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
17
from cStringIO import StringIO
21
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
22
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
44
from tempfile import (
58
from bzrlib.readdir import read_dir
59
from bzrlib import symbol_versioning
60
from bzrlib.symbol_versioning import (
63
from bzrlib.trace import mutter
66
# On win32, O_BINARY is used to indicate the file should
67
# be opened in binary mode, rather than text mode.
68
# On other platforms, O_BINARY doesn't exist, because
69
# they always open in binary mode, so it is okay to
70
# OR with 0 on those platforms
71
O_BINARY = getattr(os, 'O_BINARY', 0)
74
def make_readonly(filename):
75
"""Make a filename read-only."""
76
mod = os.lstat(filename).st_mode
77
if not stat.S_ISLNK(mod):
79
os.chmod(filename, mod)
82
def make_writable(filename):
83
mod = os.lstat(filename).st_mode
84
if not stat.S_ISLNK(mod):
86
os.chmod(filename, mod)
89
def minimum_path_selection(paths):
90
"""Return the smallset subset of paths which are outside paths.
92
:param paths: A container (and hence not None) of paths.
93
:return: A set of paths sufficient to include everything in paths via
94
is_inside_any, drawn from the paths parameter.
99
other_paths = paths.difference([path])
100
if not is_inside_any(other_paths, path):
101
# this is a top level path, we must check it.
102
search_paths.add(path)
110
"""Return a quoted filename filename
112
This previously used backslash quoting, but that works poorly on
114
# TODO: I'm not really sure this is the best format either.x
116
if _QUOTE_RE is None:
117
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
119
if _QUOTE_RE.search(f):
125
_directory_kind = 'directory'
128
stat.S_IFDIR:_directory_kind,
129
stat.S_IFCHR:'chardev',
130
stat.S_IFBLK:'block',
133
stat.S_IFLNK:'symlink',
134
stat.S_IFSOCK:'socket',
138
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
139
"""Generate a file kind from a stat mode. This is used in walkdirs.
141
Its performance is critical: Do not mutate without careful benchmarking.
144
return _formats[stat_mode & 0170000]
149
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
151
return _mapper(_lstat(f).st_mode)
153
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
154
raise errors.NoSuchFile(f)
159
"""Return the current umask"""
160
# Assume that people aren't messing with the umask while running
161
# XXX: This is not thread safe, but there is no way to get the
162
# umask without setting it
170
_directory_kind: "/",
172
'tree-reference': '+',
176
def kind_marker(kind):
178
return _kind_marker_map[kind]
180
raise errors.BzrError('invalid file kind %r' % kind)
183
lexists = getattr(os.path, 'lexists', None)
187
stat = getattr(os, 'lstat', os.stat)
191
if e.errno == errno.ENOENT:
194
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
197
def fancy_rename(old, new, rename_func, unlink_func):
198
"""A fancy rename, when you don't have atomic rename.
200
:param old: The old path, to rename from
201
:param new: The new path, to rename to
202
:param rename_func: The potentially non-atomic rename function
203
:param unlink_func: A way to delete the target file if the full rename succeeds
206
# sftp rename doesn't allow overwriting, so play tricks:
208
base = os.path.basename(new)
209
dirname = os.path.dirname(new)
210
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
211
tmp_name = pathjoin(dirname, tmp_name)
213
# Rename the file out of the way, but keep track if it didn't exist
214
# We don't want to grab just any exception
215
# something like EACCES should prevent us from continuing
216
# The downside is that the rename_func has to throw an exception
217
# with an errno = ENOENT, or NoSuchFile
220
rename_func(new, tmp_name)
221
except (errors.NoSuchFile,), e:
224
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
225
# function raises an IOError with errno is None when a rename fails.
226
# This then gets caught here.
227
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
230
if (getattr(e, 'errno', None) is None
231
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
239
# This may throw an exception, in which case success will
241
rename_func(old, new)
243
except (IOError, OSError), e:
244
# source and target may be aliases of each other (e.g. on a
245
# case-insensitive filesystem), so we may have accidentally renamed
246
# source by when we tried to rename target
247
if not (file_existed and e.errno in (None, errno.ENOENT)):
251
# If the file used to exist, rename it back into place
252
# otherwise just delete it from the tmp location
254
unlink_func(tmp_name)
256
rename_func(tmp_name, new)
259
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
260
# choke on a Unicode string containing a relative path if
261
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
263
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
264
def _posix_abspath(path):
265
# jam 20060426 rather than encoding to fsencoding
266
# copy posixpath.abspath, but use os.getcwdu instead
267
if not posixpath.isabs(path):
268
path = posixpath.join(getcwd(), path)
269
return posixpath.normpath(path)
272
def _posix_realpath(path):
273
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
276
def _win32_fixdrive(path):
277
"""Force drive letters to be consistent.
279
win32 is inconsistent whether it returns lower or upper case
280
and even if it was consistent the user might type the other
281
so we force it to uppercase
282
running python.exe under cmd.exe return capital C:\\
283
running win32 python inside a cygwin shell returns lowercase c:\\
285
drive, path = _nt_splitdrive(path)
286
return drive.upper() + path
289
def _win32_abspath(path):
290
# Real _nt_abspath doesn't have a problem with a unicode cwd
291
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
294
def _win98_abspath(path):
295
"""Return the absolute version of a path.
296
Windows 98 safe implementation (python reimplementation
297
of Win32 API function GetFullPathNameW)
302
# \\HOST\path => //HOST/path
303
# //HOST/path => //HOST/path
304
# path => C:/cwd/path
307
# check for absolute path
308
drive = _nt_splitdrive(path)[0]
309
if drive == '' and path[:2] not in('//','\\\\'):
311
# we cannot simply os.path.join cwd and path
312
# because os.path.join('C:','/path') produce '/path'
313
# and this is incorrect
314
if path[:1] in ('/','\\'):
315
cwd = _nt_splitdrive(cwd)[0]
317
path = cwd + '\\' + path
318
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
320
if win32utils.winver == 'Windows 98':
321
_win32_abspath = _win98_abspath
324
def _win32_realpath(path):
325
# Real _nt_realpath doesn't have a problem with a unicode cwd
326
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
329
def _win32_pathjoin(*args):
330
return _nt_join(*args).replace('\\', '/')
333
def _win32_normpath(path):
334
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
338
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
341
def _win32_mkdtemp(*args, **kwargs):
342
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
345
def _win32_rename(old, new):
346
"""We expect to be able to atomically replace 'new' with old.
348
On win32, if new exists, it must be moved out of the way first,
352
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
354
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
355
# If we try to rename a non-existant file onto cwd, we get
356
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
357
# if the old path doesn't exist, sometimes we get EACCES
358
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
364
return unicodedata.normalize('NFC', os.getcwdu())
367
# Default is to just use the python builtins, but these can be rebound on
368
# particular platforms.
369
abspath = _posix_abspath
370
realpath = _posix_realpath
371
pathjoin = os.path.join
372
normpath = os.path.normpath
375
dirname = os.path.dirname
376
basename = os.path.basename
377
split = os.path.split
378
splitext = os.path.splitext
379
# These were already imported into local scope
380
# mkdtemp = tempfile.mkdtemp
381
# rmtree = shutil.rmtree
383
MIN_ABS_PATHLENGTH = 1
386
if sys.platform == 'win32':
387
abspath = _win32_abspath
388
realpath = _win32_realpath
389
pathjoin = _win32_pathjoin
390
normpath = _win32_normpath
391
getcwd = _win32_getcwd
392
mkdtemp = _win32_mkdtemp
393
rename = _win32_rename
395
MIN_ABS_PATHLENGTH = 3
397
def _win32_delete_readonly(function, path, excinfo):
398
"""Error handler for shutil.rmtree function [for win32]
399
Helps to remove files and dirs marked as read-only.
401
exception = excinfo[1]
402
if function in (os.remove, os.rmdir) \
403
and isinstance(exception, OSError) \
404
and exception.errno == errno.EACCES:
410
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
411
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
412
return shutil.rmtree(path, ignore_errors, onerror)
413
elif sys.platform == 'darwin':
417
def get_terminal_encoding():
418
"""Find the best encoding for printing to the screen.
420
This attempts to check both sys.stdout and sys.stdin to see
421
what encoding they are in, and if that fails it falls back to
422
bzrlib.user_encoding.
423
The problem is that on Windows, locale.getpreferredencoding()
424
is not the same encoding as that used by the console:
425
http://mail.python.org/pipermail/python-list/2003-May/162357.html
427
On my standard US Windows XP, the preferred encoding is
428
cp1252, but the console is cp437
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 = bzrlib.user_encoding
435
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
437
output_encoding = input_encoding
438
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
440
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
441
if output_encoding == 'cp0':
442
# invalid encoding (cp0 means 'no codepage' on Windows)
443
output_encoding = bzrlib.user_encoding
444
mutter('cp0 is invalid encoding.'
445
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
448
codecs.lookup(output_encoding)
450
sys.stderr.write('bzr: warning:'
451
' unknown terminal encoding %s.\n'
452
' Using encoding %s instead.\n'
453
% (output_encoding, bzrlib.user_encoding)
455
output_encoding = bzrlib.user_encoding
457
return output_encoding
460
def normalizepath(f):
461
if getattr(os.path, 'realpath', None) is not None:
465
[p,e] = os.path.split(f)
466
if e == "" or e == "." or e == "..":
469
return pathjoin(F(p), e)
473
"""True if f is an accessible directory."""
475
return S_ISDIR(os.lstat(f)[ST_MODE])
481
"""True if f is a regular file."""
483
return S_ISREG(os.lstat(f)[ST_MODE])
488
"""True if f is a symlink."""
490
return S_ISLNK(os.lstat(f)[ST_MODE])
494
def is_inside(dir, fname):
495
"""True if fname is inside dir.
497
The parameters should typically be passed to osutils.normpath first, so
498
that . and .. and repeated slashes are eliminated, and the separators
499
are canonical for the platform.
501
The empty string as a dir name is taken as top-of-tree and matches
504
# XXX: Most callers of this can actually do something smarter by
505
# looking at the inventory
515
return fname.startswith(dir)
518
def is_inside_any(dir_list, fname):
519
"""True if fname is inside any of given dirs."""
520
for dirname in dir_list:
521
if is_inside(dirname, fname):
526
def is_inside_or_parent_of_any(dir_list, fname):
527
"""True if fname is a child or a parent of any of the given files."""
528
for dirname in dir_list:
529
if is_inside(dirname, fname) or is_inside(fname, dirname):
534
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768):
535
"""Copy contents of one file to another.
537
The read_length can either be -1 to read to end-of-file (EOF) or
538
it can specify the maximum number of bytes to read.
540
The buff_size represents the maximum size for each read operation
541
performed on from_file.
543
:return: The number of bytes copied.
547
# read specified number of bytes
549
while read_length > 0:
550
num_bytes_to_read = min(read_length, buff_size)
552
block = from_file.read(num_bytes_to_read)
558
actual_bytes_read = len(block)
559
read_length -= actual_bytes_read
560
length += actual_bytes_read
564
block = from_file.read(buff_size)
573
def file_iterator(input_file, readsize=32768):
575
b = input_file.read(readsize)
582
"""Calculate the hexdigest of an open file.
584
The file cursor should be already at the start.
596
def sha_file_by_name(fname):
597
"""Calculate the SHA1 of a file by reading the full text"""
599
f = os.open(fname, os.O_RDONLY | O_BINARY)
602
b = os.read(f, 1<<16)
610
def sha_strings(strings, _factory=sha.new):
611
"""Return the sha-1 of concatenation of strings"""
613
map(s.update, strings)
617
def sha_string(f, _factory=sha.new):
618
return _factory(f).hexdigest()
621
def fingerprint_file(f):
623
return {'size': len(b),
624
'sha1': sha.new(b).hexdigest()}
627
def compare_files(a, b):
628
"""Returns true if equal in contents"""
639
def local_time_offset(t=None):
640
"""Return offset of local zone from GMT, either at present or at time t."""
643
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
644
return offset.days * 86400 + offset.seconds
646
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
648
def format_date(t, offset=0, timezone='original', date_fmt=None,
650
"""Return a formatted date string.
652
:param t: Seconds since the epoch.
653
:param offset: Timezone offset in seconds east of utc.
654
:param timezone: How to display the time: 'utc', 'original' for the
655
timezone specified by offset, or 'local' for the process's current
657
:param show_offset: Whether to append the timezone.
658
:param date_fmt: strftime format.
660
if timezone == 'utc':
663
elif timezone == 'original':
666
tt = time.gmtime(t + offset)
667
elif timezone == 'local':
668
tt = time.localtime(t)
669
offset = local_time_offset(t)
671
raise errors.UnsupportedTimezoneFormat(timezone)
673
date_fmt = "%a %Y-%m-%d %H:%M:%S"
675
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
678
# day of week depends on locale, so we do this ourself
679
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
680
return (time.strftime(date_fmt, tt) + offset_str)
683
def compact_date(when):
684
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
687
def format_delta(delta):
688
"""Get a nice looking string for a time delta.
690
:param delta: The time difference in seconds, can be positive or negative.
691
positive indicates time in the past, negative indicates time in the
692
future. (usually time.time() - stored_time)
693
:return: String formatted to show approximate resolution
699
direction = 'in the future'
703
if seconds < 90: # print seconds up to 90 seconds
705
return '%d second %s' % (seconds, direction,)
707
return '%d seconds %s' % (seconds, direction)
709
minutes = int(seconds / 60)
710
seconds -= 60 * minutes
715
if minutes < 90: # print minutes, seconds up to 90 minutes
717
return '%d minute, %d second%s %s' % (
718
minutes, seconds, plural_seconds, direction)
720
return '%d minutes, %d second%s %s' % (
721
minutes, seconds, plural_seconds, direction)
723
hours = int(minutes / 60)
724
minutes -= 60 * hours
731
return '%d hour, %d minute%s %s' % (hours, minutes,
732
plural_minutes, direction)
733
return '%d hours, %d minute%s %s' % (hours, minutes,
734
plural_minutes, direction)
737
"""Return size of given open file."""
738
return os.fstat(f.fileno())[ST_SIZE]
741
# Define rand_bytes based on platform.
743
# Python 2.4 and later have os.urandom,
744
# but it doesn't work on some arches
746
rand_bytes = os.urandom
747
except (NotImplementedError, AttributeError):
748
# If python doesn't have os.urandom, or it doesn't work,
749
# then try to first pull random data from /dev/urandom
751
rand_bytes = file('/dev/urandom', 'rb').read
752
# Otherwise, use this hack as a last resort
753
except (IOError, OSError):
754
# not well seeded, but better than nothing
759
s += chr(random.randint(0, 255))
764
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
766
"""Return a random string of num alphanumeric characters
768
The result only contains lowercase chars because it may be used on
769
case-insensitive filesystems.
772
for raw_byte in rand_bytes(num):
773
s += ALNUM[ord(raw_byte) % 36]
777
## TODO: We could later have path objects that remember their list
778
## decomposition (might be too tricksy though.)
781
"""Turn string into list of parts."""
782
# split on either delimiter because people might use either on
784
ps = re.split(r'[\\/]', p)
789
raise errors.BzrError("sorry, %r not allowed in path" % f)
790
elif (f == '.') or (f == ''):
798
if (f == '..') or (f is None) or (f == ''):
799
raise errors.BzrError("sorry, %r not allowed in path" % f)
804
"""Split s into lines, but without removing the newline characters."""
805
lines = s.split('\n')
806
result = [line + '\n' for line in lines[:-1]]
808
result.append(lines[-1])
812
def hardlinks_good():
813
return sys.platform not in ('win32', 'cygwin', 'darwin')
816
def link_or_copy(src, dest):
817
"""Hardlink a file, or copy it if it can't be hardlinked."""
818
if not hardlinks_good():
819
shutil.copyfile(src, dest)
823
except (OSError, IOError), e:
824
if e.errno != errno.EXDEV:
826
shutil.copyfile(src, dest)
829
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
830
# Forgiveness than Permission (EAFP) because:
831
# - root can damage a solaris file system by using unlink,
832
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
833
# EACCES, OSX: EPERM) when invoked on a directory.
834
def delete_any(path):
835
"""Delete a file or directory."""
836
if isdir(path): # Takes care of symlinks
843
if getattr(os, 'symlink', None) is not None:
850
if getattr(os, 'link', None) is not None:
856
def host_os_dereferences_symlinks():
857
return (has_symlinks()
858
and sys.platform not in ('cygwin', 'win32'))
861
def contains_whitespace(s):
862
"""True if there are any whitespace characters in s."""
863
# string.whitespace can include '\xa0' in certain locales, because it is
864
# considered "non-breaking-space" as part of ISO-8859-1. But it
865
# 1) Isn't a breaking whitespace
866
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
868
# 3) '\xa0' isn't unicode safe since it is >128.
870
# This should *not* be a unicode set of characters in case the source
871
# string is not a Unicode string. We can auto-up-cast the characters since
872
# they are ascii, but we don't want to auto-up-cast the string in case it
874
for ch in ' \t\n\r\v\f':
881
def contains_linebreaks(s):
882
"""True if there is any vertical whitespace in s."""
890
def relpath(base, path):
891
"""Return path relative to base, or raise exception.
893
The path may be either an absolute path or a path relative to the
894
current working directory.
896
os.path.commonprefix (python2.4) has a bad bug that it works just
897
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
901
if len(base) < MIN_ABS_PATHLENGTH:
902
# must have space for e.g. a drive letter
903
raise ValueError('%r is too short to calculate a relative path'
910
while len(head) >= len(base):
913
head, tail = os.path.split(head)
917
raise errors.PathNotChild(rp, base)
925
def safe_unicode(unicode_or_utf8_string):
926
"""Coerce unicode_or_utf8_string into unicode.
928
If it is unicode, it is returned.
929
Otherwise it is decoded from utf-8. If a decoding error
930
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
931
as a BzrBadParameter exception.
933
if isinstance(unicode_or_utf8_string, unicode):
934
return unicode_or_utf8_string
936
return unicode_or_utf8_string.decode('utf8')
937
except UnicodeDecodeError:
938
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
941
def safe_utf8(unicode_or_utf8_string):
942
"""Coerce unicode_or_utf8_string to a utf8 string.
944
If it is a str, it is returned.
945
If it is Unicode, it is encoded into a utf-8 string.
947
if isinstance(unicode_or_utf8_string, str):
948
# TODO: jam 20070209 This is overkill, and probably has an impact on
949
# performance if we are dealing with lots of apis that want a
952
# Make sure it is a valid utf-8 string
953
unicode_or_utf8_string.decode('utf-8')
954
except UnicodeDecodeError:
955
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
956
return unicode_or_utf8_string
957
return unicode_or_utf8_string.encode('utf-8')
960
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
961
' Revision id generators should be creating utf8'
965
def safe_revision_id(unicode_or_utf8_string, warn=True):
966
"""Revision ids should now be utf8, but at one point they were unicode.
968
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
970
:param warn: Functions that are sanitizing user data can set warn=False
971
:return: None or a utf8 revision id.
973
if (unicode_or_utf8_string is None
974
or unicode_or_utf8_string.__class__ == str):
975
return unicode_or_utf8_string
977
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
979
return cache_utf8.encode(unicode_or_utf8_string)
982
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
983
' generators should be creating utf8 file ids.')
986
def safe_file_id(unicode_or_utf8_string, warn=True):
987
"""File ids should now be utf8, but at one point they were unicode.
989
This is the same as safe_utf8, except it uses the cached encode functions
990
to save a little bit of performance.
992
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
994
:param warn: Functions that are sanitizing user data can set warn=False
995
:return: None or a utf8 file id.
997
if (unicode_or_utf8_string is None
998
or unicode_or_utf8_string.__class__ == str):
999
return unicode_or_utf8_string
1001
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1003
return cache_utf8.encode(unicode_or_utf8_string)
1006
_platform_normalizes_filenames = False
1007
if sys.platform == 'darwin':
1008
_platform_normalizes_filenames = True
1011
def normalizes_filenames():
1012
"""Return True if this platform normalizes unicode filenames.
1014
Mac OSX does, Windows/Linux do not.
1016
return _platform_normalizes_filenames
1019
def _accessible_normalized_filename(path):
1020
"""Get the unicode normalized path, and if you can access the file.
1022
On platforms where the system normalizes filenames (Mac OSX),
1023
you can access a file by any path which will normalize correctly.
1024
On platforms where the system does not normalize filenames
1025
(Windows, Linux), you have to access a file by its exact path.
1027
Internally, bzr only supports NFC normalization, since that is
1028
the standard for XML documents.
1030
So return the normalized path, and a flag indicating if the file
1031
can be accessed by that path.
1034
return unicodedata.normalize('NFC', unicode(path)), True
1037
def _inaccessible_normalized_filename(path):
1038
__doc__ = _accessible_normalized_filename.__doc__
1040
normalized = unicodedata.normalize('NFC', unicode(path))
1041
return normalized, normalized == path
1044
if _platform_normalizes_filenames:
1045
normalized_filename = _accessible_normalized_filename
1047
normalized_filename = _inaccessible_normalized_filename
1050
def terminal_width():
1051
"""Return estimated terminal width."""
1052
if sys.platform == 'win32':
1053
return win32utils.get_console_size()[0]
1056
import struct, fcntl, termios
1057
s = struct.pack('HHHH', 0, 0, 0, 0)
1058
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1059
width = struct.unpack('HHHH', x)[1]
1064
width = int(os.environ['COLUMNS'])
1073
def supports_executable():
1074
return sys.platform != "win32"
1077
def supports_posix_readonly():
1078
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1080
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1081
directory controls creation/deletion, etc.
1083
And under win32, readonly means that the directory itself cannot be
1084
deleted. The contents of a readonly directory can be changed, unlike POSIX
1085
where files in readonly directories cannot be added, deleted or renamed.
1087
return sys.platform != "win32"
1090
def set_or_unset_env(env_variable, value):
1091
"""Modify the environment, setting or removing the env_variable.
1093
:param env_variable: The environment variable in question
1094
:param value: The value to set the environment to. If None, then
1095
the variable will be removed.
1096
:return: The original value of the environment variable.
1098
orig_val = os.environ.get(env_variable)
1100
if orig_val is not None:
1101
del os.environ[env_variable]
1103
if isinstance(value, unicode):
1104
value = value.encode(bzrlib.user_encoding)
1105
os.environ[env_variable] = value
1109
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1112
def check_legal_path(path):
1113
"""Check whether the supplied path is legal.
1114
This is only required on Windows, so we don't test on other platforms
1117
if sys.platform != "win32":
1119
if _validWin32PathRE.match(path) is None:
1120
raise errors.IllegalPath(path)
1123
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1125
def _is_error_enotdir(e):
1126
"""Check if this exception represents ENOTDIR.
1128
Unfortunately, python is very inconsistent about the exception
1129
here. The cases are:
1130
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1131
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1132
which is the windows error code.
1133
3) Windows, Python2.5 uses errno == EINVAL and
1134
winerror == ERROR_DIRECTORY
1136
:param e: An Exception object (expected to be OSError with an errno
1137
attribute, but we should be able to cope with anything)
1138
:return: True if this represents an ENOTDIR error. False otherwise.
1140
en = getattr(e, 'errno', None)
1141
if (en == errno.ENOTDIR
1142
or (sys.platform == 'win32'
1143
and (en == _WIN32_ERROR_DIRECTORY
1144
or (en == errno.EINVAL
1145
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1151
def walkdirs(top, prefix=""):
1152
"""Yield data about all the directories in a tree.
1154
This yields all the data about the contents of a directory at a time.
1155
After each directory has been yielded, if the caller has mutated the list
1156
to exclude some directories, they are then not descended into.
1158
The data yielded is of the form:
1159
((directory-relpath, directory-path-from-top),
1160
[(relpath, basename, kind, lstat, path-from-top), ...]),
1161
- directory-relpath is the relative path of the directory being returned
1162
with respect to top. prefix is prepended to this.
1163
- directory-path-from-root is the path including top for this directory.
1164
It is suitable for use with os functions.
1165
- relpath is the relative path within the subtree being walked.
1166
- basename is the basename of the path
1167
- kind is the kind of the file now. If unknown then the file is not
1168
present within the tree - but it may be recorded as versioned. See
1170
- lstat is the stat data *if* the file was statted.
1171
- planned, not implemented:
1172
path_from_tree_root is the path from the root of the tree.
1174
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1175
allows one to walk a subtree but get paths that are relative to a tree
1177
:return: an iterator over the dirs.
1179
#TODO there is a bit of a smell where the results of the directory-
1180
# summary in this, and the path from the root, may not agree
1181
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1182
# potentially confusing output. We should make this more robust - but
1183
# not at a speed cost. RBC 20060731
1185
_directory = _directory_kind
1186
_listdir = os.listdir
1187
_kind_from_mode = _formats.get
1188
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1190
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1191
relroot, _, _, _, top = pending.pop()
1193
relprefix = relroot + u'/'
1196
top_slash = top + u'/'
1199
append = dirblock.append
1201
names = sorted(_listdir(top))
1203
if not _is_error_enotdir(e):
1207
abspath = top_slash + name
1208
statvalue = _lstat(abspath)
1209
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1210
append((relprefix + name, name, kind, statvalue, abspath))
1211
yield (relroot, top), dirblock
1213
# push the user specified dirs from dirblock
1214
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1217
_real_walkdirs_utf8 = None
1219
def _walkdirs_utf8(top, prefix=""):
1220
"""Yield data about all the directories in a tree.
1222
This yields the same information as walkdirs() only each entry is yielded
1223
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1224
are returned as exact byte-strings.
1226
:return: yields a tuple of (dir_info, [file_info])
1227
dir_info is (utf8_relpath, path-from-top)
1228
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1229
if top is an absolute path, path-from-top is also an absolute path.
1230
path-from-top might be unicode or utf8, but it is the correct path to
1231
pass to os functions to affect the file in question. (such as os.lstat)
1233
global _real_walkdirs_utf8
1234
if _real_walkdirs_utf8 is None:
1235
fs_encoding = _fs_enc.upper()
1236
if win32utils.winver == 'Windows NT':
1237
# Win98 doesn't have unicode apis like FindFirstFileW
1238
# TODO: We possibly could support Win98 by falling back to the
1239
# original FindFirstFile, and using TCHAR instead of WCHAR,
1240
# but that gets a bit tricky, and requires custom compiling
1243
from bzrlib._walkdirs_win32 import _walkdirs_utf8_win32_find_file
1245
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1247
_real_walkdirs_utf8 = _walkdirs_utf8_win32_find_file
1248
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1249
# ANSI_X3.4-1968 is a form of ASCII
1250
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1252
_real_walkdirs_utf8 = _walkdirs_fs_utf8
1253
return _real_walkdirs_utf8(top, prefix=prefix)
1256
def _walkdirs_fs_utf8(top, prefix=""):
1257
"""See _walkdirs_utf8.
1259
This sub-function is called when we know the filesystem is already in utf8
1260
encoding. So we don't need to transcode filenames.
1263
_directory = _directory_kind
1264
# Use C accelerated directory listing.
1266
_kind_from_mode = _formats.get
1268
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1269
# But we don't actually uses 1-3 in pending, so set them to None
1270
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1272
relroot, _, _, _, top = pending.pop()
1274
relprefix = relroot + '/'
1277
top_slash = top + '/'
1280
append = dirblock.append
1281
for name, kind in sorted(_listdir(top)):
1282
abspath = top_slash + name
1283
if kind == 'unknown':
1284
statvalue = _lstat(abspath)
1285
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1288
statvalue = _lstat(abspath)
1289
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1290
append((relprefix + name, name, kind, statvalue, abspath))
1291
yield (relroot, top), dirblock
1293
# push the user specified dirs from dirblock
1294
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1297
def _walkdirs_unicode_to_utf8(top, prefix=""):
1298
"""See _walkdirs_utf8
1300
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1302
This is currently the fallback code path when the filesystem encoding is
1303
not UTF-8. It may be better to implement an alternative so that we can
1304
safely handle paths that are not properly decodable in the current
1307
_utf8_encode = codecs.getencoder('utf8')
1309
_directory = _directory_kind
1310
_listdir = os.listdir
1311
_kind_from_mode = _formats.get
1313
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1315
relroot, _, _, _, top = pending.pop()
1317
relprefix = relroot + '/'
1320
top_slash = top + u'/'
1323
append = dirblock.append
1324
for name in sorted(_listdir(top)):
1325
name_utf8 = _utf8_encode(name)[0]
1326
abspath = top_slash + name
1327
statvalue = _lstat(abspath)
1328
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1329
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1330
yield (relroot, top), dirblock
1332
# push the user specified dirs from dirblock
1333
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1336
def copy_tree(from_path, to_path, handlers={}):
1337
"""Copy all of the entries in from_path into to_path.
1339
:param from_path: The base directory to copy.
1340
:param to_path: The target directory. If it does not exist, it will
1342
:param handlers: A dictionary of functions, which takes a source and
1343
destinations for files, directories, etc.
1344
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1345
'file', 'directory', and 'symlink' should always exist.
1346
If they are missing, they will be replaced with 'os.mkdir()',
1347
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1349
# Now, just copy the existing cached tree to the new location
1350
# We use a cheap trick here.
1351
# Absolute paths are prefixed with the first parameter
1352
# relative paths are prefixed with the second.
1353
# So we can get both the source and target returned
1354
# without any extra work.
1356
def copy_dir(source, dest):
1359
def copy_link(source, dest):
1360
"""Copy the contents of a symlink"""
1361
link_to = os.readlink(source)
1362
os.symlink(link_to, dest)
1364
real_handlers = {'file':shutil.copy2,
1365
'symlink':copy_link,
1366
'directory':copy_dir,
1368
real_handlers.update(handlers)
1370
if not os.path.exists(to_path):
1371
real_handlers['directory'](from_path, to_path)
1373
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1374
for relpath, name, kind, st, abspath in entries:
1375
real_handlers[kind](abspath, relpath)
1378
def path_prefix_key(path):
1379
"""Generate a prefix-order path key for path.
1381
This can be used to sort paths in the same way that walkdirs does.
1383
return (dirname(path) , path)
1386
def compare_paths_prefix_order(path_a, path_b):
1387
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1388
key_a = path_prefix_key(path_a)
1389
key_b = path_prefix_key(path_b)
1390
return cmp(key_a, key_b)
1393
_cached_user_encoding = None
1396
def get_user_encoding(use_cache=True):
1397
"""Find out what the preferred user encoding is.
1399
This is generally the encoding that is used for command line parameters
1400
and file contents. This may be different from the terminal encoding
1401
or the filesystem encoding.
1403
:param use_cache: Enable cache for detected encoding.
1404
(This parameter is turned on by default,
1405
and required only for selftesting)
1407
:return: A string defining the preferred user encoding
1409
global _cached_user_encoding
1410
if _cached_user_encoding is not None and use_cache:
1411
return _cached_user_encoding
1413
if sys.platform == 'darwin':
1414
# work around egregious python 2.4 bug
1415
sys.platform = 'posix'
1419
sys.platform = 'darwin'
1424
user_encoding = locale.getpreferredencoding()
1425
except locale.Error, e:
1426
sys.stderr.write('bzr: warning: %s\n'
1427
' Could not determine what text encoding to use.\n'
1428
' This error usually means your Python interpreter\n'
1429
' doesn\'t support the locale set by $LANG (%s)\n'
1430
" Continuing with ascii encoding.\n"
1431
% (e, os.environ.get('LANG')))
1432
user_encoding = 'ascii'
1434
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1435
# treat that as ASCII, and not support printing unicode characters to the
1438
# For python scripts run under vim, we get '', so also treat that as ASCII
1439
if user_encoding in (None, 'cp0', ''):
1440
user_encoding = 'ascii'
1444
codecs.lookup(user_encoding)
1446
sys.stderr.write('bzr: warning:'
1447
' unknown encoding %s.'
1448
' Continuing with ascii encoding.\n'
1451
user_encoding = 'ascii'
1454
_cached_user_encoding = user_encoding
1456
return user_encoding
1459
def recv_all(socket, bytes):
1460
"""Receive an exact number of bytes.
1462
Regular Socket.recv() may return less than the requested number of bytes,
1463
dependning on what's in the OS buffer. MSG_WAITALL is not available
1464
on all platforms, but this should work everywhere. This will return
1465
less than the requested amount if the remote end closes.
1467
This isn't optimized and is intended mostly for use in testing.
1470
while len(b) < bytes:
1471
new = socket.recv(bytes - len(b))
1478
def send_all(socket, bytes):
1479
"""Send all bytes on a socket.
1481
Regular socket.sendall() can give socket error 10053 on Windows. This
1482
implementation sends no more than 64k at a time, which avoids this problem.
1485
for pos in xrange(0, len(bytes), chunk_size):
1486
socket.sendall(bytes[pos:pos+chunk_size])
1489
def dereference_path(path):
1490
"""Determine the real path to a file.
1492
All parent elements are dereferenced. But the file itself is not
1494
:param path: The original path. May be absolute or relative.
1495
:return: the real path *to* the file
1497
parent, base = os.path.split(path)
1498
# The pathjoin for '.' is a workaround for Python bug #1213894.
1499
# (initial path components aren't dereferenced)
1500
return pathjoin(realpath(pathjoin('.', parent)), base)
1503
def supports_mapi():
1504
"""Return True if we can use MAPI to launch a mail client."""
1505
return sys.platform == "win32"
1508
def resource_string(package, resource_name):
1509
"""Load a resource from a package and return it as a string.
1511
Note: Only packages that start with bzrlib are currently supported.
1513
This is designed to be a lightweight implementation of resource
1514
loading in a way which is API compatible with the same API from
1516
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1517
If and when pkg_resources becomes a standard library, this routine
1520
# Check package name is within bzrlib
1521
if package == "bzrlib":
1522
resource_relpath = resource_name
1523
elif package.startswith("bzrlib."):
1524
package = package[len("bzrlib."):].replace('.', os.sep)
1525
resource_relpath = pathjoin(package, resource_name)
1527
raise errors.BzrError('resource package %s not in bzrlib' % package)
1529
# Map the resource to a file and read its contents
1530
base = dirname(bzrlib.__file__)
1531
if getattr(sys, 'frozen', None): # bzr.exe
1532
base = abspath(pathjoin(base, '..', '..'))
1533
filename = pathjoin(base, resource_relpath)
1534
return open(filename, 'rU').read()
added
removed
bzrlib/osutils.py
