feat: add unified worker pool for recursive downloads
This commit is contained in:
+82
-101
@@ -16,10 +16,10 @@ import (
|
||||
"path/filepath"
|
||||
"strconv"
|
||||
"strings"
|
||||
"sync"
|
||||
"time"
|
||||
|
||||
"codeberg.org/petrbalvin/goget/internal/output"
|
||||
"codeberg.org/petrbalvin/goget/internal/pool"
|
||||
)
|
||||
|
||||
type Client struct {
|
||||
@@ -698,6 +698,76 @@ func (p *clientPool) close() {
|
||||
}
|
||||
}
|
||||
|
||||
// ftpTask represents a unit of work for the concurrent recursive
|
||||
// download: either a directory to list or a file to download.
|
||||
type ftpTask struct {
|
||||
remotePath string
|
||||
localPath string
|
||||
isDir bool
|
||||
name string
|
||||
}
|
||||
|
||||
// ftpTaskHandler processes one ftpTask. For directories it lists the
|
||||
// contents and submits child tasks for each entry. For files it
|
||||
// acquires a client from the pool and downloads.
|
||||
func ftpTaskHandler(cpool *clientPool, progressFunc func(file string, current, total int64)) func(ctx context.Context, t ftpTask) {
|
||||
return func(ctx context.Context, t ftpTask) {
|
||||
if ctx.Err() != nil {
|
||||
return
|
||||
}
|
||||
|
||||
if err := os.MkdirAll(t.localPath, 0755); err != nil {
|
||||
pool.Get[ftpTask](ctx).AddError(fmt.Errorf("failed to create local dir %s: %w", t.localPath, err))
|
||||
return
|
||||
}
|
||||
|
||||
if !t.isDir {
|
||||
c := cpool.acquire()
|
||||
defer cpool.release(c)
|
||||
if progressFunc != nil {
|
||||
progressFunc(t.name, 0, 0)
|
||||
}
|
||||
var fileProgress func(current, total int64)
|
||||
if progressFunc != nil {
|
||||
fileProgress = func(current, total int64) {
|
||||
progressFunc(t.name, current, total)
|
||||
}
|
||||
}
|
||||
if err := c.DownloadFile(ctx, t.remotePath, t.localPath, 0, fileProgress); err != nil {
|
||||
pool.Get[ftpTask](ctx).AddError(fmt.Errorf("download %s: %w", t.remotePath, err))
|
||||
}
|
||||
return
|
||||
}
|
||||
|
||||
c := cpool.acquire()
|
||||
entries, err := c.ListRemoteDir(t.remotePath)
|
||||
cpool.release(c)
|
||||
if err != nil {
|
||||
pool.Get[ftpTask](ctx).AddError(fmt.Errorf("list %s: %w", t.remotePath, err))
|
||||
return
|
||||
}
|
||||
|
||||
p := pool.Get[ftpTask](ctx)
|
||||
for _, entry := range entries {
|
||||
fields := strings.Fields(entry)
|
||||
if len(fields) < 9 {
|
||||
continue
|
||||
}
|
||||
name := fields[len(fields)-1]
|
||||
if name == "." || name == ".." {
|
||||
continue
|
||||
}
|
||||
childIsDir := fields[0][0] == 'd'
|
||||
p.Submit(ftpTask{
|
||||
remotePath: filepath.Join(t.remotePath, name),
|
||||
localPath: filepath.Join(t.localPath, name),
|
||||
isDir: childIsDir,
|
||||
name: name,
|
||||
})
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// DownloadDirectoryConcurrent downloads a remote directory to localPath
|
||||
// using a pool of N FTP connections. All operations (LIST, RETR) use
|
||||
// absolute paths so concurrent workers never share or mutate CWD state
|
||||
@@ -708,110 +778,21 @@ func DownloadDirectoryConcurrent(ctx context.Context, rawURL string, timeout tim
|
||||
if parallel < 1 {
|
||||
parallel = 1
|
||||
}
|
||||
pool, err := newClientPool(rawURL, timeout, parallel)
|
||||
cpool, err := newClientPool(rawURL, timeout, parallel)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
defer pool.close()
|
||||
defer cpool.close()
|
||||
|
||||
sem := make(chan struct{}, parallel)
|
||||
var wg sync.WaitGroup
|
||||
var firstErr error
|
||||
var errMu sync.Mutex
|
||||
p := pool.New(ctx, parallel, ftpTaskHandler(cpool, progressFunc))
|
||||
p.Submit(ftpTask{
|
||||
remotePath: remotePath,
|
||||
localPath: localPath,
|
||||
isDir: true,
|
||||
})
|
||||
|
||||
setErr := func(err error) {
|
||||
errMu.Lock()
|
||||
if firstErr == nil {
|
||||
firstErr = err
|
||||
}
|
||||
errMu.Unlock()
|
||||
}
|
||||
|
||||
processEntry(ctx, pool, sem, &wg, setErr, remotePath, localPath, progressFunc)
|
||||
|
||||
wg.Wait()
|
||||
return firstErr
|
||||
}
|
||||
|
||||
// processEntry processes one remote path: lists it, then dispatches
|
||||
// its child entries (file → download, subdir → recursive processEntry)
|
||||
// through the shared semaphore-bounded goroutine pool. A single
|
||||
// sem is threaded through the recursion so the total in-flight
|
||||
// goroutines never exceed `parallel`, regardless of directory depth.
|
||||
func processEntry(ctx context.Context, pool *clientPool, sem chan struct{}, wg *sync.WaitGroup, setErr func(error), remotePath, localPath string, progressFunc func(file string, current, total int64)) {
|
||||
defer wg.Done()
|
||||
defer func() { <-sem }()
|
||||
|
||||
if ctx.Err() != nil {
|
||||
return
|
||||
}
|
||||
|
||||
if err := os.MkdirAll(localPath, 0755); err != nil {
|
||||
setErr(fmt.Errorf("failed to create local dir %s: %w", localPath, err))
|
||||
return
|
||||
}
|
||||
|
||||
c := pool.acquire()
|
||||
entries, err := c.ListRemoteDir(remotePath)
|
||||
pool.release(c)
|
||||
if err != nil {
|
||||
setErr(fmt.Errorf("list %s: %w", remotePath, err))
|
||||
return
|
||||
}
|
||||
|
||||
for _, entry := range entries {
|
||||
fields := strings.Fields(entry)
|
||||
if len(fields) < 9 {
|
||||
continue
|
||||
}
|
||||
name := fields[len(fields)-1]
|
||||
if name == "." || name == ".." {
|
||||
continue
|
||||
}
|
||||
isDir := fields[0][0] == 'd'
|
||||
remoteFilePath := filepath.Join(remotePath, name)
|
||||
localFilePath := filepath.Join(localPath, name)
|
||||
|
||||
if isDir {
|
||||
wg.Add(1)
|
||||
select {
|
||||
case sem <- struct{}{}:
|
||||
go processEntry(ctx, pool, sem, wg, setErr, remoteFilePath, localFilePath, progressFunc)
|
||||
case <-ctx.Done():
|
||||
return
|
||||
}
|
||||
} else {
|
||||
wg.Add(1)
|
||||
select {
|
||||
case sem <- struct{}{}:
|
||||
go func(remoteFilePath, localFilePath, name string) {
|
||||
defer wg.Done()
|
||||
defer func() { <-sem }()
|
||||
if ctx.Err() != nil {
|
||||
return
|
||||
}
|
||||
c := pool.acquire()
|
||||
defer pool.release(c)
|
||||
if progressFunc != nil {
|
||||
progressFunc(name, 0, 0)
|
||||
}
|
||||
// DownloadFile's progress callback has a different
|
||||
// signature than our directory-level progressFunc, so
|
||||
// we wrap it here to keep the existing per-file
|
||||
// progress behaviour.
|
||||
var fileProgress func(current, total int64)
|
||||
if progressFunc != nil {
|
||||
fileProgress = func(current, total int64) {
|
||||
progressFunc(name, current, total)
|
||||
}
|
||||
}
|
||||
if err := c.DownloadFile(ctx, remoteFilePath, localFilePath, 0, fileProgress); err != nil {
|
||||
setErr(fmt.Errorf("download %s: %w", remoteFilePath, err))
|
||||
}
|
||||
}(remoteFilePath, localFilePath, name)
|
||||
case <-ctx.Done():
|
||||
return
|
||||
}
|
||||
}
|
||||
if errs := p.Wait(); len(errs) > 0 {
|
||||
return errs[0]
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
+88
-139
@@ -20,6 +20,7 @@ import (
|
||||
|
||||
"codeberg.org/petrbalvin/goget/internal/core"
|
||||
"codeberg.org/petrbalvin/goget/internal/output"
|
||||
"codeberg.org/petrbalvin/goget/internal/pool"
|
||||
"codeberg.org/petrbalvin/goget/internal/protocol"
|
||||
"golang.org/x/crypto/ssh"
|
||||
)
|
||||
@@ -862,56 +863,113 @@ func downloadRecursiveSequential(ctx context.Context, req *core.DownloadRequest)
|
||||
return sftpDownloadDirStandalone(ctx, c, req)
|
||||
}
|
||||
|
||||
// sftpTask represents a unit of work for the concurrent recursive
|
||||
// download: either a directory to list or a file to download.
|
||||
type sftpTask struct {
|
||||
req *core.DownloadRequest
|
||||
isDir bool
|
||||
size uint64
|
||||
}
|
||||
|
||||
// downloadRecursiveParallel downloads a remote directory using a pool
|
||||
// of N SFTP connections. All workers share a single semaphore so the
|
||||
// total in-flight goroutines never exceed N, regardless of directory
|
||||
// depth.
|
||||
// of N SFTP connections. All workers share a single pool so the total
|
||||
// in-flight goroutines never exceed N, regardless of directory depth.
|
||||
func downloadRecursiveParallel(ctx context.Context, req *core.DownloadRequest) (*core.DownloadResult, error) {
|
||||
parallel := req.RecursiveParallel
|
||||
if parallel < 1 {
|
||||
parallel = 1
|
||||
}
|
||||
pool, err := newSFTPClientPool(ctx, req, parallel)
|
||||
cpool, err := newSFTPClientPool(ctx, req, parallel)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
defer pool.close()
|
||||
defer cpool.close()
|
||||
|
||||
var (
|
||||
statsMu sync.Mutex
|
||||
totalBytes int64
|
||||
totalChunks int
|
||||
firstErr error
|
||||
)
|
||||
|
||||
sem := make(chan struct{}, parallel)
|
||||
var wg sync.WaitGroup
|
||||
|
||||
setErr := func(err error) {
|
||||
statsMu.Lock()
|
||||
if firstErr == nil {
|
||||
firstErr = err
|
||||
}
|
||||
statsMu.Unlock()
|
||||
}
|
||||
|
||||
wg.Add(1)
|
||||
sem <- struct{}{}
|
||||
go func() {
|
||||
defer wg.Done()
|
||||
defer func() { <-sem }()
|
||||
if ctx.Err() != nil {
|
||||
p := pool.New(ctx, parallel, func(taskCtx context.Context, t sftpTask) {
|
||||
if taskCtx.Err() != nil {
|
||||
return
|
||||
}
|
||||
c := pool.acquire()
|
||||
defer pool.release(c)
|
||||
sftpDownloadDirWorker(ctx, c, req, &wg, sem, setErr, &statsMu, &totalBytes, &totalChunks, pool)
|
||||
}()
|
||||
|
||||
wg.Wait()
|
||||
if firstErr != nil {
|
||||
return nil, firstErr
|
||||
c := cpool.acquire()
|
||||
defer cpool.release(c)
|
||||
|
||||
if !t.isDir {
|
||||
if err := sftpDownloadFile(taskCtx, c, t.req, t.size); err != nil {
|
||||
pool.Get[sftpTask](taskCtx).AddError(fmt.Errorf("sftp download %s: %w", t.req.URL.Path, err))
|
||||
return
|
||||
}
|
||||
statsMu.Lock()
|
||||
totalBytes += int64(t.size)
|
||||
totalChunks++
|
||||
statsMu.Unlock()
|
||||
return
|
||||
}
|
||||
|
||||
// Directory: list and submit child tasks.
|
||||
remotePath := t.req.URL.Path
|
||||
if remotePath == "" {
|
||||
remotePath = "."
|
||||
}
|
||||
|
||||
absPath, err := c.realpath(remotePath)
|
||||
if err != nil {
|
||||
absPath = remotePath
|
||||
}
|
||||
|
||||
outputDir := t.req.Output
|
||||
if outputDir == "" {
|
||||
outputDir = filepath.Base(absPath)
|
||||
if outputDir == "" || outputDir == "/" {
|
||||
outputDir = "download"
|
||||
}
|
||||
}
|
||||
|
||||
entries, err := sftpListDir(taskCtx, c, absPath)
|
||||
if err != nil {
|
||||
pool.Get[sftpTask](taskCtx).AddError(fmt.Errorf("sftp list %s: %w", absPath, err))
|
||||
return
|
||||
}
|
||||
|
||||
if err := os.MkdirAll(outputDir, 0755); err != nil {
|
||||
pool.Get[sftpTask](taskCtx).AddError(fmt.Errorf("mkdir %s: %w", outputDir, err))
|
||||
return
|
||||
}
|
||||
|
||||
pp := pool.Get[sftpTask](taskCtx)
|
||||
for _, e := range entries {
|
||||
name := e.Name
|
||||
if name == "." || name == ".." {
|
||||
continue
|
||||
}
|
||||
entryIsDir := e.Attr != nil && sftpIsDir(e.Attr.Mode)
|
||||
remoteChild := absPath + "/" + name
|
||||
localChild := filepath.Join(outputDir, name)
|
||||
subReq := *t.req
|
||||
subReq.URL = &url.URL{Scheme: t.req.URL.Scheme, User: t.req.URL.User, Host: t.req.URL.Host, Path: remoteChild}
|
||||
subReq.Output = localChild
|
||||
if entryIsDir {
|
||||
pp.Submit(sftpTask{req: &subReq, isDir: true})
|
||||
} else {
|
||||
subReq.Recursive = false
|
||||
subReq.RecursiveParallel = 0
|
||||
pp.Submit(sftpTask{req: &subReq, isDir: false, size: e.Attr.Size})
|
||||
}
|
||||
}
|
||||
})
|
||||
|
||||
p.Submit(sftpTask{req: req, isDir: true})
|
||||
|
||||
if errs := p.Wait(); len(errs) > 0 {
|
||||
return nil, errs[0]
|
||||
}
|
||||
statsMu.Lock()
|
||||
defer statsMu.Unlock()
|
||||
return &core.DownloadResult{
|
||||
BytesDownloaded: totalBytes,
|
||||
Duration: 0,
|
||||
@@ -958,115 +1016,6 @@ func sftpDownloadDirStandalone(ctx context.Context, c *client, req *core.Downloa
|
||||
}, nil
|
||||
}
|
||||
|
||||
// sftpDownloadDirWorker is the worker-pool variant. It lists the
|
||||
// directory, then dispatches file downloads and subdirectory recursion
|
||||
// through the shared semaphore. Subdirectory workers reuse the same
|
||||
// pool to acquire clients. The return value is always nil; results
|
||||
// and errors are routed through the shared fields.
|
||||
func sftpDownloadDirWorker(
|
||||
ctx context.Context,
|
||||
c *client,
|
||||
req *core.DownloadRequest,
|
||||
wg *sync.WaitGroup,
|
||||
sem chan struct{},
|
||||
setErr func(error),
|
||||
statsMu *sync.Mutex,
|
||||
totalBytes *int64,
|
||||
totalChunks *int,
|
||||
pool *sftpClientPool,
|
||||
) {
|
||||
remotePath := req.URL.Path
|
||||
if remotePath == "" {
|
||||
remotePath = "."
|
||||
}
|
||||
|
||||
absPath, err := c.realpath(remotePath)
|
||||
if err != nil {
|
||||
absPath = remotePath
|
||||
}
|
||||
|
||||
outputDir := req.Output
|
||||
if outputDir == "" {
|
||||
outputDir = filepath.Base(absPath)
|
||||
if outputDir == "" || outputDir == "/" {
|
||||
outputDir = "download"
|
||||
}
|
||||
}
|
||||
|
||||
entries, err := sftpListDir(ctx, c, absPath)
|
||||
if err != nil {
|
||||
setErr(fmt.Errorf("sftp list %s: %w", absPath, err))
|
||||
return
|
||||
}
|
||||
|
||||
if err := os.MkdirAll(outputDir, 0755); err != nil {
|
||||
setErr(fmt.Errorf("mkdir %s: %w", outputDir, err))
|
||||
return
|
||||
}
|
||||
|
||||
for i, e := range entries {
|
||||
name := e.Name
|
||||
if name == "." || name == ".." {
|
||||
continue
|
||||
}
|
||||
isDir := e.Attr != nil && sftpIsDir(e.Attr.Mode)
|
||||
remoteChild := absPath + "/" + name
|
||||
localChild := filepath.Join(outputDir, name)
|
||||
_ = i
|
||||
|
||||
if isDir {
|
||||
wg.Add(1)
|
||||
select {
|
||||
case sem <- struct{}{}:
|
||||
go func(remoteChild, localChild string) {
|
||||
defer wg.Done()
|
||||
defer func() { <-sem }()
|
||||
if ctx.Err() != nil {
|
||||
return
|
||||
}
|
||||
cc := pool.acquire()
|
||||
defer pool.release(cc)
|
||||
subReq := *req
|
||||
subReq.URL = &url.URL{Scheme: req.URL.Scheme, User: req.URL.User, Host: req.URL.Host, Path: remoteChild}
|
||||
subReq.Output = localChild
|
||||
sftpDownloadDirWorker(ctx, cc, &subReq, wg, sem, setErr, statsMu, totalBytes, totalChunks, pool)
|
||||
}(remoteChild, localChild)
|
||||
case <-ctx.Done():
|
||||
return
|
||||
}
|
||||
} else {
|
||||
wg.Add(1)
|
||||
select {
|
||||
case sem <- struct{}{}:
|
||||
go func(remoteChild, localChild string, size uint64) {
|
||||
defer wg.Done()
|
||||
defer func() { <-sem }()
|
||||
if ctx.Err() != nil {
|
||||
return
|
||||
}
|
||||
cc := pool.acquire()
|
||||
defer pool.release(cc)
|
||||
subReq := *req
|
||||
subReq.URL = &url.URL{Scheme: req.URL.Scheme, User: req.URL.User, Host: req.URL.Host, Path: remoteChild}
|
||||
subReq.Output = localChild
|
||||
subReq.Recursive = false
|
||||
subReq.RecursiveParallel = 0
|
||||
if err := sftpDownloadFile(ctx, cc, &subReq, size); err != nil {
|
||||
setErr(fmt.Errorf("sftp download %s: %w", remoteChild, err))
|
||||
return
|
||||
}
|
||||
statsMu.Lock()
|
||||
*totalBytes += int64(size)
|
||||
*totalChunks++
|
||||
statsMu.Unlock()
|
||||
}(remoteChild, localChild, e.Attr.Size)
|
||||
case <-ctx.Done():
|
||||
return
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// sftpWalkDir recursively walks one remote directory on `c`,
|
||||
// downloading each file and recursing into subdirectories. Stats are
|
||||
// accumulated inline. Used by the standalone mode.
|
||||
|
||||
@@ -21,6 +21,7 @@ import (
|
||||
|
||||
"codeberg.org/petrbalvin/goget/internal/core"
|
||||
"codeberg.org/petrbalvin/goget/internal/output"
|
||||
"codeberg.org/petrbalvin/goget/internal/pool"
|
||||
"codeberg.org/petrbalvin/goget/internal/protocol"
|
||||
"codeberg.org/petrbalvin/goget/internal/transport"
|
||||
)
|
||||
@@ -982,9 +983,18 @@ func (p *Protocol) downloadRecursiveSequential(ctx context.Context, req *core.Do
|
||||
}, nil
|
||||
}
|
||||
|
||||
// webdavTask represents a unit of work for the concurrent recursive
|
||||
// download: a file or subdirectory entry with its pre-built request.
|
||||
type webdavTask struct {
|
||||
entry WebDAVEntry
|
||||
subReq *core.DownloadRequest
|
||||
outPath string
|
||||
isDir bool
|
||||
}
|
||||
|
||||
// downloadRecursiveParallel is the worker-pool implementation used when
|
||||
// req.RecursiveParallel > 1. Top-level entries are processed concurrently,
|
||||
// bounded by a semaphore. Subdirectory recursion is delegated back to
|
||||
// bounded by the pool. Subdirectory recursion is delegated back to
|
||||
// downloadRecursive, so the same RecursiveParallel limit applies at every
|
||||
// level of the tree. Stats are aggregated under a mutex.
|
||||
func (p *Protocol) downloadRecursiveParallel(ctx context.Context, req *core.DownloadRequest) (*core.DownloadResult, error) {
|
||||
@@ -1021,24 +1031,52 @@ func (p *Protocol) downloadRecursiveParallel(ctx context.Context, req *core.Down
|
||||
// these counters, so they reflect the entire subtree processed by
|
||||
// this call.
|
||||
var (
|
||||
statsMu sync.Mutex
|
||||
totalBytes int64
|
||||
totalChunks int
|
||||
processedDirs int
|
||||
statsMu sync.Mutex
|
||||
totalBytes int64
|
||||
totalChunks int
|
||||
)
|
||||
|
||||
processedDirs++
|
||||
processedDirs := 1
|
||||
|
||||
// Pre-build sub-requests and output paths so the goroutine body has
|
||||
// no per-entry allocation cost.
|
||||
type task struct {
|
||||
entry WebDAVEntry
|
||||
subReq *core.DownloadRequest
|
||||
outPath string
|
||||
isDir bool
|
||||
}
|
||||
wp := pool.New(ctx, req.RecursiveParallel, func(taskCtx context.Context, t webdavTask) {
|
||||
if taskCtx.Err() != nil {
|
||||
return
|
||||
}
|
||||
|
||||
if t.isDir {
|
||||
subResult, err := p.downloadRecursive(taskCtx, t.subReq)
|
||||
if err != nil {
|
||||
if req.Verbose {
|
||||
fmt.Fprintf(os.Stderr, "[webdav] Warning: failed to %s sub-directory %s: %v\n", dryRunVerb(req.DryRun), t.entry.URL.String(), err)
|
||||
}
|
||||
return
|
||||
}
|
||||
statsMu.Lock()
|
||||
totalBytes += subResult.BytesDownloaded
|
||||
totalChunks += subResult.ChunksCount
|
||||
statsMu.Unlock()
|
||||
return
|
||||
}
|
||||
|
||||
subResult, err := p.downloadFile(taskCtx, t.subReq, t.entry.Size, t.entry.LastModified)
|
||||
if err != nil {
|
||||
if req.Verbose {
|
||||
fmt.Fprintf(os.Stderr, "[webdav] Warning: failed to download file %s: %v\n", t.entry.URL.String(), err)
|
||||
}
|
||||
return
|
||||
}
|
||||
|
||||
// Apply keep-timestamps if requested.
|
||||
if req.KeepTimestamps {
|
||||
setFileTimestamp(t.outPath, t.entry.LastModified)
|
||||
}
|
||||
|
||||
statsMu.Lock()
|
||||
totalBytes += subResult.BytesDownloaded
|
||||
totalChunks++
|
||||
statsMu.Unlock()
|
||||
})
|
||||
|
||||
tasks := make([]task, 0, len(filteredEntries))
|
||||
for _, entry := range filteredEntries {
|
||||
// Respect MaxDepth to avoid infinite recursion.
|
||||
if entry.IsDir && req.MaxDepth == 0 {
|
||||
@@ -1065,7 +1103,7 @@ func (p *Protocol) downloadRecursiveParallel(ctx context.Context, req *core.Down
|
||||
subReq.MaxDepth = req.MaxDepth - 1
|
||||
subReq.RecursiveParallel = req.RecursiveParallel
|
||||
subReq.DryRun = true
|
||||
tasks = append(tasks, task{entry: entry, subReq: &subReq, outPath: itemOutputPath, isDir: true})
|
||||
wp.Submit(webdavTask{entry: entry, subReq: &subReq, outPath: itemOutputPath, isDir: true})
|
||||
} else {
|
||||
fmt.Fprintf(os.Stderr, "[webdav] [dry-run] would download: %s (%d bytes)\n", entry.URL.String(), entry.Size)
|
||||
statsMu.Lock()
|
||||
@@ -1106,7 +1144,7 @@ func (p *Protocol) downloadRecursiveParallel(ctx context.Context, req *core.Down
|
||||
RejectPatterns: req.RejectPatterns,
|
||||
Ctx: req.Ctx,
|
||||
}
|
||||
tasks = append(tasks, task{entry: entry, subReq: subReq, outPath: itemOutputPath, isDir: true})
|
||||
wp.Submit(webdavTask{entry: entry, subReq: subReq, outPath: itemOutputPath, isDir: true})
|
||||
} else {
|
||||
if req.Verbose {
|
||||
fmt.Fprintf(os.Stderr, "[webdav] Downloading file %s -> %s (depth=%d)\n", entry.URL.String(), itemOutputPath, req.MaxDepth)
|
||||
@@ -1126,71 +1164,11 @@ func (p *Protocol) downloadRecursiveParallel(ctx context.Context, req *core.Down
|
||||
MaxDepth: req.MaxDepth - 1,
|
||||
Ctx: req.Ctx,
|
||||
}
|
||||
tasks = append(tasks, task{entry: entry, subReq: subReq, outPath: itemOutputPath, isDir: false})
|
||||
wp.Submit(webdavTask{entry: entry, subReq: subReq, outPath: itemOutputPath, isDir: false})
|
||||
}
|
||||
}
|
||||
|
||||
// Worker pool: bounded by req.RecursiveParallel. We use a semaphore
|
||||
// channel + WaitGroup so total in-flight workers never exceed the
|
||||
// requested count, regardless of how many tasks are enqueued.
|
||||
sem := make(chan struct{}, req.RecursiveParallel)
|
||||
var wg sync.WaitGroup
|
||||
|
||||
for _, t := range tasks {
|
||||
// Honour context cancellation before launching a new task.
|
||||
if ctx.Err() != nil {
|
||||
break
|
||||
}
|
||||
|
||||
wg.Add(1)
|
||||
sem <- struct{}{}
|
||||
|
||||
go func(t task) {
|
||||
defer wg.Done()
|
||||
defer func() { <-sem }()
|
||||
|
||||
// If the context was cancelled while we were waiting for
|
||||
// the semaphore, just exit cleanly.
|
||||
if ctx.Err() != nil {
|
||||
return
|
||||
}
|
||||
|
||||
if t.isDir {
|
||||
subResult, err := p.downloadRecursive(ctx, t.subReq)
|
||||
if err != nil {
|
||||
if req.Verbose {
|
||||
fmt.Fprintf(os.Stderr, "[webdav] Warning: failed to %s sub-directory %s: %v\n", dryRunVerb(req.DryRun), t.entry.URL.String(), err)
|
||||
}
|
||||
return
|
||||
}
|
||||
statsMu.Lock()
|
||||
totalBytes += subResult.BytesDownloaded
|
||||
totalChunks += subResult.ChunksCount
|
||||
statsMu.Unlock()
|
||||
return
|
||||
}
|
||||
|
||||
subResult, err := p.downloadFile(ctx, t.subReq, t.entry.Size, t.entry.LastModified)
|
||||
if err != nil {
|
||||
if req.Verbose {
|
||||
fmt.Fprintf(os.Stderr, "[webdav] Warning: failed to download file %s: %v\n", t.entry.URL.String(), err)
|
||||
}
|
||||
return
|
||||
}
|
||||
|
||||
// Apply keep-timestamps if requested.
|
||||
if req.KeepTimestamps {
|
||||
setFileTimestamp(t.outPath, t.entry.LastModified)
|
||||
}
|
||||
|
||||
statsMu.Lock()
|
||||
totalBytes += subResult.BytesDownloaded
|
||||
totalChunks++
|
||||
statsMu.Unlock()
|
||||
}(t)
|
||||
}
|
||||
|
||||
wg.Wait()
|
||||
wp.Wait()
|
||||
|
||||
_ = processedDirs // reserved for future per-directory accounting
|
||||
|
||||
|
||||
Reference in New Issue
Block a user