base32.go

Documentation: encoding/base32

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