104
_directory_kind = 'directory'
107
stat.S_IFDIR:_directory_kind,
108
stat.S_IFCHR:'chardev',
109
stat.S_IFBLK:'block',
112
stat.S_IFLNK:'symlink',
113
stat.S_IFSOCK:'socket',
117
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
118
"""Generate a file kind from a stat mode. This is used in walkdirs.
120
Its performance is critical: Do not mutate without careful benchmarking.
123
return _formats[stat_mode & 0170000]
128
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
130
return _mapper(_lstat(f).st_mode)
132
if getattr(e, 'errno', None) == errno.ENOENT:
133
raise errors.NoSuchFile(f)
138
"""Return the current umask"""
139
# Assume that people aren't messing with the umask while running
140
# XXX: This is not thread safe, but there is no way to get the
141
# umask without setting it
69
mode = os.lstat(f)[ST_MODE]
147
88
def kind_marker(kind):
148
89
if kind == 'file':
150
elif kind == _directory_kind:
91
elif kind == 'directory':
152
93
elif kind == 'symlink':
155
raise errors.BzrError('invalid file kind %r' % kind)
157
lexists = getattr(os.path, 'lexists', None)
161
if getattr(os, 'lstat') is not None:
167
if e.errno == errno.ENOENT:
170
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
173
def fancy_rename(old, new, rename_func, unlink_func):
174
"""A fancy rename, when you don't have atomic rename.
176
:param old: The old path, to rename from
177
:param new: The new path, to rename to
178
:param rename_func: The potentially non-atomic rename function
179
:param unlink_func: A way to delete the target file if the full rename succeeds
182
# sftp rename doesn't allow overwriting, so play tricks:
184
base = os.path.basename(new)
185
dirname = os.path.dirname(new)
186
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
187
tmp_name = pathjoin(dirname, tmp_name)
189
# Rename the file out of the way, but keep track if it didn't exist
190
# We don't want to grab just any exception
191
# something like EACCES should prevent us from continuing
192
# The downside is that the rename_func has to throw an exception
193
# with an errno = ENOENT, or NoSuchFile
196
rename_func(new, tmp_name)
197
except (errors.NoSuchFile,), e:
200
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
201
# function raises an IOError with errno is None when a rename fails.
202
# This then gets caught here.
203
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
206
if (getattr(e, 'errno', None) is None
207
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
214
# This may throw an exception, in which case success will
216
rename_func(old, new)
220
# If the file used to exist, rename it back into place
221
# otherwise just delete it from the tmp location
223
unlink_func(tmp_name)
225
rename_func(tmp_name, new)
228
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
229
# choke on a Unicode string containing a relative path if
230
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
232
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
233
def _posix_abspath(path):
234
# jam 20060426 rather than encoding to fsencoding
235
# copy posixpath.abspath, but use os.getcwdu instead
236
if not posixpath.isabs(path):
237
path = posixpath.join(getcwd(), path)
238
return posixpath.normpath(path)
241
def _posix_realpath(path):
242
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
245
def _win32_fixdrive(path):
246
"""Force drive letters to be consistent.
248
win32 is inconsistent whether it returns lower or upper case
249
and even if it was consistent the user might type the other
250
so we force it to uppercase
251
running python.exe under cmd.exe return capital C:\\
252
running win32 python inside a cygwin shell returns lowercase c:\\
254
drive, path = _nt_splitdrive(path)
255
return drive.upper() + path
258
def _win32_abspath(path):
259
# Real _nt_abspath doesn't have a problem with a unicode cwd
260
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
263
def _win98_abspath(path):
264
"""Return the absolute version of a path.
265
Windows 98 safe implementation (python reimplementation
266
of Win32 API function GetFullPathNameW)
271
# \\HOST\path => //HOST/path
272
# //HOST/path => //HOST/path
273
# path => C:/cwd/path
276
# check for absolute path
277
drive = _nt_splitdrive(path)[0]
278
if drive == '' and path[:2] not in('//','\\\\'):
280
# we cannot simply os.path.join cwd and path
281
# because os.path.join('C:','/path') produce '/path'
282
# and this is incorrect
283
if path[:1] in ('/','\\'):
284
cwd = _nt_splitdrive(cwd)[0]
286
path = cwd + '\\' + path
287
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
289
if win32utils.winver == 'Windows 98':
290
_win32_abspath = _win98_abspath
293
def _win32_realpath(path):
294
# Real _nt_realpath doesn't have a problem with a unicode cwd
295
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
298
def _win32_pathjoin(*args):
299
return _nt_join(*args).replace('\\', '/')
302
def _win32_normpath(path):
303
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
307
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
310
def _win32_mkdtemp(*args, **kwargs):
311
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
314
def _win32_rename(old, new):
315
"""We expect to be able to atomically replace 'new' with old.
317
On win32, if new exists, it must be moved out of the way first,
321
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
323
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
324
# If we try to rename a non-existant file onto cwd, we get
325
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
326
# if the old path doesn't exist, sometimes we get EACCES
327
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
333
return unicodedata.normalize('NFKC', os.getcwdu())
336
# Default is to just use the python builtins, but these can be rebound on
337
# particular platforms.
338
abspath = _posix_abspath
339
realpath = _posix_realpath
340
pathjoin = os.path.join
341
normpath = os.path.normpath
344
dirname = os.path.dirname
345
basename = os.path.basename
346
split = os.path.split
347
splitext = os.path.splitext
348
# These were already imported into local scope
349
# mkdtemp = tempfile.mkdtemp
350
# rmtree = shutil.rmtree
352
MIN_ABS_PATHLENGTH = 1
355
if sys.platform == 'win32':
356
abspath = _win32_abspath
357
realpath = _win32_realpath
358
pathjoin = _win32_pathjoin
359
normpath = _win32_normpath
360
getcwd = _win32_getcwd
361
mkdtemp = _win32_mkdtemp
362
rename = _win32_rename
364
MIN_ABS_PATHLENGTH = 3
366
def _win32_delete_readonly(function, path, excinfo):
367
"""Error handler for shutil.rmtree function [for win32]
368
Helps to remove files and dirs marked as read-only.
370
exception = excinfo[1]
371
if function in (os.remove, os.rmdir) \
372
and isinstance(exception, OSError) \
373
and exception.errno == errno.EACCES:
379
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
380
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
381
return shutil.rmtree(path, ignore_errors, onerror)
382
elif sys.platform == 'darwin':
386
def get_terminal_encoding():
387
"""Find the best encoding for printing to the screen.
389
This attempts to check both sys.stdout and sys.stdin to see
390
what encoding they are in, and if that fails it falls back to
391
bzrlib.user_encoding.
392
The problem is that on Windows, locale.getpreferredencoding()
393
is not the same encoding as that used by the console:
394
http://mail.python.org/pipermail/python-list/2003-May/162357.html
396
On my standard US Windows XP, the preferred encoding is
397
cp1252, but the console is cp437
399
output_encoding = getattr(sys.stdout, 'encoding', None)
400
if not output_encoding:
401
input_encoding = getattr(sys.stdin, 'encoding', None)
402
if not input_encoding:
403
output_encoding = bzrlib.user_encoding
404
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
406
output_encoding = input_encoding
407
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
409
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
410
if output_encoding == 'cp0':
411
# invalid encoding (cp0 means 'no codepage' on Windows)
412
output_encoding = bzrlib.user_encoding
413
mutter('cp0 is invalid encoding.'
414
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
417
codecs.lookup(output_encoding)
419
sys.stderr.write('bzr: warning:'
420
' unknown terminal encoding %s.\n'
421
' Using encoding %s instead.\n'
422
% (output_encoding, bzrlib.user_encoding)
424
output_encoding = bzrlib.user_encoding
426
return output_encoding
429
def normalizepath(f):
430
if getattr(os.path, 'realpath', None) is not None:
434
[p,e] = os.path.split(f)
435
if e == "" or e == "." or e == "..":
438
return pathjoin(F(p), e)
96
raise BzrError('invalid file kind %r' % kind)
441
99
def backup_file(fn):
779
505
def link_or_copy(src, dest):
780
506
"""Hardlink a file, or copy it if it can't be hardlinked."""
781
507
if not hardlinks_good():
782
shutil.copyfile(src, dest)
785
511
os.link(src, dest)
786
512
except (OSError, IOError), e:
787
513
if e.errno != errno.EXDEV:
789
shutil.copyfile(src, dest)
791
def delete_any(full_path):
792
"""Delete a file or directory."""
796
# We may be renaming a dangling inventory id
797
if e.errno not in (errno.EISDIR, errno.EACCES, errno.EPERM):
803
if getattr(os, 'symlink', None) is not None:
809
def contains_whitespace(s):
810
"""True if there are any whitespace characters in s."""
811
# string.whitespace can include '\xa0' in certain locales, because it is
812
# considered "non-breaking-space" as part of ISO-8859-1. But it
813
# 1) Isn't a breaking whitespace
814
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
816
# 3) '\xa0' isn't unicode safe since it is >128.
818
# This should *not* be a unicode set of characters in case the source
819
# string is not a Unicode string. We can auto-up-cast the characters since
820
# they are ascii, but we don't want to auto-up-cast the string in case it
822
for ch in ' \t\n\r\v\f':
829
def contains_linebreaks(s):
830
"""True if there is any vertical whitespace in s."""
838
def relpath(base, path):
839
"""Return path relative to base, or raise exception.
841
The path may be either an absolute path or a path relative to the
842
current working directory.
844
os.path.commonprefix (python2.4) has a bad bug that it works just
845
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
849
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
850
' exceed the platform minimum length (which is %d)' %
857
while len(head) >= len(base):
860
head, tail = os.path.split(head)
864
raise errors.PathNotChild(rp, base)
872
def safe_unicode(unicode_or_utf8_string):
873
"""Coerce unicode_or_utf8_string into unicode.
875
If it is unicode, it is returned.
876
Otherwise it is decoded from utf-8. If a decoding error
877
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
878
as a BzrBadParameter exception.
880
if isinstance(unicode_or_utf8_string, unicode):
881
return unicode_or_utf8_string
883
return unicode_or_utf8_string.decode('utf8')
884
except UnicodeDecodeError:
885
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
888
def safe_utf8(unicode_or_utf8_string):
889
"""Coerce unicode_or_utf8_string to a utf8 string.
891
If it is a str, it is returned.
892
If it is Unicode, it is encoded into a utf-8 string.
894
if isinstance(unicode_or_utf8_string, str):
895
# TODO: jam 20070209 This is overkill, and probably has an impact on
896
# performance if we are dealing with lots of apis that want a
899
# Make sure it is a valid utf-8 string
900
unicode_or_utf8_string.decode('utf-8')
901
except UnicodeDecodeError:
902
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
903
return unicode_or_utf8_string
904
return unicode_or_utf8_string.encode('utf-8')
907
def safe_revision_id(unicode_or_utf8_string):
908
"""Revision ids should now be utf8, but at one point they were unicode.
910
This is the same as safe_utf8, except it uses the cached encode functions
911
to save a little bit of performance.
913
if unicode_or_utf8_string is None:
915
if isinstance(unicode_or_utf8_string, str):
916
# TODO: jam 20070209 Eventually just remove this check.
918
utf8_str = cache_utf8.get_cached_utf8(unicode_or_utf8_string)
919
except UnicodeDecodeError:
920
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
922
return cache_utf8.encode(unicode_or_utf8_string)
925
# TODO: jam 20070217 We start by just re-using safe_revision_id, but ultimately
926
# we want to use a different dictionary cache, because trapping file ids
927
# and revision ids in the same dict seemed to have a noticable effect on
929
safe_file_id = safe_revision_id
932
_platform_normalizes_filenames = False
933
if sys.platform == 'darwin':
934
_platform_normalizes_filenames = True
937
def normalizes_filenames():
938
"""Return True if this platform normalizes unicode filenames.
940
Mac OSX does, Windows/Linux do not.
942
return _platform_normalizes_filenames
945
def _accessible_normalized_filename(path):
946
"""Get the unicode normalized path, and if you can access the file.
948
On platforms where the system normalizes filenames (Mac OSX),
949
you can access a file by any path which will normalize correctly.
950
On platforms where the system does not normalize filenames
951
(Windows, Linux), you have to access a file by its exact path.
953
Internally, bzr only supports NFC/NFKC normalization, since that is
954
the standard for XML documents.
956
So return the normalized path, and a flag indicating if the file
957
can be accessed by that path.
960
return unicodedata.normalize('NFKC', unicode(path)), True
963
def _inaccessible_normalized_filename(path):
964
__doc__ = _accessible_normalized_filename.__doc__
966
normalized = unicodedata.normalize('NFKC', unicode(path))
967
return normalized, normalized == path
970
if _platform_normalizes_filenames:
971
normalized_filename = _accessible_normalized_filename
973
normalized_filename = _inaccessible_normalized_filename
976
def terminal_width():
977
"""Return estimated terminal width."""
978
if sys.platform == 'win32':
979
return win32utils.get_console_size()[0]
982
import struct, fcntl, termios
983
s = struct.pack('HHHH', 0, 0, 0, 0)
984
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
985
width = struct.unpack('HHHH', x)[1]
990
width = int(os.environ['COLUMNS'])
999
def supports_executable():
1000
return sys.platform != "win32"
1003
def supports_posix_readonly():
1004
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1006
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1007
directory controls creation/deletion, etc.
1009
And under win32, readonly means that the directory itself cannot be
1010
deleted. The contents of a readonly directory can be changed, unlike POSIX
1011
where files in readonly directories cannot be added, deleted or renamed.
1013
return sys.platform != "win32"
1016
def set_or_unset_env(env_variable, value):
1017
"""Modify the environment, setting or removing the env_variable.
1019
:param env_variable: The environment variable in question
1020
:param value: The value to set the environment to. If None, then
1021
the variable will be removed.
1022
:return: The original value of the environment variable.
1024
orig_val = os.environ.get(env_variable)
1026
if orig_val is not None:
1027
del os.environ[env_variable]
1029
if isinstance(value, unicode):
1030
value = value.encode(bzrlib.user_encoding)
1031
os.environ[env_variable] = value
1035
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1038
def check_legal_path(path):
1039
"""Check whether the supplied path is legal.
1040
This is only required on Windows, so we don't test on other platforms
1043
if sys.platform != "win32":
1045
if _validWin32PathRE.match(path) is None:
1046
raise errors.IllegalPath(path)
1049
def walkdirs(top, prefix=""):
1050
"""Yield data about all the directories in a tree.
1052
This yields all the data about the contents of a directory at a time.
1053
After each directory has been yielded, if the caller has mutated the list
1054
to exclude some directories, they are then not descended into.
1056
The data yielded is of the form:
1057
((directory-relpath, directory-path-from-top),
1058
[(relpath, basename, kind, lstat), ...]),
1059
- directory-relpath is the relative path of the directory being returned
1060
with respect to top. prefix is prepended to this.
1061
- directory-path-from-root is the path including top for this directory.
1062
It is suitable for use with os functions.
1063
- relpath is the relative path within the subtree being walked.
1064
- basename is the basename of the path
1065
- kind is the kind of the file now. If unknown then the file is not
1066
present within the tree - but it may be recorded as versioned. See
1068
- lstat is the stat data *if* the file was statted.
1069
- planned, not implemented:
1070
path_from_tree_root is the path from the root of the tree.
1072
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1073
allows one to walk a subtree but get paths that are relative to a tree
1075
:return: an iterator over the dirs.
1077
#TODO there is a bit of a smell where the results of the directory-
1078
# summary in this, and the path from the root, may not agree
1079
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1080
# potentially confusing output. We should make this more robust - but
1081
# not at a speed cost. RBC 20060731
1084
_directory = _directory_kind
1085
_listdir = os.listdir
1086
pending = [(prefix, "", _directory, None, top)]
1089
currentdir = pending.pop()
1090
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1093
relroot = currentdir[0] + '/'
1096
for name in sorted(_listdir(top)):
1097
abspath = top + '/' + name
1098
statvalue = lstat(abspath)
1099
dirblock.append((relroot + name, name,
1100
file_kind_from_stat_mode(statvalue.st_mode),
1101
statvalue, abspath))
1102
yield (currentdir[0], top), dirblock
1103
# push the user specified dirs from dirblock
1104
for dir in reversed(dirblock):
1105
if dir[2] == _directory:
1109
def copy_tree(from_path, to_path, handlers={}):
1110
"""Copy all of the entries in from_path into to_path.
1112
:param from_path: The base directory to copy.
1113
:param to_path: The target directory. If it does not exist, it will
1115
:param handlers: A dictionary of functions, which takes a source and
1116
destinations for files, directories, etc.
1117
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1118
'file', 'directory', and 'symlink' should always exist.
1119
If they are missing, they will be replaced with 'os.mkdir()',
1120
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1122
# Now, just copy the existing cached tree to the new location
1123
# We use a cheap trick here.
1124
# Absolute paths are prefixed with the first parameter
1125
# relative paths are prefixed with the second.
1126
# So we can get both the source and target returned
1127
# without any extra work.
1129
def copy_dir(source, dest):
1132
def copy_link(source, dest):
1133
"""Copy the contents of a symlink"""
1134
link_to = os.readlink(source)
1135
os.symlink(link_to, dest)
1137
real_handlers = {'file':shutil.copy2,
1138
'symlink':copy_link,
1139
'directory':copy_dir,
1141
real_handlers.update(handlers)
1143
if not os.path.exists(to_path):
1144
real_handlers['directory'](from_path, to_path)
1146
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1147
for relpath, name, kind, st, abspath in entries:
1148
real_handlers[kind](abspath, relpath)
1151
def path_prefix_key(path):
1152
"""Generate a prefix-order path key for path.
1154
This can be used to sort paths in the same way that walkdirs does.
1156
return (dirname(path) , path)
1159
def compare_paths_prefix_order(path_a, path_b):
1160
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1161
key_a = path_prefix_key(path_a)
1162
key_b = path_prefix_key(path_b)
1163
return cmp(key_a, key_b)
1166
_cached_user_encoding = None
1169
def get_user_encoding(use_cache=True):
1170
"""Find out what the preferred user encoding is.
1172
This is generally the encoding that is used for command line parameters
1173
and file contents. This may be different from the terminal encoding
1174
or the filesystem encoding.
1176
:param use_cache: Enable cache for detected encoding.
1177
(This parameter is turned on by default,
1178
and required only for selftesting)
1180
:return: A string defining the preferred user encoding
1182
global _cached_user_encoding
1183
if _cached_user_encoding is not None and use_cache:
1184
return _cached_user_encoding
1186
if sys.platform == 'darwin':
1187
# work around egregious python 2.4 bug
1188
sys.platform = 'posix'
1192
sys.platform = 'darwin'
1197
user_encoding = locale.getpreferredencoding()
1198
except locale.Error, e:
1199
sys.stderr.write('bzr: warning: %s\n'
1200
' Could not determine what text encoding to use.\n'
1201
' This error usually means your Python interpreter\n'
1202
' doesn\'t support the locale set by $LANG (%s)\n'
1203
" Continuing with ascii encoding.\n"
1204
% (e, os.environ.get('LANG')))
1205
user_encoding = 'ascii'
1207
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1208
# treat that as ASCII, and not support printing unicode characters to the
1210
if user_encoding in (None, 'cp0'):
1211
user_encoding = 'ascii'
1215
codecs.lookup(user_encoding)
1217
sys.stderr.write('bzr: warning:'
1218
' unknown encoding %s.'
1219
' Continuing with ascii encoding.\n'
1222
user_encoding = 'ascii'
1225
_cached_user_encoding = user_encoding
1227
return user_encoding
1230
def recv_all(socket, bytes):
1231
"""Receive an exact number of bytes.
1233
Regular Socket.recv() may return less than the requested number of bytes,
1234
dependning on what's in the OS buffer. MSG_WAITALL is not available
1235
on all platforms, but this should work everywhere. This will return
1236
less than the requested amount if the remote end closes.
1238
This isn't optimized and is intended mostly for use in testing.
1241
while len(b) < bytes:
1242
new = socket.recv(bytes - len(b))
1248
def dereference_path(path):
1249
"""Determine the real path to a file.
1251
All parent elements are dereferenced. But the file itself is not
1253
:param path: The original path. May be absolute or relative.
1254
:return: the real path *to* the file
1256
parent, base = os.path.split(path)
1257
# The pathjoin for '.' is a workaround for Python bug #1213894.
1258
# (initial path components aren't dereferenced)
1259
return pathjoin(realpath(pathjoin('.', parent)), base)
added
removed
bzrlib/osutils.py
