// This is the simple 1MB chunk version. The rolling checksum version is below.
func WriteFileMap(bs blobserver.StatReceiver, fileMap map[string]interface{}, r io.Reader) (*blobref.BlobRef, os.Error) {
parts, size := []BytesPart{}, int64(0)
buf := new(bytes.Buffer)
for {
buf.Reset()
n, err := io.Copy(buf, io.LimitReader(r, 1<<20))
if err != nil {
return nil, err
}
if n == 0 {
break
}
hash := crypto.SHA1.New()
io.Copy(hash, bytes.NewBuffer(buf.Bytes()))
br := blobref.FromHash("sha1", hash)
hasBlob, err := serverHasBlob(bs, br)
if err != nil {
return nil, err
}
if !hasBlob {
sb, err := bs.ReceiveBlob(br, buf)
if err != nil {
return nil, err
}
if expect := (blobref.SizedBlobRef{br, n}); !expect.Equal(sb) {
return nil, fmt.Errorf("schema/filewriter: wrote %s bytes, got %s ack'd", expect, sb)
}
}
size += n
parts = append(parts, BytesPart{
BlobRef: br,
Size: uint64(n),
Offset: 0, // into BlobRef to read from (not of dest)
})
}
err := PopulateParts(fileMap, size, parts)
if err != nil {
return nil, err
}
json, err := MapToCamliJson(fileMap)
if err != nil {
return nil, err
}
br := blobref.Sha1FromString(json)
sb, err := bs.ReceiveBlob(br, strings.NewReader(json))
if err != nil {
return nil, err
}
if expect := (blobref.SizedBlobRef{br, int64(len(json))}); !expect.Equal(sb) {
return nil, fmt.Errorf("schema/filewriter: wrote %s bytes, got %s ack'd", expect, sb)
}
return br, nil
}
// Note: must not touch data after calling this.
func NewUploadHandleFromString(data string) *UploadHandle {
s1 := sha1.New()
s1.Write([]byte(data))
bref := blobref.FromHash("sha1", s1)
buf := bytes.NewBufferString(data)
return &UploadHandle{BlobRef: bref, Size: int64(len(data)), Contents: buf}
}
func (mi *Indexer) populateFile(client *mysql.Client, blobRef *blobref.BlobRef, ss *schema.Superset) (err os.Error) {
if ss.Fragment {
return nil
}
seekFetcher, err := blobref.SeekerFromStreamingFetcher(mi.BlobSource)
if err != nil {
return err
}
sha1 := sha1.New()
fr := ss.NewFileReader(seekFetcher)
mime, reader := magic.MimeTypeFromReader(fr)
n, err := io.Copy(sha1, reader)
if err != nil {
// TODO: job scheduling system to retry this spaced
// out max n times. Right now our options are
// ignoring this error (forever) or returning the
// error and making the indexing try again (likely
// forever failing). Both options suck. For now just
// log and act like all's okay.
log.Printf("mysqlindex: error indexing file %s: %v", blobRef, err)
return nil
}
attrs := []string{}
if ss.UnixPermission != "" {
attrs = append(attrs, "perm")
}
if ss.UnixOwnerId != 0 || ss.UnixOwner != "" || ss.UnixGroupId != 0 || ss.UnixGroup != "" {
attrs = append(attrs, "owner")
}
if ss.UnixMtime != "" || ss.UnixCtime != "" || ss.UnixAtime != "" {
attrs = append(attrs, "time")
}
log.Printf("file %s blobref is %s, size %d", blobRef, blobref.FromHash("sha1", sha1), n)
err = execSQL(client,
"INSERT IGNORE INTO files (fileschemaref, bytesref, size, filename, mime, setattrs) VALUES (?, ?, ?, ?, ?, ?)",
blobRef.String(),
blobref.FromHash("sha1", sha1).String(),
n,
ss.FileNameString(),
mime,
strings.Join(attrs, ","))
return
}
func blobDetails(contents io.ReadSeeker) (bref *blobref.BlobRef, size int64, err os.Error) {
s1 := sha1.New()
contents.Seek(0, 0)
size, err = io.Copy(s1, contents)
if err == nil {
bref = blobref.FromHash("sha1", s1)
}
return
}
func (d *defaultStatHasher) Hash(fileName string) (*blobref.BlobRef, os.Error) {
s1 := sha1.New()
file, err := os.Open(fileName, os.O_RDONLY, 0)
if err != nil {
return nil, err
}
defer file.Close()
_, err = io.Copy(s1, file)
if err != nil {
return nil, err
}
return blobref.FromHash("sha1", s1), nil
}
func (mi *Indexer) populateFile(blobRef *blobref.BlobRef, ss *schema.Superset) (err os.Error) {
seekFetcher, err := blobref.SeekerFromStreamingFetcher(mi.BlobSource)
if err != nil {
return err
}
sha1 := sha1.New()
fr, err := ss.NewFileReader(seekFetcher)
if err != nil {
log.Printf("mysqlindex: error indexing file %s: %v", blobRef, err)
return nil
}
mime, reader := magic.MimeTypeFromReader(fr)
n, err := io.Copy(sha1, reader)
if err != nil {
// TODO: job scheduling system to retry this spaced
// out max n times. Right now our options are
// ignoring this error (forever) or returning the
// error and making the indexing try again (likely
// forever failing). Both options suck. For now just
// log and act like all's okay.
log.Printf("mysqlindex: error indexing file %s: %v", blobRef, err)
return nil
}
log.Printf("file %s blobref is %s, size %d", blobRef, blobref.FromHash("sha1", sha1), n)
err = mi.db.Execute(
"INSERT IGNORE INTO bytesfiles (schemaref, camlitype, wholedigest, size, filename, mime) VALUES (?, ?, ?, ?, ?, ?)",
blobRef.String(),
"file",
blobref.FromHash("sha1", sha1).String(),
n,
ss.FileNameString(),
mime,
)
return
}
func (up *Uploader) UploadFileBlob(filename string) (*client.PutResult, os.Error) {
var (
err os.Error
size int64
ref *blobref.BlobRef
body io.Reader
)
if filename == "-" {
buf := bytes.NewBuffer(make([]byte, 0))
size, err = io.Copy(buf, os.Stdin)
if err != nil {
return nil, err
}
// TODO(bradfitz,mpl): limit this buffer size?
file := buf.Bytes()
s1 := sha1.New()
size, err = io.Copy(s1, buf)
if err != nil {
return nil, err
}
ref = blobref.FromHash("sha1", s1)
body = io.LimitReader(bytes.NewBuffer(file), size)
} else {
fi, err := os.Stat(filename)
if err != nil {
return nil, err
}
if !fi.IsRegular() {
return nil, fmt.Errorf("%q is not a regular file", filename)
}
file, err := os.Open(filename)
if err != nil {
return nil, err
}
ref, size, err = blobDetails(file)
if err != nil {
return nil, err
}
file.Seek(0, 0)
body = io.LimitReader(file, size)
}
handle := &client.UploadHandle{ref, size, body}
return up.Upload(handle)
}
// WriteFileFromReader creates and uploads a "file" JSON schema
// composed of chunks of r, also uploading the chunks. The returned
// BlobRef is of the JSON file schema blob.
func WriteFileFromReader(bs blobserver.Storage, filename string, r io.Reader) (*blobref.BlobRef, os.Error) {
// Naive for now. Just in 1MB chunks.
// TODO: rolling hash and hash trees.
parts, size := []ContentPart{}, int64(0)
buf := new(bytes.Buffer)
for {
buf.Reset()
n, err := io.Copy(buf, io.LimitReader(r, 1<<20))
if err != nil {
return nil, err
}
if n == 0 {
break
}
hash := crypto.SHA1.New()
io.Copy(hash, bytes.NewBuffer(buf.Bytes()))
br := blobref.FromHash("sha1", hash)
hasBlob, err := serverHasBlob(bs, br)
if err != nil {
return nil, err
}
if !hasBlob {
sb, err := bs.ReceiveBlob(br, buf)
if err != nil {
return nil, err
}
if expect := (blobref.SizedBlobRef{br, n}); !expect.Equal(sb) {
return nil, fmt.Errorf("schema/filewriter: wrote %s bytes, got %s ack'd", expect, sb)
}
}
size += n
parts = append(parts, ContentPart{
BlobRef: br,
Size: uint64(n),
Offset: 0, // into BlobRef to read from (not of dest)
})
}
m := NewCommonFilenameMap(filename)
err := PopulateRegularFileMap(m, size, parts)
if err != nil {
return nil, err
}
json, err := MapToCamliJson(m)
if err != nil {
return nil, err
}
br := blobref.Sha1FromString(json)
sb, err := bs.ReceiveBlob(br, strings.NewReader(json))
if err != nil {
return nil, err
}
if expect := (blobref.SizedBlobRef{br, int64(len(json))}); !expect.Equal(sb) {
return nil, fmt.Errorf("schema/filewriter: wrote %s bytes, got %s ack'd", expect, sb)
}
return br, nil
}
func (tb *testBlob) BlobRef() *blobref.BlobRef {
h := sha1.New()
h.Write([]byte(tb.val))
return blobref.FromHash("sha1", h)
}
func doInit() {
blobDir := path.Join(osutil.CamliConfigDir(), "keyblobs")
os.Mkdir(osutil.CamliConfigDir(), 0700)
os.Mkdir(blobDir, 0700)
keyId := *flagGpgKey
if keyId == "" {
keyId = os.Getenv("GPGKEY")
}
if keyId == "" {
// TODO: run and parse gpg --list-secret-keys and see if there's just one and suggest that? Or show
// a list of them?
log.Fatalf("Initialization requires your public GPG key. Set --gpgkey=<pubid> or set $GPGKEY in your environment. Run gpg --list-secret-keys to find their key IDs.")
}
if os.Getenv("GPG_AGENT_INFO") == "" {
log.Printf("No GPG_AGENT_INFO found in environment; you should setup gnupg-agent. camput will be annoying otherwise.")
}
// TODO: use same command-line flag as the jsonsign package.
// unify them into a shared package just for gpg-related stuff?
keyBytes, err := exec.Command("gpg", "--export", "--armor", keyId).Output()
if err != nil {
log.Fatalf("Error running gpg to export public key: %v", err)
}
hash := sha1.New()
hash.Write(keyBytes)
bref := blobref.FromHash("sha1", hash)
keyBlobPath := path.Join(blobDir, bref.String()+".camli")
if err = ioutil.WriteFile(keyBlobPath, keyBytes, 0644); err != nil {
log.Fatalf("Error writing public key blob to %q: %v", keyBlobPath, err)
}
if ok, err := jsonsign.VerifyPublicKeyFile(keyBlobPath, keyId); !ok {
log.Fatalf("Error verifying public key at %q: %v", keyBlobPath, err)
}
log.Printf("Your Camlistore identity (your GPG public key's blobref) is: %s", bref.String())
_, err = os.Stat(client.ConfigFilePath())
if err == nil {
log.Fatalf("Config file %q already exists; quitting without touching it.", client.ConfigFilePath())
}
if f, err := os.OpenFile(client.ConfigFilePath(), os.O_CREATE|os.O_EXCL|os.O_WRONLY, 0600); err == nil {
defer f.Close()
m := make(map[string]interface{})
m["publicKeyBlobref"] = bref.String()
blobPut := make([]map[string]string, 1)
blobPut[0] = map[string]string{
"alias": "local",
"host": "http://localhost:3179/",
"password": "******",
}
m["blobPut"] = blobPut
blobGet := make([]map[string]string, 2)
blobGet[0] = map[string]string{
"alias": "keyblobs",
"path": "$HOME/.camli/keyblobs",
}
blobGet[1] = map[string]string{
"alias": "local",
"host": "http://localhost:3179/",
"password": "******",
}
m["blobGet"] = blobGet
jsonBytes, err := json.MarshalIndent(m, "", " ")
if err != nil {
log.Fatalf("JSON serialization error: %v", err)
}
_, err = f.Write(jsonBytes)
if err != nil {
log.Fatalf("Error writing to %q: %v", client.ConfigFilePath(), err)
}
log.Printf("Wrote %q; modify as necessary.", client.ConfigFilePath())
}
}
func (tb *Blob) BlobRef() *blobref.BlobRef {
h := sha1.New()
h.Write([]byte(tb.Contents))
return blobref.FromHash("sha1", h)
}
