ascii85.go

     1  // Copyright 2009 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 ascii85 implements the ascii85 data encoding
     6  // as used in the btoa tool and Adobe's PostScript and PDF document formats.
     7  package ascii85
     8  
     9  import (
    10  	"io"
    11  	"strconv"
    12  )
    13  
    14  /*
    15   * Encoder
    16   */
    17  
    18  // Encode encodes src into at most [MaxEncodedLen](len(src))
    19  // bytes of dst, returning the actual number of bytes written.
    20  //
    21  // The encoding handles 4-byte chunks, using a special encoding
    22  // for the last fragment, so Encode is not appropriate for use on
    23  // individual blocks of a large data stream. Use [NewEncoder] instead.
    24  //
    25  // Often, ascii85-encoded data is wrapped in <~ and ~> symbols.
    26  // Encode does not add these.
    27  func Encode(dst, src []byte) int {
    28  	if len(src) == 0 {
    29  		return 0
    30  	}
    31  
    32  	n := 0
    33  	for len(src) > 0 {
    34  		dst[0] = 0
    35  		dst[1] = 0
    36  		dst[2] = 0
    37  		dst[3] = 0
    38  		dst[4] = 0
    39  
    40  		// Unpack 4 bytes into uint32 to repack into base 85 5-byte.
    41  		var v uint32
    42  		switch len(src) {
    43  		default:
    44  			v |= uint32(src[3])
    45  			fallthrough
    46  		case 3:
    47  			v |= uint32(src[2]) << 8
    48  			fallthrough
    49  		case 2:
    50  			v |= uint32(src[1]) << 16
    51  			fallthrough
    52  		case 1:
    53  			v |= uint32(src[0]) << 24
    54  		}
    55  
    56  		// Special case: zero (!!!!!) shortens to z.
    57  		if v == 0 && len(src) >= 4 {
    58  			dst[0] = 'z'
    59  			dst = dst[1:]
    60  			src = src[4:]
    61  			n++
    62  			continue
    63  		}
    64  
    65  		// Otherwise, 5 base 85 digits starting at !.
    66  		for i := 4; i >= 0; i-- {
    67  			dst[i] = '!' + byte(v%85)
    68  			v /= 85
    69  		}
    70  
    71  		// If src was short, discard the low destination bytes.
    72  		m := 5
    73  		if len(src) < 4 {
    74  			m -= 4 - len(src)
    75  			src = nil
    76  		} else {
    77  			src = src[4:]
    78  		}
    79  		dst = dst[m:]
    80  		n += m
    81  	}
    82  	return n
    83  }
    84  
    85  // MaxEncodedLen returns the maximum length of an encoding of n source bytes.
    86  func MaxEncodedLen(n int) int { return (n + 3) / 4 * 5 }
    87  
    88  // NewEncoder returns a new ascii85 stream encoder. Data written to
    89  // the returned writer will be encoded and then written to w.
    90  // Ascii85 encodings operate in 32-bit blocks; when finished
    91  // writing, the caller must Close the returned encoder to flush any
    92  // trailing partial block.
    93  func NewEncoder(w io.Writer) io.WriteCloser { return &encoder{w: w} }
    94  
    95  type encoder struct {
    96  	err  error
    97  	w    io.Writer
    98  	buf  [4]byte    // buffered data waiting to be encoded
    99  	nbuf int        // number of bytes in buf
   100  	out  [1024]byte // output buffer
   101  }
   102  
   103  func (e *encoder) Write(p []byte) (n int, err error) {
   104  	if e.err != nil {
   105  		return 0, e.err
   106  	}
   107  
   108  	// Leading fringe.
   109  	if e.nbuf > 0 {
   110  		var i int
   111  		for i = 0; i < len(p) && e.nbuf < 4; i++ {
   112  			e.buf[e.nbuf] = p[i]
   113  			e.nbuf++
   114  		}
   115  		n += i
   116  		p = p[i:]
   117  		if e.nbuf < 4 {
   118  			return
   119  		}
   120  		nout := Encode(e.out[0:], e.buf[0:])
   121  		if _, e.err = e.w.Write(e.out[0:nout]); e.err != nil {
   122  			return n, e.err
   123  		}
   124  		e.nbuf = 0
   125  	}
   126  
   127  	// Large interior chunks.
   128  	for len(p) >= 4 {
   129  		nn := len(e.out) / 5 * 4
   130  		if nn > len(p) {
   131  			nn = len(p)
   132  		}
   133  		nn -= nn % 4
   134  		if nn > 0 {
   135  			nout := Encode(e.out[0:], p[0:nn])
   136  			if _, e.err = e.w.Write(e.out[0:nout]); e.err != nil {
   137  				return n, e.err
   138  			}
   139  		}
   140  		n += nn
   141  		p = p[nn:]
   142  	}
   143  
   144  	// Trailing fringe.
   145  	copy(e.buf[:], p)
   146  	e.nbuf = len(p)
   147  	n += len(p)
   148  	return
   149  }
   150  
   151  // Close flushes any pending output from the encoder.
   152  // It is an error to call Write after calling Close.
   153  func (e *encoder) Close() error {
   154  	// If there's anything left in the buffer, flush it out
   155  	if e.err == nil && e.nbuf > 0 {
   156  		nout := Encode(e.out[0:], e.buf[0:e.nbuf])
   157  		e.nbuf = 0
   158  		_, e.err = e.w.Write(e.out[0:nout])
   159  	}
   160  	return e.err
   161  }
   162  
   163  /*
   164   * Decoder
   165   */
   166  
   167  type CorruptInputError int64
   168  
   169  func (e CorruptInputError) Error() string {
   170  	return "illegal ascii85 data at input byte " + strconv.FormatInt(int64(e), 10)
   171  }
   172  
   173  // Decode decodes src into dst, returning both the number
   174  // of bytes written to dst and the number consumed from src.
   175  // If src contains invalid ascii85 data, Decode will return the
   176  // number of bytes successfully written and a [CorruptInputError].
   177  // Decode ignores space and control characters in src.
   178  // Often, ascii85-encoded data is wrapped in <~ and ~> symbols.
   179  // Decode expects these to have been stripped by the caller.
   180  //
   181  // If flush is true, Decode assumes that src represents the
   182  // end of the input stream and processes it completely rather
   183  // than wait for the completion of another 32-bit block.
   184  //
   185  // [NewDecoder] wraps an [io.Reader] interface around Decode.
   186  func Decode(dst, src []byte, flush bool) (ndst, nsrc int, err error) {
   187  	var v uint32
   188  	var nb int
   189  	for i, b := range src {
   190  		if len(dst)-ndst < 4 {
   191  			return
   192  		}
   193  		switch {
   194  		case b <= ' ':
   195  			continue
   196  		case b == 'z' && nb == 0:
   197  			nb = 5
   198  			v = 0
   199  		case '!' <= b && b <= 'u':
   200  			v = v*85 + uint32(b-'!')
   201  			nb++
   202  		default:
   203  			return 0, 0, CorruptInputError(i)
   204  		}
   205  		if nb == 5 {
   206  			nsrc = i + 1
   207  			dst[ndst] = byte(v >> 24)
   208  			dst[ndst+1] = byte(v >> 16)
   209  			dst[ndst+2] = byte(v >> 8)
   210  			dst[ndst+3] = byte(v)
   211  			ndst += 4
   212  			nb = 0
   213  			v = 0
   214  		}
   215  	}
   216  	if flush {
   217  		nsrc = len(src)
   218  		if nb > 0 {
   219  			// The number of output bytes in the last fragment
   220  			// is the number of leftover input bytes - 1:
   221  			// the extra byte provides enough bits to cover
   222  			// the inefficiency of the encoding for the block.
   223  			if nb == 1 {
   224  				return 0, 0, CorruptInputError(len(src))
   225  			}
   226  			for i := nb; i < 5; i++ {
   227  				// The short encoding truncated the output value.
   228  				// We have to assume the worst case values (digit 84)
   229  				// in order to ensure that the top bits are correct.
   230  				v = v*85 + 84
   231  			}
   232  			for i := 0; i < nb-1; i++ {
   233  				dst[ndst] = byte(v >> 24)
   234  				v <<= 8
   235  				ndst++
   236  			}
   237  		}
   238  	}
   239  	return
   240  }
   241  
   242  // NewDecoder constructs a new ascii85 stream decoder.
   243  func NewDecoder(r io.Reader) io.Reader { return &decoder{r: r} }
   244  
   245  type decoder struct {
   246  	err     error
   247  	readErr error
   248  	r       io.Reader
   249  	buf     [1024]byte // leftover input
   250  	nbuf    int
   251  	out     []byte // leftover decoded output
   252  	outbuf  [1024]byte
   253  }
   254  
   255  func (d *decoder) Read(p []byte) (n int, err error) {
   256  	if len(p) == 0 {
   257  		return 0, nil
   258  	}
   259  	if d.err != nil {
   260  		return 0, d.err
   261  	}
   262  
   263  	for {
   264  		// Copy leftover output from last decode.
   265  		if len(d.out) > 0 {
   266  			n = copy(p, d.out)
   267  			d.out = d.out[n:]
   268  			return
   269  		}
   270  
   271  		// Decode leftover input from last read.
   272  		var nn, nsrc, ndst int
   273  		if d.nbuf > 0 {
   274  			ndst, nsrc, d.err = Decode(d.outbuf[0:], d.buf[0:d.nbuf], d.readErr != nil)
   275  			if ndst > 0 {
   276  				d.out = d.outbuf[0:ndst]
   277  				d.nbuf = copy(d.buf[0:], d.buf[nsrc:d.nbuf])
   278  				continue // copy out and return
   279  			}
   280  			if ndst == 0 && d.err == nil {
   281  				// Special case: input buffer is mostly filled with non-data bytes.
   282  				// Filter out such bytes to make room for more input.
   283  				off := 0
   284  				for i := 0; i < d.nbuf; i++ {
   285  					if d.buf[i] > ' ' {
   286  						d.buf[off] = d.buf[i]
   287  						off++
   288  					}
   289  				}
   290  				d.nbuf = off
   291  			}
   292  		}
   293  
   294  		// Out of input, out of decoded output. Check errors.
   295  		if d.err != nil {
   296  			return 0, d.err
   297  		}
   298  		if d.readErr != nil {
   299  			d.err = d.readErr
   300  			return 0, d.err
   301  		}
   302  
   303  		// Read more data.
   304  		nn, d.readErr = d.r.Read(d.buf[d.nbuf:])
   305  		d.nbuf += nn
   306  	}
   307  }
   308
View as plain text