stream.go

Documentation: encoding/json

     1  // Copyright 2010 The Go Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  package json
     6  
     7  import (
     8  	"bytes"
     9  	"errors"
    10  	"io"
    11  )
    12  
    13  // A Decoder reads and decodes JSON values from an input stream.
    14  type Decoder struct {
    15  	r       io.Reader
    16  	buf     []byte
    17  	d       decodeState
    18  	scanp   int   // start of unread data in buf
    19  	scanned int64 // amount of data already scanned
    20  	scan    scanner
    21  	err     error
    22  
    23  	tokenState int
    24  	tokenStack []int
    25  }
    26  
    27  // NewDecoder returns a new decoder that reads from r.
    28  //
    29  // The decoder introduces its own buffering and may
    30  // read data from r beyond the JSON values requested.
    31  func NewDecoder(r io.Reader) *Decoder {
    32  	return &Decoder{r: r}
    33  }
    34  
    35  // UseNumber causes the Decoder to unmarshal a number into an interface{} as a
    36  // [Number] instead of as a float64.
    37  func (dec *Decoder) UseNumber() { dec.d.useNumber = true }
    38  
    39  // DisallowUnknownFields causes the Decoder to return an error when the destination
    40  // is a struct and the input contains object keys which do not match any
    41  // non-ignored, exported fields in the destination.
    42  func (dec *Decoder) DisallowUnknownFields() { dec.d.disallowUnknownFields = true }
    43  
    44  // Decode reads the next JSON-encoded value from its
    45  // input and stores it in the value pointed to by v.
    46  //
    47  // See the documentation for [Unmarshal] for details about
    48  // the conversion of JSON into a Go value.
    49  func (dec *Decoder) Decode(v any) error {
    50  	if dec.err != nil {
    51  		return dec.err
    52  	}
    53  
    54  	if err := dec.tokenPrepareForDecode(); err != nil {
    55  		return err
    56  	}
    57  
    58  	if !dec.tokenValueAllowed() {
    59  		return &SyntaxError{msg: "not at beginning of value", Offset: dec.InputOffset()}
    60  	}
    61  
    62  	// Read whole value into buffer.
    63  	n, err := dec.readValue()
    64  	if err != nil {
    65  		return err
    66  	}
    67  	dec.d.init(dec.buf[dec.scanp : dec.scanp+n])
    68  	dec.scanp += n
    69  
    70  	// Don't save err from unmarshal into dec.err:
    71  	// the connection is still usable since we read a complete JSON
    72  	// object from it before the error happened.
    73  	err = dec.d.unmarshal(v)
    74  
    75  	// fixup token streaming state
    76  	dec.tokenValueEnd()
    77  
    78  	return err
    79  }
    80  
    81  // Buffered returns a reader of the data remaining in the Decoder's
    82  // buffer. The reader is valid until the next call to [Decoder.Decode].
    83  func (dec *Decoder) Buffered() io.Reader {
    84  	return bytes.NewReader(dec.buf[dec.scanp:])
    85  }
    86  
    87  // readValue reads a JSON value into dec.buf.
    88  // It returns the length of the encoding.
    89  func (dec *Decoder) readValue() (int, error) {
    90  	dec.scan.reset()
    91  
    92  	scanp := dec.scanp
    93  	var err error
    94  Input:
    95  	// help the compiler see that scanp is never negative, so it can remove
    96  	// some bounds checks below.
    97  	for scanp >= 0 {
    98  
    99  		// Look in the buffer for a new value.
   100  		for ; scanp < len(dec.buf); scanp++ {
   101  			c := dec.buf[scanp]
   102  			dec.scan.bytes++
   103  			switch dec.scan.step(&dec.scan, c) {
   104  			case scanEnd:
   105  				// scanEnd is delayed one byte so we decrement
   106  				// the scanner bytes count by 1 to ensure that
   107  				// this value is correct in the next call of Decode.
   108  				dec.scan.bytes--
   109  				break Input
   110  			case scanEndObject, scanEndArray:
   111  				// scanEnd is delayed one byte.
   112  				// We might block trying to get that byte from src,
   113  				// so instead invent a space byte.
   114  				if stateEndValue(&dec.scan, ' ') == scanEnd {
   115  					scanp++
   116  					break Input
   117  				}
   118  			case scanError:
   119  				dec.err = dec.scan.err
   120  				return 0, dec.scan.err
   121  			}
   122  		}
   123  
   124  		// Did the last read have an error?
   125  		// Delayed until now to allow buffer scan.
   126  		if err != nil {
   127  			if err == io.EOF {
   128  				if dec.scan.step(&dec.scan, ' ') == scanEnd {
   129  					break Input
   130  				}
   131  				if nonSpace(dec.buf) {
   132  					err = io.ErrUnexpectedEOF
   133  				}
   134  			}
   135  			dec.err = err
   136  			return 0, err
   137  		}
   138  
   139  		n := scanp - dec.scanp
   140  		err = dec.refill()
   141  		scanp = dec.scanp + n
   142  	}
   143  	return scanp - dec.scanp, nil
   144  }
   145  
   146  func (dec *Decoder) refill() error {
   147  	// Make room to read more into the buffer.
   148  	// First slide down data already consumed.
   149  	if dec.scanp > 0 {
   150  		dec.scanned += int64(dec.scanp)
   151  		n := copy(dec.buf, dec.buf[dec.scanp:])
   152  		dec.buf = dec.buf[:n]
   153  		dec.scanp = 0
   154  	}
   155  
   156  	// Grow buffer if not large enough.
   157  	const minRead = 512
   158  	if cap(dec.buf)-len(dec.buf) < minRead {
   159  		newBuf := make([]byte, len(dec.buf), 2*cap(dec.buf)+minRead)
   160  		copy(newBuf, dec.buf)
   161  		dec.buf = newBuf
   162  	}
   163  
   164  	// Read. Delay error for next iteration (after scan).
   165  	n, err := dec.r.Read(dec.buf[len(dec.buf):cap(dec.buf)])
   166  	dec.buf = dec.buf[0 : len(dec.buf)+n]
   167  
   168  	return err
   169  }
   170  
   171  func nonSpace(b []byte) bool {
   172  	for _, c := range b {
   173  		if !isSpace(c) {
   174  			return true
   175  		}
   176  	}
   177  	return false
   178  }
   179  
   180  // An Encoder writes JSON values to an output stream.
   181  type Encoder struct {
   182  	w          io.Writer
   183  	err        error
   184  	escapeHTML bool
   185  
   186  	indentBuf    []byte
   187  	indentPrefix string
   188  	indentValue  string
   189  }
   190  
   191  // NewEncoder returns a new encoder that writes to w.
   192  func NewEncoder(w io.Writer) *Encoder {
   193  	return &Encoder{w: w, escapeHTML: true}
   194  }
   195  
   196  // Encode writes the JSON encoding of v to the stream,
   197  // followed by a newline character.
   198  //
   199  // See the documentation for [Marshal] for details about the
   200  // conversion of Go values to JSON.
   201  func (enc *Encoder) Encode(v any) error {
   202  	if enc.err != nil {
   203  		return enc.err
   204  	}
   205  
   206  	e := newEncodeState()
   207  	defer encodeStatePool.Put(e)
   208  
   209  	err := e.marshal(v, encOpts{escapeHTML: enc.escapeHTML})
   210  	if err != nil {
   211  		return err
   212  	}
   213  
   214  	// Terminate each value with a newline.
   215  	// This makes the output look a little nicer
   216  	// when debugging, and some kind of space
   217  	// is required if the encoded value was a number,
   218  	// so that the reader knows there aren't more
   219  	// digits coming.
   220  	e.WriteByte('\n')
   221  
   222  	b := e.Bytes()
   223  	if enc.indentPrefix != "" || enc.indentValue != "" {
   224  		enc.indentBuf, err = appendIndent(enc.indentBuf[:0], b, enc.indentPrefix, enc.indentValue)
   225  		if err != nil {
   226  			return err
   227  		}
   228  		b = enc.indentBuf
   229  	}
   230  	if _, err = enc.w.Write(b); err != nil {
   231  		enc.err = err
   232  	}
   233  	return err
   234  }
   235  
   236  // SetIndent instructs the encoder to format each subsequent encoded
   237  // value as if indented by the package-level function Indent(dst, src, prefix, indent).
   238  // Calling SetIndent("", "") disables indentation.
   239  func (enc *Encoder) SetIndent(prefix, indent string) {
   240  	enc.indentPrefix = prefix
   241  	enc.indentValue = indent
   242  }
   243  
   244  // SetEscapeHTML specifies whether problematic HTML characters
   245  // should be escaped inside JSON quoted strings.
   246  // The default behavior is to escape &, <, and > to \u0026, \u003c, and \u003e
   247  // to avoid certain safety problems that can arise when embedding JSON in HTML.
   248  //
   249  // In non-HTML settings where the escaping interferes with the readability
   250  // of the output, SetEscapeHTML(false) disables this behavior.
   251  func (enc *Encoder) SetEscapeHTML(on bool) {
   252  	enc.escapeHTML = on
   253  }
   254  
   255  // RawMessage is a raw encoded JSON value.
   256  // It implements [Marshaler] and [Unmarshaler] and can
   257  // be used to delay JSON decoding or precompute a JSON encoding.
   258  type RawMessage []byte
   259  
   260  // MarshalJSON returns m as the JSON encoding of m.
   261  func (m RawMessage) MarshalJSON() ([]byte, error) {
   262  	if m == nil {
   263  		return []byte("null"), nil
   264  	}
   265  	return m, nil
   266  }
   267  
   268  // UnmarshalJSON sets *m to a copy of data.
   269  func (m *RawMessage) UnmarshalJSON(data []byte) error {
   270  	if m == nil {
   271  		return errors.New("json.RawMessage: UnmarshalJSON on nil pointer")
   272  	}
   273  	*m = append((*m)[0:0], data...)
   274  	return nil
   275  }
   276  
   277  var _ Marshaler = (*RawMessage)(nil)
   278  var _ Unmarshaler = (*RawMessage)(nil)
   279  
   280  // A Token holds a value of one of these types:
   281  //
   282  //   - [Delim], for the four JSON delimiters [ ] { }
   283  //   - bool, for JSON booleans
   284  //   - float64, for JSON numbers
   285  //   - [Number], for JSON numbers
   286  //   - string, for JSON string literals
   287  //   - nil, for JSON null
   288  type Token any
   289  
   290  const (
   291  	tokenTopValue = iota
   292  	tokenArrayStart
   293  	tokenArrayValue
   294  	tokenArrayComma
   295  	tokenObjectStart
   296  	tokenObjectKey
   297  	tokenObjectColon
   298  	tokenObjectValue
   299  	tokenObjectComma
   300  )
   301  
   302  // advance tokenstate from a separator state to a value state
   303  func (dec *Decoder) tokenPrepareForDecode() error {
   304  	// Note: Not calling peek before switch, to avoid
   305  	// putting peek into the standard Decode path.
   306  	// peek is only called when using the Token API.
   307  	switch dec.tokenState {
   308  	case tokenArrayComma:
   309  		c, err := dec.peek()
   310  		if err != nil {
   311  			return err
   312  		}
   313  		if c != ',' {
   314  			return &SyntaxError{"expected comma after array element", dec.InputOffset()}
   315  		}
   316  		dec.scanp++
   317  		dec.tokenState = tokenArrayValue
   318  	case tokenObjectColon:
   319  		c, err := dec.peek()
   320  		if err != nil {
   321  			return err
   322  		}
   323  		if c != ':' {
   324  			return &SyntaxError{"expected colon after object key", dec.InputOffset()}
   325  		}
   326  		dec.scanp++
   327  		dec.tokenState = tokenObjectValue
   328  	}
   329  	return nil
   330  }
   331  
   332  func (dec *Decoder) tokenValueAllowed() bool {
   333  	switch dec.tokenState {
   334  	case tokenTopValue, tokenArrayStart, tokenArrayValue, tokenObjectValue:
   335  		return true
   336  	}
   337  	return false
   338  }
   339  
   340  func (dec *Decoder) tokenValueEnd() {
   341  	switch dec.tokenState {
   342  	case tokenArrayStart, tokenArrayValue:
   343  		dec.tokenState = tokenArrayComma
   344  	case tokenObjectValue:
   345  		dec.tokenState = tokenObjectComma
   346  	}
   347  }
   348  
   349  // A Delim is a JSON array or object delimiter, one of [ ] { or }.
   350  type Delim rune
   351  
   352  func (d Delim) String() string {
   353  	return string(d)
   354  }
   355  
   356  // Token returns the next JSON token in the input stream.
   357  // At the end of the input stream, Token returns nil, [io.EOF].
   358  //
   359  // Token guarantees that the delimiters [ ] { } it returns are
   360  // properly nested and matched: if Token encounters an unexpected
   361  // delimiter in the input, it will return an error.
   362  //
   363  // The input stream consists of basic JSON values—bool, string,
   364  // number, and null—along with delimiters [ ] { } of type [Delim]
   365  // to mark the start and end of arrays and objects.
   366  // Commas and colons are elided.
   367  func (dec *Decoder) Token() (Token, error) {
   368  	for {
   369  		c, err := dec.peek()
   370  		if err != nil {
   371  			return nil, err
   372  		}
   373  		switch c {
   374  		case '[':
   375  			if !dec.tokenValueAllowed() {
   376  				return dec.tokenError(c)
   377  			}
   378  			dec.scanp++
   379  			dec.tokenStack = append(dec.tokenStack, dec.tokenState)
   380  			dec.tokenState = tokenArrayStart
   381  			return Delim('['), nil
   382  
   383  		case ']':
   384  			if dec.tokenState != tokenArrayStart && dec.tokenState != tokenArrayComma {
   385  				return dec.tokenError(c)
   386  			}
   387  			dec.scanp++
   388  			dec.tokenState = dec.tokenStack[len(dec.tokenStack)-1]
   389  			dec.tokenStack = dec.tokenStack[:len(dec.tokenStack)-1]
   390  			dec.tokenValueEnd()
   391  			return Delim(']'), nil
   392  
   393  		case '{':
   394  			if !dec.tokenValueAllowed() {
   395  				return dec.tokenError(c)
   396  			}
   397  			dec.scanp++
   398  			dec.tokenStack = append(dec.tokenStack, dec.tokenState)
   399  			dec.tokenState = tokenObjectStart
   400  			return Delim('{'), nil
   401  
   402  		case '}':
   403  			if dec.tokenState != tokenObjectStart && dec.tokenState != tokenObjectComma {
   404  				return dec.tokenError(c)
   405  			}
   406  			dec.scanp++
   407  			dec.tokenState = dec.tokenStack[len(dec.tokenStack)-1]
   408  			dec.tokenStack = dec.tokenStack[:len(dec.tokenStack)-1]
   409  			dec.tokenValueEnd()
   410  			return Delim('}'), nil
   411  
   412  		case ':':
   413  			if dec.tokenState != tokenObjectColon {
   414  				return dec.tokenError(c)
   415  			}
   416  			dec.scanp++
   417  			dec.tokenState = tokenObjectValue
   418  			continue
   419  
   420  		case ',':
   421  			if dec.tokenState == tokenArrayComma {
   422  				dec.scanp++
   423  				dec.tokenState = tokenArrayValue
   424  				continue
   425  			}
   426  			if dec.tokenState == tokenObjectComma {
   427  				dec.scanp++
   428  				dec.tokenState = tokenObjectKey
   429  				continue
   430  			}
   431  			return dec.tokenError(c)
   432  
   433  		case '"':
   434  			if dec.tokenState == tokenObjectStart || dec.tokenState == tokenObjectKey {
   435  				var x string
   436  				old := dec.tokenState
   437  				dec.tokenState = tokenTopValue
   438  				err := dec.Decode(&x)
   439  				dec.tokenState = old
   440  				if err != nil {
   441  					return nil, err
   442  				}
   443  				dec.tokenState = tokenObjectColon
   444  				return x, nil
   445  			}
   446  			fallthrough
   447  
   448  		default:
   449  			if !dec.tokenValueAllowed() {
   450  				return dec.tokenError(c)
   451  			}
   452  			var x any
   453  			if err := dec.Decode(&x); err != nil {
   454  				return nil, err
   455  			}
   456  			return x, nil
   457  		}
   458  	}
   459  }
   460  
   461  func (dec *Decoder) tokenError(c byte) (Token, error) {
   462  	var context string
   463  	switch dec.tokenState {
   464  	case tokenTopValue:
   465  		context = " looking for beginning of value"
   466  	case tokenArrayStart, tokenArrayValue, tokenObjectValue:
   467  		context = " looking for beginning of value"
   468  	case tokenArrayComma:
   469  		context = " after array element"
   470  	case tokenObjectKey:
   471  		context = " looking for beginning of object key string"
   472  	case tokenObjectColon:
   473  		context = " after object key"
   474  	case tokenObjectComma:
   475  		context = " after object key:value pair"
   476  	}
   477  	return nil, &SyntaxError{"invalid character " + quoteChar(c) + context, dec.InputOffset()}
   478  }
   479  
   480  // More reports whether there is another element in the
   481  // current array or object being parsed.
   482  func (dec *Decoder) More() bool {
   483  	c, err := dec.peek()
   484  	return err == nil && c != ']' && c != '}'
   485  }
   486  
   487  func (dec *Decoder) peek() (byte, error) {
   488  	var err error
   489  	for {
   490  		for i := dec.scanp; i < len(dec.buf); i++ {
   491  			c := dec.buf[i]
   492  			if isSpace(c) {
   493  				continue
   494  			}
   495  			dec.scanp = i
   496  			return c, nil
   497  		}
   498  		// buffer has been scanned, now report any error
   499  		if err != nil {
   500  			return 0, err
   501  		}
   502  		err = dec.refill()
   503  	}
   504  }
   505  
   506  // InputOffset returns the input stream byte offset of the current decoder position.
   507  // The offset gives the location of the end of the most recently returned token
   508  // and the beginning of the next token.
   509  func (dec *Decoder) InputOffset() int64 {
   510  	return dec.scanned + int64(dec.scanp)
   511  }
   512
View as plain text