error) error
+
 type Walker interface {
 	Root() string
-	Walk(ctx context.Context, fn filepath.WalkFunc) error
+	Walk(ctx context.Context, fn WalkFunc) error
 }
 
 func New(walkerType Type, root string, paths []string) (Walker, error) {