171
229
rename_func(tmp_name, new)
173
# Default is to just use the python builtins
174
abspath = os.path.abspath
175
realpath = os.path.realpath
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
176
344
pathjoin = os.path.join
177
345
normpath = os.path.normpath
178
346
getcwd = os.getcwdu
179
mkdtemp = tempfile.mkdtemp
180
347
rename = os.rename
181
348
dirname = os.path.dirname
182
349
basename = os.path.basename
184
if os.name == "posix":
185
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
186
# choke on a Unicode string containing a relative path if
187
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
189
_fs_enc = sys.getfilesystemencoding()
191
return os.path.abspath(path.encode(_fs_enc)).decode(_fs_enc)
194
return os.path.realpath(path.encode(_fs_enc)).decode(_fs_enc)
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
196
359
if sys.platform == 'win32':
197
# We need to use the Unicode-aware os.path.abspath and
198
# os.path.realpath on Windows systems.
200
return os.path.abspath(path).replace('\\', '/')
203
return os.path.realpath(path).replace('\\', '/')
206
return os.path.join(*args).replace('\\', '/')
209
return os.path.normpath(path).replace('\\', '/')
212
return os.getcwdu().replace('\\', '/')
214
def mkdtemp(*args, **kwargs):
215
return tempfile.mkdtemp(*args, **kwargs).replace('\\', '/')
217
def rename(old, new):
218
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
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
221
433
def normalizepath(f):
222
if hasattr(os.path, 'realpath'):
434
if getattr(os.path, 'realpath', None) is not None:
636
891
return unicode_or_utf8_string.decode('utf8')
637
892
except UnicodeDecodeError:
638
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
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
641
1005
def terminal_width():
642
1006
"""Return estimated terminal width."""
644
# TODO: Do something smart on Windows?
646
# TODO: Is there anything that gets a better update when the window
647
# is resized while the program is running? We could use the Python termcap
1007
if sys.platform == 'win32':
1008
return win32utils.get_console_size()[0]
650
return int(os.environ['COLUMNS'])
651
except (IndexError, KeyError, ValueError):
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'])
654
1028
def supports_executable():
655
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
