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

View as plain text