gunzip.mx raw

   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 gzip implements reading and writing of gzip format compressed files,
   6  // as specified in RFC 1952.
   7  package gzip
   8  
   9  import (
  10  	"bufio"
  11  	"compress/flate"
  12  	"encoding/binary"
  13  	"errors"
  14  	"hash/crc32"
  15  	"io"
  16  	"time"
  17  )
  18  
  19  const (
  20  	gzipID1     = 0x1f
  21  	gzipID2     = 0x8b
  22  	gzipDeflate = 8
  23  	flagText    = 1 << 0
  24  	flagHdrCrc  = 1 << 1
  25  	flagExtra   = 1 << 2
  26  	flagName    = 1 << 3
  27  	flagComment = 1 << 4
  28  )
  29  
  30  var (
  31  	// ErrChecksum is returned when reading GZIP data that has an invalid checksum.
  32  	ErrChecksum = errors.New("gzip: invalid checksum")
  33  	// ErrHeader is returned when reading GZIP data that has an invalid header.
  34  	ErrHeader = errors.New("gzip: invalid header")
  35  )
  36  
  37  var le = binary.LittleEndian
  38  
  39  // noEOF converts io.EOF to io.ErrUnexpectedEOF.
  40  func noEOF(err error) error {
  41  	if err == io.EOF {
  42  		return io.ErrUnexpectedEOF
  43  	}
  44  	return err
  45  }
  46  
  47  // The gzip file stores a header giving metadata about the compressed file.
  48  // That header is exposed as the fields of the [Writer] and [Reader] structs.
  49  //
  50  // Strings must be UTF-8 encoded and may only contain Unicode code points
  51  // U+0001 through U+00FF, due to limitations of the GZIP file format.
  52  type Header struct {
  53  	Comment []byte    // comment
  54  	Extra   []byte    // "extra data"
  55  	ModTime time.Time // modification time
  56  	Name    []byte    // file name
  57  	OS      byte      // operating system type
  58  }
  59  
  60  // A Reader is an [io.Reader] that can be read to retrieve
  61  // uncompressed data from a gzip-format compressed file.
  62  //
  63  // In general, a gzip file can be a concatenation of gzip files,
  64  // each with its own header. Reads from the Reader
  65  // return the concatenation of the uncompressed data of each.
  66  // Only the first header is recorded in the Reader fields.
  67  //
  68  // Gzip files store a length and checksum of the uncompressed data.
  69  // The Reader will return an [ErrChecksum] when [Reader.Read]
  70  // reaches the end of the uncompressed data if it does not
  71  // have the expected length or checksum. Clients should treat data
  72  // returned by [Reader.Read] as tentative until they receive the [io.EOF]
  73  // marking the end of the data.
  74  type Reader struct {
  75  	Header       // valid after NewReader or Reader.Reset
  76  	r            flate.Reader
  77  	decompressor io.ReadCloser
  78  	digest       uint32 // CRC-32, IEEE polynomial (section 8)
  79  	size         uint32 // Uncompressed size (section 2.3.1)
  80  	buf          [512]byte
  81  	err          error
  82  	multistream  bool
  83  }
  84  
  85  // NewReader creates a new [Reader] reading the given reader.
  86  // If r does not also implement [io.ByteReader],
  87  // the decompressor may read more data than necessary from r.
  88  //
  89  // It is the caller's responsibility to call [Reader.Close] when done.
  90  //
  91  // The Reader.[Header] fields will be valid in the [Reader] returned.
  92  func NewReader(r io.Reader) (*Reader, error) {
  93  	z := &Reader{}
  94  	if err := z.Reset(r); err != nil {
  95  		return nil, err
  96  	}
  97  	return z, nil
  98  }
  99  
 100  // Reset discards the [Reader] z's state and makes it equivalent to the
 101  // result of its original state from [NewReader], but reading from r instead.
 102  // This permits reusing a [Reader] rather than allocating a new one.
 103  func (z *Reader) Reset(r io.Reader) error {
 104  	*z = Reader{
 105  		decompressor: z.decompressor,
 106  		multistream:  true,
 107  	}
 108  	if rr, ok := r.(flate.Reader); ok {
 109  		z.r = rr
 110  	} else {
 111  		z.r = bufio.NewReader(r)
 112  	}
 113  	z.Header, z.err = z.readHeader()
 114  	return z.err
 115  }
 116  
 117  // Multistream controls whether the reader supports multistream files.
 118  //
 119  // If enabled (the default), the [Reader] expects the input to be a sequence
 120  // of individually gzipped data streams, each with its own header and
 121  // trailer, ending at EOF. The effect is that the concatenation of a sequence
 122  // of gzipped files is treated as equivalent to the gzip of the concatenation
 123  // of the sequence. This is standard behavior for gzip readers.
 124  //
 125  // Calling Multistream(false) disables this behavior; disabling the behavior
 126  // can be useful when reading file formats that distinguish individual gzip
 127  // data streams or mix gzip data streams with other data streams.
 128  // In this mode, when the [Reader] reaches the end of the data stream,
 129  // [Reader.Read] returns [io.EOF]. The underlying reader must implement [io.ByteReader]
 130  // in order to be left positioned just after the gzip stream.
 131  // To start the next stream, call z.Reset(r) followed by z.Multistream(false).
 132  // If there is no next stream, z.Reset(r) will return [io.EOF].
 133  func (z *Reader) Multistream(ok bool) {
 134  	z.multistream = ok
 135  }
 136  
 137  // readString reads a NUL-terminated string from z.r.
 138  // It treats the bytes read as being encoded as ISO 8859-1 (Latin-1) and
 139  // will output a string encoded using UTF-8.
 140  // This method always updates z.digest with the data read.
 141  func (z *Reader) readString() ([]byte, error) {
 142  	var err error
 143  	needConv := false
 144  	for i := 0; ; i++ {
 145  		if i >= len(z.buf) {
 146  			return nil, ErrHeader
 147  		}
 148  		z.buf[i], err = z.r.ReadByte()
 149  		if err != nil {
 150  			return nil, err
 151  		}
 152  		if z.buf[i] > 0x7f {
 153  			needConv = true
 154  		}
 155  		if z.buf[i] == 0 {
 156  			// Digest covers the NUL terminator.
 157  			z.digest = crc32.Update(z.digest, crc32.IEEETable, z.buf[:i+1])
 158  
 159  			// Strings are ISO 8859-1, Latin-1 (RFC 1952, section 2.3.1).
 160  			if needConv {
 161  				s := []rune{:0:i}
 162  				for _, v := range z.buf[:i] {
 163  					s = append(s, rune(v))
 164  				}
 165  				return []byte(string(s)), nil
 166  			}
 167  			return append([]byte(nil), z.buf[:i]...), nil
 168  		}
 169  	}
 170  }
 171  
 172  // readHeader reads the GZIP header according to section 2.3.1.
 173  // This method does not set z.err.
 174  func (z *Reader) readHeader() (hdr Header, err error) {
 175  	if _, err = io.ReadFull(z.r, z.buf[:10]); err != nil {
 176  		// RFC 1952, section 2.2, says the following:
 177  		//	A gzip file consists of a series of "members" (compressed data sets).
 178  		//
 179  		// Other than this, the specification does not clarify whether a
 180  		// "series" is defined as "one or more" or "zero or more". To err on the
 181  		// side of caution, Go interprets this to mean "zero or more".
 182  		// Thus, it is okay to return io.EOF here.
 183  		return hdr, err
 184  	}
 185  	if z.buf[0] != gzipID1 || z.buf[1] != gzipID2 || z.buf[2] != gzipDeflate {
 186  		return hdr, ErrHeader
 187  	}
 188  	flg := z.buf[3]
 189  	if t := int64(le.Uint32(z.buf[4:8])); t > 0 {
 190  		// Section 2.3.1, the zero value for MTIME means that the
 191  		// modified time is not set.
 192  		hdr.ModTime = time.Unix(t, 0)
 193  	}
 194  	// z.buf[8] is XFL and is currently ignored.
 195  	hdr.OS = z.buf[9]
 196  	z.digest = crc32.ChecksumIEEE(z.buf[:10])
 197  
 198  	if flg&flagExtra != 0 {
 199  		if _, err = io.ReadFull(z.r, z.buf[:2]); err != nil {
 200  			return hdr, noEOF(err)
 201  		}
 202  		z.digest = crc32.Update(z.digest, crc32.IEEETable, z.buf[:2])
 203  		data := []byte{:le.Uint16(z.buf[:2])}
 204  		if _, err = io.ReadFull(z.r, data); err != nil {
 205  			return hdr, noEOF(err)
 206  		}
 207  		z.digest = crc32.Update(z.digest, crc32.IEEETable, data)
 208  		hdr.Extra = data
 209  	}
 210  
 211  	var s []byte
 212  	if flg&flagName != 0 {
 213  		if s, err = z.readString(); err != nil {
 214  			return hdr, noEOF(err)
 215  		}
 216  		hdr.Name = s
 217  	}
 218  
 219  	if flg&flagComment != 0 {
 220  		if s, err = z.readString(); err != nil {
 221  			return hdr, noEOF(err)
 222  		}
 223  		hdr.Comment = s
 224  	}
 225  
 226  	if flg&flagHdrCrc != 0 {
 227  		if _, err = io.ReadFull(z.r, z.buf[:2]); err != nil {
 228  			return hdr, noEOF(err)
 229  		}
 230  		digest := le.Uint16(z.buf[:2])
 231  		if digest != uint16(z.digest) {
 232  			return hdr, ErrHeader
 233  		}
 234  	}
 235  
 236  	z.digest = 0
 237  	if z.decompressor == nil {
 238  		z.decompressor = flate.NewReader(z.r)
 239  	} else {
 240  		z.decompressor.(flate.Resetter).Reset(z.r, nil)
 241  	}
 242  	return hdr, nil
 243  }
 244  
 245  // Read implements [io.Reader], reading uncompressed bytes from its underlying reader.
 246  func (z *Reader) Read(p []byte) (n int, err error) {
 247  	if z.err != nil {
 248  		return 0, z.err
 249  	}
 250  
 251  	for n == 0 {
 252  		n, z.err = z.decompressor.Read(p)
 253  		z.digest = crc32.Update(z.digest, crc32.IEEETable, p[:n])
 254  		z.size += uint32(n)
 255  		if z.err != io.EOF {
 256  			// In the normal case we return here.
 257  			return n, z.err
 258  		}
 259  
 260  		// Finished file; check checksum and size.
 261  		if _, err := io.ReadFull(z.r, z.buf[:8]); err != nil {
 262  			z.err = noEOF(err)
 263  			return n, z.err
 264  		}
 265  		digest := le.Uint32(z.buf[:4])
 266  		size := le.Uint32(z.buf[4:8])
 267  		if digest != z.digest || size != z.size {
 268  			z.err = ErrChecksum
 269  			return n, z.err
 270  		}
 271  		z.digest, z.size = 0, 0
 272  
 273  		// File is ok; check if there is another.
 274  		if !z.multistream {
 275  			return n, io.EOF
 276  		}
 277  		z.err = nil // Remove io.EOF
 278  
 279  		if _, z.err = z.readHeader(); z.err != nil {
 280  			return n, z.err
 281  		}
 282  	}
 283  
 284  	return n, nil
 285  }
 286  
 287  // Close closes the [Reader]. It does not close the underlying reader.
 288  // In order for the GZIP checksum to be verified, the reader must be
 289  // fully consumed until the [io.EOF].
 290  func (z *Reader) Close() error { return z.decompressor.Close() }
 291