Files
navidrome/scanner/walk_dir_tree.go
Deluan Quintão 28d5299ffc feat(scanner): implement selective folder scanning and file system watcher improvements (#4674)
* feat: Add selective folder scanning capability

Implement targeted scanning of specific library/folder pairs without
full recursion. This enables efficient rescanning of individual folders
when changes are detected, significantly reducing scan time for large
libraries.

Key changes:
- Add ScanTarget struct and ScanFolders API to Scanner interface
- Implement CLI flag --targets for specifying libraryID:folderPath pairs
- Add FolderRepository.GetByPaths() for batch folder info retrieval
- Create loadSpecificFolders() for non-recursive directory loading
- Scope GC operations to affected libraries only (with TODO for full impl)
- Add comprehensive tests for selective scanning behavior

The selective scan:
- Only processes specified folders (no subdirectory recursion)
- Maintains library isolation
- Runs full maintenance pipeline scoped to affected libraries
- Supports both full and quick scan modes

Examples:
  navidrome scan --targets "1:Music/Rock,1:Music/Jazz"
  navidrome scan --full --targets "2:Classical"

* feat(folder): replace GetByPaths with GetFolderUpdateInfo for improved folder updates retrieval

Signed-off-by: Deluan <deluan@navidrome.org>

* test: update parseTargets test to handle folder names with spaces

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(folder): remove unused LibraryPath struct and update GC logging message

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(folder): enhance external scanner to support target-specific scanning

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): simplify scanner methods

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(watcher): implement folder scanning notifications with deduplication

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(watcher): add resolveFolderPath function for testability

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(watcher): implement path ignoring based on .ndignore patterns

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): implement IgnoreChecker for managing .ndignore patterns

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(ignore_checker): rename scanner to lineScanner for clarity

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): enhance ScanTarget struct with String method for better target representation

Signed-off-by: Deluan <deluan@navidrome.org>

* fix(scanner): validate library ID to prevent negative values

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): simplify GC method by removing library ID parameter

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(scanner): update folder scanning to include all descendants of specified folders

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(subsonic): allow selective scan in the /startScan endpoint

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): update CallScan to handle specific library/folder pairs

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): streamline scanning logic by removing scanAll method

Signed-off-by: Deluan <deluan@navidrome.org>

* test: enhance mockScanner for thread safety and improve test reliability

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): move scanner.ScanTarget to model.ScanTarget

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor: move scanner types to model,implement MockScanner

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): update scanner interface and implementations to use model.Scanner

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(folder_repository): normalize target path handling by using filepath.Clean

Signed-off-by: Deluan <deluan@navidrome.org>

* test(folder_repository): add comprehensive tests for folder retrieval and child exclusion

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): simplify selective scan logic using slice.Filter

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): streamline phase folder and album creation by removing unnecessary library parameter

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): move initialization logic from phase_1 to the scanner itself

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(tests): rename selective scan test file to scanner_selective_test.go

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(configuration): add DevSelectiveWatcher configuration option

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(watcher): enhance .ndignore handling for folder deletions and file changes

Signed-off-by: Deluan <deluan@navidrome.org>

* docs(scanner): comments

Signed-off-by: Deluan <deluan@navidrome.org>

* refactor(scanner): enhance walkDirTree to support target folder scanning

Signed-off-by: Deluan <deluan@navidrome.org>

* fix(scanner, watcher): handle errors when pushing ignore patterns for folders

Signed-off-by: Deluan <deluan@navidrome.org>

* Update scanner/phase_1_folders.go

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* refactor(scanner): replace parseTargets function with direct call to scanner.ParseTargets

Signed-off-by: Deluan <deluan@navidrome.org>

* test(scanner): add tests for ScanBegin and ScanEnd functionality

Signed-off-by: Deluan <deluan@navidrome.org>

* fix(library): update PRAGMA optimize to check table sizes without ANALYZE

Signed-off-by: Deluan <deluan@navidrome.org>

* test(scanner): refactor tests

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(ui): add selective scan options and update translations

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(ui): add quick and full scan options for individual libraries

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(ui): add Scan buttonsto the LibraryList

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(scan): update scanning parameters from 'path' to 'target' for selective scans.

* refactor(scan): move ParseTargets function to model package

* test(scan): suppress unused return value from SetUserLibraries in tests

* feat(gc): enhance garbage collection to support selective library purging

Signed-off-by: Deluan <deluan@navidrome.org>

* fix(scanner): prevent race condition when scanning deleted folders

When the watcher detects changes in a folder that gets deleted before
the scanner runs (due to the 10-second delay), the scanner was
prematurely removing these folders from the tracking map, preventing
them from being marked as missing.

The issue occurred because `newFolderEntry` was calling `popLastUpdate`
before verifying the folder actually exists on the filesystem.

Changes:
- Move fs.Stat check before newFolderEntry creation in loadDir to
  ensure deleted folders remain in lastUpdates for finalize() to handle
- Add early existence check in walkDirTree to skip non-existent target
  folders with a warning log
- Add unit test verifying non-existent folders aren't removed from
  lastUpdates prematurely
- Add integration test for deleted folder scenario with ScanFolders

Fixes the issue where deleting entire folders (e.g., /music/AC_DC)
wouldn't mark tracks as missing when using selective folder scanning.

* refactor(scan): streamline folder entry creation and update handling

Signed-off-by: Deluan <deluan@navidrome.org>

* feat(scan): add '@Recycle' (QNAP) to ignored directories list

Signed-off-by: Deluan <deluan@navidrome.org>

* fix(log): improve thread safety in logging level management

* test(scan): move unit tests for ParseTargets function

Signed-off-by: Deluan <deluan@navidrome.org>

* review

Signed-off-by: Deluan <deluan@navidrome.org>

---------

Signed-off-by: Deluan <deluan@navidrome.org>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: deluan <deluan.quintao@mechanical-orchard.com>
2025-11-14 22:15:43 -05:00

255 lines
8.2 KiB
Go

package scanner
import (
"context"
"io/fs"
"maps"
"path"
"slices"
"sort"
"strings"
"github.com/navidrome/navidrome/conf"
"github.com/navidrome/navidrome/log"
"github.com/navidrome/navidrome/model"
"github.com/navidrome/navidrome/utils"
)
// walkDirTree recursively walks the directory tree starting from the given targetFolders.
// If no targetFolders are provided, it starts from the root folder (".").
// It returns a channel of folderEntry pointers representing each folder found.
func walkDirTree(ctx context.Context, job *scanJob, targetFolders ...string) (<-chan *folderEntry, error) {
results := make(chan *folderEntry)
folders := targetFolders
if len(targetFolders) == 0 {
// No specific folders provided, scan the root folder
folders = []string{"."}
}
go func() {
defer close(results)
for _, folderPath := range folders {
if utils.IsCtxDone(ctx) {
return
}
// Check if target folder exists before walking it
// If it doesn't exist (e.g., deleted between watcher detection and scan execution),
// skip it so it remains in job.lastUpdates and gets handled in following steps
_, err := fs.Stat(job.fs, folderPath)
if err != nil {
log.Warn(ctx, "Scanner: Target folder does not exist.", "path", folderPath, err)
continue
}
// Create checker and push patterns from root to this folder
checker := newIgnoreChecker(job.fs)
err = checker.PushAllParents(ctx, folderPath)
if err != nil {
log.Error(ctx, "Scanner: Error pushing ignore patterns for target folder", "path", folderPath, err)
continue
}
// Recursively walk this folder and all its children
err = walkFolder(ctx, job, folderPath, checker, results)
if err != nil {
log.Error(ctx, "Scanner: Error walking target folder", "path", folderPath, err)
continue
}
}
log.Debug(ctx, "Scanner: Finished reading target folders", "lib", job.lib.Name, "path", job.lib.Path, "numFolders", job.numFolders.Load())
}()
return results, nil
}
func walkFolder(ctx context.Context, job *scanJob, currentFolder string, checker *IgnoreChecker, results chan<- *folderEntry) error {
// Push patterns for this folder onto the stack
_ = checker.Push(ctx, currentFolder)
defer checker.Pop() // Pop patterns when leaving this folder
folder, children, err := loadDir(ctx, job, currentFolder, checker)
if err != nil {
log.Warn(ctx, "Scanner: Error loading dir. Skipping", "path", currentFolder, err)
return nil
}
for _, c := range children {
err := walkFolder(ctx, job, c, checker, results)
if err != nil {
return err
}
}
dir := path.Clean(currentFolder)
log.Trace(ctx, "Scanner: Found directory", " path", dir, "audioFiles", maps.Keys(folder.audioFiles),
"images", maps.Keys(folder.imageFiles), "playlists", folder.numPlaylists, "imagesUpdatedAt", folder.imagesUpdatedAt,
"updTime", folder.updTime, "modTime", folder.modTime, "numChildren", len(children))
folder.path = dir
folder.elapsed.Start()
results <- folder
return nil
}
func loadDir(ctx context.Context, job *scanJob, dirPath string, checker *IgnoreChecker) (folder *folderEntry, children []string, err error) {
// Check if directory exists before creating the folder entry
// This is important to avoid removing the folder from lastUpdates if it doesn't exist
dirInfo, err := fs.Stat(job.fs, dirPath)
if err != nil {
log.Warn(ctx, "Scanner: Error stating dir", "path", dirPath, err)
return nil, nil, err
}
// Now that we know the folder exists, create the entry (which removes it from lastUpdates)
folder = job.createFolderEntry(dirPath)
folder.modTime = dirInfo.ModTime()
dir, err := job.fs.Open(dirPath)
if err != nil {
log.Warn(ctx, "Scanner: Error in Opening directory", "path", dirPath, err)
return folder, children, err
}
defer dir.Close()
dirFile, ok := dir.(fs.ReadDirFile)
if !ok {
log.Error(ctx, "Not a directory", "path", dirPath)
return folder, children, err
}
entries := fullReadDir(ctx, dirFile)
children = make([]string, 0, len(entries))
for _, entry := range entries {
entryPath := path.Join(dirPath, entry.Name())
if checker.ShouldIgnore(ctx, entryPath) {
log.Trace(ctx, "Scanner: Ignoring entry", "path", entryPath)
continue
}
if isEntryIgnored(entry.Name()) {
continue
}
if ctx.Err() != nil {
return folder, children, ctx.Err()
}
isDir, err := isDirOrSymlinkToDir(job.fs, dirPath, entry)
// Skip invalid symlinks
if err != nil {
log.Warn(ctx, "Scanner: Invalid symlink", "dir", entryPath, err)
continue
}
if isDir && !isDirIgnored(entry.Name()) && isDirReadable(ctx, job.fs, entryPath) {
children = append(children, entryPath)
folder.numSubFolders++
} else {
fileInfo, err := entry.Info()
if err != nil {
log.Warn(ctx, "Scanner: Error getting fileInfo", "name", entry.Name(), err)
return folder, children, err
}
if fileInfo.ModTime().After(folder.modTime) {
folder.modTime = fileInfo.ModTime()
}
switch {
case model.IsAudioFile(entry.Name()):
folder.audioFiles[entry.Name()] = entry
case model.IsValidPlaylist(entry.Name()):
folder.numPlaylists++
case model.IsImageFile(entry.Name()):
folder.imageFiles[entry.Name()] = entry
folder.imagesUpdatedAt = utils.TimeNewest(folder.imagesUpdatedAt, fileInfo.ModTime(), folder.modTime)
}
}
}
return folder, children, nil
}
// fullReadDir reads all files in the folder, skipping the ones with errors.
// It also detects when it is "stuck" with an error in the same directory over and over.
// In this case, it stops and returns whatever it was able to read until it got stuck.
// See discussion here: https://github.com/navidrome/navidrome/issues/1164#issuecomment-881922850
func fullReadDir(ctx context.Context, dir fs.ReadDirFile) []fs.DirEntry {
var allEntries []fs.DirEntry
var prevErrStr = ""
for {
if ctx.Err() != nil {
return nil
}
entries, err := dir.ReadDir(-1)
allEntries = append(allEntries, entries...)
if err == nil {
break
}
log.Warn(ctx, "Skipping DirEntry", err)
if prevErrStr == err.Error() {
log.Error(ctx, "Scanner: Duplicate DirEntry failure, bailing", err)
break
}
prevErrStr = err.Error()
}
sort.Slice(allEntries, func(i, j int) bool { return allEntries[i].Name() < allEntries[j].Name() })
return allEntries
}
// isDirOrSymlinkToDir returns true if and only if the dirEnt represents a file
// system directory, or a symbolic link to a directory. Note that if the dirEnt
// is not a directory but is a symbolic link, this method will resolve by
// sending a request to the operating system to follow the symbolic link.
// originally copied from github.com/karrick/godirwalk, modified to use dirEntry for
// efficiency for go 1.16 and beyond
func isDirOrSymlinkToDir(fsys fs.FS, baseDir string, dirEnt fs.DirEntry) (bool, error) {
if dirEnt.IsDir() {
return true, nil
}
if dirEnt.Type()&fs.ModeSymlink == 0 {
return false, nil
}
// If symlinks are disabled, return false for symlinks
if !conf.Server.Scanner.FollowSymlinks {
return false, nil
}
// Does this symlink point to a directory?
fileInfo, err := fs.Stat(fsys, path.Join(baseDir, dirEnt.Name()))
if err != nil {
return false, err
}
return fileInfo.IsDir(), nil
}
// isDirReadable returns true if the directory represented by dirEnt is readable
func isDirReadable(ctx context.Context, fsys fs.FS, dirPath string) bool {
dir, err := fsys.Open(dirPath)
if err != nil {
log.Warn("Scanner: Skipping unreadable directory", "path", dirPath, err)
return false
}
err = dir.Close()
if err != nil {
log.Warn(ctx, "Scanner: Error closing directory", "path", dirPath, err)
}
return true
}
// List of special directories to ignore
var ignoredDirs = []string{
"$RECYCLE.BIN",
"#snapshot",
"@Recycle",
"@Recently-Snapshot",
".streams",
"lost+found",
}
// isDirIgnored returns true if the directory represented by dirEnt should be ignored
func isDirIgnored(name string) bool {
// allows Album folders for albums which eg start with ellipses
if strings.HasPrefix(name, ".") && !strings.HasPrefix(name, "..") {
return true
}
if slices.ContainsFunc(ignoredDirs, func(s string) bool { return strings.EqualFold(s, name) }) {
return true
}
return false
}
func isEntryIgnored(name string) bool {
return strings.HasPrefix(name, ".") && !strings.HasPrefix(name, "..")
}