1
# Bazaar-NG -- distributed version control
3
# Copyright (C) 2005 by Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
5
3
# This program is free software; you can redistribute it and/or modify
6
4
# it under the terms of the GNU General Public License as published by
7
5
# the Free Software Foundation; either version 2 of the License, or
8
6
# (at your option) any later version.
10
8
# This program is distributed in the hope that it will be useful,
11
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
12
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13
11
# GNU General Public License for more details.
15
13
# You should have received a copy of the GNU General Public License
16
14
# along with this program; if not, write to the Free Software
17
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
import os, types, re, time, errno
20
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
22
from errors import bailout, BzrError
23
from trace import mutter
20
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
21
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
25
from bzrlib.lazy_import import lazy_import
26
lazy_import(globals(), """
28
from datetime import datetime
30
from ntpath import (abspath as _nt_abspath,
32
normpath as _nt_normpath,
33
realpath as _nt_realpath,
34
splitdrive as _nt_splitdrive,
42
from tempfile import (
54
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
56
if sys.version_info < (2, 5):
57
import md5 as _mod_md5
59
import sha as _mod_sha
69
from bzrlib import symbol_versioning
72
# On win32, O_BINARY is used to indicate the file should
73
# be opened in binary mode, rather than text mode.
74
# On other platforms, O_BINARY doesn't exist, because
75
# they always open in binary mode, so it is okay to
76
# OR with 0 on those platforms
77
O_BINARY = getattr(os, 'O_BINARY', 0)
26
80
def make_readonly(filename):
27
81
"""Make a filename read-only."""
28
# TODO: probably needs to be fixed for windows
29
mod = os.stat(filename).st_mode
31
os.chmod(filename, mod)
82
mod = os.lstat(filename).st_mode
83
if not stat.S_ISLNK(mod):
85
os.chmod(filename, mod)
34
88
def make_writable(filename):
35
mod = os.stat(filename).st_mode
37
os.chmod(filename, mod)
40
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/_~-])')
89
mod = os.lstat(filename).st_mode
90
if not stat.S_ISLNK(mod):
92
os.chmod(filename, mod)
95
def minimum_path_selection(paths):
96
"""Return the smallset subset of paths which are outside paths.
98
:param paths: A container (and hence not None) of paths.
99
:return: A set of paths sufficient to include everything in paths via
100
is_inside_any, drawn from the paths parameter.
105
other_paths = paths.difference([path])
106
if not is_inside_any(other_paths, path):
107
# this is a top level path, we must check it.
108
search_paths.add(path)
42
"""Return shell-quoted filename"""
43
## We could be a bit more terse by using double-quotes etc
44
f = _QUOTE_RE.sub(r'\\\1', f)
51
mode = os.lstat(f)[ST_MODE]
59
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
116
"""Return a quoted filename filename
118
This previously used backslash quoting, but that works poorly on
120
# TODO: I'm not really sure this is the best format either.x
122
if _QUOTE_RE is None:
123
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
125
if _QUOTE_RE.search(f):
131
_directory_kind = 'directory'
134
"""Return the current umask"""
135
# Assume that people aren't messing with the umask while running
136
# XXX: This is not thread safe, but there is no way to get the
137
# umask without setting it
145
_directory_kind: "/",
147
'tree-reference': '+',
151
def kind_marker(kind):
153
return _kind_marker_map[kind]
155
raise errors.BzrError('invalid file kind %r' % kind)
158
lexists = getattr(os.path, 'lexists', None)
162
stat = getattr(os, 'lstat', os.stat)
166
if e.errno == errno.ENOENT:
169
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
172
def fancy_rename(old, new, rename_func, unlink_func):
173
"""A fancy rename, when you don't have atomic rename.
175
:param old: The old path, to rename from
176
:param new: The new path, to rename to
177
:param rename_func: The potentially non-atomic rename function
178
:param unlink_func: A way to delete the target file if the full rename succeeds
181
# sftp rename doesn't allow overwriting, so play tricks:
182
base = os.path.basename(new)
183
dirname = os.path.dirname(new)
184
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
185
tmp_name = pathjoin(dirname, tmp_name)
187
# Rename the file out of the way, but keep track if it didn't exist
188
# We don't want to grab just any exception
189
# something like EACCES should prevent us from continuing
190
# The downside is that the rename_func has to throw an exception
191
# with an errno = ENOENT, or NoSuchFile
194
rename_func(new, tmp_name)
195
except (errors.NoSuchFile,), e:
198
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
199
# function raises an IOError with errno is None when a rename fails.
200
# This then gets caught here.
201
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
204
if (getattr(e, 'errno', None) is None
205
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
213
# This may throw an exception, in which case success will
215
rename_func(old, new)
217
except (IOError, OSError), e:
218
# source and target may be aliases of each other (e.g. on a
219
# case-insensitive filesystem), so we may have accidentally renamed
220
# source by when we tried to rename target
221
if not (file_existed and e.errno in (None, errno.ENOENT)):
225
# If the file used to exist, rename it back into place
226
# otherwise just delete it from the tmp location
228
unlink_func(tmp_name)
230
rename_func(tmp_name, new)
233
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
234
# choke on a Unicode string containing a relative path if
235
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
237
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
238
def _posix_abspath(path):
239
# jam 20060426 rather than encoding to fsencoding
240
# copy posixpath.abspath, but use os.getcwdu instead
241
if not posixpath.isabs(path):
242
path = posixpath.join(getcwd(), path)
243
return posixpath.normpath(path)
246
def _posix_realpath(path):
247
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
250
def _win32_fixdrive(path):
251
"""Force drive letters to be consistent.
253
win32 is inconsistent whether it returns lower or upper case
254
and even if it was consistent the user might type the other
255
so we force it to uppercase
256
running python.exe under cmd.exe return capital C:\\
257
running win32 python inside a cygwin shell returns lowercase c:\\
259
drive, path = _nt_splitdrive(path)
260
return drive.upper() + path
263
def _win32_abspath(path):
264
# Real _nt_abspath doesn't have a problem with a unicode cwd
265
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
268
def _win98_abspath(path):
269
"""Return the absolute version of a path.
270
Windows 98 safe implementation (python reimplementation
271
of Win32 API function GetFullPathNameW)
276
# \\HOST\path => //HOST/path
277
# //HOST/path => //HOST/path
278
# path => C:/cwd/path
281
# check for absolute path
282
drive = _nt_splitdrive(path)[0]
283
if drive == '' and path[:2] not in('//','\\\\'):
285
# we cannot simply os.path.join cwd and path
286
# because os.path.join('C:','/path') produce '/path'
287
# and this is incorrect
288
if path[:1] in ('/','\\'):
289
cwd = _nt_splitdrive(cwd)[0]
291
path = cwd + '\\' + path
292
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
295
def _win32_realpath(path):
296
# Real _nt_realpath doesn't have a problem with a unicode cwd
297
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
300
def _win32_pathjoin(*args):
301
return _nt_join(*args).replace('\\', '/')
304
def _win32_normpath(path):
305
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
309
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
312
def _win32_mkdtemp(*args, **kwargs):
313
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
316
def _win32_rename(old, new):
317
"""We expect to be able to atomically replace 'new' with old.
319
On win32, if new exists, it must be moved out of the way first,
323
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
325
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
326
# If we try to rename a non-existant file onto cwd, we get
327
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
328
# if the old path doesn't exist, sometimes we get EACCES
329
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
335
return unicodedata.normalize('NFC', os.getcwdu())
338
# Default is to just use the python builtins, but these can be rebound on
339
# particular platforms.
340
abspath = _posix_abspath
341
realpath = _posix_realpath
342
pathjoin = os.path.join
343
normpath = os.path.normpath
346
dirname = os.path.dirname
347
basename = os.path.basename
348
split = os.path.split
349
splitext = os.path.splitext
350
# These were already imported into local scope
351
# mkdtemp = tempfile.mkdtemp
352
# rmtree = shutil.rmtree
354
MIN_ABS_PATHLENGTH = 1
357
if sys.platform == 'win32':
358
if win32utils.winver == 'Windows 98':
359
abspath = _win98_abspath
361
abspath = _win32_abspath
362
realpath = _win32_realpath
363
pathjoin = _win32_pathjoin
364
normpath = _win32_normpath
365
getcwd = _win32_getcwd
366
mkdtemp = _win32_mkdtemp
367
rename = _win32_rename
369
MIN_ABS_PATHLENGTH = 3
371
def _win32_delete_readonly(function, path, excinfo):
372
"""Error handler for shutil.rmtree function [for win32]
373
Helps to remove files and dirs marked as read-only.
375
exception = excinfo[1]
376
if function in (os.remove, os.rmdir) \
377
and isinstance(exception, OSError) \
378
and exception.errno == errno.EACCES:
384
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
385
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
386
return shutil.rmtree(path, ignore_errors, onerror)
387
elif sys.platform == 'darwin':
391
def get_terminal_encoding():
392
"""Find the best encoding for printing to the screen.
394
This attempts to check both sys.stdout and sys.stdin to see
395
what encoding they are in, and if that fails it falls back to
396
osutils.get_user_encoding().
397
The problem is that on Windows, locale.getpreferredencoding()
398
is not the same encoding as that used by the console:
399
http://mail.python.org/pipermail/python-list/2003-May/162357.html
401
On my standard US Windows XP, the preferred encoding is
402
cp1252, but the console is cp437
404
from bzrlib.trace import mutter
405
output_encoding = getattr(sys.stdout, 'encoding', None)
406
if not output_encoding:
407
input_encoding = getattr(sys.stdin, 'encoding', None)
408
if not input_encoding:
409
output_encoding = get_user_encoding()
410
mutter('encoding stdout as osutils.get_user_encoding() %r',
413
output_encoding = input_encoding
414
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
416
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
417
if output_encoding == 'cp0':
418
# invalid encoding (cp0 means 'no codepage' on Windows)
419
output_encoding = get_user_encoding()
420
mutter('cp0 is invalid encoding.'
421
' encoding stdout as osutils.get_user_encoding() %r',
425
codecs.lookup(output_encoding)
427
sys.stderr.write('bzr: warning:'
428
' unknown terminal encoding %s.\n'
429
' Using encoding %s instead.\n'
430
% (output_encoding, get_user_encoding())
432
output_encoding = get_user_encoding()
434
return output_encoding
437
def normalizepath(f):
438
if getattr(os.path, 'realpath', None) is not None:
442
[p,e] = os.path.split(f)
443
if e == "" or e == "." or e == "..":
446
return pathjoin(F(p), e)
241
643
def local_time_offset(t=None):
242
644
"""Return offset of local zone from GMT, either at present or at time t."""
243
# python2.3 localtime() can't take None
247
if time.localtime(t).tm_isdst and time.daylight:
250
return -time.timezone
253
def format_date(t, offset=0, timezone='original'):
254
## TODO: Perhaps a global option to use either universal or local time?
255
## Or perhaps just let people set $TZ?
256
assert isinstance(t, float)
647
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
648
return offset.days * 86400 + offset.seconds
650
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
652
def format_date(t, offset=0, timezone='original', date_fmt=None,
654
"""Return a formatted date string.
656
:param t: Seconds since the epoch.
657
:param offset: Timezone offset in seconds east of utc.
658
:param timezone: How to display the time: 'utc', 'original' for the
659
timezone specified by offset, or 'local' for the process's current
661
:param date_fmt: strftime format.
662
:param show_offset: Whether to append the timezone.
664
(date_fmt, tt, offset_str) = \
665
_format_date(t, offset, timezone, date_fmt, show_offset)
666
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
667
date_str = time.strftime(date_fmt, tt)
668
return date_str + offset_str
670
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
672
"""Return an unicode date string formatted according to the current locale.
674
:param t: Seconds since the epoch.
675
:param offset: Timezone offset in seconds east of utc.
676
:param timezone: How to display the time: 'utc', 'original' for the
677
timezone specified by offset, or 'local' for the process's current
679
:param date_fmt: strftime format.
680
:param show_offset: Whether to append the timezone.
682
(date_fmt, tt, offset_str) = \
683
_format_date(t, offset, timezone, date_fmt, show_offset)
684
date_str = time.strftime(date_fmt, tt)
685
if not isinstance(date_str, unicode):
686
date_str = date_str.decode(bzrlib.user_encoding, 'replace')
687
return date_str + offset_str
689
def _format_date(t, offset, timezone, date_fmt, show_offset):
258
690
if timezone == 'utc':
259
691
tt = time.gmtime(t)
261
693
elif timezone == 'original':
264
696
tt = time.gmtime(t + offset)
265
697
elif timezone == 'local':
266
698
tt = time.localtime(t)
267
699
offset = local_time_offset(t)
269
bailout("unsupported timezone format %r",
270
['options are "utc", "original", "local"'])
272
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
273
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
701
raise errors.UnsupportedTimezoneFormat(timezone)
703
date_fmt = "%a %Y-%m-%d %H:%M:%S"
705
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
708
return (date_fmt, tt, offset_str)
276
711
def compact_date(when):
277
712
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
715
def format_delta(delta):
716
"""Get a nice looking string for a time delta.
718
:param delta: The time difference in seconds, can be positive or negative.
719
positive indicates time in the past, negative indicates time in the
720
future. (usually time.time() - stored_time)
721
:return: String formatted to show approximate resolution
727
direction = 'in the future'
731
if seconds < 90: # print seconds up to 90 seconds
733
return '%d second %s' % (seconds, direction,)
735
return '%d seconds %s' % (seconds, direction)
737
minutes = int(seconds / 60)
738
seconds -= 60 * minutes
743
if minutes < 90: # print minutes, seconds up to 90 minutes
745
return '%d minute, %d second%s %s' % (
746
minutes, seconds, plural_seconds, direction)
748
return '%d minutes, %d second%s %s' % (
749
minutes, seconds, plural_seconds, direction)
751
hours = int(minutes / 60)
752
minutes -= 60 * hours
759
return '%d hour, %d minute%s %s' % (hours, minutes,
760
plural_minutes, direction)
761
return '%d hours, %d minute%s %s' % (hours, minutes,
762
plural_minutes, direction)
282
765
"""Return size of given open file."""
283
766
return os.fstat(f.fileno())[ST_SIZE]
286
if hasattr(os, 'urandom'): # python 2.4 and later
769
# Define rand_bytes based on platform.
771
# Python 2.4 and later have os.urandom,
772
# but it doesn't work on some arches
287
774
rand_bytes = os.urandom
289
# FIXME: No good on non-Linux
290
_rand_file = file('/dev/urandom', 'rb')
291
rand_bytes = _rand_file.read
775
except (NotImplementedError, AttributeError):
776
# If python doesn't have os.urandom, or it doesn't work,
777
# then try to first pull random data from /dev/urandom
779
rand_bytes = file('/dev/urandom', 'rb').read
780
# Otherwise, use this hack as a last resort
781
except (IOError, OSError):
782
# not well seeded, but better than nothing
787
s += chr(random.randint(0, 255))
792
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
794
"""Return a random string of num alphanumeric characters
796
The result only contains lowercase chars because it may be used on
797
case-insensitive filesystems.
800
for raw_byte in rand_bytes(num):
801
s += ALNUM[ord(raw_byte) % 36]
294
805
## TODO: We could later have path objects that remember their list
295
806
## decomposition (might be too tricksy though.)
297
808
def splitpath(p):
298
"""Turn string into list of parts.
304
>>> splitpath('a/./b')
306
>>> splitpath('a/.b')
308
>>> splitpath('a/../b')
309
Traceback (most recent call last):
311
BzrError: ("sorry, '..' not allowed in path", [])
313
assert isinstance(p, types.StringTypes)
809
"""Turn string into list of parts."""
315
810
# split on either delimiter because people might use either on
317
812
ps = re.split(r'[\\/]', p)
322
bailout("sorry, %r not allowed in path" % f)
817
raise errors.BzrError("sorry, %r not allowed in path" % f)
323
818
elif (f == '.') or (f == ''):
330
assert isinstance(p, list)
332
if (f == '..') or (f == None) or (f == ''):
333
bailout("sorry, %r not allowed in path" % f)
334
return os.path.join(*p)
337
def appendpath(p1, p2):
341
return os.path.join(p1, p2)
344
def extern_command(cmd, ignore_errors = False):
345
mutter('external command: %s' % `cmd`)
347
if not ignore_errors:
348
bailout('command failed')
827
if (f == '..') or (f is None) or (f == ''):
828
raise errors.BzrError("sorry, %r not allowed in path" % f)
833
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
835
from bzrlib._chunks_to_lines_py import chunks_to_lines
839
"""Split s into lines, but without removing the newline characters."""
840
# Trivially convert a fulltext into a 'chunked' representation, and let
841
# chunks_to_lines do the heavy lifting.
842
if isinstance(s, str):
843
# chunks_to_lines only supports 8-bit strings
844
return chunks_to_lines([s])
846
return _split_lines(s)
850
"""Split s into lines, but without removing the newline characters.
852
This supports Unicode or plain string objects.
854
lines = s.split('\n')
855
result = [line + '\n' for line in lines[:-1]]
857
result.append(lines[-1])
861
def hardlinks_good():
862
return sys.platform not in ('win32', 'cygwin', 'darwin')
865
def link_or_copy(src, dest):
866
"""Hardlink a file, or copy it if it can't be hardlinked."""
867
if not hardlinks_good():
868
shutil.copyfile(src, dest)
872
except (OSError, IOError), e:
873
if e.errno != errno.EXDEV:
875
shutil.copyfile(src, dest)
878
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
879
# Forgiveness than Permission (EAFP) because:
880
# - root can damage a solaris file system by using unlink,
881
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
882
# EACCES, OSX: EPERM) when invoked on a directory.
883
def delete_any(path):
884
"""Delete a file or directory."""
885
if isdir(path): # Takes care of symlinks
892
if getattr(os, 'symlink', None) is not None:
899
if getattr(os, 'link', None) is not None:
905
def host_os_dereferences_symlinks():
906
return (has_symlinks()
907
and sys.platform not in ('cygwin', 'win32'))
910
def contains_whitespace(s):
911
"""True if there are any whitespace characters in s."""
912
# string.whitespace can include '\xa0' in certain locales, because it is
913
# considered "non-breaking-space" as part of ISO-8859-1. But it
914
# 1) Isn't a breaking whitespace
915
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
917
# 3) '\xa0' isn't unicode safe since it is >128.
919
# This should *not* be a unicode set of characters in case the source
920
# string is not a Unicode string. We can auto-up-cast the characters since
921
# they are ascii, but we don't want to auto-up-cast the string in case it
923
for ch in ' \t\n\r\v\f':
930
def contains_linebreaks(s):
931
"""True if there is any vertical whitespace in s."""
939
def relpath(base, path):
940
"""Return path relative to base, or raise exception.
942
The path may be either an absolute path or a path relative to the
943
current working directory.
945
os.path.commonprefix (python2.4) has a bad bug that it works just
946
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
950
if len(base) < MIN_ABS_PATHLENGTH:
951
# must have space for e.g. a drive letter
952
raise ValueError('%r is too short to calculate a relative path'
959
while len(head) >= len(base):
962
head, tail = os.path.split(head)
966
raise errors.PathNotChild(rp, base)
974
def _cicp_canonical_relpath(base, path):
975
"""Return the canonical path relative to base.
977
Like relpath, but on case-insensitive-case-preserving file-systems, this
978
will return the relpath as stored on the file-system rather than in the
979
case specified in the input string, for all existing portions of the path.
981
This will cause O(N) behaviour if called for every path in a tree; if you
982
have a number of paths to convert, you should use canonical_relpaths().
984
# TODO: it should be possible to optimize this for Windows by using the
985
# win32 API FindFiles function to look for the specified name - but using
986
# os.listdir() still gives us the correct, platform agnostic semantics in
989
rel = relpath(base, path)
990
# '.' will have been turned into ''
994
abs_base = abspath(base)
996
_listdir = os.listdir
998
# use an explicit iterator so we can easily consume the rest on early exit.
999
bit_iter = iter(rel.split('/'))
1000
for bit in bit_iter:
1002
for look in _listdir(current):
1003
if lbit == look.lower():
1004
current = pathjoin(current, look)
1007
# got to the end, nothing matched, so we just return the
1008
# non-existing bits as they were specified (the filename may be
1009
# the target of a move, for example).
1010
current = pathjoin(current, bit, *list(bit_iter))
1012
return current[len(abs_base)+1:]
1014
# XXX - TODO - we need better detection/integration of case-insensitive
1015
# file-systems; Linux often sees FAT32 devices, for example, so could
1016
# probably benefit from the same basic support there. For now though, only
1017
# Windows gets that support, and it gets it for *all* file-systems!
1018
if sys.platform == "win32":
1019
canonical_relpath = _cicp_canonical_relpath
1021
canonical_relpath = relpath
1023
def canonical_relpaths(base, paths):
1024
"""Create an iterable to canonicalize a sequence of relative paths.
1026
The intent is for this implementation to use a cache, vastly speeding
1027
up multiple transformations in the same directory.
1029
# but for now, we haven't optimized...
1030
return [canonical_relpath(base, p) for p in paths]
1032
def safe_unicode(unicode_or_utf8_string):
1033
"""Coerce unicode_or_utf8_string into unicode.
1035
If it is unicode, it is returned.
1036
Otherwise it is decoded from utf-8. If a decoding error
1037
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
1038
as a BzrBadParameter exception.
1040
if isinstance(unicode_or_utf8_string, unicode):
1041
return unicode_or_utf8_string
1043
return unicode_or_utf8_string.decode('utf8')
1044
except UnicodeDecodeError:
1045
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1048
def safe_utf8(unicode_or_utf8_string):
1049
"""Coerce unicode_or_utf8_string to a utf8 string.
1051
If it is a str, it is returned.
1052
If it is Unicode, it is encoded into a utf-8 string.
1054
if isinstance(unicode_or_utf8_string, str):
1055
# TODO: jam 20070209 This is overkill, and probably has an impact on
1056
# performance if we are dealing with lots of apis that want a
1059
# Make sure it is a valid utf-8 string
1060
unicode_or_utf8_string.decode('utf-8')
1061
except UnicodeDecodeError:
1062
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1063
return unicode_or_utf8_string
1064
return unicode_or_utf8_string.encode('utf-8')
1067
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1068
' Revision id generators should be creating utf8'
1072
def safe_revision_id(unicode_or_utf8_string, warn=True):
1073
"""Revision ids should now be utf8, but at one point they were unicode.
1075
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1077
:param warn: Functions that are sanitizing user data can set warn=False
1078
:return: None or a utf8 revision id.
1080
if (unicode_or_utf8_string is None
1081
or unicode_or_utf8_string.__class__ == str):
1082
return unicode_or_utf8_string
1084
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1086
return cache_utf8.encode(unicode_or_utf8_string)
1089
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1090
' generators should be creating utf8 file ids.')
1093
def safe_file_id(unicode_or_utf8_string, warn=True):
1094
"""File ids should now be utf8, but at one point they were unicode.
1096
This is the same as safe_utf8, except it uses the cached encode functions
1097
to save a little bit of performance.
1099
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1101
:param warn: Functions that are sanitizing user data can set warn=False
1102
:return: None or a utf8 file id.
1104
if (unicode_or_utf8_string is None
1105
or unicode_or_utf8_string.__class__ == str):
1106
return unicode_or_utf8_string
1108
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1110
return cache_utf8.encode(unicode_or_utf8_string)
1113
_platform_normalizes_filenames = False
1114
if sys.platform == 'darwin':
1115
_platform_normalizes_filenames = True
1118
def normalizes_filenames():
1119
"""Return True if this platform normalizes unicode filenames.
1121
Mac OSX does, Windows/Linux do not.
1123
return _platform_normalizes_filenames
1126
def _accessible_normalized_filename(path):
1127
"""Get the unicode normalized path, and if you can access the file.
1129
On platforms where the system normalizes filenames (Mac OSX),
1130
you can access a file by any path which will normalize correctly.
1131
On platforms where the system does not normalize filenames
1132
(Windows, Linux), you have to access a file by its exact path.
1134
Internally, bzr only supports NFC normalization, since that is
1135
the standard for XML documents.
1137
So return the normalized path, and a flag indicating if the file
1138
can be accessed by that path.
1141
return unicodedata.normalize('NFC', unicode(path)), True
1144
def _inaccessible_normalized_filename(path):
1145
__doc__ = _accessible_normalized_filename.__doc__
1147
normalized = unicodedata.normalize('NFC', unicode(path))
1148
return normalized, normalized == path
1151
if _platform_normalizes_filenames:
1152
normalized_filename = _accessible_normalized_filename
1154
normalized_filename = _inaccessible_normalized_filename
1157
def terminal_width():
1158
"""Return estimated terminal width."""
1159
if sys.platform == 'win32':
1160
return win32utils.get_console_size()[0]
1163
import struct, fcntl, termios
1164
s = struct.pack('HHHH', 0, 0, 0, 0)
1165
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1166
width = struct.unpack('HHHH', x)[1]
1171
width = int(os.environ['COLUMNS'])
1180
def supports_executable():
1181
return sys.platform != "win32"
1184
def supports_posix_readonly():
1185
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1187
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1188
directory controls creation/deletion, etc.
1190
And under win32, readonly means that the directory itself cannot be
1191
deleted. The contents of a readonly directory can be changed, unlike POSIX
1192
where files in readonly directories cannot be added, deleted or renamed.
1194
return sys.platform != "win32"
1197
def set_or_unset_env(env_variable, value):
1198
"""Modify the environment, setting or removing the env_variable.
1200
:param env_variable: The environment variable in question
1201
:param value: The value to set the environment to. If None, then
1202
the variable will be removed.
1203
:return: The original value of the environment variable.
1205
orig_val = os.environ.get(env_variable)
1207
if orig_val is not None:
1208
del os.environ[env_variable]
1210
if isinstance(value, unicode):
1211
value = value.encode(get_user_encoding())
1212
os.environ[env_variable] = value
1216
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1219
def check_legal_path(path):
1220
"""Check whether the supplied path is legal.
1221
This is only required on Windows, so we don't test on other platforms
1224
if sys.platform != "win32":
1226
if _validWin32PathRE.match(path) is None:
1227
raise errors.IllegalPath(path)
1230
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1232
def _is_error_enotdir(e):
1233
"""Check if this exception represents ENOTDIR.
1235
Unfortunately, python is very inconsistent about the exception
1236
here. The cases are:
1237
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1238
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1239
which is the windows error code.
1240
3) Windows, Python2.5 uses errno == EINVAL and
1241
winerror == ERROR_DIRECTORY
1243
:param e: An Exception object (expected to be OSError with an errno
1244
attribute, but we should be able to cope with anything)
1245
:return: True if this represents an ENOTDIR error. False otherwise.
1247
en = getattr(e, 'errno', None)
1248
if (en == errno.ENOTDIR
1249
or (sys.platform == 'win32'
1250
and (en == _WIN32_ERROR_DIRECTORY
1251
or (en == errno.EINVAL
1252
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1258
def walkdirs(top, prefix=""):
1259
"""Yield data about all the directories in a tree.
1261
This yields all the data about the contents of a directory at a time.
1262
After each directory has been yielded, if the caller has mutated the list
1263
to exclude some directories, they are then not descended into.
1265
The data yielded is of the form:
1266
((directory-relpath, directory-path-from-top),
1267
[(relpath, basename, kind, lstat, path-from-top), ...]),
1268
- directory-relpath is the relative path of the directory being returned
1269
with respect to top. prefix is prepended to this.
1270
- directory-path-from-root is the path including top for this directory.
1271
It is suitable for use with os functions.
1272
- relpath is the relative path within the subtree being walked.
1273
- basename is the basename of the path
1274
- kind is the kind of the file now. If unknown then the file is not
1275
present within the tree - but it may be recorded as versioned. See
1277
- lstat is the stat data *if* the file was statted.
1278
- planned, not implemented:
1279
path_from_tree_root is the path from the root of the tree.
1281
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1282
allows one to walk a subtree but get paths that are relative to a tree
1284
:return: an iterator over the dirs.
1286
#TODO there is a bit of a smell where the results of the directory-
1287
# summary in this, and the path from the root, may not agree
1288
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1289
# potentially confusing output. We should make this more robust - but
1290
# not at a speed cost. RBC 20060731
1292
_directory = _directory_kind
1293
_listdir = os.listdir
1294
_kind_from_mode = file_kind_from_stat_mode
1295
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1297
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1298
relroot, _, _, _, top = pending.pop()
1300
relprefix = relroot + u'/'
1303
top_slash = top + u'/'
1306
append = dirblock.append
1308
names = sorted(_listdir(top))
1310
if not _is_error_enotdir(e):
1314
abspath = top_slash + name
1315
statvalue = _lstat(abspath)
1316
kind = _kind_from_mode(statvalue.st_mode)
1317
append((relprefix + name, name, kind, statvalue, abspath))
1318
yield (relroot, top), dirblock
1320
# push the user specified dirs from dirblock
1321
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1324
class DirReader(object):
1325
"""An interface for reading directories."""
1327
def top_prefix_to_starting_dir(self, top, prefix=""):
1328
"""Converts top and prefix to a starting dir entry
1330
:param top: A utf8 path
1331
:param prefix: An optional utf8 path to prefix output relative paths
1333
:return: A tuple starting with prefix, and ending with the native
1336
raise NotImplementedError(self.top_prefix_to_starting_dir)
1338
def read_dir(self, prefix, top):
1339
"""Read a specific dir.
1341
:param prefix: A utf8 prefix to be preprended to the path basenames.
1342
:param top: A natively encoded path to read.
1343
:return: A list of the directories contents. Each item contains:
1344
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1346
raise NotImplementedError(self.read_dir)
1349
_selected_dir_reader = None
1352
def _walkdirs_utf8(top, prefix=""):
1353
"""Yield data about all the directories in a tree.
1355
This yields the same information as walkdirs() only each entry is yielded
1356
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1357
are returned as exact byte-strings.
1359
:return: yields a tuple of (dir_info, [file_info])
1360
dir_info is (utf8_relpath, path-from-top)
1361
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1362
if top is an absolute path, path-from-top is also an absolute path.
1363
path-from-top might be unicode or utf8, but it is the correct path to
1364
pass to os functions to affect the file in question. (such as os.lstat)
1366
global _selected_dir_reader
1367
if _selected_dir_reader is None:
1368
fs_encoding = _fs_enc.upper()
1369
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1370
# Win98 doesn't have unicode apis like FindFirstFileW
1371
# TODO: We possibly could support Win98 by falling back to the
1372
# original FindFirstFile, and using TCHAR instead of WCHAR,
1373
# but that gets a bit tricky, and requires custom compiling
1376
from bzrlib._walkdirs_win32 import Win32ReadDir
1378
_selected_dir_reader = UnicodeDirReader()
1380
_selected_dir_reader = Win32ReadDir()
1381
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1382
# ANSI_X3.4-1968 is a form of ASCII
1383
_selected_dir_reader = UnicodeDirReader()
1386
from bzrlib._readdir_pyx import UTF8DirReader
1388
# No optimised code path
1389
_selected_dir_reader = UnicodeDirReader()
1391
_selected_dir_reader = UTF8DirReader()
1392
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1393
# But we don't actually uses 1-3 in pending, so set them to None
1394
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1395
read_dir = _selected_dir_reader.read_dir
1396
_directory = _directory_kind
1398
relroot, _, _, _, top = pending[-1].pop()
1401
dirblock = sorted(read_dir(relroot, top))
1402
yield (relroot, top), dirblock
1403
# push the user specified dirs from dirblock
1404
next = [d for d in reversed(dirblock) if d[2] == _directory]
1406
pending.append(next)
1409
class UnicodeDirReader(DirReader):
1410
"""A dir reader for non-utf8 file systems, which transcodes."""
1412
__slots__ = ['_utf8_encode']
1415
self._utf8_encode = codecs.getencoder('utf8')
1417
def top_prefix_to_starting_dir(self, top, prefix=""):
1418
"""See DirReader.top_prefix_to_starting_dir."""
1419
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1421
def read_dir(self, prefix, top):
1422
"""Read a single directory from a non-utf8 file system.
1424
top, and the abspath element in the output are unicode, all other paths
1425
are utf8. Local disk IO is done via unicode calls to listdir etc.
1427
This is currently the fallback code path when the filesystem encoding is
1428
not UTF-8. It may be better to implement an alternative so that we can
1429
safely handle paths that are not properly decodable in the current
1432
See DirReader.read_dir for details.
1434
_utf8_encode = self._utf8_encode
1436
_listdir = os.listdir
1437
_kind_from_mode = file_kind_from_stat_mode
1440
relprefix = prefix + '/'
1443
top_slash = top + u'/'
1446
append = dirblock.append
1447
for name in sorted(_listdir(top)):
1449
name_utf8 = _utf8_encode(name)[0]
1450
except UnicodeDecodeError:
1451
raise errors.BadFilenameEncoding(
1452
_utf8_encode(relprefix)[0] + name, _fs_enc)
1453
abspath = top_slash + name
1454
statvalue = _lstat(abspath)
1455
kind = _kind_from_mode(statvalue.st_mode)
1456
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1460
def copy_tree(from_path, to_path, handlers={}):
1461
"""Copy all of the entries in from_path into to_path.
1463
:param from_path: The base directory to copy.
1464
:param to_path: The target directory. If it does not exist, it will
1466
:param handlers: A dictionary of functions, which takes a source and
1467
destinations for files, directories, etc.
1468
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1469
'file', 'directory', and 'symlink' should always exist.
1470
If they are missing, they will be replaced with 'os.mkdir()',
1471
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1473
# Now, just copy the existing cached tree to the new location
1474
# We use a cheap trick here.
1475
# Absolute paths are prefixed with the first parameter
1476
# relative paths are prefixed with the second.
1477
# So we can get both the source and target returned
1478
# without any extra work.
1480
def copy_dir(source, dest):
1483
def copy_link(source, dest):
1484
"""Copy the contents of a symlink"""
1485
link_to = os.readlink(source)
1486
os.symlink(link_to, dest)
1488
real_handlers = {'file':shutil.copy2,
1489
'symlink':copy_link,
1490
'directory':copy_dir,
1492
real_handlers.update(handlers)
1494
if not os.path.exists(to_path):
1495
real_handlers['directory'](from_path, to_path)
1497
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1498
for relpath, name, kind, st, abspath in entries:
1499
real_handlers[kind](abspath, relpath)
1502
def path_prefix_key(path):
1503
"""Generate a prefix-order path key for path.
1505
This can be used to sort paths in the same way that walkdirs does.
1507
return (dirname(path) , path)
1510
def compare_paths_prefix_order(path_a, path_b):
1511
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1512
key_a = path_prefix_key(path_a)
1513
key_b = path_prefix_key(path_b)
1514
return cmp(key_a, key_b)
1517
_cached_user_encoding = None
1520
def get_user_encoding(use_cache=True):
1521
"""Find out what the preferred user encoding is.
1523
This is generally the encoding that is used for command line parameters
1524
and file contents. This may be different from the terminal encoding
1525
or the filesystem encoding.
1527
:param use_cache: Enable cache for detected encoding.
1528
(This parameter is turned on by default,
1529
and required only for selftesting)
1531
:return: A string defining the preferred user encoding
1533
global _cached_user_encoding
1534
if _cached_user_encoding is not None and use_cache:
1535
return _cached_user_encoding
1537
if sys.platform == 'darwin':
1538
# python locale.getpreferredencoding() always return
1539
# 'mac-roman' on darwin. That's a lie.
1540
sys.platform = 'posix'
1542
if os.environ.get('LANG', None) is None:
1543
# If LANG is not set, we end up with 'ascii', which is bad
1544
# ('mac-roman' is more than ascii), so we set a default which
1545
# will give us UTF-8 (which appears to work in all cases on
1546
# OSX). Users are still free to override LANG of course, as
1547
# long as it give us something meaningful. This work-around
1548
# *may* not be needed with python 3k and/or OSX 10.5, but will
1549
# work with them too -- vila 20080908
1550
os.environ['LANG'] = 'en_US.UTF-8'
1553
sys.platform = 'darwin'
1558
user_encoding = locale.getpreferredencoding()
1559
except locale.Error, e:
1560
sys.stderr.write('bzr: warning: %s\n'
1561
' Could not determine what text encoding to use.\n'
1562
' This error usually means your Python interpreter\n'
1563
' doesn\'t support the locale set by $LANG (%s)\n'
1564
" Continuing with ascii encoding.\n"
1565
% (e, os.environ.get('LANG')))
1566
user_encoding = 'ascii'
1568
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1569
# treat that as ASCII, and not support printing unicode characters to the
1572
# For python scripts run under vim, we get '', so also treat that as ASCII
1573
if user_encoding in (None, 'cp0', ''):
1574
user_encoding = 'ascii'
1578
codecs.lookup(user_encoding)
1580
sys.stderr.write('bzr: warning:'
1581
' unknown encoding %s.'
1582
' Continuing with ascii encoding.\n'
1585
user_encoding = 'ascii'
1588
_cached_user_encoding = user_encoding
1590
return user_encoding
1593
def get_host_name():
1594
"""Return the current unicode host name.
1596
This is meant to be used in place of socket.gethostname() because that
1597
behaves inconsistently on different platforms.
1599
if sys.platform == "win32":
1601
return win32utils.get_host_name()
1604
return socket.gethostname().decode(get_user_encoding())
1607
def recv_all(socket, bytes):
1608
"""Receive an exact number of bytes.
1610
Regular Socket.recv() may return less than the requested number of bytes,
1611
dependning on what's in the OS buffer. MSG_WAITALL is not available
1612
on all platforms, but this should work everywhere. This will return
1613
less than the requested amount if the remote end closes.
1615
This isn't optimized and is intended mostly for use in testing.
1618
while len(b) < bytes:
1619
new = until_no_eintr(socket.recv, bytes - len(b))
1626
def send_all(socket, bytes):
1627
"""Send all bytes on a socket.
1629
Regular socket.sendall() can give socket error 10053 on Windows. This
1630
implementation sends no more than 64k at a time, which avoids this problem.
1633
for pos in xrange(0, len(bytes), chunk_size):
1634
until_no_eintr(socket.sendall, bytes[pos:pos+chunk_size])
1637
def dereference_path(path):
1638
"""Determine the real path to a file.
1640
All parent elements are dereferenced. But the file itself is not
1642
:param path: The original path. May be absolute or relative.
1643
:return: the real path *to* the file
1645
parent, base = os.path.split(path)
1646
# The pathjoin for '.' is a workaround for Python bug #1213894.
1647
# (initial path components aren't dereferenced)
1648
return pathjoin(realpath(pathjoin('.', parent)), base)
1651
def supports_mapi():
1652
"""Return True if we can use MAPI to launch a mail client."""
1653
return sys.platform == "win32"
1656
def resource_string(package, resource_name):
1657
"""Load a resource from a package and return it as a string.
1659
Note: Only packages that start with bzrlib are currently supported.
1661
This is designed to be a lightweight implementation of resource
1662
loading in a way which is API compatible with the same API from
1664
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1665
If and when pkg_resources becomes a standard library, this routine
1668
# Check package name is within bzrlib
1669
if package == "bzrlib":
1670
resource_relpath = resource_name
1671
elif package.startswith("bzrlib."):
1672
package = package[len("bzrlib."):].replace('.', os.sep)
1673
resource_relpath = pathjoin(package, resource_name)
1675
raise errors.BzrError('resource package %s not in bzrlib' % package)
1677
# Map the resource to a file and read its contents
1678
base = dirname(bzrlib.__file__)
1679
if getattr(sys, 'frozen', None): # bzr.exe
1680
base = abspath(pathjoin(base, '..', '..'))
1681
filename = pathjoin(base, resource_relpath)
1682
return open(filename, 'rU').read()
1685
def file_kind_from_stat_mode_thunk(mode):
1686
global file_kind_from_stat_mode
1687
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1689
from bzrlib._readdir_pyx import UTF8DirReader
1690
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1692
from bzrlib._readdir_py import (
1693
_kind_from_mode as file_kind_from_stat_mode
1695
return file_kind_from_stat_mode(mode)
1696
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1699
def file_kind(f, _lstat=os.lstat):
1701
return file_kind_from_stat_mode(_lstat(f).st_mode)
1703
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1704
raise errors.NoSuchFile(f)
1708
def until_no_eintr(f, *a, **kw):
1709
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1710
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1714
except (IOError, OSError), e:
1715
if e.errno == errno.EINTR:
1720
if sys.platform == "win32":
1723
return msvcrt.getch()
1728
fd = sys.stdin.fileno()
1729
settings = termios.tcgetattr(fd)
1732
ch = sys.stdin.read(1)
1734
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
added
removed
bzrlib/osutils.py
