// List function with retry and support for listing all keys in a prefix
func List(bucket *s3.Bucket, prefix string, delimiter string) <-chan *s3.ListResp {
ch := make(chan *s3.ListResp, 100)
go func(pfix string, del string) {
defer close(ch)
isTruncated := true
nextMarker := ""
for isTruncated {
attempts := 0
for {
attempts++
res, err := bucket.List(pfix, del, nextMarker, 1000)
if err != nil {
if err.Error() == "runtime error: index out of range" {
break
}
if attempts >= maxRetries {
log.Panic(err)
}
time.Sleep(time.Second * 3)
} else {
ch <- res
if len(res.Contents) > 0 {
nextMarker = res.Contents[len(res.Contents)-1].Key
} else if len(res.CommonPrefixes) > 0 {
nextMarker = res.CommonPrefixes[len(res.CommonPrefixes)-1]
}
isTruncated = res.IsTruncated
break
}
}
}
}(prefix, delimiter)
return ch
}
// Recursively descend into an S3 directory tree, filtering based on the given
// schema, and sending results on the given channel. The `level` parameter
// indicates how far down the tree we are, and is used to determine which schema
// field we use for filtering.
func FilterS3(bucket *s3.Bucket, prefix string, level int, schema Schema, kc chan S3ListResult) {
// Update the marker as we encounter keys / prefixes. If a response is
// truncated, the next `List` request will start from the next item after
// the marker.
marker := ""
// Keep listing if the response is incomplete (there are more than
// `listBatchSize` entries or prefixes)
done := false
for !done {
response, err := bucket.List(prefix, "/", marker, listBatchSize)
if err != nil {
fmt.Printf("Error listing: %s\n", err)
// TODO: retry?
kc <- S3ListResult{s3.Key{}, err}
break
}
if !response.IsTruncated {
// Response is not truncated, so we're done.
done = true
}
if level >= len(schema.Fields) {
// We are past all the dimensions - encountered items are now
// S3 key names. We ignore any further prefixes and assume that the
// specified schema is correct/complete.
for _, k := range response.Contents {
marker = k.Key
kc <- S3ListResult{k, nil}
}
} else {
// We are still looking at prefixes. Recursively list each one that
// matches the specified schema's allowed values.
for _, pf := range response.CommonPrefixes {
// Get just the last piece of the prefix to check it as a
// dimension. If we have '/foo/bar/baz', we just want 'baz'.
stripped := pf[len(prefix) : len(pf)-1]
allowed := schema.Dims[schema.Fields[level]].IsAllowed(stripped)
marker = pf
if allowed {
FilterS3(bucket, pf, level+1, schema, kc)
}
}
}
}
if level == 0 {
// We traverse the tree in depth-first order, so once we've reached the
// end at the root (level 0), we know we're done.
// Note that things could be made faster by parallelizing the recursive
// listing, but we would need some other mechanism to know when to close
// the channel?
close(kc)
}
return
}
// Read a single client record using a partial read from S3 using the given
// headers, which should contain a "Range: bytes=M-N" header.
func getClientRecord(bucket *s3.Bucket, o *MessageLocation, headers map[string][]string) ([]byte, error) {
resp, err := bucket.GetResponseWithHeaders(o.Key, headers)
if err != nil {
return nil, err
}
defer resp.Body.Close()
body, err := ioutil.ReadAll(resp.Body)
if err == nil && len(body) != int(o.Length) {
err = fmt.Errorf("Unexpected body length: %d != %d\n", len(body), o.Length)
}
return body, err
}
func killBucket(b *s3.Bucket) {
var err error
for attempt := attempts.Start(); attempt.Next(); {
err = b.DelBucket()
if err == nil {
return
}
if _, ok := err.(*net.DNSError); ok {
return
}
e, ok := err.(*s3.Error)
if ok && e.Code == "NoSuchBucket" {
return
}
if ok && e.Code == "BucketNotEmpty" {
// Errors are ignored here. Just retry.
resp, err := b.List("", "", "", 1000)
if err == nil {
for _, key := range resp.Contents {
_ = b.Del(key.Key)
}
}
multis, _, _ := b.ListMulti("", "")
for _, m := range multis {
_ = m.Abort()
}
}
}
message := "cannot delete test bucket"
if err != nil {
message += ": " + err.Error()
}
panic(message)
}
func Get(bucket *s3.Bucket, key string) ([]byte, error) {
attempts := 0
for {
attempts++
buff, err := bucket.Get(key)
if err == nil {
return buff, nil
}
if attempts >= maxRetries && err != nil {
return nil, err
}
}
}
func Put(bucket *s3.Bucket, key string, contents []byte, contentType string, permissions s3.ACL, options s3.Options) error {
attempts := 0
for {
attempts++
err := bucket.Put(key, contents, contentType, permissions, options)
if err == nil {
return nil
}
if attempts >= maxRetries && err != nil {
return err
}
time.Sleep(time.Second * 3)
}
}
// DeleteMulti deletes multiple keys
func DeleteMulti(bucket *s3.Bucket, keys []string) error {
attempts := 0
for {
attempts++
err := bucket.DelMulti(toDeleteStruct(keys))
if err != nil {
if attempts >= maxRetries {
return err
}
time.Sleep(time.Second * 3)
} else {
break
}
}
return nil
}
// Callers must call Close() on rc.
func getS3Reader(bucket *s3.Bucket, s3Key string, offset uint64) (rc io.ReadCloser, err error) {
if offset == 0 {
rc, err = bucket.GetReader(s3Key)
return
}
headers := map[string][]string{
"Range": []string{fmt.Sprintf("bytes=%d-", offset)},
}
resp, err := bucket.GetResponseWithHeaders(s3Key, headers)
if resp != nil {
rc = resp.Body
}
return
}
// Generate a function to use for uploading files as we walk the file tree from
// `base`.
func makeupload(base string, pattern *regexp.Regexp, bucket *s3.Bucket, bucketPrefix string, dryRun bool, progress *Progress) func(string, os.FileInfo, error) error {
// Remove any excess path separators from the bucket prefix.
bucketPrefix = strings.Trim(bucketPrefix, "/")
// Get a canonical version of the base dir
cleanBase := filepath.Clean(base)
// Create a closure for the upload function.
return func(path string, fi os.FileInfo, err error) (errOut error) {
if err != nil {
return err
}
if fi.IsDir() {
return nil
}
//fmt.Printf("Found an item: %s\n", path)
baseName := filepath.Base(path)
if !pattern.MatchString(baseName) {
//fmt.Printf("Item does not match pattern\n")
return nil
}
// Make sure we're comparing apples to apples when stripping off the
// base path. Use the canonical versions of both.
cleanPath := filepath.Clean(path)
relPath := cleanPath[len(cleanBase)+1:]
// If we encounter a file
// /path/to/base/foo/bar/baz
// <--- base ---><-- rel -->
// and our bucket prefix is `hello/files`, our file in S3 will be at
// s3://bucket-name/hello/files/foo/bar/baz
// <- prefix -><-- rel -->
s3Path := fmt.Sprintf("%s/%s", bucketPrefix, relPath)
// Update progress Count whether we were successful or not.
progress.Count += 1
if dryRun {
fmt.Printf("Dry run. Not uploading item to %s\n", s3Path)
return
}
fmt.Printf("Uploading item to: %s\n", s3Path)
reader, err := os.Open(path)
if err != nil {
fmt.Printf("Error opening %s for reading: %s\n", path, err)
progress.Errors++
return err
}
err = bucket.PutReader(s3Path, reader, fi.Size(), "binary/octet-stream", s3.BucketOwnerFull, s3.Options{})
if err != nil {
progress.Errors++
return err
}
// Count the bytes for this file as progress if there were
// no upoad errors.
progress.Bytes += fi.Size()
err = reader.Close()
if err != nil {
fmt.Printf("Error closing %s: %s\n", path, err)
progress.Errors++
errOut = err
}
// Now remove the file locally.
err = os.Remove(path)
if err != nil {
fmt.Printf("Error removing %s: %s\n", path, err)
progress.Errors++
errOut = err
}
return
}
}
// Stream takes a set of prefixes lists them and
// streams the contents by line
func Stream(prefixes []string, searchDepth int, keyRegex string, includeKeyName bool) {
if len(prefixes) == 0 {
fmt.Printf("No prefixes provided\n Usage: fasts3 get <prefix>")
return
}
keys := make(chan string, len(prefixes)*2+1)
var keyRegexFilter *regexp.Regexp
if keyRegex != "" {
keyRegexFilter = regexp.MustCompile(keyRegex)
} else {
keyRegexFilter = nil
}
var b *s3.Bucket = nil
go func() {
for _, prefix := range prefixes {
bucket, prefix := parseS3Uri(prefix)
if b == nil {
b = GetBucket(bucket)
}
keyExists, err := b.Exists(prefix)
if err != nil {
log.Fatalln(err)
}
if keyExists {
if keyRegexFilter != nil && !keyRegexFilter.MatchString(prefix) {
continue
}
keys <- prefix
} else {
for key := range s3wrapper.ListRecurse(b, prefix, searchDepth) {
if keyRegexFilter != nil && !keyRegexFilter.MatchString(key.Key) {
continue
}
keys <- key.Key
}
}
}
close(keys)
}()
var wg sync.WaitGroup
msgs := make(chan string, 1000)
for i := 1; i <= 10; i++ {
wg.Add(1)
go func() {
for key := range keys {
bts, err := s3wrapper.Get(b, key)
reader, err := getReaderByExt(bts, key)
if err != nil {
panic(err)
}
for {
line, _, err := reader.ReadLine()
if err != nil {
if err.Error() == "EOF" {
break
} else {
log.Fatalln(err)
}
}
msg := fmt.Sprintf("%s\n", string(line))
if includeKeyName {
msg = fmt.Sprintf("[%s] %s", key, msg)
}
msgs <- msg
}
}
wg.Done()
}()
}
go func() {
wg.Wait()
close(msgs)
}()
for msg := range msgs {
fmt.Print(msg)
}
}
// Get lists and retrieves s3 keys given a list of prefixes
// searchDepth can also be specified to increase speed of listing
func Get(prefixes []string, searchDepth int) {
if len(prefixes) == 0 {
fmt.Printf("No prefixes provided\n Usage: fasts3 get <prefix>")
return
}
getRequests := make(chan GetRequest, len(prefixes)*2+1)
var b *s3.Bucket = nil
go func() {
for _, prefix := range prefixes {
bucket, prefix := parseS3Uri(prefix)
if b == nil {
b = GetBucket(bucket)
}
keyExists, err := b.Exists(prefix)
if err != nil {
log.Fatalln(err)
}
if keyExists {
keyParts := strings.Split(prefix, "/")
ogPrefix := strings.Join(keyParts[0:len(keyParts)-1], "/") + "/"
getRequests <- GetRequest{Key: prefix, OriginalPrefix: ogPrefix}
} else {
for key := range s3wrapper.ListRecurse(b, prefix, searchDepth) {
getRequests <- GetRequest{Key: key.Key, OriginalPrefix: prefix}
}
}
}
close(getRequests)
}()
var wg sync.WaitGroup
msgs := make(chan string, 1000)
workingDirectory, err := os.Getwd()
if err != nil {
log.Fatalln(err)
}
for i := 1; i <= 10; i++ {
wg.Add(1)
go func() {
for rq := range getRequests {
dest := path.Join(workingDirectory, strings.Replace(rq.Key, rq.OriginalPrefix, "", 1))
msgs <- fmt.Sprintf("Getting %s -> %s\n", rq.Key, dest)
err := s3wrapper.GetToFile(b, rq.Key, dest)
if err != nil {
log.Fatalln(err)
}
}
wg.Done()
}()
}
go func() {
wg.Wait()
close(msgs)
}()
for msg := range msgs {
fmt.Print(msg)
}
}
// Del deletes a set of prefixes(s3 keys or partial keys
func Del(prefixes []string, searchDepth int, isRecursive bool) {
if len(*delPrefixes) == 0 {
fmt.Printf("No prefixes provided\n Usage: fasts3 del <prefix>")
return
}
keys := make(chan string, len(prefixes)*2+1)
var b *s3.Bucket = nil
go func() {
for _, delPrefix := range prefixes {
bucket, prefix := parseS3Uri(delPrefix)
if b == nil {
b = GetBucket(bucket)
}
keys <- prefix
if *delRecurse {
keyExists, err := b.Exists(prefix)
if err != nil {
log.Fatalln(err)
}
if keyExists {
keys <- prefix
} else if *delRecurse {
for key := range s3wrapper.ListRecurse(b, prefix, searchDepth) {
keys <- key.Key
}
} else {
fmt.Printf("trying to delete a prefix, please add --recursive or -r to proceed\n")
}
}
}
close(keys)
}()
var wg sync.WaitGroup
msgs := make(chan string, 1000)
for i := 1; i <= 10; i++ {
wg.Add(1)
go func() {
batch := make([]string, 0, 100)
for key := range keys {
batch = append(batch, key)
if len(batch) >= 100 {
err := s3wrapper.DeleteMulti(b, batch)
if err != nil {
log.Fatalln(err)
}
for _, k := range batch {
msgs <- fmt.Sprintf("File %s Deleted\n", k)
}
batch = batch[:0]
}
}
if len(batch) > 0 {
err := s3wrapper.DeleteMulti(b, batch)
if err != nil {
log.Fatalln(err)
}
for _, k := range batch {
msgs <- fmt.Sprintf("File %s Deleted\n", k)
}
}
wg.Done()
}()
}
go func() {
wg.Wait()
close(msgs)
}()
for msg := range msgs {
fmt.Print(msg)
}
}