fix(watch): debounce fsnotify events, backfill archives, resilient walk

Co-authored-by: Belphemur <197810+Belphemur@users.noreply.github.com>

cmd/cbzoptimizer/commands/watch_command.go

@@ -7,6 +7,8 @@ import (
 	"path/filepath"
 	"runtime"
 	"strings"
+	"sync"
+	"time"
 
 	utils2 "github.com/belphemur/CBZOptimizer/v2/internal/utils"
 	"github.com/belphemur/CBZOptimizer/v2/pkg/converter"
@@ -17,6 +19,16 @@ import (
 	"github.com/spf13/viper"
 )
 
+// debounceDelay is the quiet period we wait, per path, after the last fsnotify
+// event before triggering an optimization. This coalesces the bursts of
+// Write/Create/Rename events fsnotify emits while a file is still being
+// written (e.g. copied or downloaded into the watched folder).
+const debounceDelay = 2 * time.Second
+
+// watchWorkerCount controls how many optimization jobs can run concurrently
+// so the fsnotify event loop is never blocked waiting on a conversion.
+var watchWorkerCount = runtime.NumCPU()
+
 func init() {
 	if runtime.GOOS != "linux" {
 		return
@@ -58,12 +70,12 @@ func WatchCommand(_ *cobra.Command, args []string) error {
 	converterType := constant.FindConversionFormat(viper.GetString("format"))
 	chapterConverter, err := converter.Get(converterType)
 	if err != nil {
-		return fmt.Errorf("failed to get chapterConverter: %v", err)
+		return fmt.Errorf("failed to get chapterConverter: %w", err)
 	}
 
 	err = chapterConverter.PrepareConverter()
 	if err != nil {
-		return fmt.Errorf("failed to prepare converter: %v", err)
+		return fmt.Errorf("failed to prepare converter: %w", err)
 	}
 	log.Info().Str("path", path).Bool("override", override).Uint8("quality", quality).Str("format", converterType.String()).Bool("split", split).Msg("Watching directory")
 
@@ -71,11 +83,28 @@ func WatchCommand(_ *cobra.Command, args []string) error {
 	if err != nil {
 		return fmt.Errorf("failed to create file watcher: %w", err)
 	}
-	defer watcher.Close()
+	defer func() {
+		if closeErr := watcher.Close(); closeErr != nil {
+			log.Error().Err(closeErr).Msg("Failed to close file watcher")
+		}
+	}()
+
+	queue := newOptimizeQueue(watchWorkerCount, &utils2.OptimizeOptions{
+		ChapterConverter: chapterConverter,
+		Quality:          quality,
+		Override:         override,
+		Split:            split,
+		Timeout:          timeout,
+	})
+	defer queue.Stop()
+
+	debouncer := newEventDebouncer(debounceDelay, queue.Enqueue)
+	defer debouncer.Stop()
 
 	if err := addRecursiveWatch(watcher, path); err != nil {
 		return fmt.Errorf("failed to watch path %s: %w", path, err)
 	}
+	backfillExistingArchives(path, debouncer.Trigger)
 
 	for {
 		select {
@@ -91,6 +120,11 @@ func WatchCommand(_ *cobra.Command, args []string) error {
 					if err := addRecursiveWatch(watcher, event.Name); err != nil {
 						log.Error().Err(err).Str("path", event.Name).Msg("Failed to watch created directory")
 					}
+					// The newly discovered directory may already contain
+					// archives (e.g. a folder moved/copied in); back-fill them
+					// since no further fsnotify event will target them.
+					backfillExistingArchives(event.Name, debouncer.Trigger)
+					continue
 				}
 			}
 
@@ -102,16 +136,7 @@ func WatchCommand(_ *cobra.Command, args []string) error {
 				continue
 			}
 
-			if err := utils2.Optimize(&utils2.OptimizeOptions{
-				ChapterConverter: chapterConverter,
-				Path:             event.Name,
-				Quality:          quality,
-				Override:         override,
-				Split:            split,
-				Timeout:          timeout,
-			}); err != nil {
-				log.Error().Err(err).Str("file", event.Name).Msg("Error processing file")
-			}
+			debouncer.Trigger(event.Name)
 		case err, ok := <-watcher.Errors:
 			if !ok {
 				return nil
@@ -121,10 +146,15 @@ func WatchCommand(_ *cobra.Command, args []string) error {
 	}
 }
 
+// addRecursiveWatch registers a watch on rootPath and every subdirectory
+// beneath it. Directories that can't be read (e.g. permission errors) are
+// logged and skipped instead of aborting the whole walk, so a single bad
+// subtree doesn't prevent the rest of the folder from being watched.
 func addRecursiveWatch(watcher *fsnotify.Watcher, rootPath string) error {
 	return filepath.WalkDir(rootPath, func(path string, entry fs.DirEntry, err error) error {
 		if err != nil {
-			return err
+			log.Warn().Err(err).Str("path", path).Msg("Skipping path while setting up watch")
+			return nil
 		}
 		if !entry.IsDir() {
 			return nil
@@ -136,6 +166,31 @@ func addRecursiveWatch(watcher *fsnotify.Watcher, rootPath string) error {
 	})
 }
 
+// backfillExistingArchives walks rootPath for comic archives that already
+// exist on disk and hands them to process. This covers the case where a
+// directory (potentially already containing archives) is created/moved into
+// the watched tree: only the directory itself generates an fsnotify event,
+// so the files inside it would otherwise never be picked up.
+func backfillExistingArchives(rootPath string, process func(path string)) {
+	err := filepath.WalkDir(rootPath, func(path string, entry fs.DirEntry, err error) error {
+		if err != nil {
+			log.Warn().Err(err).Str("path", path).Msg("Skipping path while scanning for existing archives")
+			return nil
+		}
+		if entry.IsDir() {
+			return nil
+		}
+		if !isComicArchive(path) {
+			return nil
+		}
+		process(path)
+		return nil
+	})
+	if err != nil {
+		log.Error().Err(err).Str("path", rootPath).Msg("Failed to scan directory for existing archives")
+	}
+}
+
 func shouldProcessWatchEvent(event fsnotify.Event) bool {
 	return event.Has(fsnotify.Create) || event.Has(fsnotify.Write) || event.Has(fsnotify.Rename)
 }
@@ -144,3 +199,118 @@ func isComicArchive(path string) bool {
 	filename := strings.ToLower(path)
 	return strings.HasSuffix(filename, ".cbz") || strings.HasSuffix(filename, ".cbr")
 }
+
+// eventDebouncer coalesces bursts of fsnotify events targeting the same path
+// into a single call to onQuiet, fired after the path has been quiet for
+// delay. This prevents repeated conversions being triggered while a file is
+// still being written, and avoids running Optimize against a Rename event
+// that reports the *old* (already gone) file name once the quiet period lets
+// us re-check the path.
+type eventDebouncer struct {
+	delay   time.Duration
+	onQuiet func(path string)
+
+	mu     sync.Mutex
+	timers map[string]*time.Timer
+}
+
+func newEventDebouncer(delay time.Duration, onQuiet func(path string)) *eventDebouncer {
+	return &eventDebouncer{
+		delay:   delay,
+		onQuiet: onQuiet,
+		timers:  make(map[string]*time.Timer),
+	}
+}
+
+// Trigger (re)schedules onQuiet to run for path after the debounce delay,
+// resetting any previously pending timer for the same path.
+func (d *eventDebouncer) Trigger(path string) {
+	d.mu.Lock()
+	defer d.mu.Unlock()
+
+	if timer, exists := d.timers[path]; exists {
+		timer.Reset(d.delay)
+		return
+	}
+
+	d.timers[path] = time.AfterFunc(d.delay, func() {
+		d.mu.Lock()
+		delete(d.timers, path)
+		d.mu.Unlock()
+		d.onQuiet(path)
+	})
+}
+
+// Stop cancels all pending timers.
+func (d *eventDebouncer) Stop() {
+	d.mu.Lock()
+	defer d.mu.Unlock()
+	for path, timer := range d.timers {
+		timer.Stop()
+		delete(d.timers, path)
+	}
+}
+
+// optimizeQueue is a small worker pool that runs Optimize jobs off of the
+// fsnotify event loop, so a slow conversion never blocks event draining.
+type optimizeQueue struct {
+	jobs    chan string
+	wg      sync.WaitGroup
+	options *utils2.OptimizeOptions
+}
+
+func newOptimizeQueue(workerCount int, options *utils2.OptimizeOptions) *optimizeQueue {
+	if workerCount < 1 {
+		workerCount = 1
+	}
+	q := &optimizeQueue{
+		jobs:    make(chan string, 64),
+		options: options,
+	}
+	q.wg.Add(workerCount)
+	for i := 0; i < workerCount; i++ {
+		go q.worker()
+	}
+	return q
+}
+
+func (q *optimizeQueue) worker() {
+	defer q.wg.Done()
+	for path := range q.jobs {
+		q.process(path)
+	}
+}
+
+// process re-checks the path right before optimizing: fsnotify Rename events
+// commonly report the *old* file name (already gone by the time we act on
+// it), and Write can fire while a file is still being written elsewhere. In
+// override mode, Optimize may overwrite/delete the source file, so skipping
+// paths that no longer exist (or that are no longer regular files) avoids
+// noisy failures.
+func (q *optimizeQueue) process(path string) {
+	info, err := os.Stat(path)
+	if err != nil {
+		log.Debug().Err(err).Str("file", path).Msg("Skipping watch event: path no longer accessible")
+		return
+	}
+	if info.IsDir() {
+		return
+	}
+
+	options := *q.options
+	options.Path = path
+	if err := utils2.Optimize(&options); err != nil {
+		log.Error().Err(err).Str("file", path).Msg("Error processing file")
+	}
+}
+
+// Enqueue submits path for processing by the worker pool.
+func (q *optimizeQueue) Enqueue(path string) {
+	q.jobs <- path
+}
+
+// Stop closes the job queue and waits for in-flight jobs to finish.
+func (q *optimizeQueue) Stop() {
+	close(q.jobs)
+	q.wg.Wait()
+}