1
0
Fork 1
mirror of https://github.com/vbatts/tar-split.git synced 2024-12-29 16:46:31 +00:00
tar-split/tar/storage/packer.go

141 lines
3.1 KiB
Go
Raw Normal View History

package storage
import (
"bufio"
"encoding/json"
"errors"
"io"
"path"
)
2015-03-09 18:11:11 +00:00
// ErrDuplicatePath is occured when a tar archive has more than one entry for
// the same file path
var ErrDuplicatePath = errors.New("duplicates of file paths not supported")
2015-02-24 20:36:21 +00:00
// Packer describes the methods to pack Entries to a storage destination
type Packer interface {
2015-02-24 20:36:21 +00:00
// AddEntry packs the Entry and returns its position
AddEntry(e Entry) (int, error)
}
2015-02-24 20:36:21 +00:00
// Unpacker describes the methods to read Entries from a source
type Unpacker interface {
2015-02-24 20:36:21 +00:00
// Next returns the next Entry being unpacked, or error, until io.EOF
Next() (*Entry, error)
}
2015-02-24 20:36:21 +00:00
/* TODO(vbatts) figure out a good model for this
type PackUnpacker interface {
Packer
Unpacker
}
2015-02-24 20:36:21 +00:00
*/
type jsonUnpacker struct {
r io.Reader
b *bufio.Reader
isEOF bool
seen seenNames
}
func (jup *jsonUnpacker) Next() (*Entry, error) {
var e Entry
if jup.isEOF {
// since ReadBytes() will return read bytes AND an EOF, we handle it this
// round-a-bout way so we can Unmarshal the tail with relevant errors, but
// still get an io.EOF when the stream is ended.
return nil, io.EOF
}
line, err := jup.b.ReadBytes('\n')
if err != nil && err != io.EOF {
return nil, err
} else if err == io.EOF {
jup.isEOF = true
}
err = json.Unmarshal(line, &e)
if err != nil && jup.isEOF {
// if the remainder actually _wasn't_ a remaining json structure, then just EOF
return nil, io.EOF
}
// check for dup name
if e.Type == FileType {
cName := path.Clean(e.Name)
if _, ok := jup.seen[cName]; ok {
return nil, ErrDuplicatePath
}
jup.seen[cName] = emptyByte
}
return &e, err
}
2015-03-09 18:11:11 +00:00
// NewJSONUnpacker provides an Unpacker that reads Entries (SegmentType and
2015-02-24 20:36:21 +00:00
// FileType) as a json document.
//
// Each Entry read are expected to be delimited by new line.
2015-03-09 18:11:11 +00:00
func NewJSONUnpacker(r io.Reader) Unpacker {
return &jsonUnpacker{
r: r,
b: bufio.NewReader(r),
seen: seenNames{},
}
}
type jsonPacker struct {
w io.Writer
e *json.Encoder
pos int
seen seenNames
}
type seenNames map[string]byte
// used in the seenNames map. byte is a uint8, and we'll re-use the same one
// for minimalism.
const emptyByte byte = 0
func (jp *jsonPacker) AddEntry(e Entry) (int, error) {
// check early for dup name
if e.Type == FileType {
cName := path.Clean(e.Name)
if _, ok := jp.seen[cName]; ok {
return -1, ErrDuplicatePath
}
jp.seen[cName] = emptyByte
}
e.Position = jp.pos
err := jp.e.Encode(e)
if err != nil {
return -1, err
}
// made it this far, increment now
jp.pos++
return e.Position, nil
}
2015-03-09 18:11:11 +00:00
// NewJSONPacker provides an Packer that writes each Entry (SegmentType and
2015-02-24 20:36:21 +00:00
// FileType) as a json document.
//
// The Entries are delimited by new line.
2015-03-09 18:11:11 +00:00
func NewJSONPacker(w io.Writer) Packer {
return &jsonPacker{
w: w,
e: json.NewEncoder(w),
seen: seenNames{},
}
}
2015-02-24 20:49:06 +00:00
/*
TODO(vbatts) perhaps have a more compact packer/unpacker, maybe using msgapck
(https://github.com/ugorji/go)
Even though, since our jsonUnpacker and jsonPacker just take
io.Reader/io.Writer, then we can get away with passing them a
gzip.Reader/gzip.Writer
*/