zoneinfo_read.go

Documentation: time

     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  // Parse "zoneinfo" time zone file.
     6  // This is a fairly standard file format used on OS X, Linux, BSD, Sun, and others.
     7  // See tzfile(5), http://en.wikipedia.org/wiki/Zoneinfo,
     8  // and ftp://munnari.oz.au/pub/oldtz/
     9  
    10  package time
    11  
    12  import (
    13  	"errors"
    14  	"syscall"
    15  )
    16  
    17  // maxFileSize is the max permitted size of files read by readFile.
    18  // As reference, the zoneinfo.zip distributed by Go is ~350 KB,
    19  // so 10MB is overkill.
    20  const maxFileSize = 10 << 20
    21  
    22  type fileSizeError string
    23  
    24  func (f fileSizeError) Error() string {
    25  	return "time: file " + string(f) + " is too large"
    26  }
    27  
    28  // Copies of io.Seek* constants to avoid importing "io":
    29  const (
    30  	seekStart   = 0
    31  	seekCurrent = 1
    32  	seekEnd     = 2
    33  )
    34  
    35  // Simple I/O interface to binary blob of data.
    36  type dataIO struct {
    37  	p     []byte
    38  	error bool
    39  }
    40  
    41  func (d *dataIO) read(n int) []byte {
    42  	if len(d.p) < n {
    43  		d.p = nil
    44  		d.error = true
    45  		return nil
    46  	}
    47  	p := d.p[0:n]
    48  	d.p = d.p[n:]
    49  	return p
    50  }
    51  
    52  func (d *dataIO) big4() (n uint32, ok bool) {
    53  	p := d.read(4)
    54  	if len(p) < 4 {
    55  		d.error = true
    56  		return 0, false
    57  	}
    58  	return uint32(p[0])<<24 | uint32(p[1])<<16 | uint32(p[2])<<8 | uint32(p[3]), true
    59  }
    60  
    61  func (d *dataIO) byte() (n byte, ok bool) {
    62  	p := d.read(1)
    63  	if len(p) < 1 {
    64  		d.error = true
    65  		return 0, false
    66  	}
    67  	return p[0], true
    68  }
    69  
    70  // Make a string by stopping at the first NUL
    71  func byteString(p []byte) string {
    72  	for i := 0; i < len(p); i++ {
    73  		if p[i] == 0 {
    74  			return string(p[0:i])
    75  		}
    76  	}
    77  	return string(p)
    78  }
    79  
    80  var badData = errors.New("malformed time zone information")
    81  
    82  // LoadLocationFromTZData returns a Location with the given name
    83  // initialized from the IANA Time Zone database-formatted data.
    84  // The data should be in the format of a standard IANA time zone file
    85  // (for example, the content of /etc/localtime on Unix systems).
    86  func LoadLocationFromTZData(name string, data []byte) (*Location, error) {
    87  	d := dataIO{data, false}
    88  
    89  	// 4-byte magic "TZif"
    90  	if magic := d.read(4); string(magic) != "TZif" {
    91  		return nil, badData
    92  	}
    93  
    94  	// 1-byte version, then 15 bytes of padding
    95  	var p []byte
    96  	if p = d.read(16); len(p) != 16 || p[0] != 0 && p[0] != '2' && p[0] != '3' {
    97  		return nil, badData
    98  	}
    99  
   100  	// six big-endian 32-bit integers:
   101  	//	number of UTC/local indicators
   102  	//	number of standard/wall indicators
   103  	//	number of leap seconds
   104  	//	number of transition times
   105  	//	number of local time zones
   106  	//	number of characters of time zone abbrev strings
   107  	const (
   108  		NUTCLocal = iota
   109  		NStdWall
   110  		NLeap
   111  		NTime
   112  		NZone
   113  		NChar
   114  	)
   115  	var n [6]int
   116  	for i := 0; i < 6; i++ {
   117  		nn, ok := d.big4()
   118  		if !ok {
   119  			return nil, badData
   120  		}
   121  		n[i] = int(nn)
   122  	}
   123  
   124  	// Transition times.
   125  	txtimes := dataIO{d.read(n[NTime] * 4), false}
   126  
   127  	// Time zone indices for transition times.
   128  	txzones := d.read(n[NTime])
   129  
   130  	// Zone info structures
   131  	zonedata := dataIO{d.read(n[NZone] * 6), false}
   132  
   133  	// Time zone abbreviations.
   134  	abbrev := d.read(n[NChar])
   135  
   136  	// Leap-second time pairs
   137  	d.read(n[NLeap] * 8)
   138  
   139  	// Whether tx times associated with local time types
   140  	// are specified as standard time or wall time.
   141  	isstd := d.read(n[NStdWall])
   142  
   143  	// Whether tx times associated with local time types
   144  	// are specified as UTC or local time.
   145  	isutc := d.read(n[NUTCLocal])
   146  
   147  	if d.error { // ran out of data
   148  		return nil, badData
   149  	}
   150  
   151  	// If version == 2 or 3, the entire file repeats, this time using
   152  	// 8-byte ints for txtimes and leap seconds.
   153  	// We won't need those until 2106.
   154  
   155  	// Now we can build up a useful data structure.
   156  	// First the zone information.
   157  	//	utcoff[4] isdst[1] nameindex[1]
   158  	zone := make([]zone, n[NZone])
   159  	for i := range zone {
   160  		var ok bool
   161  		var n uint32
   162  		if n, ok = zonedata.big4(); !ok {
   163  			return nil, badData
   164  		}
   165  		zone[i].offset = int(int32(n))
   166  		var b byte
   167  		if b, ok = zonedata.byte(); !ok {
   168  			return nil, badData
   169  		}
   170  		zone[i].isDST = b != 0
   171  		if b, ok = zonedata.byte(); !ok || int(b) >= len(abbrev) {
   172  			return nil, badData
   173  		}
   174  		zone[i].name = byteString(abbrev[b:])
   175  	}
   176  
   177  	// Now the transition time info.
   178  	tx := make([]zoneTrans, n[NTime])
   179  	for i := range tx {
   180  		var ok bool
   181  		var n uint32
   182  		if n, ok = txtimes.big4(); !ok {
   183  			return nil, badData
   184  		}
   185  		tx[i].when = int64(int32(n))
   186  		if int(txzones[i]) >= len(zone) {
   187  			return nil, badData
   188  		}
   189  		tx[i].index = txzones[i]
   190  		if i < len(isstd) {
   191  			tx[i].isstd = isstd[i] != 0
   192  		}
   193  		if i < len(isutc) {
   194  			tx[i].isutc = isutc[i] != 0
   195  		}
   196  	}
   197  
   198  	if len(tx) == 0 {
   199  		// Build fake transition to cover all time.
   200  		// This happens in fixed locations like "Etc/GMT0".
   201  		tx = append(tx, zoneTrans{when: alpha, index: 0})
   202  	}
   203  
   204  	// Committed to succeed.
   205  	l := &Location{zone: zone, tx: tx, name: name}
   206  
   207  	// Fill in the cache with information about right now,
   208  	// since that will be the most common lookup.
   209  	sec, _, _ := now()
   210  	for i := range tx {
   211  		if tx[i].when <= sec && (i+1 == len(tx) || sec < tx[i+1].when) {
   212  			l.cacheStart = tx[i].when
   213  			l.cacheEnd = omega
   214  			if i+1 < len(tx) {
   215  				l.cacheEnd = tx[i+1].when
   216  			}
   217  			l.cacheZone = &l.zone[tx[i].index]
   218  		}
   219  	}
   220  
   221  	return l, nil
   222  }
   223  
   224  // loadTzinfoFromDirOrZip returns the contents of the file with the given name
   225  // in dir. dir can either be an uncompressed zip file, or a directory.
   226  func loadTzinfoFromDirOrZip(dir, name string) ([]byte, error) {
   227  	if len(dir) > 4 && dir[len(dir)-4:] == ".zip" {
   228  		return loadTzinfoFromZip(dir, name)
   229  	}
   230  	if dir != "" {
   231  		name = dir + "/" + name
   232  	}
   233  	return readFile(name)
   234  }
   235  
   236  // There are 500+ zoneinfo files. Rather than distribute them all
   237  // individually, we ship them in an uncompressed zip file.
   238  // Used this way, the zip file format serves as a commonly readable
   239  // container for the individual small files. We choose zip over tar
   240  // because zip files have a contiguous table of contents, making
   241  // individual file lookups faster, and because the per-file overhead
   242  // in a zip file is considerably less than tar's 512 bytes.
   243  
   244  // get4 returns the little-endian 32-bit value in b.
   245  func get4(b []byte) int {
   246  	if len(b) < 4 {
   247  		return 0
   248  	}
   249  	return int(b[0]) | int(b[1])<<8 | int(b[2])<<16 | int(b[3])<<24
   250  }
   251  
   252  // get2 returns the little-endian 16-bit value in b.
   253  func get2(b []byte) int {
   254  	if len(b) < 2 {
   255  		return 0
   256  	}
   257  	return int(b[0]) | int(b[1])<<8
   258  }
   259  
   260  // loadTzinfoFromZip returns the contents of the file with the given name
   261  // in the given uncompressed zip file.
   262  func loadTzinfoFromZip(zipfile, name string) ([]byte, error) {
   263  	fd, err := open(zipfile)
   264  	if err != nil {
   265  		return nil, errors.New("open " + zipfile + ": " + err.Error())
   266  	}
   267  	defer closefd(fd)
   268  
   269  	const (
   270  		zecheader = 0x06054b50
   271  		zcheader  = 0x02014b50
   272  		ztailsize = 22
   273  
   274  		zheadersize = 30
   275  		zheader     = 0x04034b50
   276  	)
   277  
   278  	buf := make([]byte, ztailsize)
   279  	if err := preadn(fd, buf, -ztailsize); err != nil || get4(buf) != zecheader {
   280  		return nil, errors.New("corrupt zip file " + zipfile)
   281  	}
   282  	n := get2(buf[10:])
   283  	size := get4(buf[12:])
   284  	off := get4(buf[16:])
   285  
   286  	buf = make([]byte, size)
   287  	if err := preadn(fd, buf, off); err != nil {
   288  		return nil, errors.New("corrupt zip file " + zipfile)
   289  	}
   290  
   291  	for i := 0; i < n; i++ {
   292  		// zip entry layout:
   293  		//	0	magic[4]
   294  		//	4	madevers[1]
   295  		//	5	madeos[1]
   296  		//	6	extvers[1]
   297  		//	7	extos[1]
   298  		//	8	flags[2]
   299  		//	10	meth[2]
   300  		//	12	modtime[2]
   301  		//	14	moddate[2]
   302  		//	16	crc[4]
   303  		//	20	csize[4]
   304  		//	24	uncsize[4]
   305  		//	28	namelen[2]
   306  		//	30	xlen[2]
   307  		//	32	fclen[2]
   308  		//	34	disknum[2]
   309  		//	36	iattr[2]
   310  		//	38	eattr[4]
   311  		//	42	off[4]
   312  		//	46	name[namelen]
   313  		//	46+namelen+xlen+fclen - next header
   314  		//
   315  		if get4(buf) != zcheader {
   316  			break
   317  		}
   318  		meth := get2(buf[10:])
   319  		size := get4(buf[24:])
   320  		namelen := get2(buf[28:])
   321  		xlen := get2(buf[30:])
   322  		fclen := get2(buf[32:])
   323  		off := get4(buf[42:])
   324  		zname := buf[46 : 46+namelen]
   325  		buf = buf[46+namelen+xlen+fclen:]
   326  		if string(zname) != name {
   327  			continue
   328  		}
   329  		if meth != 0 {
   330  			return nil, errors.New("unsupported compression for " + name + " in " + zipfile)
   331  		}
   332  
   333  		// zip per-file header layout:
   334  		//	0	magic[4]
   335  		//	4	extvers[1]
   336  		//	5	extos[1]
   337  		//	6	flags[2]
   338  		//	8	meth[2]
   339  		//	10	modtime[2]
   340  		//	12	moddate[2]
   341  		//	14	crc[4]
   342  		//	18	csize[4]
   343  		//	22	uncsize[4]
   344  		//	26	namelen[2]
   345  		//	28	xlen[2]
   346  		//	30	name[namelen]
   347  		//	30+namelen+xlen - file data
   348  		//
   349  		buf = make([]byte, zheadersize+namelen)
   350  		if err := preadn(fd, buf, off); err != nil ||
   351  			get4(buf) != zheader ||
   352  			get2(buf[8:]) != meth ||
   353  			get2(buf[26:]) != namelen ||
   354  			string(buf[30:30+namelen]) != name {
   355  			return nil, errors.New("corrupt zip file " + zipfile)
   356  		}
   357  		xlen = get2(buf[28:])
   358  
   359  		buf = make([]byte, size)
   360  		if err := preadn(fd, buf, off+30+namelen+xlen); err != nil {
   361  			return nil, errors.New("corrupt zip file " + zipfile)
   362  		}
   363  
   364  		return buf, nil
   365  	}
   366  
   367  	return nil, errors.New("cannot find " + name + " in zip file " + zipfile)
   368  }
   369  
   370  // loadTzinfoFromTzdata returns the time zone information of the time zone
   371  // with the given name, from a tzdata database file as they are typically
   372  // found on android.
   373  var loadTzinfoFromTzdata func(file, name string) ([]byte, error)
   374  
   375  // loadTzinfo returns the time zone information of the time zone
   376  // with the given name, from a given source. A source may be a
   377  // timezone database directory, tzdata database file or an uncompressed
   378  // zip file, containing the contents of such a directory.
   379  func loadTzinfo(name string, source string) ([]byte, error) {
   380  	if len(source) >= 6 && source[len(source)-6:] == "tzdata" {
   381  		return loadTzinfoFromTzdata(source, name)
   382  	}
   383  	return loadTzinfoFromDirOrZip(source, name)
   384  }
   385  
   386  // loadLocation returns the Location with the given name from one of
   387  // the specified sources. See loadTzinfo for a list of supported sources.
   388  // The first timezone data matching the given name that is successfully loaded
   389  // and parsed is returned as a Location.
   390  func loadLocation(name string, sources []string) (z *Location, firstErr error) {
   391  	for _, source := range sources {
   392  		var zoneData, err = loadTzinfo(name, source)
   393  		if err == nil {
   394  			if z, err = LoadLocationFromTZData(name, zoneData); err == nil {
   395  				return z, nil
   396  			}
   397  		}
   398  		if firstErr == nil && err != syscall.ENOENT {
   399  			firstErr = err
   400  		}
   401  	}
   402  	if firstErr != nil {
   403  		return nil, firstErr
   404  	}
   405  	return nil, errors.New("unknown time zone " + name)
   406  }
   407  
   408  // readFile reads and returns the content of the named file.
   409  // It is a trivial implementation of ioutil.ReadFile, reimplemented
   410  // here to avoid depending on io/ioutil or os.
   411  // It returns an error if name exceeds maxFileSize bytes.
   412  func readFile(name string) ([]byte, error) {
   413  	f, err := open(name)
   414  	if err != nil {
   415  		return nil, err
   416  	}
   417  	defer closefd(f)
   418  	var (
   419  		buf [4096]byte
   420  		ret []byte
   421  		n   int
   422  	)
   423  	for {
   424  		n, err = read(f, buf[:])
   425  		if n > 0 {
   426  			ret = append(ret, buf[:n]...)
   427  		}
   428  		if n == 0 || err != nil {
   429  			break
   430  		}
   431  		if len(ret) > maxFileSize {
   432  			return nil, fileSizeError(name)
   433  		}
   434  	}
   435  	return ret, err
   436  }
   437
View as plain text