rclone/fs/operations/operations.go

2303 lines
68 KiB
Go
Raw Normal View History

// Package operations does generic operations on filesystems and objects
package operations
import (
"bytes"
"context"
"encoding/base64"
"encoding/csv"
"encoding/hex"
"errors"
"fmt"
"io"
"io/ioutil"
"net/http"
"os"
"path"
"path/filepath"
"sort"
2018-01-06 14:39:31 +00:00
"strconv"
"strings"
"sync"
"sync/atomic"
"time"
"github.com/rclone/rclone/fs"
"github.com/rclone/rclone/fs/accounting"
"github.com/rclone/rclone/fs/cache"
"github.com/rclone/rclone/fs/config"
"github.com/rclone/rclone/fs/filter"
"github.com/rclone/rclone/fs/fserrors"
"github.com/rclone/rclone/fs/fshttp"
"github.com/rclone/rclone/fs/hash"
"github.com/rclone/rclone/fs/object"
"github.com/rclone/rclone/fs/walk"
"github.com/rclone/rclone/lib/atexit"
"github.com/rclone/rclone/lib/pacer"
"github.com/rclone/rclone/lib/random"
"github.com/rclone/rclone/lib/readers"
"golang.org/x/sync/errgroup"
)
// CheckHashes checks the two files to see if they have common
// known hash types and compares them
//
// Returns
//
// equal - which is equality of the hashes
//
// hash - the HashType. This is HashNone if either of the hashes were
// unset or a compatible hash couldn't be found.
//
// err - may return an error which will already have been logged
//
// If an error is returned it will return equal as false
func CheckHashes(ctx context.Context, src fs.ObjectInfo, dst fs.Object) (equal bool, ht hash.Type, err error) {
common := src.Fs().Hashes().Overlap(dst.Fs().Hashes())
// fs.Debugf(nil, "Shared hashes: %v", common)
if common.Count() == 0 {
return true, hash.None, nil
}
equal, ht, _, _, err = checkHashes(ctx, src, dst, common.GetOne())
return equal, ht, err
}
// checkHashes does the work of CheckHashes but takes a hash.Type and
// returns the effective hash type used.
func checkHashes(ctx context.Context, src fs.ObjectInfo, dst fs.Object, ht hash.Type) (equal bool, htOut hash.Type, srcHash, dstHash string, err error) {
// Calculate hashes in parallel
g, ctx := errgroup.WithContext(ctx)
g.Go(func() (err error) {
srcHash, err = src.Hash(ctx, ht)
if err != nil {
err = fs.CountError(err)
fs.Errorf(src, "Failed to calculate src hash: %v", err)
}
return err
})
g.Go(func() (err error) {
dstHash, err = dst.Hash(ctx, ht)
if err != nil {
err = fs.CountError(err)
fs.Errorf(dst, "Failed to calculate dst hash: %v", err)
}
return err
})
err = g.Wait()
if err != nil {
return false, ht, srcHash, dstHash, err
}
if srcHash == "" {
return true, hash.None, srcHash, dstHash, nil
}
if dstHash == "" {
return true, hash.None, srcHash, dstHash, nil
}
if srcHash != dstHash {
fs.Debugf(src, "%v = %s (%v)", ht, srcHash, src.Fs())
fs.Debugf(dst, "%v = %s (%v)", ht, dstHash, dst.Fs())
} else {
fs.Debugf(src, "%v = %s OK", ht, srcHash)
}
return srcHash == dstHash, ht, srcHash, dstHash, nil
}
2015-09-22 17:47:16 +00:00
// Equal checks to see if the src and dst objects are equal by looking at
// size, mtime and hash
//
// If the src and dst size are different then it is considered to be
// not equal. If --size-only is in effect then this is the only check
2016-06-17 16:20:08 +00:00
// that is done. If --ignore-size is in effect then this check is
// skipped and the files are considered the same size.
//
// If the size is the same and the mtime is the same then it is
// considered to be equal. This check is skipped if using --checksum.
//
// If the size is the same and mtime is different, unreadable or
// --checksum is set and the hash is the same then the file is
// considered to be equal. In this case the mtime on the dst is
// updated if --checksum is not set.
//
// Otherwise the file is considered to be not equal including if there
// were errors reading info.
func Equal(ctx context.Context, src fs.ObjectInfo, dst fs.Object) bool {
return equal(ctx, src, dst, defaultEqualOpt(ctx))
}
// sizeDiffers compare the size of src and dst taking into account the
// various ways of ignoring sizes
func sizeDiffers(ctx context.Context, src, dst fs.ObjectInfo) bool {
ci := fs.GetConfig(ctx)
if ci.IgnoreSize || src.Size() < 0 || dst.Size() < 0 {
return false
}
return src.Size() != dst.Size()
}
var checksumWarning sync.Once
// options for equal function()
type equalOpt struct {
sizeOnly bool // if set only check size
checkSum bool // if set check checksum+size instead of modtime+size
updateModTime bool // if set update the modtime if hashes identical and checking with modtime+size
forceModTimeMatch bool // if set assume modtimes match
}
// default set of options for equal()
func defaultEqualOpt(ctx context.Context) equalOpt {
ci := fs.GetConfig(ctx)
return equalOpt{
sizeOnly: ci.SizeOnly,
checkSum: ci.CheckSum,
updateModTime: !ci.NoUpdateModTime,
forceModTimeMatch: false,
}
}
var modTimeUploadOnce sync.Once
// emit a log if we are about to upload a file to set its modification time
func logModTimeUpload(dst fs.Object) {
modTimeUploadOnce.Do(func() {
fs.Logf(dst.Fs(), "Forced to upload files to set modification times on this backend.")
})
}
func equal(ctx context.Context, src fs.ObjectInfo, dst fs.Object, opt equalOpt) bool {
ci := fs.GetConfig(ctx)
if sizeDiffers(ctx, src, dst) {
fs.Debugf(src, "Sizes differ (src %d vs dst %d)", src.Size(), dst.Size())
return false
}
if opt.sizeOnly {
fs.Debugf(src, "Sizes identical")
return true
}
// Assert: Size is equal or being ignored
// If checking checksum and not modtime
if opt.checkSum {
// Check the hash
same, ht, _ := CheckHashes(ctx, src, dst)
if !same {
fs.Debugf(src, "%v differ", ht)
return false
}
if ht == hash.None {
common := src.Fs().Hashes().Overlap(dst.Fs().Hashes())
if common.Count() == 0 {
checksumWarning.Do(func() {
fs.Logf(dst.Fs(), "--checksum is in use but the source and destination have no hashes in common; falling back to --size-only")
})
}
fs.Debugf(src, "Size of src and dst objects identical")
} else {
fs.Debugf(src, "Size and %v of src and dst objects identical", ht)
}
return true
}
srcModTime := src.ModTime(ctx)
if !opt.forceModTimeMatch {
// Sizes the same so check the mtime
modifyWindow := fs.GetModifyWindow(ctx, src.Fs(), dst.Fs())
if modifyWindow == fs.ModTimeNotSupported {
fs.Debugf(src, "Sizes identical")
return true
}
dstModTime := dst.ModTime(ctx)
dt := dstModTime.Sub(srcModTime)
if dt < modifyWindow && dt > -modifyWindow {
fs.Debugf(src, "Size and modification time the same (differ by %s, within tolerance %s)", dt, modifyWindow)
return true
}
fs.Debugf(src, "Modification times differ by %s: %v, %v", dt, srcModTime, dstModTime)
}
// Check if the hashes are the same
same, ht, _ := CheckHashes(ctx, src, dst)
if !same {
fs.Debugf(src, "%v differ", ht)
return false
}
if ht == hash.None && !ci.RefreshTimes {
// if couldn't check hash, return that they differ
return false
}
// mod time differs but hash is the same to reset mod time if required
if opt.updateModTime {
if !SkipDestructive(ctx, src, "update modification time") {
2017-09-02 08:29:01 +00:00
// Size and hash the same but mtime different
// Error if objects are treated as immutable
if ci.Immutable {
fs.Errorf(dst, "Timestamp mismatch between immutable objects")
2017-09-02 08:29:01 +00:00
return false
}
// Update the mtime of the dst object here
err := dst.SetModTime(ctx, srcModTime)
if err == fs.ErrorCantSetModTime {
logModTimeUpload(dst)
fs.Infof(dst, "src and dst identical but can't set mod time without re-uploading")
return false
} else if err == fs.ErrorCantSetModTimeWithoutDelete {
logModTimeUpload(dst)
fs.Infof(dst, "src and dst identical but can't set mod time without deleting and re-uploading")
// Remove the file if BackupDir isn't set. If BackupDir is set we would rather have the old file
// put in the BackupDir than deleted which is what will happen if we don't delete it.
if ci.BackupDir == "" {
err = dst.Remove(ctx)
if err != nil {
fs.Errorf(dst, "failed to delete before re-upload: %v", err)
}
}
return false
} else if err != nil {
err = fs.CountError(err)
fs.Errorf(dst, "Failed to set modification time: %v", err)
} else {
fs.Infof(src, "Updated modification time in destination")
}
}
}
return true
}
// Used to remove a failed copy
//
2019-04-30 12:06:24 +00:00
// Returns whether the file was successfully removed or not
func removeFailedCopy(ctx context.Context, dst fs.Object) bool {
if dst == nil {
return false
}
fs.Infof(dst, "Removing failed copy")
removeErr := dst.Remove(ctx)
if removeErr != nil {
fs.Infof(dst, "Failed to remove failed copy: %s", removeErr)
return false
}
return true
}
// OverrideRemote is a wrapper to override the Remote for an
// ObjectInfo
type OverrideRemote struct {
fs.ObjectInfo
remote string
}
// NewOverrideRemote returns an OverrideRemoteObject which will
// return the remote specified
func NewOverrideRemote(oi fs.ObjectInfo, remote string) *OverrideRemote {
return &OverrideRemote{
ObjectInfo: oi,
remote: remote,
}
}
2019-04-30 12:06:24 +00:00
// Remote returns the overridden remote name
func (o *OverrideRemote) Remote() string {
return o.remote
}
// MimeType returns the mime type of the underlying object or "" if it
// can't be worked out
func (o *OverrideRemote) MimeType(ctx context.Context) string {
if do, ok := o.ObjectInfo.(fs.MimeTyper); ok {
return do.MimeType(ctx)
}
return ""
}
// ID returns the ID of the Object if known, or "" if not
func (o *OverrideRemote) ID() string {
if do, ok := o.ObjectInfo.(fs.IDer); ok {
return do.ID()
}
return ""
}
// UnWrap returns the Object that this Object is wrapping or nil if it
// isn't wrapping anything
func (o *OverrideRemote) UnWrap() fs.Object {
if o, ok := o.ObjectInfo.(fs.Object); ok {
return o
}
return nil
}
// GetTier returns storage tier or class of the Object
func (o *OverrideRemote) GetTier() string {
if do, ok := o.ObjectInfo.(fs.GetTierer); ok {
return do.GetTier()
}
return ""
}
// Check all optional interfaces satisfied
var _ fs.FullObjectInfo = (*OverrideRemote)(nil)
// CommonHash returns a single hash.Type and a HashOption with that
// type which is in common between the two fs.Fs.
func CommonHash(ctx context.Context, fa, fb fs.Info) (hash.Type, *fs.HashesOption) {
ci := fs.GetConfig(ctx)
// work out which hash to use - limit to 1 hash in common
var common hash.Set
hashType := hash.None
if !ci.IgnoreChecksum {
common = fb.Hashes().Overlap(fa.Hashes())
if common.Count() > 0 {
hashType = common.GetOne()
common = hash.Set(hashType)
}
}
return hashType, &fs.HashesOption{Hashes: common}
}
// Copy src object to dst or f if nil. If dst is nil then it uses
// remote as the name of the new object.
//
// It returns the destination object if possible. Note that this may
// be nil.
func Copy(ctx context.Context, f fs.Fs, dst fs.Object, remote string, src fs.Object) (newDst fs.Object, err error) {
ci := fs.GetConfig(ctx)
tr := accounting.Stats(ctx).NewTransfer(src)
defer func() {
tr.Done(ctx, err)
}()
newDst = dst
if SkipDestructive(ctx, src, "copy") {
in := tr.Account(ctx, nil)
in.DryRun(src.Size())
return newDst, nil
}
maxTries := ci.LowLevelRetries
tries := 0
doUpdate := dst != nil
hashType, hashOption := CommonHash(ctx, f, src.Fs())
var actionTaken string
for {
// Try server-side copy first - if has optional interface and
// is same underlying remote
actionTaken = "Copied (server-side copy)"
if ci.MaxTransfer >= 0 {
var bytesSoFar int64
if ci.CutoffMode == fs.CutoffModeCautious {
bytesSoFar = accounting.Stats(ctx).GetBytesWithPending() + src.Size()
} else {
bytesSoFar = accounting.Stats(ctx).GetBytes()
}
if bytesSoFar >= int64(ci.MaxTransfer) {
if ci.CutoffMode == fs.CutoffModeHard {
return nil, accounting.ErrorMaxTransferLimitReachedFatal
}
return nil, accounting.ErrorMaxTransferLimitReachedGraceful
}
}
if doCopy := f.Features().Copy; doCopy != nil && (SameConfig(src.Fs(), f) || (SameRemoteType(src.Fs(), f) && f.Features().ServerSideAcrossConfigs)) {
in := tr.Account(ctx, nil) // account the transfer
in.ServerSideCopyStart()
newDst, err = doCopy(ctx, src, remote)
if err == nil {
dst = newDst
in.ServerSideCopyEnd(dst.Size()) // account the bytes for the server-side transfer
_ = in.Close()
} else {
_ = in.Close()
}
if err == fs.ErrorCantCopy {
tr.Reset(ctx) // skip incomplete accounting - will be overwritten by the manual copy below
}
} else {
err = fs.ErrorCantCopy
}
// If can't server-side copy, do it manually
if err == fs.ErrorCantCopy {
if doMultiThreadCopy(ctx, f, src) {
// Number of streams proportional to size
streams := src.Size() / int64(ci.MultiThreadCutoff)
// With maximum
if streams > int64(ci.MultiThreadStreams) {
streams = int64(ci.MultiThreadStreams)
}
if streams < 2 {
streams = 2
}
dst, err = multiThreadCopy(ctx, f, remote, src, int(streams), tr)
if doUpdate {
actionTaken = "Multi-thread Copied (replaced existing)"
} else {
actionTaken = "Multi-thread Copied (new)"
}
} else {
var in0 io.ReadCloser
options := []fs.OpenOption{hashOption}
for _, option := range ci.DownloadHeaders {
options = append(options, option)
}
in0, err = NewReOpen(ctx, src, ci.LowLevelRetries, options...)
if err != nil {
err = fmt.Errorf("failed to open source object: %w", err)
} else {
if src.Size() == -1 {
// -1 indicates unknown size. Use Rcat to handle both remotes supporting and not supporting PutStream.
if doUpdate {
actionTaken = "Copied (Rcat, replaced existing)"
} else {
actionTaken = "Copied (Rcat, new)"
}
// NB Rcat closes in0
dst, err = Rcat(ctx, f, remote, in0, src.ModTime(ctx))
newDst = dst
} else {
in := tr.Account(ctx, in0).WithBuffer() // account and buffer the transfer
var wrappedSrc fs.ObjectInfo = src
// We try to pass the original object if possible
if src.Remote() != remote {
wrappedSrc = NewOverrideRemote(src, remote)
}
options := []fs.OpenOption{hashOption}
for _, option := range ci.UploadHeaders {
options = append(options, option)
}
if doUpdate {
actionTaken = "Copied (replaced existing)"
err = dst.Update(ctx, in, wrappedSrc, options...)
} else {
actionTaken = "Copied (new)"
dst, err = f.Put(ctx, in, wrappedSrc, options...)
}
closeErr := in.Close()
if err == nil {
newDst = dst
err = closeErr
}
}
}
}
}
tries++
if tries >= maxTries {
break
}
// Retry if err returned a retry error
if fserrors.ContextError(ctx, &err) {
break
}
var retry bool
if fserrors.IsRetryError(err) || fserrors.ShouldRetry(err) {
retry = true
} else if t, ok := pacer.IsRetryAfter(err); ok {
fs.Debugf(src, "Sleeping for %v (as indicated by the server) to obey Retry-After error: %v", t, err)
time.Sleep(t)
retry = true
}
if retry {
fs.Debugf(src, "Received error: %v - low level retry %d/%d", err, tries, maxTries)
tr.Reset(ctx) // skip incomplete accounting - will be overwritten by retry
continue
}
// otherwise finish
break
}
if err != nil {
err = fs.CountError(err)
fs.Errorf(src, "Failed to copy: %v", err)
return newDst, err
}
// Verify sizes are the same after transfer
if sizeDiffers(ctx, src, dst) {
err = fmt.Errorf("corrupted on transfer: sizes differ %d vs %d", src.Size(), dst.Size())
fs.Errorf(dst, "%v", err)
err = fs.CountError(err)
removeFailedCopy(ctx, dst)
return newDst, err
}
// Verify hashes are the same after transfer - ignoring blank hashes
if hashType != hash.None {
// checkHashes has logged and counted errors
equal, _, srcSum, dstSum, _ := checkHashes(ctx, src, dst, hashType)
if !equal {
err = fmt.Errorf("corrupted on transfer: %v hash differ %q vs %q", hashType, srcSum, dstSum)
fs.Errorf(dst, "%v", err)
err = fs.CountError(err)
removeFailedCopy(ctx, dst)
return newDst, err
}
}
if newDst != nil && src.String() != newDst.String() {
fs.Infof(src, "%s to: %s", actionTaken, newDst.String())
} else {
fs.Infof(src, actionTaken)
}
return newDst, err
}
// SameObject returns true if src and dst could be pointing to the
// same object.
func SameObject(src, dst fs.Object) bool {
srcFs, dstFs := src.Fs(), dst.Fs()
if !SameConfig(srcFs, dstFs) {
// If same remote type then check ID of objects if available
doSrcID, srcIDOK := src.(fs.IDer)
doDstID, dstIDOK := dst.(fs.IDer)
if srcIDOK && dstIDOK && SameRemoteType(srcFs, dstFs) {
srcID, dstID := doSrcID.ID(), doDstID.ID()
if srcID != "" && srcID == dstID {
return true
}
}
return false
}
srcPath := path.Join(srcFs.Root(), src.Remote())
dstPath := path.Join(dstFs.Root(), dst.Remote())
if dst.Fs().Features().CaseInsensitive {
srcPath = strings.ToLower(srcPath)
dstPath = strings.ToLower(dstPath)
}
return srcPath == dstPath
}
// Move src object to dst or fdst if nil. If dst is nil then it uses
// remote as the name of the new object.
//
// Note that you must check the destination does not exist before
// calling this and pass it as dst. If you pass dst=nil and the
// destination does exist then this may create duplicates or return
// errors.
//
// It returns the destination object if possible. Note that this may
// be nil.
func Move(ctx context.Context, fdst fs.Fs, dst fs.Object, remote string, src fs.Object) (newDst fs.Object, err error) {
tr := accounting.Stats(ctx).NewCheckingTransfer(src)
defer func() {
if err == nil {
accounting.Stats(ctx).Renames(1)
}
tr.Done(ctx, err)
}()
newDst = dst
if SkipDestructive(ctx, src, "move") {
in := tr.Account(ctx, nil)
in.DryRun(src.Size())
return newDst, nil
}
// See if we have Move available
if doMove := fdst.Features().Move; doMove != nil && (SameConfig(src.Fs(), fdst) || (SameRemoteType(src.Fs(), fdst) && fdst.Features().ServerSideAcrossConfigs)) {
// Delete destination if it exists and is not the same file as src (could be same file while seemingly different if the remote is case insensitive)
if dst != nil && !SameObject(src, dst) {
err = DeleteFile(ctx, dst)
if err != nil {
return newDst, err
}
}
// Move dst <- src
in := tr.Account(ctx, nil) // account the transfer
in.ServerSideCopyStart()
newDst, err = doMove(ctx, src, remote)
switch err {
case nil:
if newDst != nil && src.String() != newDst.String() {
fs.Infof(src, "Moved (server-side) to: %s", newDst.String())
} else {
fs.Infof(src, "Moved (server-side)")
}
in.ServerSideCopyEnd(newDst.Size()) // account the bytes for the server-side transfer
_ = in.Close()
return newDst, nil
case fs.ErrorCantMove:
fs.Debugf(src, "Can't move, switching to copy")
_ = in.Close()
default:
err = fs.CountError(err)
fs.Errorf(src, "Couldn't move: %v", err)
_ = in.Close()
return newDst, err
}
}
// Move not found or didn't work so copy dst <- src
newDst, err = Copy(ctx, fdst, dst, remote, src)
if err != nil {
fs.Errorf(src, "Not deleting source as copy failed: %v", err)
return newDst, err
}
// Delete src if no error on copy
return newDst, DeleteFile(ctx, src)
}
// CanServerSideMove returns true if fdst support server-side moves or
// server-side copies
2017-01-10 20:03:55 +00:00
//
// Some remotes simulate rename by server-side copy and delete, so include
// remotes that implements either Mover or Copier.
func CanServerSideMove(fdst fs.Fs) bool {
canMove := fdst.Features().Move != nil
canCopy := fdst.Features().Copy != nil
2017-01-10 20:03:55 +00:00
return canMove || canCopy
}
// SuffixName adds the current --suffix to the remote, obeying
// --suffix-keep-extension if set
func SuffixName(ctx context.Context, remote string) string {
ci := fs.GetConfig(ctx)
if ci.Suffix == "" {
return remote
}
if ci.SuffixKeepExtension {
ext := path.Ext(remote)
base := remote[:len(remote)-len(ext)]
return base + ci.Suffix + ext
}
return remote + ci.Suffix
}
// DeleteFileWithBackupDir deletes a single file respecting --dry-run
// and accumulating stats and errors.
//
// If backupDir is set then it moves the file to there instead of
// deleting
func DeleteFileWithBackupDir(ctx context.Context, dst fs.Object, backupDir fs.Fs) (err error) {
ci := fs.GetConfig(ctx)
tr := accounting.Stats(ctx).NewCheckingTransfer(dst)
defer func() {
tr.Done(ctx, err)
}()
numDeletes := accounting.Stats(ctx).Deletes(1)
if ci.MaxDelete != -1 && numDeletes > ci.MaxDelete {
return fserrors.FatalError(errors.New("--max-delete threshold reached"))
}
action, actioned := "delete", "Deleted"
if backupDir != nil {
action, actioned = "move into backup dir", "Moved into backup dir"
}
skip := SkipDestructive(ctx, dst, action)
if skip {
// do nothing
} else if backupDir != nil {
err = MoveBackupDir(ctx, backupDir, dst)
} else {
err = dst.Remove(ctx)
}
if err != nil {
fs.Errorf(dst, "Couldn't %s: %v", action, err)
err = fs.CountError(err)
} else if !skip {
fs.Infof(dst, actioned)
}
return err
}
// DeleteFile deletes a single file respecting --dry-run and accumulating stats and errors.
//
// If useBackupDir is set and --backup-dir is in effect then it moves
// the file to there instead of deleting
func DeleteFile(ctx context.Context, dst fs.Object) (err error) {
return DeleteFileWithBackupDir(ctx, dst, nil)
}
// DeleteFilesWithBackupDir removes all the files passed in the
// channel
//
// If backupDir is set the files will be placed into that directory
// instead of being deleted.
func DeleteFilesWithBackupDir(ctx context.Context, toBeDeleted fs.ObjectsChan, backupDir fs.Fs) error {
var wg sync.WaitGroup
ci := fs.GetConfig(ctx)
wg.Add(ci.Transfers)
var errorCount int32
var fatalErrorCount int32
for i := 0; i < ci.Transfers; i++ {
go func() {
defer wg.Done()
2015-09-22 17:47:16 +00:00
for dst := range toBeDeleted {
err := DeleteFileWithBackupDir(ctx, dst, backupDir)
if err != nil {
atomic.AddInt32(&errorCount, 1)
if fserrors.IsFatalError(err) {
fs.Errorf(nil, "Got fatal error on delete: %s", err)
atomic.AddInt32(&fatalErrorCount, 1)
return
}
}
}
}()
}
fs.Debugf(nil, "Waiting for deletions to finish")
wg.Wait()
if errorCount > 0 {
err := fmt.Errorf("failed to delete %d files", errorCount)
if fatalErrorCount > 0 {
return fserrors.FatalError(err)
}
return err
}
return nil
}
// DeleteFiles removes all the files passed in the channel
func DeleteFiles(ctx context.Context, toBeDeleted fs.ObjectsChan) error {
return DeleteFilesWithBackupDir(ctx, toBeDeleted, nil)
}
// SameRemoteType returns true if fdst and fsrc are the same type
func SameRemoteType(fdst, fsrc fs.Info) bool {
return fmt.Sprintf("%T", fdst) == fmt.Sprintf("%T", fsrc)
}
// SameConfig returns true if fdst and fsrc are using the same config
// file entry
func SameConfig(fdst, fsrc fs.Info) bool {
return fdst.Name() == fsrc.Name()
}
// SameConfigArr returns true if any of []fsrcs has same config file entry with fdst
func SameConfigArr(fdst fs.Info, fsrcs []fs.Fs) bool {
for _, fsrc := range fsrcs {
if fdst.Name() == fsrc.Name() {
return true
}
}
return false
}
2015-09-22 17:47:16 +00:00
// Same returns true if fdst and fsrc point to the same underlying Fs
func Same(fdst, fsrc fs.Info) bool {
return SameConfig(fdst, fsrc) && strings.Trim(fdst.Root(), "/") == strings.Trim(fsrc.Root(), "/")
}
// fixRoot returns the Root with a trailing / if not empty. It is
// aware of case insensitive filesystems.
func fixRoot(f fs.Info) string {
s := strings.Trim(filepath.ToSlash(f.Root()), "/")
if s != "" {
s += "/"
}
if f.Features().CaseInsensitive {
s = strings.ToLower(s)
}
return s
}
// Overlapping returns true if fdst and fsrc point to the same
// underlying Fs and they overlap.
func Overlapping(fdst, fsrc fs.Info) bool {
if !SameConfig(fdst, fsrc) {
return false
}
fdstRoot := fixRoot(fdst)
fsrcRoot := fixRoot(fsrc)
return strings.HasPrefix(fdstRoot, fsrcRoot) || strings.HasPrefix(fsrcRoot, fdstRoot)
}
// SameDir returns true if fdst and fsrc point to the same
// underlying Fs and they are the same directory.
func SameDir(fdst, fsrc fs.Info) bool {
if !SameConfig(fdst, fsrc) {
return false
}
fdstRoot := fixRoot(fdst)
fsrcRoot := fixRoot(fsrc)
return fdstRoot == fsrcRoot
}
// Retry runs fn up to maxTries times if it returns a retriable error
func Retry(ctx context.Context, o interface{}, maxTries int, fn func() error) (err error) {
for tries := 1; tries <= maxTries; tries++ {
// Call the function which might error
err = fn()
if err == nil {
break
}
// Retry if err returned a retry error
if fserrors.ContextError(ctx, &err) {
break
}
if fserrors.IsRetryError(err) || fserrors.ShouldRetry(err) {
fs.Debugf(o, "Received error: %v - low level retry %d/%d", err, tries, maxTries)
continue
}
break
}
return err
}
2015-09-22 17:47:16 +00:00
// ListFn lists the Fs to the supplied function
//
// Lists in parallel which may get them out of order
func ListFn(ctx context.Context, f fs.Fs, fn func(fs.Object)) error {
ci := fs.GetConfig(ctx)
return walk.ListR(ctx, f, "", false, ci.MaxDepth, walk.ListObjects, func(entries fs.DirEntries) error {
entries.ForObject(fn)
return nil
})
}
// mutex for synchronized output
var outMutex sync.Mutex
// SyncPrintf is a global var holding the Printf function used in syncFprintf so that it can be overridden
// Note, despite name, does not provide sync and should not be called directly
// Call syncFprintf, which provides sync
var SyncPrintf = func(format string, a ...interface{}) {
fmt.Printf(format, a...)
}
// Synchronized fmt.Fprintf
2015-09-22 06:31:12 +00:00
//
// Ignores errors from Fprintf
//
// Updated to print to terminal if no writer is defined
// This special behavior is used to allow easier replacement of the print to terminal code by progress
2015-09-22 06:31:12 +00:00
func syncFprintf(w io.Writer, format string, a ...interface{}) {
outMutex.Lock()
defer outMutex.Unlock()
if w == nil || w == os.Stdout {
SyncPrintf(format, a...)
} else {
_, _ = fmt.Fprintf(w, format, a...)
}
}
// SizeString make string representation of size for output
//
// Optional human-readable format including a binary suffix
func SizeString(size int64, humanReadable bool) string {
if humanReadable {
if size < 0 {
return "-" + fs.SizeSuffix(-size).String()
}
return fs.SizeSuffix(size).String()
}
return strconv.FormatInt(size, 10)
}
// SizeStringField make string representation of size for output in fixed width field
//
// Optional human-readable format including a binary suffix
// Argument rawWidth is used to format field with of raw value. When humanReadable
// option the width is hard coded to 9, since SizeSuffix strings have precision 3
// and longest value will be "999.999Ei". This way the width can be optimized
// depending to the humanReadable option. To always use a longer width the return
// value can always be fed into another format string with a specific field with.
func SizeStringField(size int64, humanReadable bool, rawWidth int) string {
str := SizeString(size, humanReadable)
if humanReadable {
return fmt.Sprintf("%9s", str)
}
return fmt.Sprintf("%[2]*[1]s", str, rawWidth)
}
// CountString make string representation of count for output
//
// Optional human-readable format including a decimal suffix
func CountString(count int64, humanReadable bool) string {
if humanReadable {
if count < 0 {
return "-" + fs.CountSuffix(-count).String()
}
return fs.CountSuffix(count).String()
}
return strconv.FormatInt(count, 10)
}
// CountStringField make string representation of count for output in fixed width field
//
// Similar to SizeStringField, but human readable with decimal prefix and field width 8
// since there is no 'i' in the decimal prefix symbols (e.g. "999.999E")
func CountStringField(count int64, humanReadable bool, rawWidth int) string {
str := CountString(count, humanReadable)
if humanReadable {
return fmt.Sprintf("%8s", str)
}
return fmt.Sprintf("%[2]*[1]s", str, rawWidth)
}
// List the Fs to the supplied writer
//
// Shows size and path - obeys includes and excludes
//
// Lists in parallel which may get them out of order
func List(ctx context.Context, f fs.Fs, w io.Writer) error {
ci := fs.GetConfig(ctx)
return ListFn(ctx, f, func(o fs.Object) {
syncFprintf(w, "%s %s\n", SizeStringField(o.Size(), ci.HumanReadable, 9), o.Remote())
})
}
2015-09-22 17:47:16 +00:00
// ListLong lists the Fs to the supplied writer
//
// Shows size, mod time and path - obeys includes and excludes
//
// Lists in parallel which may get them out of order
func ListLong(ctx context.Context, f fs.Fs, w io.Writer) error {
ci := fs.GetConfig(ctx)
return ListFn(ctx, f, func(o fs.Object) {
tr := accounting.Stats(ctx).NewCheckingTransfer(o)
defer func() {
tr.Done(ctx, nil)
}()
modTime := o.ModTime(ctx)
syncFprintf(w, "%s %s %s\n", SizeStringField(o.Size(), ci.HumanReadable, 9), modTime.Local().Format("2006-01-02 15:04:05.000000000"), o.Remote())
})
}
// hashSum returns the human-readable hash for ht passed in. This may
// be UNSUPPORTED or ERROR. If it isn't returning a valid hash it will
// return an error.
func hashSum(ctx context.Context, ht hash.Type, base64Encoded bool, downloadFlag bool, o fs.Object) (string, error) {
var sum string
var err error
// If downloadFlag is true, download and hash the file.
// If downloadFlag is false, call o.Hash asking the remote for the hash
if downloadFlag {
// Setup: Define accounting, open the file with NewReOpen to provide restarts, account for the transfer, and setup a multi-hasher with the appropriate type
// Execution: io.Copy file to hasher, get hash and encode in hex
tr := accounting.Stats(ctx).NewTransfer(o)
defer func() {
tr.Done(ctx, err)
}()
// Open with NewReOpen to provide restarts
var options []fs.OpenOption
for _, option := range fs.GetConfig(ctx).DownloadHeaders {
options = append(options, option)
}
in, err := NewReOpen(ctx, o, fs.GetConfig(ctx).LowLevelRetries, options...)
if err != nil {
return "ERROR", fmt.Errorf("failed to open file %v: %w", o, err)
}
// Account and buffer the transfer
in = tr.Account(ctx, in).WithBuffer()
// Setup hasher
hasher, err := hash.NewMultiHasherTypes(hash.NewHashSet(ht))
if err != nil {
return "UNSUPPORTED", fmt.Errorf("hash unsupported: %w", err)
}
// Copy to hasher, downloading the file and passing directly to hash
_, err = io.Copy(hasher, in)
if err != nil {
return "ERROR", fmt.Errorf("failed to copy file to hasher: %w", err)
}
// Get hash as hex or base64 encoded string
sum, err = hasher.SumString(ht, base64Encoded)
if err != nil {
return "ERROR", fmt.Errorf("hasher returned an error: %w", err)
}
} else {
tr := accounting.Stats(ctx).NewCheckingTransfer(o)
defer func() {
tr.Done(ctx, err)
}()
sum, err = o.Hash(ctx, ht)
if base64Encoded {
hexBytes, _ := hex.DecodeString(sum)
sum = base64.URLEncoding.EncodeToString(hexBytes)
}
if err == hash.ErrUnsupported {
return "", fmt.Errorf("hash unsupported: %w", err)
}
if err != nil {
return "", fmt.Errorf("failed to get hash %v from backend: %w", ht, err)
}
2018-01-06 17:53:37 +00:00
}
return sum, nil
}
// HashLister does an md5sum equivalent for the hash type passed in
// Updated to handle both standard hex encoding and base64
// Updated to perform multiple hashes concurrently
func HashLister(ctx context.Context, ht hash.Type, outputBase64 bool, downloadFlag bool, f fs.Fs, w io.Writer) error {
width := hash.Width(ht, outputBase64)
concurrencyControl := make(chan struct{}, fs.GetConfig(ctx).Transfers)
var wg sync.WaitGroup
err := ListFn(ctx, f, func(o fs.Object) {
wg.Add(1)
concurrencyControl <- struct{}{}
go func() {
defer func() {
<-concurrencyControl
wg.Done()
}()
sum, err := hashSum(ctx, ht, outputBase64, downloadFlag, o)
if err != nil {
fs.Errorf(o, "%v", fs.CountError(err))
return
}
syncFprintf(w, "%*s %s\n", width, sum, o.Remote())
}()
})
wg.Wait()
return err
}
// HashSumStream outputs a line compatible with md5sum to w based on the
// input stream in and the hash type ht passed in. If outputBase64 is
// set then the hash will be base64 instead of hexadecimal.
func HashSumStream(ht hash.Type, outputBase64 bool, in io.ReadCloser, w io.Writer) error {
hasher, err := hash.NewMultiHasherTypes(hash.NewHashSet(ht))
if err != nil {
return fmt.Errorf("hash unsupported: %w", err)
}
written, err := io.Copy(hasher, in)
fs.Debugf(nil, "Creating %s hash of %d bytes read from input stream", ht, written)
if err != nil {
return fmt.Errorf("failed to copy input to hasher: %w", err)
}
sum, err := hasher.SumString(ht, outputBase64)
if err != nil {
return fmt.Errorf("hasher returned an error: %w", err)
}
width := hash.Width(ht, outputBase64)
syncFprintf(w, "%*s -\n", width, sum)
return nil
}
// Count counts the objects and their sizes in the Fs
//
// Obeys includes and excludes
func Count(ctx context.Context, f fs.Fs) (objects int64, size int64, err error) {
err = ListFn(ctx, f, func(o fs.Object) {
atomic.AddInt64(&objects, 1)
objectSize := o.Size()
if objectSize > 0 {
atomic.AddInt64(&size, objectSize)
}
})
return
}
// ConfigMaxDepth returns the depth to use for a recursive or non recursive listing.
func ConfigMaxDepth(ctx context.Context, recursive bool) int {
ci := fs.GetConfig(ctx)
depth := ci.MaxDepth
if !recursive && depth < 0 {
depth = 1
}
return depth
}
2015-09-22 17:47:16 +00:00
// ListDir lists the directories/buckets/containers in the Fs to the supplied writer
func ListDir(ctx context.Context, f fs.Fs, w io.Writer) error {
ci := fs.GetConfig(ctx)
return walk.ListR(ctx, f, "", false, ConfigMaxDepth(ctx, false), walk.ListDirs, func(entries fs.DirEntries) error {
entries.ForDir(func(dir fs.Directory) {
if dir != nil {
syncFprintf(w, "%s %13s %s %s\n", SizeStringField(dir.Size(), ci.HumanReadable, 12), dir.ModTime(ctx).Local().Format("2006-01-02 15:04:05"), CountStringField(dir.Items(), ci.HumanReadable, 9), dir.Remote())
}
})
return nil
})
}
2015-09-22 17:47:16 +00:00
// Mkdir makes a destination directory or container
func Mkdir(ctx context.Context, f fs.Fs, dir string) error {
if SkipDestructive(ctx, fs.LogDirName(f, dir), "make directory") {
return nil
}
fs.Debugf(fs.LogDirName(f, dir), "Making directory")
err := f.Mkdir(ctx, dir)
if err != nil {
err = fs.CountError(err)
return err
}
return nil
}
// TryRmdir removes a container but not if not empty. It doesn't
// count errors but may return one.
func TryRmdir(ctx context.Context, f fs.Fs, dir string) error {
accounting.Stats(ctx).DeletedDirs(1)
if SkipDestructive(ctx, fs.LogDirName(f, dir), "remove directory") {
return nil
}
fs.Infof(fs.LogDirName(f, dir), "Removing directory")
return f.Rmdir(ctx, dir)
}
// Rmdir removes a container but not if not empty
func Rmdir(ctx context.Context, f fs.Fs, dir string) error {
err := TryRmdir(ctx, f, dir)
if err != nil {
err = fs.CountError(err)
return err
}
return err
}
// Purge removes a directory and all of its contents
func Purge(ctx context.Context, f fs.Fs, dir string) (err error) {
doFallbackPurge := true
if doPurge := f.Features().Purge; doPurge != nil {
doFallbackPurge = false
accounting.Stats(ctx).DeletedDirs(1)
if SkipDestructive(ctx, fs.LogDirName(f, dir), "purge directory") {
return nil
}
err = doPurge(ctx, dir)
if err == fs.ErrorCantPurge {
doFallbackPurge = true
}
}
if doFallbackPurge {
2014-07-25 17:19:49 +00:00
// DeleteFiles and Rmdir observe --dry-run
err = DeleteFiles(ctx, listToChan(ctx, f, dir))
if err != nil {
return err
}
err = Rmdirs(ctx, f, dir, false)
2014-07-25 17:19:49 +00:00
}
if err != nil {
err = fs.CountError(err)
2014-07-25 17:19:49 +00:00
return err
}
return nil
}
// Delete removes all the contents of a container. Unlike Purge, it
// obeys includes and excludes.
func Delete(ctx context.Context, f fs.Fs) error {
ci := fs.GetConfig(ctx)
delChan := make(fs.ObjectsChan, ci.Transfers)
delErr := make(chan error, 1)
go func() {
delErr <- DeleteFiles(ctx, delChan)
}()
err := ListFn(ctx, f, func(o fs.Object) {
2018-08-04 10:16:43 +00:00
delChan <- o
})
2018-08-04 10:16:43 +00:00
close(delChan)
delError := <-delErr
if err == nil {
err = delError
}
return err
}
// listToChan will transfer all objects in the listing to the output
//
// If an error occurs, the error will be logged, and it will close the
// channel.
//
// If the error was ErrorDirNotFound then it will be ignored
func listToChan(ctx context.Context, f fs.Fs, dir string) fs.ObjectsChan {
ci := fs.GetConfig(ctx)
o := make(fs.ObjectsChan, ci.Checkers)
go func() {
defer close(o)
err := walk.ListR(ctx, f, dir, true, ci.MaxDepth, walk.ListObjects, func(entries fs.DirEntries) error {
entries.ForObject(func(obj fs.Object) {
o <- obj
})
return nil
})
if err != nil && err != fs.ErrorDirNotFound {
err = fmt.Errorf("failed to list: %w", err)
err = fs.CountError(err)
fs.Errorf(nil, "%v", err)
}
}()
return o
}
// CleanUp removes the trash for the Fs
func CleanUp(ctx context.Context, f fs.Fs) error {
doCleanUp := f.Features().CleanUp
if doCleanUp == nil {
return fmt.Errorf("%v doesn't support cleanup", f)
}
if SkipDestructive(ctx, f, "clean up old files") {
return nil
}
return doCleanUp(ctx)
}
2016-08-18 21:43:02 +00:00
2017-02-09 11:25:36 +00:00
// wrap a Reader and a Closer together into a ReadCloser
type readCloser struct {
io.Reader
io.Closer
2017-02-09 11:25:36 +00:00
}
2016-08-18 21:43:02 +00:00
// Cat any files to the io.Writer
//
// if offset == 0 it will be ignored
// if offset > 0 then the file will be seeked to that offset
// if offset < 0 then the file will be seeked that far from the end
//
// if count < 0 then it will be ignored
// if count >= 0 then only that many characters will be output
func Cat(ctx context.Context, f fs.Fs, w io.Writer, offset, count int64) error {
2016-08-18 21:43:02 +00:00
var mu sync.Mutex
ci := fs.GetConfig(ctx)
return ListFn(ctx, f, func(o fs.Object) {
var err error
tr := accounting.Stats(ctx).NewTransfer(o)
defer func() {
tr.Done(ctx, err)
}()
opt := fs.RangeOption{Start: offset, End: -1}
size := o.Size()
if opt.Start < 0 {
opt.Start += size
}
if count >= 0 {
opt.End = opt.Start + count - 1
}
var options []fs.OpenOption
if opt.Start > 0 || opt.End >= 0 {
options = append(options, &opt)
}
for _, option := range ci.DownloadHeaders {
options = append(options, option)
}
in, err := o.Open(ctx, options...)
2016-08-18 21:43:02 +00:00
if err != nil {
err = fs.CountError(err)
fs.Errorf(o, "Failed to open: %v", err)
2016-08-18 21:43:02 +00:00
return
}
if count >= 0 {
2017-02-09 11:25:36 +00:00
in = &readCloser{Reader: &io.LimitedReader{R: in, N: count}, Closer: in}
}
in = tr.Account(ctx, in).WithBuffer() // account and buffer the transfer
// take the lock just before we output stuff, so at the last possible moment
mu.Lock()
defer mu.Unlock()
2017-02-09 11:25:36 +00:00
_, err = io.Copy(w, in)
2016-08-18 21:43:02 +00:00
if err != nil {
err = fs.CountError(err)
fs.Errorf(o, "Failed to send to output: %v", err)
2016-08-18 21:43:02 +00:00
}
})
}
// Rcat reads data from the Reader until EOF and uploads it to a file on remote
func Rcat(ctx context.Context, fdst fs.Fs, dstFileName string, in io.ReadCloser, modTime time.Time) (dst fs.Object, err error) {
ci := fs.GetConfig(ctx)
tr := accounting.Stats(ctx).NewTransferRemoteSize(dstFileName, -1)
defer func() {
tr.Done(ctx, err)
}()
in = tr.Account(ctx, in).WithBuffer()
readCounter := readers.NewCountingReader(in)
var trackingIn io.Reader
var hasher *hash.MultiHasher
var options []fs.OpenOption
if !ci.IgnoreChecksum {
hashes := hash.NewHashSet(fdst.Hashes().GetOne()) // just pick one hash
hashOption := &fs.HashesOption{Hashes: hashes}
options = append(options, hashOption)
hasher, err = hash.NewMultiHasherTypes(hashes)
if err != nil {
return nil, err
}
trackingIn = io.TeeReader(readCounter, hasher)
} else {
trackingIn = readCounter
}
for _, option := range ci.UploadHeaders {
options = append(options, option)
}
compare := func(dst fs.Object) error {
var sums map[hash.Type]string
if hasher != nil {
sums = hasher.Sums()
}
src := object.NewStaticObjectInfo(dstFileName, modTime, int64(readCounter.BytesRead()), false, sums, fdst)
if !Equal(ctx, src, dst) {
err = fmt.Errorf("corrupted on transfer")
err = fs.CountError(err)
fs.Errorf(dst, "%v", err)
return err
}
return nil
}
// check if file small enough for direct upload
buf := make([]byte, ci.StreamingUploadCutoff)
if n, err := io.ReadFull(trackingIn, buf); err == io.EOF || err == io.ErrUnexpectedEOF {
fs.Debugf(fdst, "File to upload is small (%d bytes), uploading instead of streaming", n)
src := object.NewMemoryObject(dstFileName, modTime, buf[:n])
return Copy(ctx, fdst, nil, dstFileName, src)
}
// Make a new ReadCloser with the bits we've already read
in = &readCloser{
Reader: io.MultiReader(bytes.NewReader(buf), trackingIn),
Closer: in,
}
fStreamTo := fdst
canStream := fdst.Features().PutStream != nil
if !canStream {
fs.Debugf(fdst, "Target remote doesn't support streaming uploads, creating temporary local FS to spool file")
tmpLocalFs, err := fs.TemporaryLocalFs(ctx)
if err != nil {
return nil, fmt.Errorf("Failed to create temporary local FS to spool file: %w", err)
}
defer func() {
err := Purge(ctx, tmpLocalFs, "")
if err != nil {
fs.Infof(tmpLocalFs, "Failed to cleanup temporary FS: %v", err)
}
}()
fStreamTo = tmpLocalFs
}
if SkipDestructive(ctx, dstFileName, "upload from pipe") {
// prevents "broken pipe" errors
_, err = io.Copy(ioutil.Discard, in)
return nil, err
}
objInfo := object.NewStaticObjectInfo(dstFileName, modTime, -1, false, nil, nil)
if dst, err = fStreamTo.Features().PutStream(ctx, in, objInfo, options...); err != nil {
return dst, err
}
if err = compare(dst); err != nil {
return dst, err
}
if !canStream {
// copy dst (which is the local object we have just streamed to) to the remote
return Copy(ctx, fdst, nil, dstFileName, dst)
}
return dst, nil
}
// PublicLink adds a "readable by anyone with link" permission on the given file or folder.
func PublicLink(ctx context.Context, f fs.Fs, remote string, expire fs.Duration, unlink bool) (string, error) {
doPublicLink := f.Features().PublicLink
if doPublicLink == nil {
return "", fmt.Errorf("%v doesn't support public links", f)
}
return doPublicLink(ctx, remote, expire, unlink)
}
// Rmdirs removes any empty directories (or directories only
// containing empty directories) under f, including f.
//
// Rmdirs obeys the filters
func Rmdirs(ctx context.Context, f fs.Fs, dir string, leaveRoot bool) error {
ci := fs.GetConfig(ctx)
fi := filter.GetConfig(ctx)
dirEmpty := make(map[string]bool)
dirEmpty[dir] = !leaveRoot
err := walk.Walk(ctx, f, dir, false, ci.MaxDepth, func(dirPath string, entries fs.DirEntries, err error) error {
if err != nil {
err = fs.CountError(err)
fs.Errorf(f, "Failed to list %q: %v", dirPath, err)
return nil
}
for _, entry := range entries {
switch x := entry.(type) {
case fs.Directory:
// add a new directory as empty
dir := x.Remote()
_, found := dirEmpty[dir]
if !found {
dirEmpty[dir] = true
}
case fs.Object:
// mark the parents of the file as being non-empty
dir := x.Remote()
for dir != "" {
dir = path.Dir(dir)
if dir == "." || dir == "/" {
dir = ""
}
empty, found := dirEmpty[dir]
// End if we reach a directory which is non-empty
if found && !empty {
break
}
dirEmpty[dir] = false
}
}
}
return nil
})
if err != nil {
return fmt.Errorf("failed to rmdirs: %w", err)
}
// Now delete the empty directories, starting from the longest path
var toDelete []string
for dir, empty := range dirEmpty {
if empty {
toDelete = append(toDelete, dir)
}
}
sort.Strings(toDelete)
for i := len(toDelete) - 1; i >= 0; i-- {
dir := toDelete[i]
// If a filter matches the directory then that
// directory is a candidate for deletion
if !fi.Include(dir+"/", 0, time.Now()) {
continue
}
err = TryRmdir(ctx, f, dir)
if err != nil {
err = fs.CountError(err)
fs.Errorf(dir, "Failed to rmdir: %v", err)
return err
}
}
return nil
}
// GetCompareDest sets up --compare-dest
func GetCompareDest(ctx context.Context) (CompareDest []fs.Fs, err error) {
ci := fs.GetConfig(ctx)
CompareDest, err = cache.GetArr(ctx, ci.CompareDest)
if err != nil {
return nil, fserrors.FatalError(fmt.Errorf("Failed to make fs for --compare-dest %q: %v", ci.CompareDest, err))
}
return CompareDest, nil
}
// compareDest checks --compare-dest to see if src needs to
// be copied
//
// Returns True if src is in --compare-dest
func compareDest(ctx context.Context, dst, src fs.Object, CompareDest fs.Fs) (NoNeedTransfer bool, err error) {
var remote string
if dst == nil {
remote = src.Remote()
} else {
remote = dst.Remote()
}
CompareDestFile, err := CompareDest.NewObject(ctx, remote)
switch err {
case fs.ErrorObjectNotFound:
return false, nil
case nil:
break
default:
return false, err
}
opt := defaultEqualOpt(ctx)
opt.updateModTime = false
if equal(ctx, src, CompareDestFile, opt) {
fs.Debugf(src, "Destination found in --compare-dest, skipping")
return true, nil
}
return false, nil
}
// GetCopyDest sets up --copy-dest
func GetCopyDest(ctx context.Context, fdst fs.Fs) (CopyDest []fs.Fs, err error) {
ci := fs.GetConfig(ctx)
CopyDest, err = cache.GetArr(ctx, ci.CopyDest)
if err != nil {
return nil, fserrors.FatalError(fmt.Errorf("Failed to make fs for --copy-dest %q: %v", ci.CopyDest, err))
}
if !SameConfigArr(fdst, CopyDest) {
return nil, fserrors.FatalError(errors.New("parameter to --copy-dest has to be on the same remote as destination"))
}
for _, cf := range CopyDest {
if cf.Features().Copy == nil {
return nil, fserrors.FatalError(errors.New("can't use --copy-dest on a remote which doesn't support server side copy"))
}
}
return CopyDest, nil
}
// copyDest checks --copy-dest to see if src needs to
// be copied
//
// Returns True if src was copied from --copy-dest
func copyDest(ctx context.Context, fdst fs.Fs, dst, src fs.Object, CopyDest, backupDir fs.Fs) (NoNeedTransfer bool, err error) {
var remote string
if dst == nil {
remote = src.Remote()
} else {
remote = dst.Remote()
}
CopyDestFile, err := CopyDest.NewObject(ctx, remote)
switch err {
case fs.ErrorObjectNotFound:
return false, nil
case nil:
break
default:
return false, err
}
opt := defaultEqualOpt(ctx)
opt.updateModTime = false
if equal(ctx, src, CopyDestFile, opt) {
if dst == nil || !Equal(ctx, src, dst) {
if dst != nil && backupDir != nil {
err = MoveBackupDir(ctx, backupDir, dst)
if err != nil {
return false, fmt.Errorf("moving to --backup-dir failed: %w", err)
}
// If successful zero out the dstObj as it is no longer there
dst = nil
}
_, err := Copy(ctx, fdst, dst, remote, CopyDestFile)
if err != nil {
fs.Errorf(src, "Destination found in --copy-dest, error copying")
return false, nil
}
fs.Debugf(src, "Destination found in --copy-dest, using server-side copy")
return true, nil
}
fs.Debugf(src, "Unchanged skipping")
return true, nil
}
fs.Debugf(src, "Destination not found in --copy-dest")
return false, nil
}
// CompareOrCopyDest checks --compare-dest and --copy-dest to see if src
// does not need to be copied
//
// Returns True if src does not need to be copied
func CompareOrCopyDest(ctx context.Context, fdst fs.Fs, dst, src fs.Object, CompareOrCopyDest []fs.Fs, backupDir fs.Fs) (NoNeedTransfer bool, err error) {
ci := fs.GetConfig(ctx)
if len(ci.CompareDest) > 0 {
for _, compareF := range CompareOrCopyDest {
NoNeedTransfer, err := compareDest(ctx, dst, src, compareF)
if NoNeedTransfer || err != nil {
return NoNeedTransfer, err
}
}
} else if len(ci.CopyDest) > 0 {
for _, copyF := range CompareOrCopyDest {
NoNeedTransfer, err := copyDest(ctx, fdst, dst, src, copyF, backupDir)
if NoNeedTransfer || err != nil {
return NoNeedTransfer, err
}
}
}
return false, nil
}
// NeedTransfer checks to see if src needs to be copied to dst using
// the current config.
//
// Returns a flag which indicates whether the file needs to be
// transferred or not.
func NeedTransfer(ctx context.Context, dst, src fs.Object) bool {
ci := fs.GetConfig(ctx)
if dst == nil {
fs.Debugf(src, "Need to transfer - File not found at Destination")
return true
}
// If we should ignore existing files, don't transfer
if ci.IgnoreExisting {
fs.Debugf(src, "Destination exists, skipping")
return false
}
// If we should upload unconditionally
if ci.IgnoreTimes {
fs.Debugf(src, "Transferring unconditionally as --ignore-times is in use")
return true
}
// If UpdateOlder is in effect, skip if dst is newer than src
if ci.UpdateOlder {
srcModTime := src.ModTime(ctx)
dstModTime := dst.ModTime(ctx)
dt := dstModTime.Sub(srcModTime)
// If have a mutually agreed precision then use that
modifyWindow := fs.GetModifyWindow(ctx, dst.Fs(), src.Fs())
if modifyWindow == fs.ModTimeNotSupported {
// Otherwise use 1 second as a safe default as
// the resolution of the time a file was
// uploaded.
modifyWindow = time.Second
}
switch {
case dt >= modifyWindow:
fs.Debugf(src, "Destination is newer than source, skipping")
return false
case dt <= -modifyWindow:
// force --checksum on for the check and do update modtimes by default
opt := defaultEqualOpt(ctx)
opt.forceModTimeMatch = true
if equal(ctx, src, dst, opt) {
fs.Debugf(src, "Unchanged skipping")
return false
}
default:
// Do a size only compare unless --checksum is set
opt := defaultEqualOpt(ctx)
opt.sizeOnly = !ci.CheckSum
if equal(ctx, src, dst, opt) {
fs.Debugf(src, "Destination mod time is within %v of source and files identical, skipping", modifyWindow)
return false
}
fs.Debugf(src, "Destination mod time is within %v of source but files differ, transferring", modifyWindow)
}
} else {
// Check to see if changed or not
if Equal(ctx, src, dst) {
fs.Debugf(src, "Unchanged skipping")
return false
}
}
return true
}
2018-08-30 15:45:41 +00:00
// RcatSize reads data from the Reader until EOF and uploads it to a file on remote.
// Pass in size >=0 if known, <0 if not known
func RcatSize(ctx context.Context, fdst fs.Fs, dstFileName string, in io.ReadCloser, size int64, modTime time.Time) (dst fs.Object, err error) {
2018-08-30 15:45:41 +00:00
var obj fs.Object
if size >= 0 {
var err error
2018-08-30 15:45:41 +00:00
// Size known use Put
tr := accounting.Stats(ctx).NewTransferRemoteSize(dstFileName, size)
defer func() {
tr.Done(ctx, err)
}()
body := ioutil.NopCloser(in) // we let the server close the body
in := tr.Account(ctx, body) // account the transfer (no buffering)
2019-01-04 19:31:09 +00:00
if SkipDestructive(ctx, dstFileName, "upload from pipe") {
2019-01-04 19:31:09 +00:00
// prevents "broken pipe" errors
_, err = io.Copy(ioutil.Discard, in)
return nil, err
}
2018-08-30 15:45:41 +00:00
info := object.NewStaticObjectInfo(dstFileName, modTime, size, true, nil, fdst)
obj, err = fdst.Put(ctx, in, info)
2018-08-30 15:45:41 +00:00
if err != nil {
fs.Errorf(dstFileName, "Post request put error: %v", err)
return nil, err
}
} else {
// Size unknown use Rcat
obj, err = Rcat(ctx, fdst, dstFileName, in, modTime)
2018-08-30 15:45:41 +00:00
if err != nil {
fs.Errorf(dstFileName, "Post request rcat error: %v", err)
return nil, err
}
}
return obj, nil
}
// copyURLFunc is called from CopyURLFn
type copyURLFunc func(ctx context.Context, dstFileName string, in io.ReadCloser, size int64, modTime time.Time) (err error)
// copyURLFn copies the data from the url to the function supplied
func copyURLFn(ctx context.Context, dstFileName string, url string, dstFileNameFromURL bool, fn copyURLFunc) (err error) {
client := fshttp.NewClient(ctx)
2019-03-08 20:33:22 +00:00
resp, err := client.Get(url)
if err != nil {
return err
}
defer fs.CheckClose(resp.Body, &err)
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
return fmt.Errorf("CopyURL failed: %s", resp.Status)
}
modTime, err := http.ParseTime(resp.Header.Get("Last-Modified"))
if err != nil {
modTime = time.Now()
}
if dstFileNameFromURL {
dstFileName = path.Base(resp.Request.URL.Path)
if dstFileName == "." || dstFileName == "/" {
return fmt.Errorf("CopyURL failed: file name wasn't found in url")
}
fs.Debugf(dstFileName, "File name found in url")
}
return fn(ctx, dstFileName, resp.Body, resp.ContentLength, modTime)
}
// CopyURL copies the data from the url to (fdst, dstFileName)
func CopyURL(ctx context.Context, fdst fs.Fs, dstFileName string, url string, dstFileNameFromURL bool, noClobber bool) (dst fs.Object, err error) {
err = copyURLFn(ctx, dstFileName, url, dstFileNameFromURL, func(ctx context.Context, dstFileName string, in io.ReadCloser, size int64, modTime time.Time) (err error) {
if noClobber {
_, err = fdst.NewObject(ctx, dstFileName)
if err == nil {
return errors.New("CopyURL failed: file already exist")
}
}
dst, err = RcatSize(ctx, fdst, dstFileName, in, size, modTime)
return err
})
return dst, err
}
// CopyURLToWriter copies the data from the url to the io.Writer supplied
func CopyURLToWriter(ctx context.Context, url string, out io.Writer) (err error) {
return copyURLFn(ctx, "", url, false, func(ctx context.Context, dstFileName string, in io.ReadCloser, size int64, modTime time.Time) (err error) {
_, err = io.Copy(out, in)
return err
})
}
// BackupDir returns the correctly configured --backup-dir
func BackupDir(ctx context.Context, fdst fs.Fs, fsrc fs.Fs, srcFileName string) (backupDir fs.Fs, err error) {
ci := fs.GetConfig(ctx)
if ci.BackupDir != "" {
backupDir, err = cache.Get(ctx, ci.BackupDir)
if err != nil {
return nil, fserrors.FatalError(fmt.Errorf("Failed to make fs for --backup-dir %q: %v", ci.BackupDir, err))
}
if !SameConfig(fdst, backupDir) {
return nil, fserrors.FatalError(errors.New("parameter to --backup-dir has to be on the same remote as destination"))
}
if srcFileName == "" {
if Overlapping(fdst, backupDir) {
return nil, fserrors.FatalError(errors.New("destination and parameter to --backup-dir mustn't overlap"))
}
if Overlapping(fsrc, backupDir) {
return nil, fserrors.FatalError(errors.New("source and parameter to --backup-dir mustn't overlap"))
}
} else {
if ci.Suffix == "" {
if SameDir(fdst, backupDir) {
return nil, fserrors.FatalError(errors.New("destination and parameter to --backup-dir mustn't be the same"))
}
if SameDir(fsrc, backupDir) {
return nil, fserrors.FatalError(errors.New("source and parameter to --backup-dir mustn't be the same"))
}
}
}
} else if ci.Suffix != "" {
// --backup-dir is not set but --suffix is - use the destination as the backupDir
backupDir = fdst
} else {
return nil, fserrors.FatalError(errors.New("internal error: BackupDir called when --backup-dir and --suffix both empty"))
}
if !CanServerSideMove(backupDir) {
return nil, fserrors.FatalError(errors.New("can't use --backup-dir on a remote which doesn't support server-side move or copy"))
}
return backupDir, nil
}
// MoveBackupDir moves a file to the backup dir
func MoveBackupDir(ctx context.Context, backupDir fs.Fs, dst fs.Object) (err error) {
remoteWithSuffix := SuffixName(ctx, dst.Remote())
overwritten, _ := backupDir.NewObject(ctx, remoteWithSuffix)
_, err = Move(ctx, backupDir, overwritten, remoteWithSuffix, dst)
return err
}
// moveOrCopyFile moves or copies a single file possibly to a new name
func moveOrCopyFile(ctx context.Context, fdst fs.Fs, fsrc fs.Fs, dstFileName string, srcFileName string, cp bool) (err error) {
ci := fs.GetConfig(ctx)
dstFilePath := path.Join(fdst.Root(), dstFileName)
srcFilePath := path.Join(fsrc.Root(), srcFileName)
if fdst.Name() == fsrc.Name() && dstFilePath == srcFilePath {
fs.Debugf(fdst, "don't need to copy/move %s, it is already at target location", dstFileName)
return nil
}
// Choose operations
Op := Move
if cp {
Op = Copy
}
// Find src object
srcObj, err := fsrc.NewObject(ctx, srcFileName)
if err != nil {
return err
}
// Find dst object if it exists
var dstObj fs.Object
if !ci.NoCheckDest {
dstObj, err = fdst.NewObject(ctx, dstFileName)
if err == fs.ErrorObjectNotFound {
dstObj = nil
} else if err != nil {
return err
}
}
// Special case for changing case of a file on a case insensitive remote
// This will move the file to a temporary name then
// move it back to the intended destination. This is required
// to avoid issues with certain remotes and avoid file deletion.
if !cp && fdst.Name() == fsrc.Name() && fdst.Features().CaseInsensitive && dstFileName != srcFileName && strings.EqualFold(dstFilePath, srcFilePath) {
// Create random name to temporarily move file to
tmpObjName := dstFileName + "-rclone-move-" + random.String(8)
_, err := fdst.NewObject(ctx, tmpObjName)
if err != fs.ErrorObjectNotFound {
if err == nil {
return errors.New("found an already existing file with a randomly generated name. Try the operation again")
}
return fmt.Errorf("error while attempting to move file to a temporary location: %w", err)
}
tr := accounting.Stats(ctx).NewTransfer(srcObj)
defer func() {
tr.Done(ctx, err)
}()
tmpObj, err := Op(ctx, fdst, nil, tmpObjName, srcObj)
if err != nil {
return fmt.Errorf("error while moving file to temporary location: %w", err)
}
_, err = Op(ctx, fdst, nil, dstFileName, tmpObj)
return err
}
var backupDir fs.Fs
var copyDestDir []fs.Fs
if ci.BackupDir != "" || ci.Suffix != "" {
backupDir, err = BackupDir(ctx, fdst, fsrc, srcFileName)
if err != nil {
return fmt.Errorf("creating Fs for --backup-dir failed: %w", err)
}
}
if len(ci.CompareDest) > 0 {
copyDestDir, err = GetCompareDest(ctx)
if err != nil {
return err
}
} else if len(ci.CopyDest) > 0 {
copyDestDir, err = GetCopyDest(ctx, fdst)
if err != nil {
return err
}
}
NoNeedTransfer, err := CompareOrCopyDest(ctx, fdst, dstObj, srcObj, copyDestDir, backupDir)
if err != nil {
return err
}
if !NoNeedTransfer && NeedTransfer(ctx, dstObj, srcObj) {
// If destination already exists, then we must move it into --backup-dir if required
if dstObj != nil && backupDir != nil {
err = MoveBackupDir(ctx, backupDir, dstObj)
if err != nil {
return fmt.Errorf("moving to --backup-dir failed: %w", err)
}
// If successful zero out the dstObj as it is no longer there
dstObj = nil
}
_, err = Op(ctx, fdst, dstObj, dstFileName, srcObj)
} else {
tr := accounting.Stats(ctx).NewCheckingTransfer(srcObj)
if !cp {
if ci.IgnoreExisting {
fs.Debugf(srcObj, "Not removing source file as destination file exists and --ignore-existing is set")
} else {
err = DeleteFile(ctx, srcObj)
}
}
tr.Done(ctx, err)
}
return err
}
// MoveFile moves a single file possibly to a new name
func MoveFile(ctx context.Context, fdst fs.Fs, fsrc fs.Fs, dstFileName string, srcFileName string) (err error) {
return moveOrCopyFile(ctx, fdst, fsrc, dstFileName, srcFileName, false)
}
// CopyFile moves a single file possibly to a new name
func CopyFile(ctx context.Context, fdst fs.Fs, fsrc fs.Fs, dstFileName string, srcFileName string) (err error) {
return moveOrCopyFile(ctx, fdst, fsrc, dstFileName, srcFileName, true)
}
2018-01-06 14:39:31 +00:00
// SetTier changes tier of object in remote
func SetTier(ctx context.Context, fsrc fs.Fs, tier string) error {
return ListFn(ctx, fsrc, func(o fs.Object) {
objImpl, ok := o.(fs.SetTierer)
if !ok {
fs.Errorf(fsrc, "Remote object does not implement SetTier")
return
}
err := objImpl.SetTier(tier)
if err != nil {
fs.Errorf(fsrc, "Failed to do SetTier, %v", err)
}
})
}
// TouchDir touches every file in directory with time t
func TouchDir(ctx context.Context, f fs.Fs, remote string, t time.Time, recursive bool) error {
return walk.ListR(ctx, f, remote, false, ConfigMaxDepth(ctx, recursive), walk.ListObjects, func(entries fs.DirEntries) error {
entries.ForObject(func(o fs.Object) {
if !SkipDestructive(ctx, o, "touch") {
fs.Debugf(f, "Touching %q", o.Remote())
err := o.SetModTime(ctx, t)
if err != nil {
err = fmt.Errorf("failed to touch: %w", err)
err = fs.CountError(err)
fs.Errorf(o, "%v", err)
}
}
})
return nil
})
}
2018-01-06 14:39:31 +00:00
// ListFormat defines files information print format
type ListFormat struct {
separator string
dirSlash bool
absolute bool
output []func(entry *ListJSONItem) string
csv *csv.Writer
buf bytes.Buffer
2018-01-06 14:39:31 +00:00
}
// SetSeparator changes separator in struct
func (l *ListFormat) SetSeparator(separator string) {
l.separator = separator
}
// SetDirSlash defines if slash should be printed
func (l *ListFormat) SetDirSlash(dirSlash bool) {
l.dirSlash = dirSlash
}
// SetAbsolute prints a leading slash in front of path names
func (l *ListFormat) SetAbsolute(absolute bool) {
l.absolute = absolute
}
// SetCSV defines if the output should be csv
//
// Note that you should call SetSeparator before this if you want a
// custom separator
func (l *ListFormat) SetCSV(useCSV bool) {
if useCSV {
l.csv = csv.NewWriter(&l.buf)
if l.separator != "" {
l.csv.Comma = []rune(l.separator)[0]
}
} else {
l.csv = nil
}
}
2018-01-06 14:39:31 +00:00
// SetOutput sets functions used to create files information
func (l *ListFormat) SetOutput(output []func(entry *ListJSONItem) string) {
2018-01-06 14:39:31 +00:00
l.output = output
}
// AddModTime adds file's Mod Time to output
func (l *ListFormat) AddModTime() {
l.AppendOutput(func(entry *ListJSONItem) string {
return entry.ModTime.When.Local().Format("2006-01-02 15:04:05")
})
2018-01-06 14:39:31 +00:00
}
// AddSize adds file's size to output
func (l *ListFormat) AddSize() {
l.AppendOutput(func(entry *ListJSONItem) string {
return strconv.FormatInt(entry.Size, 10)
2018-01-06 17:53:37 +00:00
})
2018-01-06 14:39:31 +00:00
}
// normalisePath makes sure the path has the correct slashes for the current mode
func (l *ListFormat) normalisePath(entry *ListJSONItem, remote string) string {
if l.absolute && !strings.HasPrefix(remote, "/") {
remote = "/" + remote
}
if entry.IsDir && l.dirSlash {
remote += "/"
}
return remote
}
2018-01-06 14:39:31 +00:00
// AddPath adds path to file to output
func (l *ListFormat) AddPath() {
l.AppendOutput(func(entry *ListJSONItem) string {
return l.normalisePath(entry, entry.Path)
})
}
// AddEncrypted adds the encrypted path to file to output
func (l *ListFormat) AddEncrypted() {
l.AppendOutput(func(entry *ListJSONItem) string {
return l.normalisePath(entry, entry.Encrypted)
2018-01-06 14:39:31 +00:00
})
}
2018-01-06 17:53:37 +00:00
// AddHash adds the hash of the type given to the output
func (l *ListFormat) AddHash(ht hash.Type) {
hashName := ht.String()
l.AppendOutput(func(entry *ListJSONItem) string {
if entry.IsDir {
2018-01-06 17:53:37 +00:00
return ""
}
return entry.Hashes[hashName]
2018-01-06 17:53:37 +00:00
})
}
// AddID adds file's ID to the output if known
func (l *ListFormat) AddID() {
l.AppendOutput(func(entry *ListJSONItem) string {
return entry.ID
})
}
// AddOrigID adds file's Original ID to the output if known
func (l *ListFormat) AddOrigID() {
l.AppendOutput(func(entry *ListJSONItem) string {
return entry.OrigID
})
}
// AddTier adds file's Tier to the output if known
func (l *ListFormat) AddTier() {
l.AppendOutput(func(entry *ListJSONItem) string {
return entry.Tier
})
}
// AddMimeType adds file's MimeType to the output if known
func (l *ListFormat) AddMimeType() {
l.AppendOutput(func(entry *ListJSONItem) string {
return entry.MimeType
})
}
2018-01-06 14:39:31 +00:00
// AppendOutput adds string generated by specific function to printed output
func (l *ListFormat) AppendOutput(functionToAppend func(item *ListJSONItem) string) {
2018-01-06 14:39:31 +00:00
l.output = append(l.output, functionToAppend)
}
// Format prints information about the DirEntry in the format defined
func (l *ListFormat) Format(entry *ListJSONItem) (result string) {
var out []string
for _, fun := range l.output {
out = append(out, fun(entry))
}
if l.csv != nil {
l.buf.Reset()
_ = l.csv.Write(out) // can't fail writing to bytes.Buffer
l.csv.Flush()
result = strings.TrimRight(l.buf.String(), "\n")
} else {
result = strings.Join(out, l.separator)
2018-01-06 14:39:31 +00:00
}
return result
2018-01-06 14:39:31 +00:00
}
// DirMove renames srcRemote to dstRemote
//
// It does this by loading the directory tree into memory (using ListR
// if available) and doing renames in parallel.
func DirMove(ctx context.Context, f fs.Fs, srcRemote, dstRemote string) (err error) {
ci := fs.GetConfig(ctx)
if SkipDestructive(ctx, srcRemote, "dirMove") {
accounting.Stats(ctx).Renames(1)
return nil
}
// Use DirMove if possible
if doDirMove := f.Features().DirMove; doDirMove != nil {
err = doDirMove(ctx, f, srcRemote, dstRemote)
if err == nil {
accounting.Stats(ctx).Renames(1)
}
return err
}
// Load the directory tree into memory
tree, err := walk.NewDirTree(ctx, f, srcRemote, true, -1)
if err != nil {
return fmt.Errorf("RenameDir tree walk: %w", err)
}
// Get the directories in sorted order
dirs := tree.Dirs()
// Make the destination directories - must be done in order not in parallel
for _, dir := range dirs {
dstPath := dstRemote + dir[len(srcRemote):]
err := f.Mkdir(ctx, dstPath)
if err != nil {
return fmt.Errorf("RenameDir mkdir: %w", err)
}
}
// Rename the files in parallel
type rename struct {
o fs.Object
newPath string
}
renames := make(chan rename, ci.Transfers)
g, gCtx := errgroup.WithContext(context.Background())
for i := 0; i < ci.Transfers; i++ {
g.Go(func() error {
for job := range renames {
dstOverwritten, _ := f.NewObject(gCtx, job.newPath)
_, err := Move(gCtx, f, dstOverwritten, job.newPath, job.o)
if err != nil {
return err
}
select {
case <-gCtx.Done():
return gCtx.Err()
default:
}
}
return nil
})
}
for dir, entries := range tree {
dstPath := dstRemote + dir[len(srcRemote):]
for _, entry := range entries {
if o, ok := entry.(fs.Object); ok {
renames <- rename{o, path.Join(dstPath, path.Base(o.Remote()))}
}
}
}
close(renames)
err = g.Wait()
if err != nil {
return fmt.Errorf("RenameDir renames: %w", err)
}
// Remove the source directories in reverse order
for i := len(dirs) - 1; i >= 0; i-- {
err := f.Rmdir(ctx, dirs[i])
if err != nil {
return fmt.Errorf("RenameDir rmdir: %w", err)
}
}
return nil
}
// FsInfo provides information about a remote
type FsInfo struct {
// Name of the remote (as passed into NewFs)
Name string
// Root of the remote (as passed into NewFs)
Root string
// String returns a description of the FS
String string
// Precision of the ModTimes in this Fs in Nanoseconds
Precision time.Duration
// Returns the supported hash types of the filesystem
Hashes []string
// Features returns the optional features of this Fs
Features map[string]bool
}
// GetFsInfo gets the information (FsInfo) about a given Fs
func GetFsInfo(f fs.Fs) *FsInfo {
info := &FsInfo{
Name: f.Name(),
Root: f.Root(),
String: f.String(),
Precision: f.Precision(),
Hashes: make([]string, 0, 4),
Features: f.Features().Enabled(),
}
for _, hashType := range f.Hashes().Array() {
info.Hashes = append(info.Hashes, hashType.String())
}
return info
}
var (
interactiveMu sync.Mutex
skipped = map[string]bool{}
)
// skipDestructiveChoose asks the user which action to take
//
// Call with interactiveMu held
func skipDestructiveChoose(ctx context.Context, subject interface{}, action string) (skip bool) {
fmt.Printf("rclone: %s \"%v\"?\n", action, subject)
switch i := config.CommandDefault([]string{
"yYes, this is OK",
"nNo, skip this",
fmt.Sprintf("sSkip all %s operations with no more questions", action),
fmt.Sprintf("!Do all %s operations with no more questions", action),
"qExit rclone now.",
}, 0); i {
case 'y':
skip = false
case 'n':
skip = true
case 's':
skip = true
skipped[action] = true
fs.Logf(nil, "Skipping all %s operations from now on without asking", action)
case '!':
skip = false
skipped[action] = false
fs.Logf(nil, "Doing all %s operations from now on without asking", action)
case 'q':
fs.Logf(nil, "Quitting rclone now")
atexit.Run()
os.Exit(0)
default:
skip = true
fs.Errorf(nil, "Bad choice %c", i)
}
return skip
}
// SkipDestructive should be called whenever rclone is about to do an destructive operation.
//
// It will check the --dry-run flag and it will ask the user if the --interactive flag is set.
//
// subject should be the object or directory in use
//
// action should be a descriptive word or short phrase
//
// Together they should make sense in this sentence: "Rclone is about
// to action subject".
func SkipDestructive(ctx context.Context, subject interface{}, action string) (skip bool) {
var flag string
ci := fs.GetConfig(ctx)
switch {
case ci.DryRun:
flag = "--dry-run"
skip = true
case ci.Interactive:
flag = "--interactive"
interactiveMu.Lock()
defer interactiveMu.Unlock()
var found bool
skip, found = skipped[action]
if !found {
skip = skipDestructiveChoose(ctx, subject, action)
}
default:
return false
}
if skip {
size := int64(-1)
if do, ok := subject.(interface{ Size() int64 }); ok {
size = do.Size()
}
if size >= 0 {
fs.Logf(subject, "Skipped %s as %s is set (size %v)", fs.LogValue("skipped", action), flag, fs.LogValue("size", fs.SizeSuffix(size)))
} else {
fs.Logf(subject, "Skipped %s as %s is set", fs.LogValue("skipped", action), flag)
}
}
return skip
}