base32.go

     1  // Copyright 2011 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 base32 implements base32 encoding as specified by RFC 4648.
     6  package base32
     7  
     8  import (
     9  	"io"
    10  	"strconv"
    11  )
    12  
    13  /*
    14   * Encodings
    15   */
    16  
    17  // An Encoding is a radix 32 encoding/decoding scheme, defined by a
    18  // 32-character alphabet. The most common is the "base32" encoding
    19  // introduced for SASL GSSAPI and standardized in RFC 4648.
    20  // The alternate "base32hex" encoding is used in DNSSEC.
    21  type Encoding struct {
    22  	encode    [32]byte
    23  	decodeMap [256]byte
    24  	padChar   rune
    25  }
    26  
    27  const (
    28  	StdPadding          rune = '=' // Standard padding character
    29  	NoPadding           rune = -1  // No padding
    30  	decodeMapInitialize      = "" +
    31  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    32  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    33  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    34  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    35  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    36  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    37  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    38  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    39  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    40  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    41  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    42  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    43  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    44  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    45  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff" +
    46  		"\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff"
    47  )
    48  
    49  const encodeStd = "ABCDEFGHIJKLMNOPQRSTUVWXYZ234567"
    50  const encodeHex = "0123456789ABCDEFGHIJKLMNOPQRSTUV"
    51  
    52  // NewEncoding returns a new Encoding defined by the given alphabet,
    53  // which must be a 32-byte string.
    54  func NewEncoding(encoder string) *Encoding {
    55  	if len(encoder) != 32 {
    56  		panic("encoding alphabet is not 32-bytes long")
    57  	}
    58  
    59  	e := new(Encoding)
    60  	e.padChar = StdPadding
    61  	copy(e.encode[:], encoder)
    62  	copy(e.decodeMap[:], decodeMapInitialize)
    63  
    64  	for i := 0; i < len(encoder); i++ {
    65  		e.decodeMap[encoder[i]] = byte(i)
    66  	}
    67  	return e
    68  }
    69  
    70  // StdEncoding is the standard base32 encoding, as defined in
    71  // RFC 4648.
    72  var StdEncoding = NewEncoding(encodeStd)
    73  
    74  // HexEncoding is the “Extended Hex Alphabet” defined in RFC 4648.
    75  // It is typically used in DNS.
    76  var HexEncoding = NewEncoding(encodeHex)
    77  
    78  // WithPadding creates a new encoding identical to enc except
    79  // with a specified padding character, or NoPadding to disable padding.
    80  // The padding character must not be '\r' or '\n', must not
    81  // be contained in the encoding's alphabet and must be a rune equal or
    82  // below '\xff'.
    83  func (enc Encoding) WithPadding(padding rune) *Encoding {
    84  	if padding == '\r' || padding == '\n' || padding > 0xff {
    85  		panic("invalid padding")
    86  	}
    87  
    88  	for i := 0; i < len(enc.encode); i++ {
    89  		if rune(enc.encode[i]) == padding {
    90  			panic("padding contained in alphabet")
    91  		}
    92  	}
    93  
    94  	enc.padChar = padding
    95  	return &enc
    96  }
    97  
    98  /*
    99   * Encoder
   100   */
   101  
   102  // Encode encodes src using the encoding enc, writing
   103  // EncodedLen(len(src)) bytes to dst.
   104  //
   105  // The encoding pads the output to a multiple of 8 bytes,
   106  // so Encode is not appropriate for use on individual blocks
   107  // of a large data stream. Use NewEncoder() instead.
   108  func (enc *Encoding) Encode(dst, src []byte) {
   109  	for len(src) > 0 {
   110  		var b [8]byte
   111  
   112  		// Unpack 8x 5-bit source blocks into a 5 byte
   113  		// destination quantum
   114  		switch len(src) {
   115  		default:
   116  			b[7] = src[4] & 0x1F
   117  			b[6] = src[4] >> 5
   118  			fallthrough
   119  		case 4:
   120  			b[6] |= (src[3] << 3) & 0x1F
   121  			b[5] = (src[3] >> 2) & 0x1F
   122  			b[4] = src[3] >> 7
   123  			fallthrough
   124  		case 3:
   125  			b[4] |= (src[2] << 1) & 0x1F
   126  			b[3] = (src[2] >> 4) & 0x1F
   127  			fallthrough
   128  		case 2:
   129  			b[3] |= (src[1] << 4) & 0x1F
   130  			b[2] = (src[1] >> 1) & 0x1F
   131  			b[1] = (src[1] >> 6) & 0x1F
   132  			fallthrough
   133  		case 1:
   134  			b[1] |= (src[0] << 2) & 0x1F
   135  			b[0] = src[0] >> 3
   136  		}
   137  
   138  		// Encode 5-bit blocks using the base32 alphabet
   139  		size := len(dst)
   140  		if size >= 8 {
   141  			// Common case, unrolled for extra performance
   142  			dst[0] = enc.encode[b[0]&31]
   143  			dst[1] = enc.encode[b[1]&31]
   144  			dst[2] = enc.encode[b[2]&31]
   145  			dst[3] = enc.encode[b[3]&31]
   146  			dst[4] = enc.encode[b[4]&31]
   147  			dst[5] = enc.encode[b[5]&31]
   148  			dst[6] = enc.encode[b[6]&31]
   149  			dst[7] = enc.encode[b[7]&31]
   150  		} else {
   151  			for i := 0; i < size; i++ {
   152  				dst[i] = enc.encode[b[i]&31]
   153  			}
   154  		}
   155  
   156  		// Pad the final quantum
   157  		if len(src) < 5 {
   158  			if enc.padChar == NoPadding {
   159  				break
   160  			}
   161  
   162  			dst[7] = byte(enc.padChar)
   163  			if len(src) < 4 {
   164  				dst[6] = byte(enc.padChar)
   165  				dst[5] = byte(enc.padChar)
   166  				if len(src) < 3 {
   167  					dst[4] = byte(enc.padChar)
   168  					if len(src) < 2 {
   169  						dst[3] = byte(enc.padChar)
   170  						dst[2] = byte(enc.padChar)
   171  					}
   172  				}
   173  			}
   174  
   175  			break
   176  		}
   177  
   178  		src = src[5:]
   179  		dst = dst[8:]
   180  	}
   181  }
   182  
   183  // EncodeToString returns the base32 encoding of src.
   184  func (enc *Encoding) EncodeToString(src []byte) string {
   185  	buf := make([]byte, enc.EncodedLen(len(src)))
   186  	enc.Encode(buf, src)
   187  	return string(buf)
   188  }
   189  
   190  type encoder struct {
   191  	err  error
   192  	enc  *Encoding
   193  	w    io.Writer
   194  	buf  [5]byte    // buffered data waiting to be encoded
   195  	nbuf int        // number of bytes in buf
   196  	out  [1024]byte // output buffer
   197  }
   198  
   199  func (e *encoder) Write(p []byte) (n int, err error) {
   200  	if e.err != nil {
   201  		return 0, e.err
   202  	}
   203  
   204  	// Leading fringe.
   205  	if e.nbuf > 0 {
   206  		var i int
   207  		for i = 0; i < len(p) && e.nbuf < 5; i++ {
   208  			e.buf[e.nbuf] = p[i]
   209  			e.nbuf++
   210  		}
   211  		n += i
   212  		p = p[i:]
   213  		if e.nbuf < 5 {
   214  			return
   215  		}
   216  		e.enc.Encode(e.out[0:], e.buf[0:])
   217  		if _, e.err = e.w.Write(e.out[0:8]); e.err != nil {
   218  			return n, e.err
   219  		}
   220  		e.nbuf = 0
   221  	}
   222  
   223  	// Large interior chunks.
   224  	for len(p) >= 5 {
   225  		nn := len(e.out) / 8 * 5
   226  		if nn > len(p) {
   227  			nn = len(p)
   228  			nn -= nn % 5
   229  		}
   230  		e.enc.Encode(e.out[0:], p[0:nn])
   231  		if _, e.err = e.w.Write(e.out[0 : nn/5*8]); e.err != nil {
   232  			return n, e.err
   233  		}
   234  		n += nn
   235  		p = p[nn:]
   236  	}
   237  
   238  	// Trailing fringe.
   239  	copy(e.buf[:], p)
   240  	e.nbuf = len(p)
   241  	n += len(p)
   242  	return
   243  }
   244  
   245  // Close flushes any pending output from the encoder.
   246  // It is an error to call Write after calling Close.
   247  func (e *encoder) Close() error {
   248  	// If there's anything left in the buffer, flush it out
   249  	if e.err == nil && e.nbuf > 0 {
   250  		e.enc.Encode(e.out[0:], e.buf[0:e.nbuf])
   251  		encodedLen := e.enc.EncodedLen(e.nbuf)
   252  		e.nbuf = 0
   253  		_, e.err = e.w.Write(e.out[0:encodedLen])
   254  	}
   255  	return e.err
   256  }
   257  
   258  // NewEncoder returns a new base32 stream encoder. Data written to
   259  // the returned writer will be encoded using enc and then written to w.
   260  // Base32 encodings operate in 5-byte blocks; when finished
   261  // writing, the caller must Close the returned encoder to flush any
   262  // partially written blocks.
   263  func NewEncoder(enc *Encoding, w io.Writer) io.WriteCloser {
   264  	return &encoder{enc: enc, w: w}
   265  }
   266  
   267  // EncodedLen returns the length in bytes of the base32 encoding
   268  // of an input buffer of length n.
   269  func (enc *Encoding) EncodedLen(n int) int {
   270  	if enc.padChar == NoPadding {
   271  		return (n*8 + 4) / 5
   272  	}
   273  	return (n + 4) / 5 * 8
   274  }
   275  
   276  /*
   277   * Decoder
   278   */
   279  
   280  type CorruptInputError int64
   281  
   282  func (e CorruptInputError) Error() string {
   283  	return "illegal base32 data at input byte " + strconv.FormatInt(int64(e), 10)
   284  }
   285  
   286  // decode is like Decode but returns an additional 'end' value, which
   287  // indicates if end-of-message padding was encountered and thus any
   288  // additional data is an error. This method assumes that src has been
   289  // stripped of all supported whitespace ('\r' and '\n').
   290  func (enc *Encoding) decode(dst, src []byte) (n int, end bool, err error) {
   291  	// Lift the nil check outside of the loop.
   292  	_ = enc.decodeMap
   293  
   294  	dsti := 0
   295  	olen := len(src)
   296  
   297  	for len(src) > 0 && !end {
   298  		// Decode quantum using the base32 alphabet
   299  		var dbuf [8]byte
   300  		dlen := 8
   301  
   302  		for j := 0; j < 8; {
   303  
   304  			if len(src) == 0 {
   305  				if enc.padChar != NoPadding {
   306  					// We have reached the end and are missing padding
   307  					return n, false, CorruptInputError(olen - len(src) - j)
   308  				}
   309  				// We have reached the end and are not expecting any padding
   310  				dlen, end = j, true
   311  				break
   312  			}
   313  			in := src[0]
   314  			src = src[1:]
   315  			if in == byte(enc.padChar) && j >= 2 && len(src) < 8 {
   316  				// We've reached the end and there's padding
   317  				if len(src)+j < 8-1 {
   318  					// not enough padding
   319  					return n, false, CorruptInputError(olen)
   320  				}
   321  				for k := 0; k < 8-1-j; k++ {
   322  					if len(src) > k && src[k] != byte(enc.padChar) {
   323  						// incorrect padding
   324  						return n, false, CorruptInputError(olen - len(src) + k - 1)
   325  					}
   326  				}
   327  				dlen, end = j, true
   328  				// 7, 5 and 2 are not valid padding lengths, and so 1, 3 and 6 are not
   329  				// valid dlen values. See RFC 4648 Section 6 "Base 32 Encoding" listing
   330  				// the five valid padding lengths, and Section 9 "Illustrations and
   331  				// Examples" for an illustration for how the 1st, 3rd and 6th base32
   332  				// src bytes do not yield enough information to decode a dst byte.
   333  				if dlen == 1 || dlen == 3 || dlen == 6 {
   334  					return n, false, CorruptInputError(olen - len(src) - 1)
   335  				}
   336  				break
   337  			}
   338  			dbuf[j] = enc.decodeMap[in]
   339  			if dbuf[j] == 0xFF {
   340  				return n, false, CorruptInputError(olen - len(src) - 1)
   341  			}
   342  			j++
   343  		}
   344  
   345  		// Pack 8x 5-bit source blocks into 5 byte destination
   346  		// quantum
   347  		switch dlen {
   348  		case 8:
   349  			dst[dsti+4] = dbuf[6]<<5 | dbuf[7]
   350  			n++
   351  			fallthrough
   352  		case 7:
   353  			dst[dsti+3] = dbuf[4]<<7 | dbuf[5]<<2 | dbuf[6]>>3
   354  			n++
   355  			fallthrough
   356  		case 5:
   357  			dst[dsti+2] = dbuf[3]<<4 | dbuf[4]>>1
   358  			n++
   359  			fallthrough
   360  		case 4:
   361  			dst[dsti+1] = dbuf[1]<<6 | dbuf[2]<<1 | dbuf[3]>>4
   362  			n++
   363  			fallthrough
   364  		case 2:
   365  			dst[dsti+0] = dbuf[0]<<3 | dbuf[1]>>2
   366  			n++
   367  		}
   368  		dsti += 5
   369  	}
   370  	return n, end, nil
   371  }
   372  
   373  // Decode decodes src using the encoding enc. It writes at most
   374  // DecodedLen(len(src)) bytes to dst and returns the number of bytes
   375  // written. If src contains invalid base32 data, it will return the
   376  // number of bytes successfully written and CorruptInputError.
   377  // New line characters (\r and \n) are ignored.
   378  func (enc *Encoding) Decode(dst, src []byte) (n int, err error) {
   379  	buf := make([]byte, len(src))
   380  	l := stripNewlines(buf, src)
   381  	n, _, err = enc.decode(dst, buf[:l])
   382  	return
   383  }
   384  
   385  // DecodeString returns the bytes represented by the base32 string s.
   386  func (enc *Encoding) DecodeString(s string) ([]byte, error) {
   387  	buf := []byte(s)
   388  	l := stripNewlines(buf, buf)
   389  	n, _, err := enc.decode(buf, buf[:l])
   390  	return buf[:n], err
   391  }
   392  
   393  type decoder struct {
   394  	err    error
   395  	enc    *Encoding
   396  	r      io.Reader
   397  	end    bool       // saw end of message
   398  	buf    [1024]byte // leftover input
   399  	nbuf   int
   400  	out    []byte // leftover decoded output
   401  	outbuf [1024 / 8 * 5]byte
   402  }
   403  
   404  func readEncodedData(r io.Reader, buf []byte, min int, expectsPadding bool) (n int, err error) {
   405  	for n < min && err == nil {
   406  		var nn int
   407  		nn, err = r.Read(buf[n:])
   408  		n += nn
   409  	}
   410  	// data was read, less than min bytes could be read
   411  	if n < min && n > 0 && err == io.EOF {
   412  		err = io.ErrUnexpectedEOF
   413  	}
   414  	// no data was read, the buffer already contains some data
   415  	// when padding is disabled this is not an error, as the message can be of
   416  	// any length
   417  	if expectsPadding && min < 8 && n == 0 && err == io.EOF {
   418  		err = io.ErrUnexpectedEOF
   419  	}
   420  	return
   421  }
   422  
   423  func (d *decoder) Read(p []byte) (n int, err error) {
   424  	// Use leftover decoded output from last read.
   425  	if len(d.out) > 0 {
   426  		n = copy(p, d.out)
   427  		d.out = d.out[n:]
   428  		if len(d.out) == 0 {
   429  			return n, d.err
   430  		}
   431  		return n, nil
   432  	}
   433  
   434  	if d.err != nil {
   435  		return 0, d.err
   436  	}
   437  
   438  	// Read a chunk.
   439  	nn := len(p) / 5 * 8
   440  	if nn < 8 {
   441  		nn = 8
   442  	}
   443  	if nn > len(d.buf) {
   444  		nn = len(d.buf)
   445  	}
   446  
   447  	// Minimum amount of bytes that needs to be read each cycle
   448  	var min int
   449  	var expectsPadding bool
   450  	if d.enc.padChar == NoPadding {
   451  		min = 1
   452  		expectsPadding = false
   453  	} else {
   454  		min = 8 - d.nbuf
   455  		expectsPadding = true
   456  	}
   457  
   458  	nn, d.err = readEncodedData(d.r, d.buf[d.nbuf:nn], min, expectsPadding)
   459  	d.nbuf += nn
   460  	if d.nbuf < min {
   461  		return 0, d.err
   462  	}
   463  	if nn > 0 && d.end {
   464  		return 0, CorruptInputError(0)
   465  	}
   466  
   467  	// Decode chunk into p, or d.out and then p if p is too small.
   468  	var nr int
   469  	if d.enc.padChar == NoPadding {
   470  		nr = d.nbuf
   471  	} else {
   472  		nr = d.nbuf / 8 * 8
   473  	}
   474  	nw := d.enc.DecodedLen(d.nbuf)
   475  
   476  	if nw > len(p) {
   477  		nw, d.end, err = d.enc.decode(d.outbuf[0:], d.buf[0:nr])
   478  		d.out = d.outbuf[0:nw]
   479  		n = copy(p, d.out)
   480  		d.out = d.out[n:]
   481  	} else {
   482  		n, d.end, err = d.enc.decode(p, d.buf[0:nr])
   483  	}
   484  	d.nbuf -= nr
   485  	for i := 0; i < d.nbuf; i++ {
   486  		d.buf[i] = d.buf[i+nr]
   487  	}
   488  
   489  	if err != nil && (d.err == nil || d.err == io.EOF) {
   490  		d.err = err
   491  	}
   492  
   493  	if len(d.out) > 0 {
   494  		// We cannot return all the decoded bytes to the caller in this
   495  		// invocation of Read, so we return a nil error to ensure that Read
   496  		// will be called again.  The error stored in d.err, if any, will be
   497  		// returned with the last set of decoded bytes.
   498  		return n, nil
   499  	}
   500  
   501  	return n, d.err
   502  }
   503  
   504  type newlineFilteringReader struct {
   505  	wrapped io.Reader
   506  }
   507  
   508  // stripNewlines removes newline characters and returns the number
   509  // of non-newline characters copied to dst.
   510  func stripNewlines(dst, src []byte) int {
   511  	offset := 0
   512  	for _, b := range src {
   513  		if b == '\r' || b == '\n' {
   514  			continue
   515  		}
   516  		dst[offset] = b
   517  		offset++
   518  	}
   519  	return offset
   520  }
   521  
   522  func (r *newlineFilteringReader) Read(p []byte) (int, error) {
   523  	n, err := r.wrapped.Read(p)
   524  	for n > 0 {
   525  		s := p[0:n]
   526  		offset := stripNewlines(s, s)
   527  		if err != nil || offset > 0 {
   528  			return offset, err
   529  		}
   530  		// Previous buffer entirely whitespace, read again
   531  		n, err = r.wrapped.Read(p)
   532  	}
   533  	return n, err
   534  }
   535  
   536  // NewDecoder constructs a new base32 stream decoder.
   537  func NewDecoder(enc *Encoding, r io.Reader) io.Reader {
   538  	return &decoder{enc: enc, r: &newlineFilteringReader{r}}
   539  }
   540  
   541  // DecodedLen returns the maximum length in bytes of the decoded data
   542  // corresponding to n bytes of base32-encoded data.
   543  func (enc *Encoding) DecodedLen(n int) int {
   544  	if enc.padChar == NoPadding {
   545  		return n * 5 / 8
   546  	}
   547  
   548  	return n / 8 * 5
   549  }
   550
View as plain text