171
233
rename_func(tmp_name, new)
173
# Default is to just use the python builtins
174
abspath = os.path.abspath
175
realpath = os.path.realpath
236
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
237
# choke on a Unicode string containing a relative path if
238
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
240
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
241
def _posix_abspath(path):
242
# jam 20060426 rather than encoding to fsencoding
243
# copy posixpath.abspath, but use os.getcwdu instead
244
if not posixpath.isabs(path):
245
path = posixpath.join(getcwd(), path)
246
return posixpath.normpath(path)
249
def _posix_realpath(path):
250
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
253
def _win32_fixdrive(path):
254
"""Force drive letters to be consistent.
256
win32 is inconsistent whether it returns lower or upper case
257
and even if it was consistent the user might type the other
258
so we force it to uppercase
259
running python.exe under cmd.exe return capital C:\\
260
running win32 python inside a cygwin shell returns lowercase c:\\
262
drive, path = _nt_splitdrive(path)
263
return drive.upper() + path
266
def _win32_abspath(path):
267
# Real _nt_abspath doesn't have a problem with a unicode cwd
268
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
271
def _win98_abspath(path):
272
"""Return the absolute version of a path.
273
Windows 98 safe implementation (python reimplementation
274
of Win32 API function GetFullPathNameW)
279
# \\HOST\path => //HOST/path
280
# //HOST/path => //HOST/path
281
# path => C:/cwd/path
284
# check for absolute path
285
drive = _nt_splitdrive(path)[0]
286
if drive == '' and path[:2] not in('//','\\\\'):
288
# we cannot simply os.path.join cwd and path
289
# because os.path.join('C:','/path') produce '/path'
290
# and this is incorrect
291
if path[:1] in ('/','\\'):
292
cwd = _nt_splitdrive(cwd)[0]
294
path = cwd + '\\' + path
295
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
297
if win32utils.winver == 'Windows 98':
298
_win32_abspath = _win98_abspath
301
def _win32_realpath(path):
302
# Real _nt_realpath doesn't have a problem with a unicode cwd
303
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
306
def _win32_pathjoin(*args):
307
return _nt_join(*args).replace('\\', '/')
310
def _win32_normpath(path):
311
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
315
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
318
def _win32_mkdtemp(*args, **kwargs):
319
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
322
def _win32_rename(old, new):
323
"""We expect to be able to atomically replace 'new' with old.
325
On win32, if new exists, it must be moved out of the way first,
329
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
331
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
332
# If we try to rename a non-existant file onto cwd, we get
333
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
334
# if the old path doesn't exist, sometimes we get EACCES
335
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
341
return unicodedata.normalize('NFKC', os.getcwdu())
344
# Default is to just use the python builtins, but these can be rebound on
345
# particular platforms.
346
abspath = _posix_abspath
347
realpath = _posix_realpath
176
348
pathjoin = os.path.join
177
349
normpath = os.path.normpath
178
350
getcwd = os.getcwdu
179
mkdtemp = tempfile.mkdtemp
180
351
rename = os.rename
181
352
dirname = os.path.dirname
182
353
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)
354
split = os.path.split
355
splitext = os.path.splitext
356
# These were already imported into local scope
357
# mkdtemp = tempfile.mkdtemp
358
# rmtree = shutil.rmtree
360
MIN_ABS_PATHLENGTH = 1
196
363
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)
364
abspath = _win32_abspath
365
realpath = _win32_realpath
366
pathjoin = _win32_pathjoin
367
normpath = _win32_normpath
368
getcwd = _win32_getcwd
369
mkdtemp = _win32_mkdtemp
370
rename = _win32_rename
372
MIN_ABS_PATHLENGTH = 3
374
def _win32_delete_readonly(function, path, excinfo):
375
"""Error handler for shutil.rmtree function [for win32]
376
Helps to remove files and dirs marked as read-only.
378
exception = excinfo[1]
379
if function in (os.remove, os.rmdir) \
380
and isinstance(exception, OSError) \
381
and exception.errno == errno.EACCES:
387
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
388
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
389
return shutil.rmtree(path, ignore_errors, onerror)
390
elif sys.platform == 'darwin':
394
def get_terminal_encoding():
395
"""Find the best encoding for printing to the screen.
397
This attempts to check both sys.stdout and sys.stdin to see
398
what encoding they are in, and if that fails it falls back to
399
bzrlib.user_encoding.
400
The problem is that on Windows, locale.getpreferredencoding()
401
is not the same encoding as that used by the console:
402
http://mail.python.org/pipermail/python-list/2003-May/162357.html
404
On my standard US Windows XP, the preferred encoding is
405
cp1252, but the console is cp437
407
output_encoding = getattr(sys.stdout, 'encoding', None)
408
if not output_encoding:
409
input_encoding = getattr(sys.stdin, 'encoding', None)
410
if not input_encoding:
411
output_encoding = bzrlib.user_encoding
412
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
414
output_encoding = input_encoding
415
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
417
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
418
if output_encoding == 'cp0':
419
# invalid encoding (cp0 means 'no codepage' on Windows)
420
output_encoding = bzrlib.user_encoding
421
mutter('cp0 is invalid encoding.'
422
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
425
codecs.lookup(output_encoding)
427
sys.stderr.write('bzr: warning:'
428
' unknown terminal encoding %s.\n'
429
' Using encoding %s instead.\n'
430
% (output_encoding, bzrlib.user_encoding)
432
output_encoding = bzrlib.user_encoding
434
return output_encoding
221
437
def normalizepath(f):
222
if hasattr(os.path, 'realpath'):
438
if getattr(os.path, 'realpath', None) is not None:
636
887
return unicode_or_utf8_string.decode('utf8')
637
888
except UnicodeDecodeError:
638
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
889
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
892
def safe_utf8(unicode_or_utf8_string):
893
"""Coerce unicode_or_utf8_string to a utf8 string.
895
If it is a str, it is returned.
896
If it is Unicode, it is encoded into a utf-8 string.
898
if isinstance(unicode_or_utf8_string, str):
899
# TODO: jam 20070209 This is overkill, and probably has an impact on
900
# performance if we are dealing with lots of apis that want a
903
# Make sure it is a valid utf-8 string
904
unicode_or_utf8_string.decode('utf-8')
905
except UnicodeDecodeError:
906
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
907
return unicode_or_utf8_string
908
return unicode_or_utf8_string.encode('utf-8')
911
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
912
' Revision id generators should be creating utf8'
916
def safe_revision_id(unicode_or_utf8_string, warn=True):
917
"""Revision ids should now be utf8, but at one point they were unicode.
919
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
921
:param warn: Functions that are sanitizing user data can set warn=False
922
:return: None or a utf8 revision id.
924
if (unicode_or_utf8_string is None
925
or unicode_or_utf8_string.__class__ == str):
926
return unicode_or_utf8_string
928
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
930
return cache_utf8.encode(unicode_or_utf8_string)
933
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
934
' generators should be creating utf8 file ids.')
937
def safe_file_id(unicode_or_utf8_string, warn=True):
938
"""File ids should now be utf8, but at one point they were unicode.
940
This is the same as safe_utf8, except it uses the cached encode functions
941
to save a little bit of performance.
943
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
945
:param warn: Functions that are sanitizing user data can set warn=False
946
:return: None or a utf8 file id.
948
if (unicode_or_utf8_string is None
949
or unicode_or_utf8_string.__class__ == str):
950
return unicode_or_utf8_string
952
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
954
return cache_utf8.encode(unicode_or_utf8_string)
957
_platform_normalizes_filenames = False
958
if sys.platform == 'darwin':
959
_platform_normalizes_filenames = True
962
def normalizes_filenames():
963
"""Return True if this platform normalizes unicode filenames.
965
Mac OSX does, Windows/Linux do not.
967
return _platform_normalizes_filenames
970
def _accessible_normalized_filename(path):
971
"""Get the unicode normalized path, and if you can access the file.
973
On platforms where the system normalizes filenames (Mac OSX),
974
you can access a file by any path which will normalize correctly.
975
On platforms where the system does not normalize filenames
976
(Windows, Linux), you have to access a file by its exact path.
978
Internally, bzr only supports NFC/NFKC normalization, since that is
979
the standard for XML documents.
981
So return the normalized path, and a flag indicating if the file
982
can be accessed by that path.
985
return unicodedata.normalize('NFKC', unicode(path)), True
988
def _inaccessible_normalized_filename(path):
989
__doc__ = _accessible_normalized_filename.__doc__
991
normalized = unicodedata.normalize('NFKC', unicode(path))
992
return normalized, normalized == path
995
if _platform_normalizes_filenames:
996
normalized_filename = _accessible_normalized_filename
998
normalized_filename = _inaccessible_normalized_filename
641
1001
def terminal_width():
642
1002
"""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
1003
if sys.platform == 'win32':
1004
return win32utils.get_console_size()[0]
650
return int(os.environ['COLUMNS'])
651
except (IndexError, KeyError, ValueError):
1007
import struct, fcntl, termios
1008
s = struct.pack('HHHH', 0, 0, 0, 0)
1009
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1010
width = struct.unpack('HHHH', x)[1]
1015
width = int(os.environ['COLUMNS'])
654
1024
def supports_executable():
655
1025
return sys.platform != "win32"
1028
def supports_posix_readonly():
1029
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1031
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1032
directory controls creation/deletion, etc.
1034
And under win32, readonly means that the directory itself cannot be
1035
deleted. The contents of a readonly directory can be changed, unlike POSIX
1036
where files in readonly directories cannot be added, deleted or renamed.
1038
return sys.platform != "win32"
1041
def set_or_unset_env(env_variable, value):
1042
"""Modify the environment, setting or removing the env_variable.
1044
:param env_variable: The environment variable in question
1045
:param value: The value to set the environment to. If None, then
1046
the variable will be removed.
1047
:return: The original value of the environment variable.
1049
orig_val = os.environ.get(env_variable)
1051
if orig_val is not None:
1052
del os.environ[env_variable]
1054
if isinstance(value, unicode):
1055
value = value.encode(bzrlib.user_encoding)
1056
os.environ[env_variable] = value
1060
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1063
def check_legal_path(path):
1064
"""Check whether the supplied path is legal.
1065
This is only required on Windows, so we don't test on other platforms
1068
if sys.platform != "win32":
1070
if _validWin32PathRE.match(path) is None:
1071
raise errors.IllegalPath(path)
1074
def walkdirs(top, prefix=""):
1075
"""Yield data about all the directories in a tree.
1077
This yields all the data about the contents of a directory at a time.
1078
After each directory has been yielded, if the caller has mutated the list
1079
to exclude some directories, they are then not descended into.
1081
The data yielded is of the form:
1082
((directory-relpath, directory-path-from-top),
1083
[(relpath, basename, kind, lstat, path-from-top), ...]),
1084
- directory-relpath is the relative path of the directory being returned
1085
with respect to top. prefix is prepended to this.
1086
- directory-path-from-root is the path including top for this directory.
1087
It is suitable for use with os functions.
1088
- relpath is the relative path within the subtree being walked.
1089
- basename is the basename of the path
1090
- kind is the kind of the file now. If unknown then the file is not
1091
present within the tree - but it may be recorded as versioned. See
1093
- lstat is the stat data *if* the file was statted.
1094
- planned, not implemented:
1095
path_from_tree_root is the path from the root of the tree.
1097
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1098
allows one to walk a subtree but get paths that are relative to a tree
1100
:return: an iterator over the dirs.
1102
#TODO there is a bit of a smell where the results of the directory-
1103
# summary in this, and the path from the root, may not agree
1104
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1105
# potentially confusing output. We should make this more robust - but
1106
# not at a speed cost. RBC 20060731
1108
_directory = _directory_kind
1109
_listdir = os.listdir
1110
_kind_from_mode = _formats.get
1111
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1113
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1114
relroot, _, _, _, top = pending.pop()
1116
relprefix = relroot + u'/'
1119
top_slash = top + u'/'
1122
append = dirblock.append
1123
for name in sorted(_listdir(top)):
1124
abspath = top_slash + name
1125
statvalue = _lstat(abspath)
1126
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1127
append((relprefix + name, name, kind, statvalue, abspath))
1128
yield (relroot, top), dirblock
1130
# push the user specified dirs from dirblock
1131
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1134
def _walkdirs_utf8(top, prefix=""):
1135
"""Yield data about all the directories in a tree.
1137
This yields the same information as walkdirs() only each entry is yielded
1138
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1139
are returned as exact byte-strings.
1141
:return: yields a tuple of (dir_info, [file_info])
1142
dir_info is (utf8_relpath, path-from-top)
1143
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1144
if top is an absolute path, path-from-top is also an absolute path.
1145
path-from-top might be unicode or utf8, but it is the correct path to
1146
pass to os functions to affect the file in question. (such as os.lstat)
1148
fs_encoding = _fs_enc.upper()
1149
if (sys.platform == 'win32' or
1150
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1151
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1153
return _walkdirs_fs_utf8(top, prefix=prefix)
1156
def _walkdirs_fs_utf8(top, prefix=""):
1157
"""See _walkdirs_utf8.
1159
This sub-function is called when we know the filesystem is already in utf8
1160
encoding. So we don't need to transcode filenames.
1163
_directory = _directory_kind
1164
_listdir = os.listdir
1165
_kind_from_mode = _formats.get
1167
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1168
# But we don't actually uses 1-3 in pending, so set them to None
1169
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1171
relroot, _, _, _, top = pending.pop()
1173
relprefix = relroot + '/'
1176
top_slash = top + '/'
1179
append = dirblock.append
1180
for name in sorted(_listdir(top)):
1181
abspath = top_slash + name
1182
statvalue = _lstat(abspath)
1183
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1184
append((relprefix + name, name, kind, statvalue, abspath))
1185
yield (relroot, top), dirblock
1187
# push the user specified dirs from dirblock
1188
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1191
def _walkdirs_unicode_to_utf8(top, prefix=""):
1192
"""See _walkdirs_utf8
1194
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1196
This is currently the fallback code path when the filesystem encoding is
1197
not UTF-8. It may be better to implement an alternative so that we can
1198
safely handle paths that are not properly decodable in the current
1201
_utf8_encode = codecs.getencoder('utf8')
1203
_directory = _directory_kind
1204
_listdir = os.listdir
1205
_kind_from_mode = _formats.get
1207
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1209
relroot, _, _, _, top = pending.pop()
1211
relprefix = relroot + '/'
1214
top_slash = top + u'/'
1217
append = dirblock.append
1218
for name in sorted(_listdir(top)):
1219
name_utf8 = _utf8_encode(name)[0]
1220
abspath = top_slash + name
1221
statvalue = _lstat(abspath)
1222
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1223
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1224
yield (relroot, top), dirblock
1226
# push the user specified dirs from dirblock
1227
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1230
def copy_tree(from_path, to_path, handlers={}):
1231
"""Copy all of the entries in from_path into to_path.
1233
:param from_path: The base directory to copy.
1234
:param to_path: The target directory. If it does not exist, it will
1236
:param handlers: A dictionary of functions, which takes a source and
1237
destinations for files, directories, etc.
1238
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1239
'file', 'directory', and 'symlink' should always exist.
1240
If they are missing, they will be replaced with 'os.mkdir()',
1241
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1243
# Now, just copy the existing cached tree to the new location
1244
# We use a cheap trick here.
1245
# Absolute paths are prefixed with the first parameter
1246
# relative paths are prefixed with the second.
1247
# So we can get both the source and target returned
1248
# without any extra work.
1250
def copy_dir(source, dest):
1253
def copy_link(source, dest):
1254
"""Copy the contents of a symlink"""
1255
link_to = os.readlink(source)
1256
os.symlink(link_to, dest)
1258
real_handlers = {'file':shutil.copy2,
1259
'symlink':copy_link,
1260
'directory':copy_dir,
1262
real_handlers.update(handlers)
1264
if not os.path.exists(to_path):
1265
real_handlers['directory'](from_path, to_path)
1267
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1268
for relpath, name, kind, st, abspath in entries:
1269
real_handlers[kind](abspath, relpath)
1272
def path_prefix_key(path):
1273
"""Generate a prefix-order path key for path.
1275
This can be used to sort paths in the same way that walkdirs does.
1277
return (dirname(path) , path)
1280
def compare_paths_prefix_order(path_a, path_b):
1281
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1282
key_a = path_prefix_key(path_a)
1283
key_b = path_prefix_key(path_b)
1284
return cmp(key_a, key_b)
1287
_cached_user_encoding = None
1290
def get_user_encoding(use_cache=True):
1291
"""Find out what the preferred user encoding is.
1293
This is generally the encoding that is used for command line parameters
1294
and file contents. This may be different from the terminal encoding
1295
or the filesystem encoding.
1297
:param use_cache: Enable cache for detected encoding.
1298
(This parameter is turned on by default,
1299
and required only for selftesting)
1301
:return: A string defining the preferred user encoding
1303
global _cached_user_encoding
1304
if _cached_user_encoding is not None and use_cache:
1305
return _cached_user_encoding
1307
if sys.platform == 'darwin':
1308
# work around egregious python 2.4 bug
1309
sys.platform = 'posix'
1313
sys.platform = 'darwin'
1318
user_encoding = locale.getpreferredencoding()
1319
except locale.Error, e:
1320
sys.stderr.write('bzr: warning: %s\n'
1321
' Could not determine what text encoding to use.\n'
1322
' This error usually means your Python interpreter\n'
1323
' doesn\'t support the locale set by $LANG (%s)\n'
1324
" Continuing with ascii encoding.\n"
1325
% (e, os.environ.get('LANG')))
1326
user_encoding = 'ascii'
1328
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1329
# treat that as ASCII, and not support printing unicode characters to the
1331
if user_encoding in (None, 'cp0'):
1332
user_encoding = 'ascii'
1336
codecs.lookup(user_encoding)
1338
sys.stderr.write('bzr: warning:'
1339
' unknown encoding %s.'
1340
' Continuing with ascii encoding.\n'
1343
user_encoding = 'ascii'
1346
_cached_user_encoding = user_encoding
1348
return user_encoding
1351
def recv_all(socket, bytes):
1352
"""Receive an exact number of bytes.
1354
Regular Socket.recv() may return less than the requested number of bytes,
1355
dependning on what's in the OS buffer. MSG_WAITALL is not available
1356
on all platforms, but this should work everywhere. This will return
1357
less than the requested amount if the remote end closes.
1359
This isn't optimized and is intended mostly for use in testing.
1362
while len(b) < bytes:
1363
new = socket.recv(bytes - len(b))
1369
def dereference_path(path):
1370
"""Determine the real path to a file.
1372
All parent elements are dereferenced. But the file itself is not
1374
:param path: The original path. May be absolute or relative.
1375
:return: the real path *to* the file
1377
parent, base = os.path.split(path)
1378
# The pathjoin for '.' is a workaround for Python bug #1213894.
1379
# (initial path components aren't dereferenced)
1380
return pathjoin(realpath(pathjoin('.', parent)), base)
1383
def supports_mapi():
1384
"""Return True if we can use MAPI to launch a mail client."""
1385
return sys.platform == "win32"
added
removed
bzrlib/osutils.py
