feat(local): add directory size support (#624)

* feat(local): add directory size support

* fix(local): fix and improve directory size calculation

* style(local): fix code style

* style(local): fix code style

* style(local): fix code style

* fix(local): refresh directory size when force refresh

Signed-off-by: 我怎么就不是一只猫呢? <26274059+dezhishen@users.noreply.github.com>

* fix:(local): Avoid traversing the parent's parent, which leads to an endless loop

Signed-off-by: 我怎么就不是一只猫呢? <26274059+dezhishen@users.noreply.github.com>

* fix(local:) refresh dir size only enabled

Signed-off-by: 我怎么就不是一只猫呢? <26274059+dezhishen@users.noreply.github.com>

* fix(local): logical error && add RecalculateDirSize && cleaner code for int64

* feat(local): add Benchmark for CalculateDirSize

* refactor(local): 优化移动中对于错误的判断。

---------

Signed-off-by: 我怎么就不是一只猫呢? <26274059+dezhishen@users.noreply.github.com>
Co-authored-by: 我怎么就不是一只猫呢? <26274059+dezhishen@users.noreply.github.com>
This commit is contained in:
LmanTW
2025-08-08 16:59:16 +08:00
committed by GitHub
parent b9b8eed285
commit 93c06213d4
4 changed files with 434 additions and 17 deletions

View File

@ -0,0 +1,92 @@
package local
// TestDirCalculateSize tests the directory size calculation
// It should be run with the local driver enabled and directory size calculation set to true
import (
"os"
"path/filepath"
"strconv"
"testing"
"github.com/OpenListTeam/OpenList/v4/internal/driver"
)
func generatedTestDir(dir string, dep, filecount int) {
if dep == 0 {
return
}
for i := 0; i < dep; i++ {
subDir := dir + "/dir" + strconv.Itoa(i)
os.Mkdir(subDir, 0755)
generatedTestDir(subDir, dep-1, filecount)
generatedFiles(subDir, filecount)
}
}
func generatedFiles(path string, count int) error {
for i := 0; i < count; i++ {
filePath := filepath.Join(path, "file"+strconv.Itoa(i)+".txt")
file, err := os.Create(filePath)
if err != nil {
return err
}
// 使用随机ascii字符填充文件
content := make([]byte, 1024) // 1KB file
for j := range content {
content[j] = byte('a' + j%26) // Fill with 'a' to 'z'
}
_, err = file.Write(content)
if err != nil {
return err
}
file.Close()
}
return nil
}
// performance tests for directory size calculation
func BenchmarkCalculateDirSize(t *testing.B) {
// 初始化t的日志
t.Logf("Starting performance test for directory size calculation")
// 确保测试目录存在
if testing.Short() {
t.Skip("Skipping performance test in short mode")
}
// 创建tmp directory for testing
testTempDir := t.TempDir()
err := os.MkdirAll(testTempDir, 0755)
if err != nil {
t.Fatalf("Failed to create test directory: %v", err)
}
defer os.RemoveAll(testTempDir) // Clean up after test
// 构建一个深度为5每层10个文件和10个目录的目录结构
generatedTestDir(testTempDir, 5, 10)
// Initialize the local driver with directory size calculation enabled
d := &Local{
directoryMap: DirectoryMap{
root: testTempDir,
},
Addition: Addition{
DirectorySize: true,
RootPath: driver.RootPath{
RootFolderPath: testTempDir,
},
},
}
//record the start time
t.StartTimer()
// Calculate the directory size
err = d.directoryMap.RecalculateDirSize()
if err != nil {
t.Fatalf("Failed to calculate directory size: %v", err)
}
//record the end time
t.StopTimer()
// Print the size and duration
node, ok := d.directoryMap.Get(d.directoryMap.root)
if !ok {
t.Fatalf("Failed to get root node from directory map")
}
t.Logf("Directory size: %d bytes", node.fileSum+node.directorySum)
t.Logf("Performance test completed successfully")
}

View File

@ -33,6 +33,9 @@ type Local struct {
Addition
mkdirPerm int32
// directory size data
directoryMap DirectoryMap
// zero means no limit
thumbConcurrency int
thumbTokenBucket TokenBucket
@ -66,6 +69,15 @@ func (d *Local) Init(ctx context.Context) error {
}
d.Addition.RootFolderPath = abs
}
if d.DirectorySize {
d.directoryMap.root = d.GetRootPath()
_, err := d.directoryMap.CalculateDirSize(d.GetRootPath())
if err != nil {
return err
}
} else {
d.directoryMap.Clear()
}
if d.ThumbCacheFolder != "" && !utils.Exists(d.ThumbCacheFolder) {
err := os.MkdirAll(d.ThumbCacheFolder, os.FileMode(d.mkdirPerm))
if err != nil {
@ -124,6 +136,9 @@ func (d *Local) GetAddition() driver.Additional {
func (d *Local) List(ctx context.Context, dir model.Obj, args model.ListArgs) ([]model.Obj, error) {
fullPath := dir.GetPath()
rawFiles, err := readDir(fullPath)
if d.DirectorySize && args.Refresh {
d.directoryMap.RecalculateDirSize()
}
if err != nil {
return nil, err
}
@ -147,7 +162,12 @@ func (d *Local) FileInfoToObj(ctx context.Context, f fs.FileInfo, reqPath string
}
isFolder := f.IsDir() || isSymlinkDir(f, fullPath)
var size int64
if !isFolder {
if isFolder {
node, ok := d.directoryMap.Get(filepath.Join(fullPath, f.Name()))
if ok {
size = node.fileSum + node.directorySum
}
} else {
size = f.Size()
}
var ctime time.Time
@ -186,7 +206,12 @@ func (d *Local) Get(ctx context.Context, path string) (model.Obj, error) {
isFolder := f.IsDir() || isSymlinkDir(f, path)
size := f.Size()
if isFolder {
size = 0
node, ok := d.directoryMap.Get(path)
if ok {
size = node.fileSum + node.directorySum
}
} else {
size = f.Size()
}
var ctime time.Time
t, err := times.Stat(path)
@ -271,23 +296,32 @@ func (d *Local) Move(ctx context.Context, srcObj, dstDir model.Obj) error {
if utils.IsSubPath(srcPath, dstPath) {
return fmt.Errorf("the destination folder is a subfolder of the source folder")
}
if err := os.Rename(srcPath, dstPath); err != nil && strings.Contains(err.Error(), "invalid cross-device link") {
// Handle cross-device file move in local driver
if err = d.Copy(ctx, srcObj, dstDir); err != nil {
err := os.Rename(srcPath, dstPath)
if err != nil && strings.Contains(err.Error(), "invalid cross-device link") {
// 跨设备移动,先复制再删除
if err := d.Copy(ctx, srcObj, dstDir); err != nil {
return err
} else {
// Directly remove file without check recycle bin if successfully copied
}
// 复制成功后直接删除源文件/文件夹
if srcObj.IsDir() {
err = os.RemoveAll(srcObj.GetPath())
} else {
err = os.Remove(srcObj.GetPath())
return os.RemoveAll(srcObj.GetPath())
}
return os.Remove(srcObj.GetPath())
}
if err == nil {
srcParent := filepath.Dir(srcPath)
dstParent := filepath.Dir(dstPath)
if d.directoryMap.Has(srcParent) {
d.directoryMap.UpdateDirSize(srcParent)
d.directoryMap.UpdateDirParents(srcParent)
}
if d.directoryMap.Has(dstParent) {
d.directoryMap.UpdateDirSize(dstParent)
d.directoryMap.UpdateDirParents(dstParent)
}
}
return err
}
} else {
return err
}
}
func (d *Local) Rename(ctx context.Context, srcObj model.Obj, newName string) error {
srcPath := srcObj.GetPath()
@ -296,6 +330,14 @@ func (d *Local) Rename(ctx context.Context, srcObj model.Obj, newName string) er
if err != nil {
return err
}
if srcObj.IsDir() {
if d.directoryMap.Has(srcPath) {
d.directoryMap.DeleteDirNode(srcPath)
d.directoryMap.CalculateDirSize(dstPath)
}
}
return nil
}
@ -306,11 +348,21 @@ func (d *Local) Copy(_ context.Context, srcObj, dstDir model.Obj) error {
return fmt.Errorf("the destination folder is a subfolder of the source folder")
}
// Copy using otiai10/copy to perform more secure & efficient copy
return cp.Copy(srcPath, dstPath, cp.Options{
err := cp.Copy(srcPath, dstPath, cp.Options{
Sync: true, // Sync file to disk after copy, may have performance penalty in filesystem such as ZFS
PreserveTimes: true,
PreserveOwner: true,
})
if err != nil {
return err
}
if d.directoryMap.Has(filepath.Dir(dstPath)) {
d.directoryMap.UpdateDirSize(filepath.Dir(dstPath))
d.directoryMap.UpdateDirParents(filepath.Dir(dstPath))
}
return nil
}
func (d *Local) Remove(ctx context.Context, obj model.Obj) error {
@ -331,6 +383,19 @@ func (d *Local) Remove(ctx context.Context, obj model.Obj) error {
if err != nil {
return err
}
if obj.IsDir() {
if d.directoryMap.Has(obj.GetPath()) {
d.directoryMap.DeleteDirNode(obj.GetPath())
d.directoryMap.UpdateDirSize(filepath.Dir(obj.GetPath()))
d.directoryMap.UpdateDirParents(filepath.Dir(obj.GetPath()))
}
} else {
if d.directoryMap.Has(filepath.Dir(obj.GetPath())) {
d.directoryMap.UpdateDirSize(filepath.Dir(obj.GetPath()))
d.directoryMap.UpdateDirParents(filepath.Dir(obj.GetPath()))
}
}
return nil
}
@ -354,6 +419,11 @@ func (d *Local) Put(ctx context.Context, dstDir model.Obj, stream model.FileStre
if err != nil {
log.Errorf("[local] failed to change time of %s: %s", fullPath, err)
}
if d.directoryMap.Has(dstDir.GetPath()) {
d.directoryMap.UpdateDirSize(dstDir.GetPath())
d.directoryMap.UpdateDirParents(dstDir.GetPath())
}
return nil
}

View File

@ -7,6 +7,7 @@ import (
type Addition struct {
driver.RootPath
DirectorySize bool `json:"directory_size" default:"false" help:"This might impact host performance"`
Thumbnail bool `json:"thumbnail" required:"true" help:"enable thumbnail"`
ThumbCacheFolder string `json:"thumb_cache_folder"`
ThumbConcurrency string `json:"thumb_concurrency" default:"16" required:"false" help:"Number of concurrent thumbnail generation goroutines. This controls how many thumbnails can be generated in parallel."`
@ -27,6 +28,8 @@ var config = driver.Config{
func init() {
op.RegisterDriver(func() driver.Driver {
return &Local{}
return &Local{
directoryMap: DirectoryMap{},
}
})
}

View File

@ -8,9 +8,11 @@ import (
"os"
"path/filepath"
"runtime"
"slices"
"sort"
"strconv"
"strings"
"sync"
"github.com/OpenListTeam/OpenList/v4/internal/conf"
"github.com/OpenListTeam/OpenList/v4/internal/model"
@ -153,3 +155,253 @@ func (d *Local) getThumb(file model.Obj) (*bytes.Buffer, *string, error) {
}
return &buf, nil, nil
}
type DirectoryMap struct {
root string
data sync.Map
}
type DirectoryNode struct {
fileSum int64
directorySum int64
children []string
}
type DirectoryTask struct {
path string
cache *DirectoryTaskCache
}
type DirectoryTaskCache struct {
fileSum int64
children []string
}
func (m *DirectoryMap) Has(path string) bool {
_, ok := m.data.Load(path)
return ok
}
func (m *DirectoryMap) Get(path string) (*DirectoryNode, bool) {
value, ok := m.data.Load(path)
if !ok {
return &DirectoryNode{}, false
}
node, ok := value.(*DirectoryNode)
if !ok {
return &DirectoryNode{}, false
}
return node, true
}
func (m *DirectoryMap) Set(path string, node *DirectoryNode) {
m.data.Store(path, node)
}
func (m *DirectoryMap) Delete(path string) {
m.data.Delete(path)
}
func (m *DirectoryMap) Clear() {
m.data.Clear()
}
func (m *DirectoryMap) RecalculateDirSize() error {
m.Clear()
if m.root == "" {
return fmt.Errorf("root path is not set")
}
size, err := m.CalculateDirSize(m.root)
if err != nil {
return err
}
if node, ok := m.Get(m.root); ok {
node.fileSum = size
node.directorySum = size
}
return nil
}
func (m *DirectoryMap) CalculateDirSize(dirname string) (int64, error) {
stack := []DirectoryTask{
{path: dirname},
}
for len(stack) > 0 {
task := stack[len(stack)-1]
stack = stack[:len(stack)-1]
if task.cache != nil {
directorySum := int64(0)
for _, filename := range task.cache.children {
child, ok := m.Get(filepath.Join(task.path, filename))
if !ok {
return 0, fmt.Errorf("child node not found")
}
directorySum += child.fileSum + child.directorySum
}
m.Set(task.path, &DirectoryNode{
fileSum: task.cache.fileSum,
directorySum: directorySum,
children: task.cache.children,
})
continue
}
files, err := readDir(task.path)
if err != nil {
return 0, err
}
fileSum := int64(0)
directorySum := int64(0)
children := []string{}
queue := []DirectoryTask{}
for _, f := range files {
fullpath := filepath.Join(task.path, f.Name())
isFolder := f.IsDir() || isSymlinkDir(f, fullpath)
if isFolder {
if node, ok := m.Get(fullpath); ok {
directorySum += node.fileSum + node.directorySum
} else {
queue = append(queue, DirectoryTask{
path: fullpath,
})
}
children = append(children, f.Name())
} else {
fileSum += f.Size()
}
}
if len(queue) > 0 {
stack = append(stack, DirectoryTask{
path: task.path,
cache: &DirectoryTaskCache{
fileSum: fileSum,
children: children,
},
})
stack = append(stack, queue...)
continue
}
m.Set(task.path, &DirectoryNode{
fileSum: fileSum,
directorySum: directorySum,
children: children,
})
}
if node, ok := m.Get(dirname); ok {
return node.fileSum + node.directorySum, nil
}
return 0, nil
}
func (m *DirectoryMap) UpdateDirSize(dirname string) (int64, error) {
node, ok := m.Get(dirname)
if !ok {
return 0, fmt.Errorf("directory node not found")
}
files, err := readDir(dirname)
if err != nil {
return 0, err
}
fileSum := int64(0)
directorySum := int64(0)
children := []string{}
for _, f := range files {
fullpath := filepath.Join(dirname, f.Name())
isFolder := f.IsDir() || isSymlinkDir(f, fullpath)
if isFolder {
if node, ok := m.Get(fullpath); ok {
directorySum += node.fileSum + node.directorySum
} else {
value, err := m.CalculateDirSize(fullpath)
if err != nil {
return 0, err
}
directorySum += value
}
children = append(children, f.Name())
} else {
fileSum += f.Size()
}
}
for _, c := range node.children {
if !slices.Contains(children, c) {
m.DeleteDirNode(filepath.Join(dirname, c))
}
}
node.fileSum = fileSum
node.directorySum = directorySum
node.children = children
return fileSum + directorySum, nil
}
func (m *DirectoryMap) UpdateDirParents(dirname string) error {
parentPath := filepath.Dir(dirname)
for parentPath != m.root && !strings.HasPrefix(m.root, parentPath) {
if node, ok := m.Get(parentPath); ok {
directorySum := int64(0)
for _, c := range node.children {
child, ok := m.Get(filepath.Join(parentPath, c))
if !ok {
return fmt.Errorf("child node not found")
}
directorySum += child.fileSum + child.directorySum
}
node.directorySum = directorySum
}
parentPath = filepath.Dir(parentPath)
}
return nil
}
func (m *DirectoryMap) DeleteDirNode(dirname string) error {
stack := []string{dirname}
for len(stack) > 0 {
current := stack[len(stack)-1]
stack = stack[:len(stack)-1]
if node, ok := m.Get(current); ok {
for _, filename := range node.children {
stack = append(stack, filepath.Join(current, filename))
}
m.Delete(current)
}
}
return nil
}