restic/internal/pipe/pipe.go

293 lines
6.9 KiB
Go
Raw Normal View History

2015-02-15 11:57:09 +00:00
package pipe
import (
2017-06-04 09:16:55 +00:00
"context"
2015-02-15 11:57:09 +00:00
"fmt"
"os"
"path/filepath"
"sort"
2015-03-02 13:48:47 +00:00
2017-07-23 12:21:03 +00:00
"github.com/restic/restic/internal/errors"
2017-07-23 12:21:03 +00:00
"github.com/restic/restic/internal/debug"
"github.com/restic/restic/internal/fs"
2015-02-15 11:57:09 +00:00
)
2015-03-07 10:53:32 +00:00
type Result interface{}
type Job interface {
Path() string
Fullpath() string
Error() error
Info() os.FileInfo
Result() chan<- Result
}
2015-02-15 11:57:09 +00:00
type Entry struct {
2015-03-07 10:53:32 +00:00
basedir string
path string
info os.FileInfo
error error
result chan<- Result
// points to the old node if available, interface{} is used to prevent
// circular import
Node interface{}
2015-02-15 11:57:09 +00:00
}
2015-03-07 10:53:32 +00:00
func (e Entry) Path() string { return e.path }
func (e Entry) Fullpath() string { return filepath.Join(e.basedir, e.path) }
func (e Entry) Error() error { return e.error }
func (e Entry) Info() os.FileInfo { return e.info }
func (e Entry) Result() chan<- Result { return e.result }
2015-02-15 11:57:09 +00:00
type Dir struct {
2015-03-07 10:53:32 +00:00
basedir string
path string
error error
info os.FileInfo
2015-02-15 11:57:09 +00:00
2015-03-07 10:53:32 +00:00
Entries [](<-chan Result)
result chan<- Result
2015-02-15 11:57:09 +00:00
}
2015-03-07 10:53:32 +00:00
func (e Dir) Path() string { return e.path }
func (e Dir) Fullpath() string { return filepath.Join(e.basedir, e.path) }
func (e Dir) Error() error { return e.error }
func (e Dir) Info() os.FileInfo { return e.info }
func (e Dir) Result() chan<- Result { return e.result }
2015-02-15 11:57:09 +00:00
// readDirNames reads the directory named by dirname and returns
// a sorted list of directory entries.
// taken from filepath/path.go
func readDirNames(dirname string) ([]string, error) {
f, err := fs.Open(dirname)
2015-02-15 11:57:09 +00:00
if err != nil {
2016-08-29 20:16:58 +00:00
return nil, errors.Wrap(err, "Open")
2015-02-15 11:57:09 +00:00
}
names, err := f.Readdirnames(-1)
_ = f.Close()
2015-02-15 11:57:09 +00:00
if err != nil {
2016-08-29 20:16:58 +00:00
return nil, errors.Wrap(err, "Readdirnames")
2015-02-15 11:57:09 +00:00
}
sort.Strings(names)
return names, nil
}
// SelectFunc returns true for all items that should be included (files and
// dirs). If false is returned, files are ignored and dirs are not even walked.
type SelectFunc func(item string, fi os.FileInfo) bool
2017-06-04 09:16:55 +00:00
func walk(ctx context.Context, basedir, dir string, selectFunc SelectFunc, jobs chan<- Job, res chan<- Result) (excluded bool) {
2016-09-27 20:35:08 +00:00
debug.Log("start on %q, basedir %q", dir, basedir)
2015-11-06 18:41:57 +00:00
relpath, err := filepath.Rel(basedir, dir)
if err != nil {
panic(err)
}
info, err := fs.Lstat(dir)
2015-02-15 11:57:09 +00:00
if err != nil {
2016-08-29 20:16:58 +00:00
err = errors.Wrap(err, "Lstat")
2016-09-27 20:35:08 +00:00
debug.Log("error for %v: %v, res %p", dir, err, res)
2015-11-06 18:41:57 +00:00
select {
case jobs <- Dir{basedir: basedir, path: relpath, info: info, error: err, result: res}:
2017-06-04 09:16:55 +00:00
case <-ctx.Done():
2015-11-06 18:41:57 +00:00
}
return
2015-02-15 11:57:09 +00:00
}
if !selectFunc(dir, info) {
2016-09-27 20:35:08 +00:00
debug.Log("file %v excluded by filter, res %p", dir, res)
excluded = true
2015-11-06 18:41:57 +00:00
return
}
2015-02-15 11:57:09 +00:00
if !info.IsDir() {
2016-09-27 20:35:08 +00:00
debug.Log("sending file job for %v, res %p", dir, res)
2015-03-02 18:44:16 +00:00
select {
2015-03-07 10:53:32 +00:00
case jobs <- Entry{info: info, basedir: basedir, path: relpath, result: res}:
2017-06-04 09:16:55 +00:00
case <-ctx.Done():
2015-03-02 18:44:16 +00:00
}
2015-11-06 18:41:57 +00:00
return
2015-02-15 11:57:09 +00:00
}
2015-11-06 21:38:34 +00:00
debug.RunHook("pipe.readdirnames", dir)
2015-03-15 13:24:58 +00:00
names, err := readDirNames(dir)
2015-02-15 11:57:09 +00:00
if err != nil {
2016-09-27 20:35:08 +00:00
debug.Log("Readdirnames(%v) returned error: %v, res %p", dir, err, res)
2015-11-06 18:41:57 +00:00
select {
2017-06-04 09:16:55 +00:00
case <-ctx.Done():
2015-11-06 18:41:57 +00:00
case jobs <- Dir{basedir: basedir, path: relpath, info: info, error: err, result: res}:
}
return
2015-02-15 11:57:09 +00:00
}
2015-03-15 13:24:58 +00:00
// Insert breakpoint to allow testing behaviour with vanishing files
// between Readdir() and lstat()
debug.RunHook("pipe.walk1", relpath)
2015-03-15 13:24:58 +00:00
2015-03-07 10:53:32 +00:00
entries := make([]<-chan Result, 0, len(names))
2015-02-15 11:57:09 +00:00
for _, name := range names {
2015-03-15 13:24:58 +00:00
subpath := filepath.Join(dir, name)
2015-02-15 13:44:54 +00:00
fi, statErr := fs.Lstat(subpath)
if !selectFunc(subpath, fi) {
2016-09-27 20:35:08 +00:00
debug.Log("file %v excluded by filter", subpath)
continue
}
2015-03-07 10:53:32 +00:00
ch := make(chan Result, 1)
2015-02-15 13:44:54 +00:00
entries = append(entries, ch)
2015-02-15 11:57:09 +00:00
if statErr != nil {
2016-08-29 20:16:58 +00:00
statErr = errors.Wrap(statErr, "Lstat")
2016-09-27 20:35:08 +00:00
debug.Log("sending file job for %v, err %v, res %p", subpath, err, res)
2015-03-02 18:44:16 +00:00
select {
case jobs <- Entry{info: fi, error: statErr, basedir: basedir, path: filepath.Join(relpath, name), result: ch}:
2017-06-04 09:16:55 +00:00
case <-ctx.Done():
2015-11-06 18:41:57 +00:00
return
2015-03-02 18:44:16 +00:00
}
continue
2015-02-15 11:57:09 +00:00
}
2015-03-15 13:24:58 +00:00
// Insert breakpoint to allow testing behaviour with vanishing files
// between walk and open
debug.RunHook("pipe.walk2", filepath.Join(relpath, name))
2015-03-15 13:24:58 +00:00
2017-06-04 09:16:55 +00:00
walk(ctx, basedir, subpath, selectFunc, jobs, ch)
2015-02-15 11:57:09 +00:00
}
2016-09-27 20:35:08 +00:00
debug.Log("sending dirjob for %q, basedir %q, res %p", dir, basedir, res)
2015-03-02 18:44:16 +00:00
select {
2015-03-07 10:53:32 +00:00
case jobs <- Dir{basedir: basedir, path: relpath, info: info, Entries: entries, result: res}:
2017-06-04 09:16:55 +00:00
case <-ctx.Done():
2015-03-02 18:44:16 +00:00
}
return
2015-02-15 11:57:09 +00:00
}
// cleanupPath is used to clean a path. For a normal path, a slice with just
// the path is returned. For special cases such as "." and "/" the list of
// names within those paths is returned.
func cleanupPath(path string) ([]string, error) {
path = filepath.Clean(path)
if filepath.Dir(path) != path {
return []string{path}, nil
}
paths, err := readDirNames(path)
if err != nil {
return nil, err
}
for i, p := range paths {
paths[i] = filepath.Join(path, p)
}
return paths, nil
}
2015-03-02 13:48:47 +00:00
// Walk sends a Job for each file and directory it finds below the paths. When
// the channel done is closed, processing stops.
2017-06-04 09:16:55 +00:00
func Walk(ctx context.Context, walkPaths []string, selectFunc SelectFunc, jobs chan<- Job, res chan<- Result) {
var paths []string
for _, p := range walkPaths {
ps, err := cleanupPath(p)
if err != nil {
fmt.Fprintf(os.Stderr, "Readdirnames(%v): %v, skipping\n", p, err)
2016-09-27 20:35:08 +00:00
debug.Log("Readdirnames(%v) returned error: %v, skipping", p, err)
continue
}
paths = append(paths, ps...)
}
2016-09-27 20:35:08 +00:00
debug.Log("start on %v", paths)
2015-03-02 13:48:47 +00:00
defer func() {
2016-09-27 20:35:08 +00:00
debug.Log("output channel closed")
2015-03-07 10:53:32 +00:00
close(jobs)
2015-03-02 13:48:47 +00:00
}()
2015-03-07 10:53:32 +00:00
entries := make([]<-chan Result, 0, len(paths))
2015-03-02 13:48:47 +00:00
for _, path := range paths {
2016-09-27 20:35:08 +00:00
debug.Log("start walker for %v", path)
2015-03-07 10:53:32 +00:00
ch := make(chan Result, 1)
2017-06-04 09:16:55 +00:00
excluded := walk(ctx, filepath.Dir(path), path, selectFunc, jobs, ch)
if excluded {
2016-09-27 20:35:08 +00:00
debug.Log("walker for %v done, it was excluded by the filter", path)
continue
}
2015-03-28 15:35:46 +00:00
entries = append(entries, ch)
2016-09-27 20:35:08 +00:00
debug.Log("walker for %v done", path)
2015-03-02 13:48:47 +00:00
}
2016-09-27 20:35:08 +00:00
debug.Log("sending root node, res %p", res)
select {
2017-06-04 09:16:55 +00:00
case <-ctx.Done():
2015-11-06 18:41:57 +00:00
return
case jobs <- Dir{Entries: entries, result: res}:
}
2016-09-27 20:35:08 +00:00
debug.Log("walker done")
2015-03-02 13:48:47 +00:00
}
// Split feeds all elements read from inChan to dirChan and entChan.
2015-03-07 10:53:32 +00:00
func Split(inChan <-chan Job, dirChan chan<- Dir, entChan chan<- Entry) {
2016-09-27 20:35:08 +00:00
debug.Log("start")
defer debug.Log("done")
2015-03-02 13:48:47 +00:00
inCh := inChan
dirCh := dirChan
entCh := entChan
var (
dir Dir
ent Entry
)
// deactivate sending until we received at least one job
dirCh = nil
entCh = nil
for {
select {
case job, ok := <-inCh:
if !ok {
// channel is closed
return
}
if job == nil {
panic("nil job received")
}
// disable receiving until the current job has been sent
inCh = nil
switch j := job.(type) {
case Dir:
dir = j
dirCh = dirChan
case Entry:
ent = j
entCh = entChan
default:
panic(fmt.Sprintf("unknown job type %v", j))
}
case dirCh <- dir:
// disable sending, re-enable receiving
dirCh = nil
inCh = inChan
case entCh <- ent:
// disable sending, re-enable receiving
entCh = nil
inCh = inChan
}
}
2015-02-15 11:57:09 +00:00
}