37
83
os.chmod(filename, mod)
40
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/_~-])')
42
"""Return shell-quoted filename"""
43
## We could be a bit more terse by using double-quotes etc
44
f = _QUOTE_RE.sub(r'\\\1', f)
51
mode = os.lstat(f)[ST_MODE]
59
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
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
148
def kind_marker(kind):
151
elif kind == _directory_kind:
153
elif kind == 'symlink':
156
raise errors.BzrError('invalid file kind %r' % kind)
158
lexists = getattr(os.path, 'lexists', None)
162
if getattr(os, 'lstat') is not None:
168
if e.errno == errno.ENOENT:
171
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
174
def fancy_rename(old, new, rename_func, unlink_func):
175
"""A fancy rename, when you don't have atomic rename.
177
:param old: The old path, to rename from
178
:param new: The new path, to rename to
179
:param rename_func: The potentially non-atomic rename function
180
:param unlink_func: A way to delete the target file if the full rename succeeds
183
# sftp rename doesn't allow overwriting, so play tricks:
185
base = os.path.basename(new)
186
dirname = os.path.dirname(new)
187
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
188
tmp_name = pathjoin(dirname, tmp_name)
190
# Rename the file out of the way, but keep track if it didn't exist
191
# We don't want to grab just any exception
192
# something like EACCES should prevent us from continuing
193
# The downside is that the rename_func has to throw an exception
194
# with an errno = ENOENT, or NoSuchFile
197
rename_func(new, tmp_name)
198
except (errors.NoSuchFile,), e:
201
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
202
# function raises an IOError with errno is None when a rename fails.
203
# This then gets caught here.
204
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
207
if (getattr(e, 'errno', None) is None
208
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
215
# This may throw an exception, in which case success will
217
rename_func(old, new)
221
# If the file used to exist, rename it back into place
222
# otherwise just delete it from the tmp location
224
unlink_func(tmp_name)
226
rename_func(tmp_name, new)
229
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
230
# choke on a Unicode string containing a relative path if
231
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
233
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
234
def _posix_abspath(path):
235
# jam 20060426 rather than encoding to fsencoding
236
# copy posixpath.abspath, but use os.getcwdu instead
237
if not posixpath.isabs(path):
238
path = posixpath.join(getcwd(), path)
239
return posixpath.normpath(path)
242
def _posix_realpath(path):
243
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
246
def _win32_fixdrive(path):
247
"""Force drive letters to be consistent.
249
win32 is inconsistent whether it returns lower or upper case
250
and even if it was consistent the user might type the other
251
so we force it to uppercase
252
running python.exe under cmd.exe return capital C:\\
253
running win32 python inside a cygwin shell returns lowercase c:\\
255
drive, path = _nt_splitdrive(path)
256
return drive.upper() + path
259
def _win32_abspath(path):
260
# Real _nt_abspath doesn't have a problem with a unicode cwd
261
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
264
def _win98_abspath(path):
265
"""Return the absolute version of a path.
266
Windows 98 safe implementation (python reimplementation
267
of Win32 API function GetFullPathNameW)
272
# \\HOST\path => //HOST/path
273
# //HOST/path => //HOST/path
274
# path => C:/cwd/path
277
# check for absolute path
278
drive = _nt_splitdrive(path)[0]
279
if drive == '' and path[:2] not in('//','\\\\'):
281
# we cannot simply os.path.join cwd and path
282
# because os.path.join('C:','/path') produce '/path'
283
# and this is incorrect
284
if path[:1] in ('/','\\'):
285
cwd = _nt_splitdrive(cwd)[0]
287
path = cwd + '\\' + path
288
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
290
if win32utils.winver == 'Windows 98':
291
_win32_abspath = _win98_abspath
294
def _win32_realpath(path):
295
# Real _nt_realpath doesn't have a problem with a unicode cwd
296
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
299
def _win32_pathjoin(*args):
300
return _nt_join(*args).replace('\\', '/')
303
def _win32_normpath(path):
304
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
308
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
311
def _win32_mkdtemp(*args, **kwargs):
312
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
315
def _win32_rename(old, new):
316
"""We expect to be able to atomically replace 'new' with old.
318
On win32, if new exists, it must be moved out of the way first,
322
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
324
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
325
# If we try to rename a non-existant file onto cwd, we get
326
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
327
# if the old path doesn't exist, sometimes we get EACCES
328
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
334
return unicodedata.normalize('NFKC', os.getcwdu())
337
# Default is to just use the python builtins, but these can be rebound on
338
# particular platforms.
339
abspath = _posix_abspath
340
realpath = _posix_realpath
341
pathjoin = os.path.join
342
normpath = os.path.normpath
345
dirname = os.path.dirname
346
basename = os.path.basename
347
split = os.path.split
348
splitext = os.path.splitext
349
# These were already imported into local scope
350
# mkdtemp = tempfile.mkdtemp
351
# rmtree = shutil.rmtree
353
MIN_ABS_PATHLENGTH = 1
356
if sys.platform == 'win32':
357
abspath = _win32_abspath
358
realpath = _win32_realpath
359
pathjoin = _win32_pathjoin
360
normpath = _win32_normpath
361
getcwd = _win32_getcwd
362
mkdtemp = _win32_mkdtemp
363
rename = _win32_rename
365
MIN_ABS_PATHLENGTH = 3
367
def _win32_delete_readonly(function, path, excinfo):
368
"""Error handler for shutil.rmtree function [for win32]
369
Helps to remove files and dirs marked as read-only.
371
exception = excinfo[1]
372
if function in (os.remove, os.rmdir) \
373
and isinstance(exception, OSError) \
374
and exception.errno == errno.EACCES:
380
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
381
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
382
return shutil.rmtree(path, ignore_errors, onerror)
383
elif sys.platform == 'darwin':
387
def get_terminal_encoding():
388
"""Find the best encoding for printing to the screen.
390
This attempts to check both sys.stdout and sys.stdin to see
391
what encoding they are in, and if that fails it falls back to
392
bzrlib.user_encoding.
393
The problem is that on Windows, locale.getpreferredencoding()
394
is not the same encoding as that used by the console:
395
http://mail.python.org/pipermail/python-list/2003-May/162357.html
397
On my standard US Windows XP, the preferred encoding is
398
cp1252, but the console is cp437
400
output_encoding = getattr(sys.stdout, 'encoding', None)
401
if not output_encoding:
402
input_encoding = getattr(sys.stdin, 'encoding', None)
403
if not input_encoding:
404
output_encoding = bzrlib.user_encoding
405
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
407
output_encoding = input_encoding
408
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
410
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
411
if output_encoding == 'cp0':
412
# invalid encoding (cp0 means 'no codepage' on Windows)
413
output_encoding = bzrlib.user_encoding
414
mutter('cp0 is invalid encoding.'
415
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
418
codecs.lookup(output_encoding)
420
sys.stderr.write('bzr: warning:'
421
' unknown terminal encoding %s.\n'
422
' Using encoding %s instead.\n'
423
% (output_encoding, bzrlib.user_encoding)
425
output_encoding = bzrlib.user_encoding
427
return output_encoding
430
def normalizepath(f):
431
if getattr(os.path, 'realpath', None) is not None:
435
[p,e] = os.path.split(f)
436
if e == "" or e == "." or e == "..":
439
return pathjoin(F(p), e)
443
"""Copy a file to a backup.
445
Backups are named in GNU-style, with a ~ suffix.
447
If the file is already a backup, it's not copied.
453
if has_symlinks() and os.path.islink(fn):
454
target = os.readlink(fn)
455
os.symlink(target, bfn)
463
outf = file(bfn, 'wb')
259
615
tt = time.gmtime(t)
261
617
elif timezone == 'original':
264
620
tt = time.gmtime(t + offset)
265
621
elif timezone == 'local':
266
622
tt = time.localtime(t)
267
623
offset = local_time_offset(t)
269
bailout("unsupported timezone format %r",
270
['options are "utc", "original", "local"'])
272
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
273
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
625
raise errors.BzrError("unsupported timezone format %r" % timezone,
626
['options are "utc", "original", "local"'])
628
date_fmt = "%a %Y-%m-%d %H:%M:%S"
630
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
633
return (time.strftime(date_fmt, tt) + offset_str)
276
636
def compact_date(when):
277
637
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
640
def format_delta(delta):
641
"""Get a nice looking string for a time delta.
643
:param delta: The time difference in seconds, can be positive or negative.
644
positive indicates time in the past, negative indicates time in the
645
future. (usually time.time() - stored_time)
646
:return: String formatted to show approximate resolution
652
direction = 'in the future'
656
if seconds < 90: # print seconds up to 90 seconds
658
return '%d second %s' % (seconds, direction,)
660
return '%d seconds %s' % (seconds, direction)
662
minutes = int(seconds / 60)
663
seconds -= 60 * minutes
668
if minutes < 90: # print minutes, seconds up to 90 minutes
670
return '%d minute, %d second%s %s' % (
671
minutes, seconds, plural_seconds, direction)
673
return '%d minutes, %d second%s %s' % (
674
minutes, seconds, plural_seconds, direction)
676
hours = int(minutes / 60)
677
minutes -= 60 * hours
684
return '%d hour, %d minute%s %s' % (hours, minutes,
685
plural_minutes, direction)
686
return '%d hours, %d minute%s %s' % (hours, minutes,
687
plural_minutes, direction)
282
690
"""Return size of given open file."""
283
691
return os.fstat(f.fileno())[ST_SIZE]
286
if hasattr(os, 'urandom'): # python 2.4 and later
694
# Define rand_bytes based on platform.
696
# Python 2.4 and later have os.urandom,
697
# but it doesn't work on some arches
287
699
rand_bytes = os.urandom
289
# FIXME: No good on non-Linux
290
_rand_file = file('/dev/urandom', 'rb')
291
rand_bytes = _rand_file.read
700
except (NotImplementedError, AttributeError):
701
# If python doesn't have os.urandom, or it doesn't work,
702
# then try to first pull random data from /dev/urandom
704
rand_bytes = file('/dev/urandom', 'rb').read
705
# Otherwise, use this hack as a last resort
706
except (IOError, OSError):
707
# not well seeded, but better than nothing
712
s += chr(random.randint(0, 255))
717
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
719
"""Return a random string of num alphanumeric characters
721
The result only contains lowercase chars because it may be used on
722
case-insensitive filesystems.
725
for raw_byte in rand_bytes(num):
726
s += ALNUM[ord(raw_byte) % 36]
294
730
## TODO: We could later have path objects that remember their list
295
731
## decomposition (might be too tricksy though.)
297
733
def splitpath(p):
298
"""Turn string into list of parts.
304
>>> splitpath('a/./b')
306
>>> splitpath('a/.b')
308
>>> splitpath('a/../b')
309
Traceback (most recent call last):
311
BzrError: ("sorry, '..' not allowed in path", [])
313
assert isinstance(p, types.StringTypes)
734
"""Turn string into list of parts."""
735
assert isinstance(p, basestring)
315
737
# split on either delimiter because people might use either on
330
assert isinstance(p, list)
752
assert isinstance(p, (list, tuple))
332
if (f == '..') or (f == None) or (f == ''):
333
bailout("sorry, %r not allowed in path" % f)
334
return os.path.join(*p)
754
if (f == '..') or (f is None) or (f == ''):
755
raise errors.BzrError("sorry, %r not allowed in path" % f)
759
@deprecated_function(zero_nine)
337
760
def appendpath(p1, p2):
341
return os.path.join(p1, p2)
344
def extern_command(cmd, ignore_errors = False):
345
mutter('external command: %s' % `cmd`)
347
if not ignore_errors:
348
bailout('command failed')
764
return pathjoin(p1, p2)
768
"""Split s into lines, but without removing the newline characters."""
769
lines = s.split('\n')
770
result = [line + '\n' for line in lines[:-1]]
772
result.append(lines[-1])
776
def hardlinks_good():
777
return sys.platform not in ('win32', 'cygwin', 'darwin')
780
def link_or_copy(src, dest):
781
"""Hardlink a file, or copy it if it can't be hardlinked."""
782
if not hardlinks_good():
783
shutil.copyfile(src, dest)
787
except (OSError, IOError), e:
788
if e.errno != errno.EXDEV:
790
shutil.copyfile(src, dest)
792
def delete_any(full_path):
793
"""Delete a file or directory."""
797
# We may be renaming a dangling inventory id
798
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
804
if getattr(os, 'symlink', None) is not None:
810
def contains_whitespace(s):
811
"""True if there are any whitespace characters in s."""
812
# string.whitespace can include '\xa0' in certain locales, because it is
813
# considered "non-breaking-space" as part of ISO-8859-1. But it
814
# 1) Isn't a breaking whitespace
815
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
817
# 3) '\xa0' isn't unicode safe since it is >128.
819
# This should *not* be a unicode set of characters in case the source
820
# string is not a Unicode string. We can auto-up-cast the characters since
821
# they are ascii, but we don't want to auto-up-cast the string in case it
823
for ch in ' \t\n\r\v\f':
830
def contains_linebreaks(s):
831
"""True if there is any vertical whitespace in s."""
839
def relpath(base, path):
840
"""Return path relative to base, or raise exception.
842
The path may be either an absolute path or a path relative to the
843
current working directory.
845
os.path.commonprefix (python2.4) has a bad bug that it works just
846
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
850
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
851
' exceed the platform minimum length (which is %d)' %
858
while len(head) >= len(base):
861
head, tail = os.path.split(head)
865
raise errors.PathNotChild(rp, base)
873
def safe_unicode(unicode_or_utf8_string):
874
"""Coerce unicode_or_utf8_string into unicode.
876
If it is unicode, it is returned.
877
Otherwise it is decoded from utf-8. If a decoding error
878
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
879
as a BzrBadParameter exception.
881
if isinstance(unicode_or_utf8_string, unicode):
882
return unicode_or_utf8_string
884
return unicode_or_utf8_string.decode('utf8')
885
except UnicodeDecodeError:
886
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
889
def safe_utf8(unicode_or_utf8_string):
890
"""Coerce unicode_or_utf8_string to a utf8 string.
892
If it is a str, it is returned.
893
If it is Unicode, it is encoded into a utf-8 string.
895
if isinstance(unicode_or_utf8_string, str):
896
# TODO: jam 20070209 This is overkill, and probably has an impact on
897
# performance if we are dealing with lots of apis that want a
900
# Make sure it is a valid utf-8 string
901
unicode_or_utf8_string.decode('utf-8')
902
except UnicodeDecodeError:
903
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
904
return unicode_or_utf8_string
905
return unicode_or_utf8_string.encode('utf-8')
908
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
909
' Revision id generators should be creating utf8'
913
def safe_revision_id(unicode_or_utf8_string, warn=True):
914
"""Revision ids should now be utf8, but at one point they were unicode.
916
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
918
:param warn: Functions that are sanitizing user data can set warn=False
919
:return: None or a utf8 revision id.
921
if (unicode_or_utf8_string is None
922
or unicode_or_utf8_string.__class__ == str):
923
return unicode_or_utf8_string
925
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
927
return cache_utf8.encode(unicode_or_utf8_string)
930
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
931
' generators should be creating utf8 file ids.')
934
def safe_file_id(unicode_or_utf8_string, warn=True):
935
"""File ids should now be utf8, but at one point they were unicode.
937
This is the same as safe_utf8, except it uses the cached encode functions
938
to save a little bit of performance.
940
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
942
:param warn: Functions that are sanitizing user data can set warn=False
943
:return: None or a utf8 file id.
945
if (unicode_or_utf8_string is None
946
or unicode_or_utf8_string.__class__ == str):
947
return unicode_or_utf8_string
949
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
951
return cache_utf8.encode(unicode_or_utf8_string)
954
_platform_normalizes_filenames = False
955
if sys.platform == 'darwin':
956
_platform_normalizes_filenames = True
959
def normalizes_filenames():
960
"""Return True if this platform normalizes unicode filenames.
962
Mac OSX does, Windows/Linux do not.
964
return _platform_normalizes_filenames
967
def _accessible_normalized_filename(path):
968
"""Get the unicode normalized path, and if you can access the file.
970
On platforms where the system normalizes filenames (Mac OSX),
971
you can access a file by any path which will normalize correctly.
972
On platforms where the system does not normalize filenames
973
(Windows, Linux), you have to access a file by its exact path.
975
Internally, bzr only supports NFC/NFKC normalization, since that is
976
the standard for XML documents.
978
So return the normalized path, and a flag indicating if the file
979
can be accessed by that path.
982
return unicodedata.normalize('NFKC', unicode(path)), True
985
def _inaccessible_normalized_filename(path):
986
__doc__ = _accessible_normalized_filename.__doc__
988
normalized = unicodedata.normalize('NFKC', unicode(path))
989
return normalized, normalized == path
992
if _platform_normalizes_filenames:
993
normalized_filename = _accessible_normalized_filename
995
normalized_filename = _inaccessible_normalized_filename
998
def terminal_width():
999
"""Return estimated terminal width."""
1000
if sys.platform == 'win32':
1001
return win32utils.get_console_size()[0]
1004
import struct, fcntl, termios
1005
s = struct.pack('HHHH', 0, 0, 0, 0)
1006
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1007
width = struct.unpack('HHHH', x)[1]
1012
width = int(os.environ['COLUMNS'])
1021
def supports_executable():
1022
return sys.platform != "win32"
1025
def supports_posix_readonly():
1026
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1028
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1029
directory controls creation/deletion, etc.
1031
And under win32, readonly means that the directory itself cannot be
1032
deleted. The contents of a readonly directory can be changed, unlike POSIX
1033
where files in readonly directories cannot be added, deleted or renamed.
1035
return sys.platform != "win32"
1038
def set_or_unset_env(env_variable, value):
1039
"""Modify the environment, setting or removing the env_variable.
1041
:param env_variable: The environment variable in question
1042
:param value: The value to set the environment to. If None, then
1043
the variable will be removed.
1044
:return: The original value of the environment variable.
1046
orig_val = os.environ.get(env_variable)
1048
if orig_val is not None:
1049
del os.environ[env_variable]
1051
if isinstance(value, unicode):
1052
value = value.encode(bzrlib.user_encoding)
1053
os.environ[env_variable] = value
1057
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1060
def check_legal_path(path):
1061
"""Check whether the supplied path is legal.
1062
This is only required on Windows, so we don't test on other platforms
1065
if sys.platform != "win32":
1067
if _validWin32PathRE.match(path) is None:
1068
raise errors.IllegalPath(path)
1071
def walkdirs(top, prefix=""):
1072
"""Yield data about all the directories in a tree.
1074
This yields all the data about the contents of a directory at a time.
1075
After each directory has been yielded, if the caller has mutated the list
1076
to exclude some directories, they are then not descended into.
1078
The data yielded is of the form:
1079
((directory-relpath, directory-path-from-top),
1080
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1081
- directory-relpath is the relative path of the directory being returned
1082
with respect to top. prefix is prepended to this.
1083
- directory-path-from-root is the path including top for this directory.
1084
It is suitable for use with os functions.
1085
- relpath is the relative path within the subtree being walked.
1086
- basename is the basename of the path
1087
- kind is the kind of the file now. If unknown then the file is not
1088
present within the tree - but it may be recorded as versioned. See
1090
- lstat is the stat data *if* the file was statted.
1091
- planned, not implemented:
1092
path_from_tree_root is the path from the root of the tree.
1094
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1095
allows one to walk a subtree but get paths that are relative to a tree
1097
:return: an iterator over the dirs.
1099
#TODO there is a bit of a smell where the results of the directory-
1100
# summary in this, and the path from the root, may not agree
1101
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1102
# potentially confusing output. We should make this more robust - but
1103
# not at a speed cost. RBC 20060731
1105
_directory = _directory_kind
1106
_listdir = os.listdir
1107
_kind_from_mode = _formats.get
1108
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1110
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1111
relroot, _, _, _, top = pending.pop()
1113
relprefix = relroot + u'/'
1116
top_slash = top + u'/'
1119
append = dirblock.append
1120
for name in sorted(_listdir(top)):
1121
abspath = top_slash + name
1122
statvalue = _lstat(abspath)
1123
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1124
append((relprefix + name, name, kind, statvalue, abspath))
1125
yield (relroot, top), dirblock
1127
# push the user specified dirs from dirblock
1128
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1131
def _walkdirs_utf8(top, prefix=""):
1132
"""Yield data about all the directories in a tree.
1134
This yields the same information as walkdirs() only each entry is yielded
1135
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1136
are returned as exact byte-strings.
1138
:return: yields a tuple of (dir_info, [file_info])
1139
dir_info is (utf8_relpath, path-from-top)
1140
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1141
if top is an absolute path, path-from-top is also an absolute path.
1142
path-from-top might be unicode or utf8, but it is the correct path to
1143
pass to os functions to affect the file in question. (such as os.lstat)
1145
fs_encoding = sys.getfilesystemencoding()
1146
if (sys.platform == 'win32' or
1147
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1148
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1150
return _walkdirs_fs_utf8(top, prefix=prefix)
1153
def _walkdirs_fs_utf8(top, prefix=""):
1154
"""See _walkdirs_utf8.
1156
This sub-function is called when we know the filesystem is already in utf8
1157
encoding. So we don't need to transcode filenames.
1160
_directory = _directory_kind
1161
_listdir = os.listdir
1162
_kind_from_mode = _formats.get
1164
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1165
# But we don't actually uses 1-3 in pending, so set them to None
1166
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1168
relroot, _, _, _, top = pending.pop()
1170
relprefix = relroot + '/'
1173
top_slash = top + '/'
1176
append = dirblock.append
1177
for name in sorted(_listdir(top)):
1178
abspath = top_slash + name
1179
statvalue = _lstat(abspath)
1180
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1181
append((relprefix + name, name, kind, statvalue, abspath))
1182
yield (relroot, top), dirblock
1184
# push the user specified dirs from dirblock
1185
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1188
def _walkdirs_unicode_to_utf8(top, prefix=""):
1189
"""See _walkdirs_utf8
1191
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1193
This is currently the fallback code path when the filesystem encoding is
1194
not UTF-8. It may be better to implement an alternative so that we can
1195
safely handle paths that are not properly decodable in the current
1198
_utf8_encode = codecs.getencoder('utf8')
1200
_directory = _directory_kind
1201
_listdir = os.listdir
1202
_kind_from_mode = _formats.get
1204
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1206
relroot, _, _, _, top = pending.pop()
1208
relprefix = relroot + '/'
1211
top_slash = top + u'/'
1214
append = dirblock.append
1215
for name in sorted(_listdir(top)):
1216
name_utf8 = _utf8_encode(name)[0]
1217
abspath = top_slash + name
1218
statvalue = _lstat(abspath)
1219
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1220
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1221
yield (relroot, top), dirblock
1223
# push the user specified dirs from dirblock
1224
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1227
def copy_tree(from_path, to_path, handlers={}):
1228
"""Copy all of the entries in from_path into to_path.
1230
:param from_path: The base directory to copy.
1231
:param to_path: The target directory. If it does not exist, it will
1233
:param handlers: A dictionary of functions, which takes a source and
1234
destinations for files, directories, etc.
1235
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1236
'file', 'directory', and 'symlink' should always exist.
1237
If they are missing, they will be replaced with 'os.mkdir()',
1238
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1240
# Now, just copy the existing cached tree to the new location
1241
# We use a cheap trick here.
1242
# Absolute paths are prefixed with the first parameter
1243
# relative paths are prefixed with the second.
1244
# So we can get both the source and target returned
1245
# without any extra work.
1247
def copy_dir(source, dest):
1250
def copy_link(source, dest):
1251
"""Copy the contents of a symlink"""
1252
link_to = os.readlink(source)
1253
os.symlink(link_to, dest)
1255
real_handlers = {'file':shutil.copy2,
1256
'symlink':copy_link,
1257
'directory':copy_dir,
1259
real_handlers.update(handlers)
1261
if not os.path.exists(to_path):
1262
real_handlers['directory'](from_path, to_path)
1264
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1265
for relpath, name, kind, st, abspath in entries:
1266
real_handlers[kind](abspath, relpath)
1269
def path_prefix_key(path):
1270
"""Generate a prefix-order path key for path.
1272
This can be used to sort paths in the same way that walkdirs does.
1274
return (dirname(path) , path)
1277
def compare_paths_prefix_order(path_a, path_b):
1278
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1279
key_a = path_prefix_key(path_a)
1280
key_b = path_prefix_key(path_b)
1281
return cmp(key_a, key_b)
1284
_cached_user_encoding = None
1287
def get_user_encoding(use_cache=True):
1288
"""Find out what the preferred user encoding is.
1290
This is generally the encoding that is used for command line parameters
1291
and file contents. This may be different from the terminal encoding
1292
or the filesystem encoding.
1294
:param use_cache: Enable cache for detected encoding.
1295
(This parameter is turned on by default,
1296
and required only for selftesting)
1298
:return: A string defining the preferred user encoding
1300
global _cached_user_encoding
1301
if _cached_user_encoding is not None and use_cache:
1302
return _cached_user_encoding
1304
if sys.platform == 'darwin':
1305
# work around egregious python 2.4 bug
1306
sys.platform = 'posix'
1310
sys.platform = 'darwin'
1315
user_encoding = locale.getpreferredencoding()
1316
except locale.Error, e:
1317
sys.stderr.write('bzr: warning: %s\n'
1318
' Could not determine what text encoding to use.\n'
1319
' This error usually means your Python interpreter\n'
1320
' doesn\'t support the locale set by $LANG (%s)\n'
1321
" Continuing with ascii encoding.\n"
1322
% (e, os.environ.get('LANG')))
1323
user_encoding = 'ascii'
1325
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1326
# treat that as ASCII, and not support printing unicode characters to the
1328
if user_encoding in (None, 'cp0'):
1329
user_encoding = 'ascii'
1333
codecs.lookup(user_encoding)
1335
sys.stderr.write('bzr: warning:'
1336
' unknown encoding %s.'
1337
' Continuing with ascii encoding.\n'
1340
user_encoding = 'ascii'
1343
_cached_user_encoding = user_encoding
1345
return user_encoding
1348
def recv_all(socket, bytes):
1349
"""Receive an exact number of bytes.
1351
Regular Socket.recv() may return less than the requested number of bytes,
1352
dependning on what's in the OS buffer. MSG_WAITALL is not available
1353
on all platforms, but this should work everywhere. This will return
1354
less than the requested amount if the remote end closes.
1356
This isn't optimized and is intended mostly for use in testing.
1359
while len(b) < bytes:
1360
new = socket.recv(bytes - len(b))
1366
def dereference_path(path):
1367
"""Determine the real path to a file.
1369
All parent elements are dereferenced. But the file itself is not
1371
:param path: The original path. May be absolute or relative.
1372
:return: the real path *to* the file
1374
parent, base = os.path.split(path)
1375
# The pathjoin for '.' is a workaround for Python bug #1213894.
1376
# (initial path components aren't dereferenced)
1377
return pathjoin(realpath(pathjoin('.', parent)), base)
added
removed
bzrlib/osutils.py
