1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
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,
43
from tempfile import (
57
from bzrlib import symbol_versioning
60
# On win32, O_BINARY is used to indicate the file should
61
# be opened in binary mode, rather than text mode.
62
# On other platforms, O_BINARY doesn't exist, because
63
# they always open in binary mode, so it is okay to
64
# OR with 0 on those platforms
65
O_BINARY = getattr(os, 'O_BINARY', 0)
68
def make_readonly(filename):
69
"""Make a filename read-only."""
70
mod = os.lstat(filename).st_mode
71
if not stat.S_ISLNK(mod):
73
os.chmod(filename, mod)
76
def make_writable(filename):
77
mod = os.lstat(filename).st_mode
78
if not stat.S_ISLNK(mod):
80
os.chmod(filename, mod)
83
def minimum_path_selection(paths):
84
"""Return the smallset subset of paths which are outside paths.
86
:param paths: A container (and hence not None) of paths.
87
:return: A set of paths sufficient to include everything in paths via
88
is_inside_any, drawn from the paths parameter.
93
other_paths = paths.difference([path])
94
if not is_inside_any(other_paths, path):
95
# this is a top level path, we must check it.
96
search_paths.add(path)
104
"""Return a quoted filename filename
106
This previously used backslash quoting, but that works poorly on
108
# TODO: I'm not really sure this is the best format either.x
110
if _QUOTE_RE is None:
111
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
113
if _QUOTE_RE.search(f):
119
_directory_kind = 'directory'
122
stat.S_IFDIR:_directory_kind,
123
stat.S_IFCHR:'chardev',
124
stat.S_IFBLK:'block',
127
stat.S_IFLNK:'symlink',
128
stat.S_IFSOCK:'socket',
132
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
133
"""Generate a file kind from a stat mode. This is used in walkdirs.
135
Its performance is critical: Do not mutate without careful benchmarking.
138
return _formats[stat_mode & 0170000]
143
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
145
return _mapper(_lstat(f).st_mode)
147
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
148
raise errors.NoSuchFile(f)
153
"""Return the current umask"""
154
# Assume that people aren't messing with the umask while running
155
# XXX: This is not thread safe, but there is no way to get the
156
# umask without setting it
164
_directory_kind: "/",
166
'tree-reference': '+',
170
def kind_marker(kind):
172
return _kind_marker_map[kind]
174
raise errors.BzrError('invalid file kind %r' % kind)
177
lexists = getattr(os.path, 'lexists', None)
181
stat = getattr(os, 'lstat', os.stat)
185
if e.errno == errno.ENOENT:
188
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
191
def fancy_rename(old, new, rename_func, unlink_func):
192
"""A fancy rename, when you don't have atomic rename.
194
:param old: The old path, to rename from
195
:param new: The new path, to rename to
196
:param rename_func: The potentially non-atomic rename function
197
:param unlink_func: A way to delete the target file if the full rename succeeds
200
# sftp rename doesn't allow overwriting, so play tricks:
201
base = os.path.basename(new)
202
dirname = os.path.dirname(new)
203
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
204
tmp_name = pathjoin(dirname, tmp_name)
206
# Rename the file out of the way, but keep track if it didn't exist
207
# We don't want to grab just any exception
208
# something like EACCES should prevent us from continuing
209
# The downside is that the rename_func has to throw an exception
210
# with an errno = ENOENT, or NoSuchFile
213
rename_func(new, tmp_name)
214
except (errors.NoSuchFile,), e:
217
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
218
# function raises an IOError with errno is None when a rename fails.
219
# This then gets caught here.
220
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
223
if (getattr(e, 'errno', None) is None
224
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
232
# This may throw an exception, in which case success will
234
rename_func(old, new)
236
except (IOError, OSError), e:
237
# source and target may be aliases of each other (e.g. on a
238
# case-insensitive filesystem), so we may have accidentally renamed
239
# source by when we tried to rename target
240
if not (file_existed and e.errno in (None, errno.ENOENT)):
244
# If the file used to exist, rename it back into place
245
# otherwise just delete it from the tmp location
247
unlink_func(tmp_name)
249
rename_func(tmp_name, new)
252
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
253
# choke on a Unicode string containing a relative path if
254
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
256
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
257
def _posix_abspath(path):
258
# jam 20060426 rather than encoding to fsencoding
259
# copy posixpath.abspath, but use os.getcwdu instead
260
if not posixpath.isabs(path):
261
path = posixpath.join(getcwd(), path)
262
return posixpath.normpath(path)
265
def _posix_realpath(path):
266
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
269
def _win32_fixdrive(path):
270
"""Force drive letters to be consistent.
272
win32 is inconsistent whether it returns lower or upper case
273
and even if it was consistent the user might type the other
274
so we force it to uppercase
275
running python.exe under cmd.exe return capital C:\\
276
running win32 python inside a cygwin shell returns lowercase c:\\
278
drive, path = _nt_splitdrive(path)
279
return drive.upper() + path
282
def _win32_abspath(path):
283
global _win32_abspath
284
if win32utils.winver == 'Windows 98':
285
_win32_abspath = _win98_abspath
287
_win32_abspath = _real_win32_abspath
288
return _win32_abspath(path)
291
def _real_win32_abspath(path):
292
# Real _nt_abspath doesn't have a problem with a unicode cwd
293
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
296
def _win98_abspath(path):
297
"""Return the absolute version of a path.
298
Windows 98 safe implementation (python reimplementation
299
of Win32 API function GetFullPathNameW)
304
# \\HOST\path => //HOST/path
305
# //HOST/path => //HOST/path
306
# path => C:/cwd/path
309
# check for absolute path
310
drive = _nt_splitdrive(path)[0]
311
if drive == '' and path[:2] not in('//','\\\\'):
313
# we cannot simply os.path.join cwd and path
314
# because os.path.join('C:','/path') produce '/path'
315
# and this is incorrect
316
if path[:1] in ('/','\\'):
317
cwd = _nt_splitdrive(cwd)[0]
319
path = cwd + '\\' + path
320
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
323
def _win32_realpath(path):
324
# Real _nt_realpath doesn't have a problem with a unicode cwd
325
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
328
def _win32_pathjoin(*args):
329
return _nt_join(*args).replace('\\', '/')
332
def _win32_normpath(path):
333
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
337
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
340
def _win32_mkdtemp(*args, **kwargs):
341
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
344
def _win32_rename(old, new):
345
"""We expect to be able to atomically replace 'new' with old.
347
On win32, if new exists, it must be moved out of the way first,
351
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
353
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
354
# If we try to rename a non-existant file onto cwd, we get
355
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
356
# if the old path doesn't exist, sometimes we get EACCES
357
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
363
return unicodedata.normalize('NFC', os.getcwdu())
366
# Default is to just use the python builtins, but these can be rebound on
367
# particular platforms.
368
abspath = _posix_abspath
369
realpath = _posix_realpath
370
pathjoin = os.path.join
371
normpath = os.path.normpath
374
dirname = os.path.dirname
375
basename = os.path.basename
376
split = os.path.split
377
splitext = os.path.splitext
378
# These were already imported into local scope
379
# mkdtemp = tempfile.mkdtemp
380
# rmtree = shutil.rmtree
382
MIN_ABS_PATHLENGTH = 1
385
if sys.platform == 'win32':
386
abspath = _win32_abspath
387
realpath = _win32_realpath
388
pathjoin = _win32_pathjoin
389
normpath = _win32_normpath
390
getcwd = _win32_getcwd
391
mkdtemp = _win32_mkdtemp
392
rename = _win32_rename
394
MIN_ABS_PATHLENGTH = 3
396
def _win32_delete_readonly(function, path, excinfo):
397
"""Error handler for shutil.rmtree function [for win32]
398
Helps to remove files and dirs marked as read-only.
400
exception = excinfo[1]
401
if function in (os.remove, os.rmdir) \
402
and isinstance(exception, OSError) \
403
and exception.errno == errno.EACCES:
409
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
410
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
411
return shutil.rmtree(path, ignore_errors, onerror)
412
elif sys.platform == 'darwin':
416
def get_terminal_encoding():
417
"""Find the best encoding for printing to the screen.
419
This attempts to check both sys.stdout and sys.stdin to see
420
what encoding they are in, and if that fails it falls back to
421
osutils.get_user_encoding().
422
The problem is that on Windows, locale.getpreferredencoding()
423
is not the same encoding as that used by the console:
424
http://mail.python.org/pipermail/python-list/2003-May/162357.html
426
On my standard US Windows XP, the preferred encoding is
427
cp1252, but the console is cp437
429
from bzrlib.trace import mutter
430
output_encoding = getattr(sys.stdout, 'encoding', None)
431
if not output_encoding:
432
input_encoding = getattr(sys.stdin, 'encoding', None)
433
if not input_encoding:
434
output_encoding = get_user_encoding()
435
mutter('encoding stdout as osutils.get_user_encoding() %r',
438
output_encoding = input_encoding
439
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
441
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
442
if output_encoding == 'cp0':
443
# invalid encoding (cp0 means 'no codepage' on Windows)
444
output_encoding = get_user_encoding()
445
mutter('cp0 is invalid encoding.'
446
' encoding stdout as osutils.get_user_encoding() %r',
450
codecs.lookup(output_encoding)
452
sys.stderr.write('bzr: warning:'
453
' unknown terminal encoding %s.\n'
454
' Using encoding %s instead.\n'
455
% (output_encoding, get_user_encoding())
457
output_encoding = get_user_encoding()
459
return output_encoding
462
def normalizepath(f):
463
if getattr(os.path, 'realpath', None) is not None:
467
[p,e] = os.path.split(f)
468
if e == "" or e == "." or e == "..":
471
return pathjoin(F(p), e)
475
"""True if f is an accessible directory."""
477
return S_ISDIR(os.lstat(f)[ST_MODE])
483
"""True if f is a regular file."""
485
return S_ISREG(os.lstat(f)[ST_MODE])
490
"""True if f is a symlink."""
492
return S_ISLNK(os.lstat(f)[ST_MODE])
496
def is_inside(dir, fname):
497
"""True if fname is inside dir.
499
The parameters should typically be passed to osutils.normpath first, so
500
that . and .. and repeated slashes are eliminated, and the separators
501
are canonical for the platform.
503
The empty string as a dir name is taken as top-of-tree and matches
506
# XXX: Most callers of this can actually do something smarter by
507
# looking at the inventory
517
return fname.startswith(dir)
520
def is_inside_any(dir_list, fname):
521
"""True if fname is inside any of given dirs."""
522
for dirname in dir_list:
523
if is_inside(dirname, fname):
528
def is_inside_or_parent_of_any(dir_list, fname):
529
"""True if fname is a child or a parent of any of the given files."""
530
for dirname in dir_list:
531
if is_inside(dirname, fname) or is_inside(fname, dirname):
536
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768):
537
"""Copy contents of one file to another.
539
The read_length can either be -1 to read to end-of-file (EOF) or
540
it can specify the maximum number of bytes to read.
542
The buff_size represents the maximum size for each read operation
543
performed on from_file.
545
:return: The number of bytes copied.
549
# read specified number of bytes
551
while read_length > 0:
552
num_bytes_to_read = min(read_length, buff_size)
554
block = from_file.read(num_bytes_to_read)
560
actual_bytes_read = len(block)
561
read_length -= actual_bytes_read
562
length += actual_bytes_read
566
block = from_file.read(buff_size)
575
def pump_string_file(bytes, file_handle, segment_size=None):
576
"""Write bytes to file_handle in many smaller writes.
578
:param bytes: The string to write.
579
:param file_handle: The file to write to.
581
# Write data in chunks rather than all at once, because very large
582
# writes fail on some platforms (e.g. Windows with SMB mounted
585
segment_size = 5242880 # 5MB
586
segments = range(len(bytes) / segment_size + 1)
587
write = file_handle.write
588
for segment_index in segments:
589
segment = buffer(bytes, segment_index * segment_size, segment_size)
593
def file_iterator(input_file, readsize=32768):
595
b = input_file.read(readsize)
602
"""Calculate the hexdigest of an open file.
604
The file cursor should be already at the start.
616
def sha_file_by_name(fname):
617
"""Calculate the SHA1 of a file by reading the full text"""
619
f = os.open(fname, os.O_RDONLY | O_BINARY)
622
b = os.read(f, 1<<16)
630
def sha_strings(strings):
631
"""Return the sha-1 of concatenation of strings"""
632
# Do some hackery here to install an optimised version of this function on
633
# the first invocation of this function. (We don't define it like this
634
# initially so that we can avoid loading the sha module, which takes up to
635
# 2ms, unless we need to.)
637
def sha_strings(strings, _factory=sha.new):
638
"""Return the sha-1 of concatenation of strings"""
640
map(s.update, strings)
642
# Now that we've installed the real version, call it.
643
return sha_strings(strings)
648
def sha_string(f, _factory=sha.new):
649
return _factory(f).hexdigest()
654
def fingerprint_file(f):
656
return {'size': len(b),
657
'sha1': sha.new(b).hexdigest()}
660
def compare_files(a, b):
661
"""Returns true if equal in contents"""
672
def local_time_offset(t=None):
673
"""Return offset of local zone from GMT, either at present or at time t."""
676
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
677
return offset.days * 86400 + offset.seconds
679
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
681
def format_date(t, offset=0, timezone='original', date_fmt=None,
683
"""Return a formatted date string.
685
:param t: Seconds since the epoch.
686
:param offset: Timezone offset in seconds east of utc.
687
:param timezone: How to display the time: 'utc', 'original' for the
688
timezone specified by offset, or 'local' for the process's current
690
:param show_offset: Whether to append the timezone.
691
:param date_fmt: strftime format.
693
if timezone == 'utc':
696
elif timezone == 'original':
699
tt = time.gmtime(t + offset)
700
elif timezone == 'local':
701
tt = time.localtime(t)
702
offset = local_time_offset(t)
704
raise errors.UnsupportedTimezoneFormat(timezone)
706
date_fmt = "%a %Y-%m-%d %H:%M:%S"
708
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
711
# day of week depends on locale, so we do this ourself
712
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
713
return (time.strftime(date_fmt, tt) + offset_str)
716
def compact_date(when):
717
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
720
def format_delta(delta):
721
"""Get a nice looking string for a time delta.
723
:param delta: The time difference in seconds, can be positive or negative.
724
positive indicates time in the past, negative indicates time in the
725
future. (usually time.time() - stored_time)
726
:return: String formatted to show approximate resolution
732
direction = 'in the future'
736
if seconds < 90: # print seconds up to 90 seconds
738
return '%d second %s' % (seconds, direction,)
740
return '%d seconds %s' % (seconds, direction)
742
minutes = int(seconds / 60)
743
seconds -= 60 * minutes
748
if minutes < 90: # print minutes, seconds up to 90 minutes
750
return '%d minute, %d second%s %s' % (
751
minutes, seconds, plural_seconds, direction)
753
return '%d minutes, %d second%s %s' % (
754
minutes, seconds, plural_seconds, direction)
756
hours = int(minutes / 60)
757
minutes -= 60 * hours
764
return '%d hour, %d minute%s %s' % (hours, minutes,
765
plural_minutes, direction)
766
return '%d hours, %d minute%s %s' % (hours, minutes,
767
plural_minutes, direction)
770
"""Return size of given open file."""
771
return os.fstat(f.fileno())[ST_SIZE]
774
# Define rand_bytes based on platform.
776
# Python 2.4 and later have os.urandom,
777
# but it doesn't work on some arches
779
rand_bytes = os.urandom
780
except (NotImplementedError, AttributeError):
781
# If python doesn't have os.urandom, or it doesn't work,
782
# then try to first pull random data from /dev/urandom
784
rand_bytes = file('/dev/urandom', 'rb').read
785
# Otherwise, use this hack as a last resort
786
except (IOError, OSError):
787
# not well seeded, but better than nothing
792
s += chr(random.randint(0, 255))
797
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
799
"""Return a random string of num alphanumeric characters
801
The result only contains lowercase chars because it may be used on
802
case-insensitive filesystems.
805
for raw_byte in rand_bytes(num):
806
s += ALNUM[ord(raw_byte) % 36]
810
## TODO: We could later have path objects that remember their list
811
## decomposition (might be too tricksy though.)
814
"""Turn string into list of parts."""
815
# split on either delimiter because people might use either on
817
ps = re.split(r'[\\/]', p)
822
raise errors.BzrError("sorry, %r not allowed in path" % f)
823
elif (f == '.') or (f == ''):
831
if (f == '..') or (f is None) or (f == ''):
832
raise errors.BzrError("sorry, %r not allowed in path" % f)
837
"""Split s into lines, but without removing the newline characters."""
838
lines = s.split('\n')
839
result = [line + '\n' for line in lines[:-1]]
841
result.append(lines[-1])
845
def hardlinks_good():
846
return sys.platform not in ('win32', 'cygwin', 'darwin')
849
def link_or_copy(src, dest):
850
"""Hardlink a file, or copy it if it can't be hardlinked."""
851
if not hardlinks_good():
852
shutil.copyfile(src, dest)
856
except (OSError, IOError), e:
857
if e.errno != errno.EXDEV:
859
shutil.copyfile(src, dest)
862
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
863
# Forgiveness than Permission (EAFP) because:
864
# - root can damage a solaris file system by using unlink,
865
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
866
# EACCES, OSX: EPERM) when invoked on a directory.
867
def delete_any(path):
868
"""Delete a file or directory."""
869
if isdir(path): # Takes care of symlinks
876
if getattr(os, 'symlink', None) is not None:
883
if getattr(os, 'link', None) is not None:
889
def host_os_dereferences_symlinks():
890
return (has_symlinks()
891
and sys.platform not in ('cygwin', 'win32'))
894
def contains_whitespace(s):
895
"""True if there are any whitespace characters in s."""
896
# string.whitespace can include '\xa0' in certain locales, because it is
897
# considered "non-breaking-space" as part of ISO-8859-1. But it
898
# 1) Isn't a breaking whitespace
899
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
901
# 3) '\xa0' isn't unicode safe since it is >128.
903
# This should *not* be a unicode set of characters in case the source
904
# string is not a Unicode string. We can auto-up-cast the characters since
905
# they are ascii, but we don't want to auto-up-cast the string in case it
907
for ch in ' \t\n\r\v\f':
914
def contains_linebreaks(s):
915
"""True if there is any vertical whitespace in s."""
923
def relpath(base, path):
924
"""Return path relative to base, or raise exception.
926
The path may be either an absolute path or a path relative to the
927
current working directory.
929
os.path.commonprefix (python2.4) has a bad bug that it works just
930
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
934
if len(base) < MIN_ABS_PATHLENGTH:
935
# must have space for e.g. a drive letter
936
raise ValueError('%r is too short to calculate a relative path'
943
while len(head) >= len(base):
946
head, tail = os.path.split(head)
950
raise errors.PathNotChild(rp, base)
958
def safe_unicode(unicode_or_utf8_string):
959
"""Coerce unicode_or_utf8_string into unicode.
961
If it is unicode, it is returned.
962
Otherwise it is decoded from utf-8. If a decoding error
963
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
964
as a BzrBadParameter exception.
966
if isinstance(unicode_or_utf8_string, unicode):
967
return unicode_or_utf8_string
969
return unicode_or_utf8_string.decode('utf8')
970
except UnicodeDecodeError:
971
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
974
def safe_utf8(unicode_or_utf8_string):
975
"""Coerce unicode_or_utf8_string to a utf8 string.
977
If it is a str, it is returned.
978
If it is Unicode, it is encoded into a utf-8 string.
980
if isinstance(unicode_or_utf8_string, str):
981
# TODO: jam 20070209 This is overkill, and probably has an impact on
982
# performance if we are dealing with lots of apis that want a
985
# Make sure it is a valid utf-8 string
986
unicode_or_utf8_string.decode('utf-8')
987
except UnicodeDecodeError:
988
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
989
return unicode_or_utf8_string
990
return unicode_or_utf8_string.encode('utf-8')
993
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
994
' Revision id generators should be creating utf8'
998
def safe_revision_id(unicode_or_utf8_string, warn=True):
999
"""Revision ids should now be utf8, but at one point they were unicode.
1001
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1003
:param warn: Functions that are sanitizing user data can set warn=False
1004
:return: None or a utf8 revision id.
1006
if (unicode_or_utf8_string is None
1007
or unicode_or_utf8_string.__class__ == str):
1008
return unicode_or_utf8_string
1010
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1012
return cache_utf8.encode(unicode_or_utf8_string)
1015
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1016
' generators should be creating utf8 file ids.')
1019
def safe_file_id(unicode_or_utf8_string, warn=True):
1020
"""File ids should now be utf8, but at one point they were unicode.
1022
This is the same as safe_utf8, except it uses the cached encode functions
1023
to save a little bit of performance.
1025
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1027
:param warn: Functions that are sanitizing user data can set warn=False
1028
:return: None or a utf8 file id.
1030
if (unicode_or_utf8_string is None
1031
or unicode_or_utf8_string.__class__ == str):
1032
return unicode_or_utf8_string
1034
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1036
return cache_utf8.encode(unicode_or_utf8_string)
1039
_platform_normalizes_filenames = False
1040
if sys.platform == 'darwin':
1041
_platform_normalizes_filenames = True
1044
def normalizes_filenames():
1045
"""Return True if this platform normalizes unicode filenames.
1047
Mac OSX does, Windows/Linux do not.
1049
return _platform_normalizes_filenames
1052
def _accessible_normalized_filename(path):
1053
"""Get the unicode normalized path, and if you can access the file.
1055
On platforms where the system normalizes filenames (Mac OSX),
1056
you can access a file by any path which will normalize correctly.
1057
On platforms where the system does not normalize filenames
1058
(Windows, Linux), you have to access a file by its exact path.
1060
Internally, bzr only supports NFC normalization, since that is
1061
the standard for XML documents.
1063
So return the normalized path, and a flag indicating if the file
1064
can be accessed by that path.
1067
return unicodedata.normalize('NFC', unicode(path)), True
1070
def _inaccessible_normalized_filename(path):
1071
__doc__ = _accessible_normalized_filename.__doc__
1073
normalized = unicodedata.normalize('NFC', unicode(path))
1074
return normalized, normalized == path
1077
if _platform_normalizes_filenames:
1078
normalized_filename = _accessible_normalized_filename
1080
normalized_filename = _inaccessible_normalized_filename
1083
def terminal_width():
1084
"""Return estimated terminal width."""
1085
if sys.platform == 'win32':
1086
return win32utils.get_console_size()[0]
1089
import struct, fcntl, termios
1090
s = struct.pack('HHHH', 0, 0, 0, 0)
1091
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1092
width = struct.unpack('HHHH', x)[1]
1097
width = int(os.environ['COLUMNS'])
1106
def supports_executable():
1107
return sys.platform != "win32"
1110
def supports_posix_readonly():
1111
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1113
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1114
directory controls creation/deletion, etc.
1116
And under win32, readonly means that the directory itself cannot be
1117
deleted. The contents of a readonly directory can be changed, unlike POSIX
1118
where files in readonly directories cannot be added, deleted or renamed.
1120
return sys.platform != "win32"
1123
def set_or_unset_env(env_variable, value):
1124
"""Modify the environment, setting or removing the env_variable.
1126
:param env_variable: The environment variable in question
1127
:param value: The value to set the environment to. If None, then
1128
the variable will be removed.
1129
:return: The original value of the environment variable.
1131
orig_val = os.environ.get(env_variable)
1133
if orig_val is not None:
1134
del os.environ[env_variable]
1136
if isinstance(value, unicode):
1137
value = value.encode(get_user_encoding())
1138
os.environ[env_variable] = value
1142
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1145
def check_legal_path(path):
1146
"""Check whether the supplied path is legal.
1147
This is only required on Windows, so we don't test on other platforms
1150
if sys.platform != "win32":
1152
if _validWin32PathRE.match(path) is None:
1153
raise errors.IllegalPath(path)
1156
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1158
def _is_error_enotdir(e):
1159
"""Check if this exception represents ENOTDIR.
1161
Unfortunately, python is very inconsistent about the exception
1162
here. The cases are:
1163
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1164
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1165
which is the windows error code.
1166
3) Windows, Python2.5 uses errno == EINVAL and
1167
winerror == ERROR_DIRECTORY
1169
:param e: An Exception object (expected to be OSError with an errno
1170
attribute, but we should be able to cope with anything)
1171
:return: True if this represents an ENOTDIR error. False otherwise.
1173
en = getattr(e, 'errno', None)
1174
if (en == errno.ENOTDIR
1175
or (sys.platform == 'win32'
1176
and (en == _WIN32_ERROR_DIRECTORY
1177
or (en == errno.EINVAL
1178
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1184
def walkdirs(top, prefix=""):
1185
"""Yield data about all the directories in a tree.
1187
This yields all the data about the contents of a directory at a time.
1188
After each directory has been yielded, if the caller has mutated the list
1189
to exclude some directories, they are then not descended into.
1191
The data yielded is of the form:
1192
((directory-relpath, directory-path-from-top),
1193
[(relpath, basename, kind, lstat, path-from-top), ...]),
1194
- directory-relpath is the relative path of the directory being returned
1195
with respect to top. prefix is prepended to this.
1196
- directory-path-from-root is the path including top for this directory.
1197
It is suitable for use with os functions.
1198
- relpath is the relative path within the subtree being walked.
1199
- basename is the basename of the path
1200
- kind is the kind of the file now. If unknown then the file is not
1201
present within the tree - but it may be recorded as versioned. See
1203
- lstat is the stat data *if* the file was statted.
1204
- planned, not implemented:
1205
path_from_tree_root is the path from the root of the tree.
1207
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1208
allows one to walk a subtree but get paths that are relative to a tree
1210
:return: an iterator over the dirs.
1212
#TODO there is a bit of a smell where the results of the directory-
1213
# summary in this, and the path from the root, may not agree
1214
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1215
# potentially confusing output. We should make this more robust - but
1216
# not at a speed cost. RBC 20060731
1218
_directory = _directory_kind
1219
_listdir = os.listdir
1220
_kind_from_mode = _formats.get
1221
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1223
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1224
relroot, _, _, _, top = pending.pop()
1226
relprefix = relroot + u'/'
1229
top_slash = top + u'/'
1232
append = dirblock.append
1234
names = sorted(_listdir(top))
1236
if not _is_error_enotdir(e):
1240
abspath = top_slash + name
1241
statvalue = _lstat(abspath)
1242
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1243
append((relprefix + name, name, kind, statvalue, abspath))
1244
yield (relroot, top), dirblock
1246
# push the user specified dirs from dirblock
1247
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1250
_real_walkdirs_utf8 = None
1252
def _walkdirs_utf8(top, prefix=""):
1253
"""Yield data about all the directories in a tree.
1255
This yields the same information as walkdirs() only each entry is yielded
1256
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1257
are returned as exact byte-strings.
1259
:return: yields a tuple of (dir_info, [file_info])
1260
dir_info is (utf8_relpath, path-from-top)
1261
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1262
if top is an absolute path, path-from-top is also an absolute path.
1263
path-from-top might be unicode or utf8, but it is the correct path to
1264
pass to os functions to affect the file in question. (such as os.lstat)
1266
global _real_walkdirs_utf8
1267
if _real_walkdirs_utf8 is None:
1268
fs_encoding = _fs_enc.upper()
1269
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1270
# Win98 doesn't have unicode apis like FindFirstFileW
1271
# TODO: We possibly could support Win98 by falling back to the
1272
# original FindFirstFile, and using TCHAR instead of WCHAR,
1273
# but that gets a bit tricky, and requires custom compiling
1276
from bzrlib._walkdirs_win32 import _walkdirs_utf8_win32_find_file
1278
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1280
_real_walkdirs_utf8 = _walkdirs_utf8_win32_find_file
1281
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1282
# ANSI_X3.4-1968 is a form of ASCII
1283
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1285
_real_walkdirs_utf8 = _walkdirs_fs_utf8
1286
return _real_walkdirs_utf8(top, prefix=prefix)
1289
def _walkdirs_fs_utf8(top, prefix=""):
1290
"""See _walkdirs_utf8.
1292
This sub-function is called when we know the filesystem is already in utf8
1293
encoding. So we don't need to transcode filenames.
1296
_directory = _directory_kind
1297
# Use C accelerated directory listing.
1298
_listdir = _read_dir
1299
_kind_from_mode = _formats.get
1301
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1302
# But we don't actually uses 1-3 in pending, so set them to None
1303
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1305
relroot, _, _, _, top = pending.pop()
1307
relprefix = relroot + '/'
1310
top_slash = top + '/'
1313
append = dirblock.append
1314
# read_dir supplies in should-stat order.
1315
for _, name in sorted(_listdir(top)):
1316
abspath = top_slash + name
1317
statvalue = _lstat(abspath)
1318
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1319
append((relprefix + name, name, kind, statvalue, abspath))
1321
yield (relroot, top), dirblock
1323
# push the user specified dirs from dirblock
1324
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1327
def _walkdirs_unicode_to_utf8(top, prefix=""):
1328
"""See _walkdirs_utf8
1330
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1332
This is currently the fallback code path when the filesystem encoding is
1333
not UTF-8. It may be better to implement an alternative so that we can
1334
safely handle paths that are not properly decodable in the current
1337
_utf8_encode = codecs.getencoder('utf8')
1339
_directory = _directory_kind
1340
_listdir = os.listdir
1341
_kind_from_mode = _formats.get
1343
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1345
relroot, _, _, _, top = pending.pop()
1347
relprefix = relroot + '/'
1350
top_slash = top + u'/'
1353
append = dirblock.append
1354
for name in sorted(_listdir(top)):
1355
name_utf8 = _utf8_encode(name)[0]
1356
abspath = top_slash + name
1357
statvalue = _lstat(abspath)
1358
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1359
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1360
yield (relroot, top), dirblock
1362
# push the user specified dirs from dirblock
1363
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1366
def copy_tree(from_path, to_path, handlers={}):
1367
"""Copy all of the entries in from_path into to_path.
1369
:param from_path: The base directory to copy.
1370
:param to_path: The target directory. If it does not exist, it will
1372
:param handlers: A dictionary of functions, which takes a source and
1373
destinations for files, directories, etc.
1374
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1375
'file', 'directory', and 'symlink' should always exist.
1376
If they are missing, they will be replaced with 'os.mkdir()',
1377
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1379
# Now, just copy the existing cached tree to the new location
1380
# We use a cheap trick here.
1381
# Absolute paths are prefixed with the first parameter
1382
# relative paths are prefixed with the second.
1383
# So we can get both the source and target returned
1384
# without any extra work.
1386
def copy_dir(source, dest):
1389
def copy_link(source, dest):
1390
"""Copy the contents of a symlink"""
1391
link_to = os.readlink(source)
1392
os.symlink(link_to, dest)
1394
real_handlers = {'file':shutil.copy2,
1395
'symlink':copy_link,
1396
'directory':copy_dir,
1398
real_handlers.update(handlers)
1400
if not os.path.exists(to_path):
1401
real_handlers['directory'](from_path, to_path)
1403
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1404
for relpath, name, kind, st, abspath in entries:
1405
real_handlers[kind](abspath, relpath)
1408
def path_prefix_key(path):
1409
"""Generate a prefix-order path key for path.
1411
This can be used to sort paths in the same way that walkdirs does.
1413
return (dirname(path) , path)
1416
def compare_paths_prefix_order(path_a, path_b):
1417
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1418
key_a = path_prefix_key(path_a)
1419
key_b = path_prefix_key(path_b)
1420
return cmp(key_a, key_b)
1423
_cached_user_encoding = None
1426
def get_user_encoding(use_cache=True):
1427
"""Find out what the preferred user encoding is.
1429
This is generally the encoding that is used for command line parameters
1430
and file contents. This may be different from the terminal encoding
1431
or the filesystem encoding.
1433
:param use_cache: Enable cache for detected encoding.
1434
(This parameter is turned on by default,
1435
and required only for selftesting)
1437
:return: A string defining the preferred user encoding
1439
global _cached_user_encoding
1440
if _cached_user_encoding is not None and use_cache:
1441
return _cached_user_encoding
1443
if sys.platform == 'darwin':
1444
# work around egregious python 2.4 bug
1445
sys.platform = 'posix'
1449
sys.platform = 'darwin'
1454
user_encoding = locale.getpreferredencoding()
1455
except locale.Error, e:
1456
sys.stderr.write('bzr: warning: %s\n'
1457
' Could not determine what text encoding to use.\n'
1458
' This error usually means your Python interpreter\n'
1459
' doesn\'t support the locale set by $LANG (%s)\n'
1460
" Continuing with ascii encoding.\n"
1461
% (e, os.environ.get('LANG')))
1462
user_encoding = 'ascii'
1464
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1465
# treat that as ASCII, and not support printing unicode characters to the
1468
# For python scripts run under vim, we get '', so also treat that as ASCII
1469
if user_encoding in (None, 'cp0', ''):
1470
user_encoding = 'ascii'
1474
codecs.lookup(user_encoding)
1476
sys.stderr.write('bzr: warning:'
1477
' unknown encoding %s.'
1478
' Continuing with ascii encoding.\n'
1481
user_encoding = 'ascii'
1484
_cached_user_encoding = user_encoding
1486
return user_encoding
1489
def get_host_name():
1490
"""Return the current unicode host name.
1492
This is meant to be used in place of socket.gethostname() because that
1493
behaves inconsistently on different platforms.
1495
if sys.platform == "win32":
1497
return win32utils.get_host_name()
1500
return socket.gethostname().decode(get_user_encoding())
1503
def recv_all(socket, bytes):
1504
"""Receive an exact number of bytes.
1506
Regular Socket.recv() may return less than the requested number of bytes,
1507
dependning on what's in the OS buffer. MSG_WAITALL is not available
1508
on all platforms, but this should work everywhere. This will return
1509
less than the requested amount if the remote end closes.
1511
This isn't optimized and is intended mostly for use in testing.
1514
while len(b) < bytes:
1515
new = socket.recv(bytes - len(b))
1522
def send_all(socket, bytes):
1523
"""Send all bytes on a socket.
1525
Regular socket.sendall() can give socket error 10053 on Windows. This
1526
implementation sends no more than 64k at a time, which avoids this problem.
1529
for pos in xrange(0, len(bytes), chunk_size):
1530
socket.sendall(bytes[pos:pos+chunk_size])
1533
def dereference_path(path):
1534
"""Determine the real path to a file.
1536
All parent elements are dereferenced. But the file itself is not
1538
:param path: The original path. May be absolute or relative.
1539
:return: the real path *to* the file
1541
parent, base = os.path.split(path)
1542
# The pathjoin for '.' is a workaround for Python bug #1213894.
1543
# (initial path components aren't dereferenced)
1544
return pathjoin(realpath(pathjoin('.', parent)), base)
1547
def supports_mapi():
1548
"""Return True if we can use MAPI to launch a mail client."""
1549
return sys.platform == "win32"
1552
def resource_string(package, resource_name):
1553
"""Load a resource from a package and return it as a string.
1555
Note: Only packages that start with bzrlib are currently supported.
1557
This is designed to be a lightweight implementation of resource
1558
loading in a way which is API compatible with the same API from
1560
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1561
If and when pkg_resources becomes a standard library, this routine
1564
# Check package name is within bzrlib
1565
if package == "bzrlib":
1566
resource_relpath = resource_name
1567
elif package.startswith("bzrlib."):
1568
package = package[len("bzrlib."):].replace('.', os.sep)
1569
resource_relpath = pathjoin(package, resource_name)
1571
raise errors.BzrError('resource package %s not in bzrlib' % package)
1573
# Map the resource to a file and read its contents
1574
base = dirname(bzrlib.__file__)
1575
if getattr(sys, 'frozen', None): # bzr.exe
1576
base = abspath(pathjoin(base, '..', '..'))
1577
filename = pathjoin(base, resource_relpath)
1578
return open(filename, 'rU').read()
1582
from bzrlib._readdir_pyx import read_dir as _read_dir
1584
from bzrlib._readdir_py import read_dir as _read_dir
added
removed
bzrlib/osutils.py
