FedP2P/bencode/decode.go

671 lines
14 KiB
Go
Raw Normal View History

2012-06-20 21:21:32 +08:00
package bencode
import (
"bytes"
"errors"
"fmt"
"io"
"math/big"
"reflect"
"runtime"
"strconv"
2018-07-23 08:50:58 +08:00
"sync"
)
2012-06-20 21:21:32 +08:00
type Decoder struct {
r interface {
io.ByteScanner
io.Reader
}
2017-11-05 12:45:30 +08:00
// Sum of bytes used to Decode values.
2017-11-05 12:42:37 +08:00
Offset int64
2012-06-20 21:21:32 +08:00
buf bytes.Buffer
}
func (d *Decoder) Decode(v interface{}) (err error) {
2012-06-20 21:34:10 +08:00
defer func() {
if err != nil {
return
}
r := recover()
_, ok := r.(runtime.Error)
if ok {
panic(r)
}
err, ok = r.(error)
if !ok && r != nil {
panic(r)
2012-06-20 21:34:10 +08:00
}
}()
2012-06-20 21:21:32 +08:00
pv := reflect.ValueOf(v)
if pv.Kind() != reflect.Ptr || pv.IsNil() {
return &UnmarshalInvalidArgError{reflect.TypeOf(v)}
}
ok, err := d.parseValue(pv.Elem())
if err != nil {
return
}
if !ok {
2017-11-05 12:42:37 +08:00
d.throwSyntaxError(d.Offset-1, errors.New("unexpected 'e'"))
}
return
2012-06-20 21:21:32 +08:00
}
func checkForUnexpectedEOF(err error, offset int64) {
2012-06-20 21:21:32 +08:00
if err == io.EOF {
panic(&SyntaxError{
Offset: offset,
What: io.ErrUnexpectedEOF,
2012-06-20 21:21:32 +08:00
})
}
}
func (d *Decoder) readByte() byte {
b, err := d.r.ReadByte()
2012-06-20 21:21:32 +08:00
if err != nil {
2017-11-05 12:42:37 +08:00
checkForUnexpectedEOF(err, d.Offset)
2012-06-20 21:21:32 +08:00
panic(err)
}
2017-11-05 12:42:37 +08:00
d.Offset++
2012-06-20 21:21:32 +08:00
return b
}
// reads data writing it to 'd.buf' until 'sep' byte is encountered, 'sep' byte
// is consumed, but not included into the 'd.buf'
func (d *Decoder) readUntil(sep byte) {
2012-06-20 21:21:32 +08:00
for {
b := d.readByte()
2012-06-20 21:21:32 +08:00
if b == sep {
return
}
d.buf.WriteByte(b)
}
}
func checkForIntParseError(err error, offset int64) {
2012-06-20 21:21:32 +08:00
if err != nil {
panic(&SyntaxError{
Offset: offset,
What: err,
2012-06-20 21:21:32 +08:00
})
}
}
func (d *Decoder) throwSyntaxError(offset int64, err error) {
panic(&SyntaxError{
Offset: offset,
What: err,
})
}
2012-06-20 21:21:32 +08:00
// called when 'i' was consumed
2016-08-26 12:51:38 +08:00
func (d *Decoder) parseInt(v reflect.Value) {
2017-11-05 12:42:37 +08:00
start := d.Offset - 1
d.readUntil('e')
2012-06-20 21:21:32 +08:00
if d.buf.Len() == 0 {
panic(&SyntaxError{
Offset: start,
What: errors.New("empty integer value"),
2012-06-20 21:21:32 +08:00
})
}
s := bytesAsString(d.buf.Bytes())
2015-08-23 15:51:24 +08:00
2012-06-20 21:21:32 +08:00
switch v.Kind() {
case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
2015-08-23 15:51:24 +08:00
n, err := strconv.ParseInt(s, 10, 64)
checkForIntParseError(err, start)
2012-06-20 21:21:32 +08:00
if v.OverflowInt(n) {
panic(&UnmarshalTypeError{
2015-08-23 15:51:24 +08:00
Value: "integer " + s,
2012-06-20 21:21:32 +08:00
Type: v.Type(),
})
}
v.SetInt(n)
case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64:
2015-08-23 15:51:24 +08:00
n, err := strconv.ParseUint(s, 10, 64)
checkForIntParseError(err, start)
2012-06-20 21:21:32 +08:00
if v.OverflowUint(n) {
panic(&UnmarshalTypeError{
2015-08-23 15:51:24 +08:00
Value: "integer " + s,
2012-06-20 21:21:32 +08:00
Type: v.Type(),
})
}
v.SetUint(n)
case reflect.Bool:
2015-08-23 15:51:24 +08:00
v.SetBool(s != "0")
2012-06-20 21:21:32 +08:00
default:
panic(&UnmarshalTypeError{
2015-08-23 15:51:24 +08:00
Value: "integer " + s,
2012-06-20 21:21:32 +08:00
Type: v.Type(),
})
}
d.buf.Reset()
}
func (d *Decoder) parseString(v reflect.Value) error {
2017-11-05 12:42:37 +08:00
start := d.Offset - 1
2012-06-20 21:21:32 +08:00
// read the string length first
d.readUntil(':')
length, err := strconv.ParseInt(bytesAsString(d.buf.Bytes()), 10, 0)
checkForIntParseError(err, start)
2012-06-20 21:21:32 +08:00
defer d.buf.Reset()
read := func(b []byte) {
n, err := io.ReadFull(d.r, b)
d.Offset += int64(n)
if err != nil {
checkForUnexpectedEOF(err, d.Offset)
panic(&SyntaxError{
Offset: d.Offset,
What: errors.New("unexpected I/O error: " + err.Error()),
})
}
2012-06-20 21:21:32 +08:00
}
switch v.Kind() {
case reflect.String:
b := make([]byte, length)
read(b)
v.SetString(bytesAsString(b))
return nil
2012-06-20 21:21:32 +08:00
case reflect.Slice:
if v.Type().Elem().Kind() != reflect.Uint8 {
break
2012-06-20 21:21:32 +08:00
}
b := make([]byte, length)
read(b)
v.SetBytes(b)
return nil
case reflect.Array:
if v.Type().Elem().Kind() != reflect.Uint8 {
break
}
d.buf.Grow(int(length))
b := d.buf.Bytes()[:length]
read(b)
reflect.Copy(v, reflect.ValueOf(b))
return nil
}
d.buf.Grow(int(length))
read(d.buf.Bytes()[:length])
// I believe we return here to support "ignore_unmarshal_type_error".
return &UnmarshalTypeError{
Value: "string",
Type: v.Type(),
2012-06-20 21:21:32 +08:00
}
}
// Info for parsing a dict value.
type dictField struct {
Type reflect.Type
Get func(value reflect.Value) func(reflect.Value)
Tags tag
}
// Returns specifics for parsing a dict field value.
func getDictField(dict reflect.Type, key string) dictField {
// get valuev as a map value or as a struct field
switch dict.Kind() {
case reflect.Map:
return dictField{
Type: dict.Elem(),
Get: func(mapValue reflect.Value) func(reflect.Value) {
return func(value reflect.Value) {
if mapValue.IsNil() {
mapValue.Set(reflect.MakeMap(dict))
}
// Assigns the value into the map.
//log.Printf("map type: %v", mapValue.Type())
mapValue.SetMapIndex(reflect.ValueOf(key).Convert(dict.Key()), value)
}
},
}
case reflect.Struct:
return getStructFieldForKey(dict, key)
//if sf.r.PkgPath != "" {
// panic(&UnmarshalFieldError{
// Key: key,
// Type: dict.Type(),
// Field: sf.r,
// })
//}
default:
panic("unimplemented")
return dictField{}
}
}
2018-07-23 08:50:58 +08:00
var (
structFieldsMu sync.Mutex
structFields = map[reflect.Type]map[string]dictField{}
2018-07-23 08:50:58 +08:00
)
func parseStructFields(struct_ reflect.Type, each func(key string, df dictField)) {
for _i, n := 0, struct_.NumField(); _i < n; _i++ {
i := _i
2018-07-23 08:50:58 +08:00
f := struct_.Field(i)
2018-07-23 09:09:42 +08:00
if f.Anonymous {
parseStructFields(f.Type.Elem(), func(key string, df dictField) {
innerGet := df.Get
df.Get = func(value reflect.Value) func(reflect.Value) {
anonPtr := value.Field(i)
if anonPtr.IsNil() {
anonPtr.Set(reflect.New(f.Type.Elem()))
}
return innerGet(anonPtr.Elem())
}
each(key, df)
})
continue
}
2018-07-23 09:09:42 +08:00
tagStr := f.Tag.Get("bencode")
if tagStr == "-" {
continue
}
2018-07-23 09:09:42 +08:00
tag := parseTag(tagStr)
key := tag.Key()
if key == "" {
key = f.Name
}
each(key, dictField{f.Type, func(value reflect.Value) func(reflect.Value) {
return value.Field(i).Set
}, tag})
2018-07-23 08:50:58 +08:00
}
}
2018-07-23 08:50:58 +08:00
func saveStructFields(struct_ reflect.Type) {
m := make(map[string]dictField)
parseStructFields(struct_, func(key string, sf dictField) {
2018-07-23 08:50:58 +08:00
m[key] = sf
})
structFields[struct_] = m
}
func getStructFieldForKey(struct_ reflect.Type, key string) (f dictField) {
2018-07-23 08:50:58 +08:00
structFieldsMu.Lock()
if _, ok := structFields[struct_]; !ok {
saveStructFields(struct_)
}
f, ok := structFields[struct_][key]
2018-07-23 08:50:58 +08:00
structFieldsMu.Unlock()
if !ok {
var discard interface{}
return dictField{
Type: reflect.TypeOf(discard),
Get: func(reflect.Value) func(reflect.Value) { return func(reflect.Value) {} },
Tags: nil,
}
}
return
}
func (d *Decoder) parseDict(v reflect.Value) error {
2012-06-20 21:21:32 +08:00
// so, at this point 'd' byte was consumed, let's just read key/value
// pairs one by one
for {
var keyStr string
keyValue := reflect.ValueOf(&keyStr).Elem()
ok, err := d.parseValue(keyValue)
if err != nil {
return fmt.Errorf("error parsing dict key: %s", err)
}
if !ok {
return nil
2012-06-20 21:21:32 +08:00
}
df := getDictField(v.Type(), keyStr)
2012-06-20 21:21:32 +08:00
// now we need to actually parse it
if df.Type == nil {
// Discard the value, there's nowhere to put it.
var if_ interface{}
if_, ok = d.parseValueInterface()
if if_ == nil {
return fmt.Errorf("error parsing value for key %q", keyStr)
2012-06-20 21:21:32 +08:00
}
if !ok {
return fmt.Errorf("missing value for key %q", keyStr)
}
continue
2012-06-20 21:21:32 +08:00
}
setValue := reflect.New(df.Type).Elem()
//log.Printf("parsing into %v", setValue.Type())
ok, err = d.parseValue(setValue)
if err != nil {
if _, ok := err.(*UnmarshalTypeError); !ok || !df.Tags.IgnoreUnmarshalTypeError() {
return fmt.Errorf("parsing value for key %q: %s", keyStr, err)
}
}
if !ok {
return fmt.Errorf("missing value for key %q", keyStr)
2012-06-20 21:21:32 +08:00
}
df.Get(v)(setValue)
2012-06-20 21:21:32 +08:00
}
}
func (d *Decoder) parseList(v reflect.Value) error {
2012-06-20 21:21:32 +08:00
switch v.Kind() {
default:
// If the list is a singleton of the expected type, use that value. See
// https://github.com/anacrolix/torrent/issues/297.
l := reflect.New(reflect.SliceOf(v.Type()))
if err := d.parseList(l.Elem()); err != nil {
return err
}
if l.Elem().Len() != 1 {
return &UnmarshalTypeError{
Value: "list",
Type: v.Type(),
}
}
v.Set(l.Elem().Index(0))
return nil
case reflect.Array, reflect.Slice:
// We can work with this. Normal case, fallthrough.
2012-06-20 21:21:32 +08:00
}
i := 0
for ; ; i++ {
2012-06-20 21:21:32 +08:00
if v.Kind() == reflect.Slice && i >= v.Len() {
v.Set(reflect.Append(v, reflect.Zero(v.Type().Elem())))
}
if i < v.Len() {
ok, err := d.parseValue(v.Index(i))
if err != nil {
return err
}
if !ok {
break
}
2012-06-20 21:21:32 +08:00
} else {
_, ok := d.parseValueInterface()
if !ok {
break
}
2012-06-20 21:21:32 +08:00
}
}
if i < v.Len() {
if v.Kind() == reflect.Array {
z := reflect.Zero(v.Type().Elem())
for n := v.Len(); i < n; i++ {
v.Index(i).Set(z)
}
} else {
v.SetLen(i)
}
}
if i == 0 && v.Kind() == reflect.Slice {
v.Set(reflect.MakeSlice(v.Type(), 0, 0))
}
return nil
2012-06-20 21:21:32 +08:00
}
func (d *Decoder) readOneValue() bool {
b, err := d.r.ReadByte()
if err != nil {
panic(err)
}
if b == 'e' {
d.r.UnreadByte()
return false
} else {
2017-11-05 12:42:37 +08:00
d.Offset++
d.buf.WriteByte(b)
}
switch b {
case 'd', 'l':
// read until there is nothing to read
for d.readOneValue() {
}
// consume 'e' as well
b = d.readByte()
d.buf.WriteByte(b)
case 'i':
d.readUntil('e')
d.buf.WriteString("e")
default:
if b >= '0' && b <= '9' {
start := d.buf.Len() - 1
d.readUntil(':')
length, err := strconv.ParseInt(bytesAsString(d.buf.Bytes()[start:]), 10, 64)
2017-11-05 12:42:37 +08:00
checkForIntParseError(err, d.Offset-1)
d.buf.WriteString(":")
n, err := io.CopyN(&d.buf, d.r, length)
2017-11-05 12:42:37 +08:00
d.Offset += n
if err != nil {
2017-11-05 12:42:37 +08:00
checkForUnexpectedEOF(err, d.Offset)
panic(&SyntaxError{
2017-11-05 12:42:37 +08:00
Offset: d.Offset,
What: errors.New("unexpected I/O error: " + err.Error()),
})
}
break
}
2017-11-05 12:42:37 +08:00
d.raiseUnknownValueType(b, d.Offset-1)
}
return true
}
func (d *Decoder) parseUnmarshaler(v reflect.Value) bool {
if !v.Type().Implements(unmarshalerType) {
if v.Addr().Type().Implements(unmarshalerType) {
v = v.Addr()
} else {
return false
}
}
d.buf.Reset()
if !d.readOneValue() {
return false
}
m := v.Interface().(Unmarshaler)
err := m.UnmarshalBencode(d.buf.Bytes())
if err != nil {
panic(&UnmarshalerError{v.Type(), err})
}
return true
}
// Returns true if there was a value and it's now stored in 'v', otherwise
// there was an end symbol ("e") and no value was stored.
func (d *Decoder) parseValue(v reflect.Value) (bool, error) {
// we support one level of indirection at the moment
if v.Kind() == reflect.Ptr {
2012-06-20 21:21:32 +08:00
// if the pointer is nil, allocate a new element of the type it
// points to
if v.IsNil() {
v.Set(reflect.New(v.Type().Elem()))
2012-06-20 21:21:32 +08:00
}
v = v.Elem()
}
if d.parseUnmarshaler(v) {
return true, nil
2012-06-20 21:21:32 +08:00
}
// common case: interface{}
if v.Kind() == reflect.Interface && v.NumMethod() == 0 {
iface, _ := d.parseValueInterface()
2012-06-20 21:21:32 +08:00
v.Set(reflect.ValueOf(iface))
return true, nil
2012-06-20 21:21:32 +08:00
}
b, err := d.r.ReadByte()
2012-06-20 21:21:32 +08:00
if err != nil {
panic(err)
}
2017-11-05 12:42:37 +08:00
d.Offset++
2012-06-20 21:21:32 +08:00
switch b {
case 'e':
return false, nil
2012-06-20 21:21:32 +08:00
case 'd':
return true, d.parseDict(v)
2012-06-20 21:21:32 +08:00
case 'l':
return true, d.parseList(v)
2012-06-20 21:21:32 +08:00
case 'i':
2016-08-26 12:51:38 +08:00
d.parseInt(v)
return true, nil
2012-06-20 21:21:32 +08:00
default:
if b >= '0' && b <= '9' {
2018-04-13 15:07:19 +08:00
// It's a string.
d.buf.Reset()
2018-07-23 08:50:58 +08:00
// Write the first digit of the length to the buffer.
2012-06-20 21:21:32 +08:00
d.buf.WriteByte(b)
return true, d.parseString(v)
2012-06-20 21:21:32 +08:00
}
2017-11-05 12:42:37 +08:00
d.raiseUnknownValueType(b, d.Offset-1)
2012-06-20 21:21:32 +08:00
}
panic("unreachable")
2012-06-20 21:21:32 +08:00
}
// An unknown bencode type character was encountered.
func (d *Decoder) raiseUnknownValueType(b byte, offset int64) {
panic(&SyntaxError{
Offset: offset,
What: fmt.Errorf("unknown value type %+q", b),
})
}
func (d *Decoder) parseValueInterface() (interface{}, bool) {
b, err := d.r.ReadByte()
2012-06-20 21:21:32 +08:00
if err != nil {
panic(err)
}
2017-11-05 12:42:37 +08:00
d.Offset++
2012-06-20 21:21:32 +08:00
switch b {
case 'e':
return nil, false
case 'd':
return d.parseDictInterface(), true
2012-06-20 21:21:32 +08:00
case 'l':
return d.parseListInterface(), true
2012-06-20 21:21:32 +08:00
case 'i':
return d.parseIntInterface(), true
2012-06-20 21:21:32 +08:00
default:
if b >= '0' && b <= '9' {
// string
// append first digit of the length to the buffer
d.buf.WriteByte(b)
return d.parseStringInterface(), true
2012-06-20 21:21:32 +08:00
}
2017-11-05 12:42:37 +08:00
d.raiseUnknownValueType(b, d.Offset-1)
panic("unreachable")
2012-06-20 21:21:32 +08:00
}
}
func (d *Decoder) parseIntInterface() (ret interface{}) {
2017-11-05 12:42:37 +08:00
start := d.Offset - 1
d.readUntil('e')
2012-06-20 21:21:32 +08:00
if d.buf.Len() == 0 {
panic(&SyntaxError{
Offset: start,
What: errors.New("empty integer value"),
2012-06-20 21:21:32 +08:00
})
}
n, err := strconv.ParseInt(d.buf.String(), 10, 64)
if ne, ok := err.(*strconv.NumError); ok && ne.Err == strconv.ErrRange {
i := new(big.Int)
_, ok := i.SetString(d.buf.String(), 10)
if !ok {
panic(&SyntaxError{
Offset: start,
What: errors.New("failed to parse integer"),
})
}
ret = i
} else {
checkForIntParseError(err, start)
ret = n
}
2012-06-20 21:21:32 +08:00
d.buf.Reset()
return
2012-06-20 21:21:32 +08:00
}
func (d *Decoder) parseStringInterface() interface{} {
2017-11-05 12:42:37 +08:00
start := d.Offset - 1
2012-06-20 21:21:32 +08:00
// read the string length first
d.readUntil(':')
2012-06-20 21:21:32 +08:00
length, err := strconv.ParseInt(d.buf.String(), 10, 64)
checkForIntParseError(err, start)
2012-06-20 21:21:32 +08:00
d.buf.Reset()
n, err := io.CopyN(&d.buf, d.r, length)
2017-11-05 12:42:37 +08:00
d.Offset += n
2012-06-20 21:21:32 +08:00
if err != nil {
2017-11-05 12:42:37 +08:00
checkForUnexpectedEOF(err, d.Offset)
2012-06-20 21:21:32 +08:00
panic(&SyntaxError{
2017-11-05 12:42:37 +08:00
Offset: d.Offset,
What: errors.New("unexpected I/O error: " + err.Error()),
2012-06-20 21:21:32 +08:00
})
}
s := d.buf.String()
d.buf.Reset()
return s
}
func (d *Decoder) parseDictInterface() interface{} {
2012-06-20 21:21:32 +08:00
dict := make(map[string]interface{})
for {
keyi, ok := d.parseValueInterface()
2012-06-20 21:21:32 +08:00
if !ok {
break
}
key, ok := keyi.(string)
if !ok {
panic(&SyntaxError{
2017-11-05 12:42:37 +08:00
Offset: d.Offset,
What: errors.New("non-string key in a dict"),
2012-06-20 21:21:32 +08:00
})
}
valuei, ok := d.parseValueInterface()
2012-06-20 21:21:32 +08:00
if !ok {
break
2012-06-20 21:21:32 +08:00
}
dict[key] = valuei
}
return dict
}
func (d *Decoder) parseListInterface() interface{} {
2012-06-20 21:21:32 +08:00
var list []interface{}
for {
valuei, ok := d.parseValueInterface()
2012-06-20 21:21:32 +08:00
if !ok {
break
}
list = append(list, valuei)
}
if list == nil {
list = make([]interface{}, 0, 0)
}
2012-06-20 21:21:32 +08:00
return list
}