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 (
57
from bzrlib import symbol_versioning
58
from bzrlib.symbol_versioning import (
62
from bzrlib.trace import mutter
65
# On win32, O_BINARY is used to indicate the file should
66
# be opened in binary mode, rather than text mode.
67
# On other platforms, O_BINARY doesn't exist, because
68
# they always open in binary mode, so it is okay to
69
# OR with 0 on those platforms
70
O_BINARY = getattr(os, 'O_BINARY', 0)
73
def make_readonly(filename):
74
"""Make a filename read-only."""
75
mod = os.stat(filename).st_mode
77
os.chmod(filename, mod)
80
def make_writable(filename):
81
mod = os.stat(filename).st_mode
83
os.chmod(filename, mod)
90
"""Return a quoted filename filename
92
This previously used backslash quoting, but that works poorly on
94
# TODO: I'm not really sure this is the best format either.x
97
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
99
if _QUOTE_RE.search(f):
105
_directory_kind = 'directory'
108
stat.S_IFDIR:_directory_kind,
109
stat.S_IFCHR:'chardev',
110
stat.S_IFBLK:'block',
113
stat.S_IFLNK:'symlink',
114
stat.S_IFSOCK:'socket',
118
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
119
"""Generate a file kind from a stat mode. This is used in walkdirs.
121
Its performance is critical: Do not mutate without careful benchmarking.
124
return _formats[stat_mode & 0170000]
129
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
131
return _mapper(_lstat(f).st_mode)
133
if getattr(e, 'errno', None) == errno.ENOENT:
134
raise errors.NoSuchFile(f)
139
"""Return the current umask"""
140
# Assume that people aren't messing with the umask while running
141
# XXX: This is not thread safe, but there is no way to get the
142
# umask without setting it
150
_directory_kind: "/",
152
'tree-reference': '+',
156
def kind_marker(kind):
158
return _kind_marker_map[kind]
160
raise errors.BzrError('invalid file kind %r' % kind)
163
lexists = getattr(os.path, 'lexists', None)
167
stat = getattr(os, 'lstat', os.stat)
171
if e.errno == errno.ENOENT:
174
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
177
def fancy_rename(old, new, rename_func, unlink_func):
178
"""A fancy rename, when you don't have atomic rename.
180
:param old: The old path, to rename from
181
:param new: The new path, to rename to
182
:param rename_func: The potentially non-atomic rename function
183
:param unlink_func: A way to delete the target file if the full rename succeeds
186
# sftp rename doesn't allow overwriting, so play tricks:
188
base = os.path.basename(new)
189
dirname = os.path.dirname(new)
190
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
191
tmp_name = pathjoin(dirname, tmp_name)
193
# Rename the file out of the way, but keep track if it didn't exist
194
# We don't want to grab just any exception
195
# something like EACCES should prevent us from continuing
196
# The downside is that the rename_func has to throw an exception
197
# with an errno = ENOENT, or NoSuchFile
200
rename_func(new, tmp_name)
201
except (errors.NoSuchFile,), e:
204
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
205
# function raises an IOError with errno is None when a rename fails.
206
# This then gets caught here.
207
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
210
if (getattr(e, 'errno', None) is None
211
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
218
# This may throw an exception, in which case success will
220
rename_func(old, new)
224
# If the file used to exist, rename it back into place
225
# otherwise just delete it from the tmp location
227
unlink_func(tmp_name)
229
rename_func(tmp_name, new)
232
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
233
# choke on a Unicode string containing a relative path if
234
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
236
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
237
def _posix_abspath(path):
238
# jam 20060426 rather than encoding to fsencoding
239
# copy posixpath.abspath, but use os.getcwdu instead
240
if not posixpath.isabs(path):
241
path = posixpath.join(getcwd(), path)
242
return posixpath.normpath(path)
245
def _posix_realpath(path):
246
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
249
def _win32_fixdrive(path):
250
"""Force drive letters to be consistent.
252
win32 is inconsistent whether it returns lower or upper case
253
and even if it was consistent the user might type the other
254
so we force it to uppercase
255
running python.exe under cmd.exe return capital C:\\
256
running win32 python inside a cygwin shell returns lowercase c:\\
258
drive, path = _nt_splitdrive(path)
259
return drive.upper() + path
262
def _win32_abspath(path):
263
# Real _nt_abspath doesn't have a problem with a unicode cwd
264
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
267
def _win98_abspath(path):
268
"""Return the absolute version of a path.
269
Windows 98 safe implementation (python reimplementation
270
of Win32 API function GetFullPathNameW)
275
# \\HOST\path => //HOST/path
276
# //HOST/path => //HOST/path
277
# path => C:/cwd/path
280
# check for absolute path
281
drive = _nt_splitdrive(path)[0]
282
if drive == '' and path[:2] not in('//','\\\\'):
284
# we cannot simply os.path.join cwd and path
285
# because os.path.join('C:','/path') produce '/path'
286
# and this is incorrect
287
if path[:1] in ('/','\\'):
288
cwd = _nt_splitdrive(cwd)[0]
290
path = cwd + '\\' + path
291
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
293
if win32utils.winver == 'Windows 98':
294
_win32_abspath = _win98_abspath
297
def _win32_realpath(path):
298
# Real _nt_realpath doesn't have a problem with a unicode cwd
299
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
302
def _win32_pathjoin(*args):
303
return _nt_join(*args).replace('\\', '/')
306
def _win32_normpath(path):
307
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
311
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
314
def _win32_mkdtemp(*args, **kwargs):
315
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
318
def _win32_rename(old, new):
319
"""We expect to be able to atomically replace 'new' with old.
321
On win32, if new exists, it must be moved out of the way first,
325
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
327
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
328
# If we try to rename a non-existant file onto cwd, we get
329
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
330
# if the old path doesn't exist, sometimes we get EACCES
331
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
337
return unicodedata.normalize('NFKC', os.getcwdu())
340
# Default is to just use the python builtins, but these can be rebound on
341
# particular platforms.
342
abspath = _posix_abspath
343
realpath = _posix_realpath
344
pathjoin = os.path.join
345
normpath = os.path.normpath
348
dirname = os.path.dirname
349
basename = os.path.basename
350
split = os.path.split
351
splitext = os.path.splitext
352
# These were already imported into local scope
353
# mkdtemp = tempfile.mkdtemp
354
# rmtree = shutil.rmtree
356
MIN_ABS_PATHLENGTH = 1
359
if sys.platform == 'win32':
360
abspath = _win32_abspath
361
realpath = _win32_realpath
362
pathjoin = _win32_pathjoin
363
normpath = _win32_normpath
364
getcwd = _win32_getcwd
365
mkdtemp = _win32_mkdtemp
366
rename = _win32_rename
368
MIN_ABS_PATHLENGTH = 3
370
def _win32_delete_readonly(function, path, excinfo):
371
"""Error handler for shutil.rmtree function [for win32]
372
Helps to remove files and dirs marked as read-only.
374
exception = excinfo[1]
375
if function in (os.remove, os.rmdir) \
376
and isinstance(exception, OSError) \
377
and exception.errno == errno.EACCES:
383
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
384
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
385
return shutil.rmtree(path, ignore_errors, onerror)
386
elif sys.platform == 'darwin':
390
def get_terminal_encoding():
391
"""Find the best encoding for printing to the screen.
393
This attempts to check both sys.stdout and sys.stdin to see
394
what encoding they are in, and if that fails it falls back to
395
bzrlib.user_encoding.
396
The problem is that on Windows, locale.getpreferredencoding()
397
is not the same encoding as that used by the console:
398
http://mail.python.org/pipermail/python-list/2003-May/162357.html
400
On my standard US Windows XP, the preferred encoding is
401
cp1252, but the console is cp437
403
output_encoding = getattr(sys.stdout, 'encoding', None)
404
if not output_encoding:
405
input_encoding = getattr(sys.stdin, 'encoding', None)
406
if not input_encoding:
407
output_encoding = bzrlib.user_encoding
408
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
410
output_encoding = input_encoding
411
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
413
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
414
if output_encoding == 'cp0':
415
# invalid encoding (cp0 means 'no codepage' on Windows)
416
output_encoding = bzrlib.user_encoding
417
mutter('cp0 is invalid encoding.'
418
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
421
codecs.lookup(output_encoding)
423
sys.stderr.write('bzr: warning:'
424
' unknown terminal encoding %s.\n'
425
' Using encoding %s instead.\n'
426
% (output_encoding, bzrlib.user_encoding)
428
output_encoding = bzrlib.user_encoding
430
return output_encoding
433
def normalizepath(f):
434
if getattr(os.path, 'realpath', None) is not None:
438
[p,e] = os.path.split(f)
439
if e == "" or e == "." or e == "..":
442
return pathjoin(F(p), e)
446
"""Copy a file to a backup.
448
Backups are named in GNU-style, with a ~ suffix.
450
If the file is already a backup, it's not copied.
456
if has_symlinks() and os.path.islink(fn):
457
target = os.readlink(fn)
458
os.symlink(target, bfn)
466
outf = file(bfn, 'wb')
474
"""True if f is an accessible directory."""
476
return S_ISDIR(os.lstat(f)[ST_MODE])
482
"""True if f is a regular file."""
484
return S_ISREG(os.lstat(f)[ST_MODE])
489
"""True if f is a symlink."""
491
return S_ISLNK(os.lstat(f)[ST_MODE])
495
def is_inside(dir, fname):
496
"""True if fname is inside dir.
498
The parameters should typically be passed to osutils.normpath first, so
499
that . and .. and repeated slashes are eliminated, and the separators
500
are canonical for the platform.
502
The empty string as a dir name is taken as top-of-tree and matches
505
# XXX: Most callers of this can actually do something smarter by
506
# looking at the inventory
516
return fname.startswith(dir)
519
def is_inside_any(dir_list, fname):
520
"""True if fname is inside any of given dirs."""
521
for dirname in dir_list:
522
if is_inside(dirname, fname):
527
def is_inside_or_parent_of_any(dir_list, fname):
528
"""True if fname is a child or a parent of any of the given files."""
529
for dirname in dir_list:
530
if is_inside(dirname, fname) or is_inside(fname, dirname):
535
def pumpfile(fromfile, tofile):
536
"""Copy contents of one file to another."""
539
b = fromfile.read(BUFSIZE)
545
def file_iterator(input_file, readsize=32768):
547
b = input_file.read(readsize)
554
if getattr(f, 'tell', None) is not None:
567
def sha_strings(strings):
568
"""Return the sha-1 of concatenation of strings"""
570
map(s.update, strings)
580
def fingerprint_file(f):
585
return {'size': size,
586
'sha1': s.hexdigest()}
589
def compare_files(a, b):
590
"""Returns true if equal in contents"""
601
def local_time_offset(t=None):
602
"""Return offset of local zone from GMT, either at present or at time t."""
605
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
606
return offset.days * 86400 + offset.seconds
609
def format_date(t, offset=0, timezone='original', date_fmt=None,
611
"""Return a formatted date string.
613
:param t: Seconds since the epoch.
614
:param offset: Timezone offset in seconds east of utc.
615
:param timezone: How to display the time: 'utc', 'original' for the
616
timezone specified by offset, or 'local' for the process's current
618
:param show_offset: Whether to append the timezone.
619
:param date_fmt: strftime format.
621
if timezone == 'utc':
624
elif timezone == 'original':
627
tt = time.gmtime(t + offset)
628
elif timezone == 'local':
629
tt = time.localtime(t)
630
offset = local_time_offset(t)
632
raise errors.BzrError("unsupported timezone format %r" % timezone,
633
['options are "utc", "original", "local"'])
635
date_fmt = "%a %Y-%m-%d %H:%M:%S"
637
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
640
return (time.strftime(date_fmt, tt) + offset_str)
643
def compact_date(when):
644
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
647
def format_delta(delta):
648
"""Get a nice looking string for a time delta.
650
:param delta: The time difference in seconds, can be positive or negative.
651
positive indicates time in the past, negative indicates time in the
652
future. (usually time.time() - stored_time)
653
:return: String formatted to show approximate resolution
659
direction = 'in the future'
663
if seconds < 90: # print seconds up to 90 seconds
665
return '%d second %s' % (seconds, direction,)
667
return '%d seconds %s' % (seconds, direction)
669
minutes = int(seconds / 60)
670
seconds -= 60 * minutes
675
if minutes < 90: # print minutes, seconds up to 90 minutes
677
return '%d minute, %d second%s %s' % (
678
minutes, seconds, plural_seconds, direction)
680
return '%d minutes, %d second%s %s' % (
681
minutes, seconds, plural_seconds, direction)
683
hours = int(minutes / 60)
684
minutes -= 60 * hours
691
return '%d hour, %d minute%s %s' % (hours, minutes,
692
plural_minutes, direction)
693
return '%d hours, %d minute%s %s' % (hours, minutes,
694
plural_minutes, direction)
697
"""Return size of given open file."""
698
return os.fstat(f.fileno())[ST_SIZE]
701
# Define rand_bytes based on platform.
703
# Python 2.4 and later have os.urandom,
704
# but it doesn't work on some arches
706
rand_bytes = os.urandom
707
except (NotImplementedError, AttributeError):
708
# If python doesn't have os.urandom, or it doesn't work,
709
# then try to first pull random data from /dev/urandom
711
rand_bytes = file('/dev/urandom', 'rb').read
712
# Otherwise, use this hack as a last resort
713
except (IOError, OSError):
714
# not well seeded, but better than nothing
719
s += chr(random.randint(0, 255))
724
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
726
"""Return a random string of num alphanumeric characters
728
The result only contains lowercase chars because it may be used on
729
case-insensitive filesystems.
732
for raw_byte in rand_bytes(num):
733
s += ALNUM[ord(raw_byte) % 36]
737
## TODO: We could later have path objects that remember their list
738
## decomposition (might be too tricksy though.)
741
"""Turn string into list of parts."""
742
assert isinstance(p, basestring)
744
# split on either delimiter because people might use either on
746
ps = re.split(r'[\\/]', p)
751
raise errors.BzrError("sorry, %r not allowed in path" % f)
752
elif (f == '.') or (f == ''):
759
assert isinstance(p, (list, tuple))
761
if (f == '..') or (f is None) or (f == ''):
762
raise errors.BzrError("sorry, %r not allowed in path" % f)
766
@deprecated_function(zero_nine)
767
def appendpath(p1, p2):
771
return pathjoin(p1, p2)
775
"""Split s into lines, but without removing the newline characters."""
776
lines = s.split('\n')
777
result = [line + '\n' for line in lines[:-1]]
779
result.append(lines[-1])
783
def hardlinks_good():
784
return sys.platform not in ('win32', 'cygwin', 'darwin')
787
def link_or_copy(src, dest):
788
"""Hardlink a file, or copy it if it can't be hardlinked."""
789
if not hardlinks_good():
790
shutil.copyfile(src, dest)
794
except (OSError, IOError), e:
795
if e.errno != errno.EXDEV:
797
shutil.copyfile(src, dest)
799
def delete_any(full_path):
800
"""Delete a file or directory."""
804
# We may be renaming a dangling inventory id
805
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
811
if getattr(os, 'symlink', None) is not None:
817
def contains_whitespace(s):
818
"""True if there are any whitespace characters in s."""
819
# string.whitespace can include '\xa0' in certain locales, because it is
820
# considered "non-breaking-space" as part of ISO-8859-1. But it
821
# 1) Isn't a breaking whitespace
822
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
824
# 3) '\xa0' isn't unicode safe since it is >128.
826
# This should *not* be a unicode set of characters in case the source
827
# string is not a Unicode string. We can auto-up-cast the characters since
828
# they are ascii, but we don't want to auto-up-cast the string in case it
830
for ch in ' \t\n\r\v\f':
837
def contains_linebreaks(s):
838
"""True if there is any vertical whitespace in s."""
846
def relpath(base, path):
847
"""Return path relative to base, or raise exception.
849
The path may be either an absolute path or a path relative to the
850
current working directory.
852
os.path.commonprefix (python2.4) has a bad bug that it works just
853
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
857
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
858
' exceed the platform minimum length (which is %d)' %
865
while len(head) >= len(base):
868
head, tail = os.path.split(head)
872
raise errors.PathNotChild(rp, base)
880
def safe_unicode(unicode_or_utf8_string):
881
"""Coerce unicode_or_utf8_string into unicode.
883
If it is unicode, it is returned.
884
Otherwise it is decoded from utf-8. If a decoding error
885
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
886
as a BzrBadParameter exception.
888
if isinstance(unicode_or_utf8_string, unicode):
889
return unicode_or_utf8_string
891
return unicode_or_utf8_string.decode('utf8')
892
except UnicodeDecodeError:
893
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
896
def safe_utf8(unicode_or_utf8_string):
897
"""Coerce unicode_or_utf8_string to a utf8 string.
899
If it is a str, it is returned.
900
If it is Unicode, it is encoded into a utf-8 string.
902
if isinstance(unicode_or_utf8_string, str):
903
# TODO: jam 20070209 This is overkill, and probably has an impact on
904
# performance if we are dealing with lots of apis that want a
907
# Make sure it is a valid utf-8 string
908
unicode_or_utf8_string.decode('utf-8')
909
except UnicodeDecodeError:
910
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
911
return unicode_or_utf8_string
912
return unicode_or_utf8_string.encode('utf-8')
915
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
916
' Revision id generators should be creating utf8'
920
def safe_revision_id(unicode_or_utf8_string, warn=True):
921
"""Revision ids should now be utf8, but at one point they were unicode.
923
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
925
:param warn: Functions that are sanitizing user data can set warn=False
926
:return: None or a utf8 revision id.
928
if (unicode_or_utf8_string is None
929
or unicode_or_utf8_string.__class__ == str):
930
return unicode_or_utf8_string
932
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
934
return cache_utf8.encode(unicode_or_utf8_string)
937
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
938
' generators should be creating utf8 file ids.')
941
def safe_file_id(unicode_or_utf8_string, warn=True):
942
"""File ids should now be utf8, but at one point they were unicode.
944
This is the same as safe_utf8, except it uses the cached encode functions
945
to save a little bit of performance.
947
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
949
:param warn: Functions that are sanitizing user data can set warn=False
950
:return: None or a utf8 file id.
952
if (unicode_or_utf8_string is None
953
or unicode_or_utf8_string.__class__ == str):
954
return unicode_or_utf8_string
956
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
958
return cache_utf8.encode(unicode_or_utf8_string)
961
_platform_normalizes_filenames = False
962
if sys.platform == 'darwin':
963
_platform_normalizes_filenames = True
966
def normalizes_filenames():
967
"""Return True if this platform normalizes unicode filenames.
969
Mac OSX does, Windows/Linux do not.
971
return _platform_normalizes_filenames
974
def _accessible_normalized_filename(path):
975
"""Get the unicode normalized path, and if you can access the file.
977
On platforms where the system normalizes filenames (Mac OSX),
978
you can access a file by any path which will normalize correctly.
979
On platforms where the system does not normalize filenames
980
(Windows, Linux), you have to access a file by its exact path.
982
Internally, bzr only supports NFC/NFKC normalization, since that is
983
the standard for XML documents.
985
So return the normalized path, and a flag indicating if the file
986
can be accessed by that path.
989
return unicodedata.normalize('NFKC', unicode(path)), True
992
def _inaccessible_normalized_filename(path):
993
__doc__ = _accessible_normalized_filename.__doc__
995
normalized = unicodedata.normalize('NFKC', unicode(path))
996
return normalized, normalized == path
999
if _platform_normalizes_filenames:
1000
normalized_filename = _accessible_normalized_filename
1002
normalized_filename = _inaccessible_normalized_filename
1005
def terminal_width():
1006
"""Return estimated terminal width."""
1007
if sys.platform == 'win32':
1008
return win32utils.get_console_size()[0]
1011
import struct, fcntl, termios
1012
s = struct.pack('HHHH', 0, 0, 0, 0)
1013
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1014
width = struct.unpack('HHHH', x)[1]
1019
width = int(os.environ['COLUMNS'])
1028
def supports_executable():
1029
return sys.platform != "win32"
1032
def supports_posix_readonly():
1033
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1035
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1036
directory controls creation/deletion, etc.
1038
And under win32, readonly means that the directory itself cannot be
1039
deleted. The contents of a readonly directory can be changed, unlike POSIX
1040
where files in readonly directories cannot be added, deleted or renamed.
1042
return sys.platform != "win32"
1045
def set_or_unset_env(env_variable, value):
1046
"""Modify the environment, setting or removing the env_variable.
1048
:param env_variable: The environment variable in question
1049
:param value: The value to set the environment to. If None, then
1050
the variable will be removed.
1051
:return: The original value of the environment variable.
1053
orig_val = os.environ.get(env_variable)
1055
if orig_val is not None:
1056
del os.environ[env_variable]
1058
if isinstance(value, unicode):
1059
value = value.encode(bzrlib.user_encoding)
1060
os.environ[env_variable] = value
1064
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1067
def check_legal_path(path):
1068
"""Check whether the supplied path is legal.
1069
This is only required on Windows, so we don't test on other platforms
1072
if sys.platform != "win32":
1074
if _validWin32PathRE.match(path) is None:
1075
raise errors.IllegalPath(path)
1078
def walkdirs(top, prefix=""):
1079
"""Yield data about all the directories in a tree.
1081
This yields all the data about the contents of a directory at a time.
1082
After each directory has been yielded, if the caller has mutated the list
1083
to exclude some directories, they are then not descended into.
1085
The data yielded is of the form:
1086
((directory-relpath, directory-path-from-top),
1087
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1088
- directory-relpath is the relative path of the directory being returned
1089
with respect to top. prefix is prepended to this.
1090
- directory-path-from-root is the path including top for this directory.
1091
It is suitable for use with os functions.
1092
- relpath is the relative path within the subtree being walked.
1093
- basename is the basename of the path
1094
- kind is the kind of the file now. If unknown then the file is not
1095
present within the tree - but it may be recorded as versioned. See
1097
- lstat is the stat data *if* the file was statted.
1098
- planned, not implemented:
1099
path_from_tree_root is the path from the root of the tree.
1101
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1102
allows one to walk a subtree but get paths that are relative to a tree
1104
:return: an iterator over the dirs.
1106
#TODO there is a bit of a smell where the results of the directory-
1107
# summary in this, and the path from the root, may not agree
1108
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1109
# potentially confusing output. We should make this more robust - but
1110
# not at a speed cost. RBC 20060731
1112
_directory = _directory_kind
1113
_listdir = os.listdir
1114
_kind_from_mode = _formats.get
1115
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1117
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1118
relroot, _, _, _, top = pending.pop()
1120
relprefix = relroot + u'/'
1123
top_slash = top + u'/'
1126
append = dirblock.append
1127
for name in sorted(_listdir(top)):
1128
abspath = top_slash + name
1129
statvalue = _lstat(abspath)
1130
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1131
append((relprefix + name, name, kind, statvalue, abspath))
1132
yield (relroot, top), dirblock
1134
# push the user specified dirs from dirblock
1135
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1138
def _walkdirs_utf8(top, prefix=""):
1139
"""Yield data about all the directories in a tree.
1141
This yields the same information as walkdirs() only each entry is yielded
1142
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1143
are returned as exact byte-strings.
1145
:return: yields a tuple of (dir_info, [file_info])
1146
dir_info is (utf8_relpath, path-from-top)
1147
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1148
if top is an absolute path, path-from-top is also an absolute path.
1149
path-from-top might be unicode or utf8, but it is the correct path to
1150
pass to os functions to affect the file in question. (such as os.lstat)
1152
fs_encoding = sys.getfilesystemencoding()
1153
if (sys.platform == 'win32' or
1154
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1155
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1157
return _walkdirs_fs_utf8(top, prefix=prefix)
1160
def _walkdirs_fs_utf8(top, prefix=""):
1161
"""See _walkdirs_utf8.
1163
This sub-function is called when we know the filesystem is already in utf8
1164
encoding. So we don't need to transcode filenames.
1167
_directory = _directory_kind
1168
_listdir = os.listdir
1169
_kind_from_mode = _formats.get
1171
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1172
# But we don't actually uses 1-3 in pending, so set them to None
1173
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1175
relroot, _, _, _, top = pending.pop()
1177
relprefix = relroot + '/'
1180
top_slash = top + '/'
1183
append = dirblock.append
1184
for name in sorted(_listdir(top)):
1185
abspath = top_slash + name
1186
statvalue = _lstat(abspath)
1187
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1188
append((relprefix + name, name, kind, statvalue, abspath))
1189
yield (relroot, top), dirblock
1191
# push the user specified dirs from dirblock
1192
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1195
def _walkdirs_unicode_to_utf8(top, prefix=""):
1196
"""See _walkdirs_utf8
1198
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1200
This is currently the fallback code path when the filesystem encoding is
1201
not UTF-8. It may be better to implement an alternative so that we can
1202
safely handle paths that are not properly decodable in the current
1205
_utf8_encode = codecs.getencoder('utf8')
1207
_directory = _directory_kind
1208
_listdir = os.listdir
1209
_kind_from_mode = _formats.get
1211
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1213
relroot, _, _, _, top = pending.pop()
1215
relprefix = relroot + '/'
1218
top_slash = top + u'/'
1221
append = dirblock.append
1222
for name in sorted(_listdir(top)):
1223
name_utf8 = _utf8_encode(name)[0]
1224
abspath = top_slash + name
1225
statvalue = _lstat(abspath)
1226
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1227
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1228
yield (relroot, top), dirblock
1230
# push the user specified dirs from dirblock
1231
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1234
def copy_tree(from_path, to_path, handlers={}):
1235
"""Copy all of the entries in from_path into to_path.
1237
:param from_path: The base directory to copy.
1238
:param to_path: The target directory. If it does not exist, it will
1240
:param handlers: A dictionary of functions, which takes a source and
1241
destinations for files, directories, etc.
1242
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1243
'file', 'directory', and 'symlink' should always exist.
1244
If they are missing, they will be replaced with 'os.mkdir()',
1245
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1247
# Now, just copy the existing cached tree to the new location
1248
# We use a cheap trick here.
1249
# Absolute paths are prefixed with the first parameter
1250
# relative paths are prefixed with the second.
1251
# So we can get both the source and target returned
1252
# without any extra work.
1254
def copy_dir(source, dest):
1257
def copy_link(source, dest):
1258
"""Copy the contents of a symlink"""
1259
link_to = os.readlink(source)
1260
os.symlink(link_to, dest)
1262
real_handlers = {'file':shutil.copy2,
1263
'symlink':copy_link,
1264
'directory':copy_dir,
1266
real_handlers.update(handlers)
1268
if not os.path.exists(to_path):
1269
real_handlers['directory'](from_path, to_path)
1271
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1272
for relpath, name, kind, st, abspath in entries:
1273
real_handlers[kind](abspath, relpath)
1276
def path_prefix_key(path):
1277
"""Generate a prefix-order path key for path.
1279
This can be used to sort paths in the same way that walkdirs does.
1281
return (dirname(path) , path)
1284
def compare_paths_prefix_order(path_a, path_b):
1285
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1286
key_a = path_prefix_key(path_a)
1287
key_b = path_prefix_key(path_b)
1288
return cmp(key_a, key_b)
1291
_cached_user_encoding = None
1294
def get_user_encoding(use_cache=True):
1295
"""Find out what the preferred user encoding is.
1297
This is generally the encoding that is used for command line parameters
1298
and file contents. This may be different from the terminal encoding
1299
or the filesystem encoding.
1301
:param use_cache: Enable cache for detected encoding.
1302
(This parameter is turned on by default,
1303
and required only for selftesting)
1305
:return: A string defining the preferred user encoding
1307
global _cached_user_encoding
1308
if _cached_user_encoding is not None and use_cache:
1309
return _cached_user_encoding
1311
if sys.platform == 'darwin':
1312
# work around egregious python 2.4 bug
1313
sys.platform = 'posix'
1317
sys.platform = 'darwin'
1322
user_encoding = locale.getpreferredencoding()
1323
except locale.Error, e:
1324
sys.stderr.write('bzr: warning: %s\n'
1325
' Could not determine what text encoding to use.\n'
1326
' This error usually means your Python interpreter\n'
1327
' doesn\'t support the locale set by $LANG (%s)\n'
1328
" Continuing with ascii encoding.\n"
1329
% (e, os.environ.get('LANG')))
1330
user_encoding = 'ascii'
1332
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1333
# treat that as ASCII, and not support printing unicode characters to the
1335
if user_encoding in (None, 'cp0'):
1336
user_encoding = 'ascii'
1340
codecs.lookup(user_encoding)
1342
sys.stderr.write('bzr: warning:'
1343
' unknown encoding %s.'
1344
' Continuing with ascii encoding.\n'
1347
user_encoding = 'ascii'
1350
_cached_user_encoding = user_encoding
1352
return user_encoding
1355
def recv_all(socket, bytes):
1356
"""Receive an exact number of bytes.
1358
Regular Socket.recv() may return less than the requested number of bytes,
1359
dependning on what's in the OS buffer. MSG_WAITALL is not available
1360
on all platforms, but this should work everywhere. This will return
1361
less than the requested amount if the remote end closes.
1363
This isn't optimized and is intended mostly for use in testing.
1366
while len(b) < bytes:
1367
new = socket.recv(bytes - len(b))
1373
def dereference_path(path):
1374
"""Determine the real path to a file.
1376
All parent elements are dereferenced. But the file itself is not
1378
:param path: The original path. May be absolute or relative.
1379
:return: the real path *to* the file
1381
parent, base = os.path.split(path)
1382
# The pathjoin for '.' is a workaround for Python bug #1213894.
1383
# (initial path components aren't dereferenced)
1384
return pathjoin(realpath(pathjoin('.', parent)), base)
added
removed
bzrlib/osutils.py
