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
881
return unicode_or_utf8_string.decode('utf8')
637
882
except UnicodeDecodeError:
638
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
883
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
886
def safe_utf8(unicode_or_utf8_string):
887
"""Coerce unicode_or_utf8_string to a utf8 string.
889
If it is a str, it is returned.
890
If it is Unicode, it is encoded into a utf-8 string.
892
if isinstance(unicode_or_utf8_string, str):
893
# TODO: jam 20070209 This is overkill, and probably has an impact on
894
# performance if we are dealing with lots of apis that want a
897
# Make sure it is a valid utf-8 string
898
unicode_or_utf8_string.decode('utf-8')
899
except UnicodeDecodeError:
900
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
901
return unicode_or_utf8_string
902
return unicode_or_utf8_string.encode('utf-8')
905
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
906
' Revision id generators should be creating utf8'
910
def safe_revision_id(unicode_or_utf8_string, warn=True):
911
"""Revision ids should now be utf8, but at one point they were unicode.
913
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
915
:param warn: Functions that are sanitizing user data can set warn=False
916
:return: None or a utf8 revision id.
918
if (unicode_or_utf8_string is None
919
or unicode_or_utf8_string.__class__ == str):
920
return unicode_or_utf8_string
922
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
924
return cache_utf8.encode(unicode_or_utf8_string)
927
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
928
' generators should be creating utf8 file ids.')
931
def safe_file_id(unicode_or_utf8_string, warn=True):
932
"""File ids should now be utf8, but at one point they were unicode.
934
This is the same as safe_utf8, except it uses the cached encode functions
935
to save a little bit of performance.
937
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
939
:param warn: Functions that are sanitizing user data can set warn=False
940
:return: None or a utf8 file id.
942
if (unicode_or_utf8_string is None
943
or unicode_or_utf8_string.__class__ == str):
944
return unicode_or_utf8_string
946
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
948
return cache_utf8.encode(unicode_or_utf8_string)
951
_platform_normalizes_filenames = False
952
if sys.platform == 'darwin':
953
_platform_normalizes_filenames = True
956
def normalizes_filenames():
957
"""Return True if this platform normalizes unicode filenames.
959
Mac OSX does, Windows/Linux do not.
961
return _platform_normalizes_filenames
964
def _accessible_normalized_filename(path):
965
"""Get the unicode normalized path, and if you can access the file.
967
On platforms where the system normalizes filenames (Mac OSX),
968
you can access a file by any path which will normalize correctly.
969
On platforms where the system does not normalize filenames
970
(Windows, Linux), you have to access a file by its exact path.
972
Internally, bzr only supports NFC/NFKC normalization, since that is
973
the standard for XML documents.
975
So return the normalized path, and a flag indicating if the file
976
can be accessed by that path.
979
return unicodedata.normalize('NFKC', unicode(path)), True
982
def _inaccessible_normalized_filename(path):
983
__doc__ = _accessible_normalized_filename.__doc__
985
normalized = unicodedata.normalize('NFKC', unicode(path))
986
return normalized, normalized == path
989
if _platform_normalizes_filenames:
990
normalized_filename = _accessible_normalized_filename
992
normalized_filename = _inaccessible_normalized_filename
641
995
def terminal_width():
642
996
"""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
997
if sys.platform == 'win32':
998
return win32utils.get_console_size()[0]
650
return int(os.environ['COLUMNS'])
651
except (IndexError, KeyError, ValueError):
1001
import struct, fcntl, termios
1002
s = struct.pack('HHHH', 0, 0, 0, 0)
1003
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1004
width = struct.unpack('HHHH', x)[1]
1009
width = int(os.environ['COLUMNS'])
654
1018
def supports_executable():
655
1019
return sys.platform != "win32"
1022
def supports_posix_readonly():
1023
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1025
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1026
directory controls creation/deletion, etc.
1028
And under win32, readonly means that the directory itself cannot be
1029
deleted. The contents of a readonly directory can be changed, unlike POSIX
1030
where files in readonly directories cannot be added, deleted or renamed.
1032
return sys.platform != "win32"
1035
def set_or_unset_env(env_variable, value):
1036
"""Modify the environment, setting or removing the env_variable.
1038
:param env_variable: The environment variable in question
1039
:param value: The value to set the environment to. If None, then
1040
the variable will be removed.
1041
:return: The original value of the environment variable.
1043
orig_val = os.environ.get(env_variable)
1045
if orig_val is not None:
1046
del os.environ[env_variable]
1048
if isinstance(value, unicode):
1049
value = value.encode(bzrlib.user_encoding)
1050
os.environ[env_variable] = value
1054
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1057
def check_legal_path(path):
1058
"""Check whether the supplied path is legal.
1059
This is only required on Windows, so we don't test on other platforms
1062
if sys.platform != "win32":
1064
if _validWin32PathRE.match(path) is None:
1065
raise errors.IllegalPath(path)
1068
def walkdirs(top, prefix=""):
1069
"""Yield data about all the directories in a tree.
1071
This yields all the data about the contents of a directory at a time.
1072
After each directory has been yielded, if the caller has mutated the list
1073
to exclude some directories, they are then not descended into.
1075
The data yielded is of the form:
1076
((directory-relpath, directory-path-from-top),
1077
[(relpath, basename, kind, lstat, path-from-top), ...]),
1078
- directory-relpath is the relative path of the directory being returned
1079
with respect to top. prefix is prepended to this.
1080
- directory-path-from-root is the path including top for this directory.
1081
It is suitable for use with os functions.
1082
- relpath is the relative path within the subtree being walked.
1083
- basename is the basename of the path
1084
- kind is the kind of the file now. If unknown then the file is not
1085
present within the tree - but it may be recorded as versioned. See
1087
- lstat is the stat data *if* the file was statted.
1088
- planned, not implemented:
1089
path_from_tree_root is the path from the root of the tree.
1091
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1092
allows one to walk a subtree but get paths that are relative to a tree
1094
:return: an iterator over the dirs.
1096
#TODO there is a bit of a smell where the results of the directory-
1097
# summary in this, and the path from the root, may not agree
1098
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1099
# potentially confusing output. We should make this more robust - but
1100
# not at a speed cost. RBC 20060731
1102
_directory = _directory_kind
1103
_listdir = os.listdir
1104
_kind_from_mode = _formats.get
1105
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1107
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1108
relroot, _, _, _, top = pending.pop()
1110
relprefix = relroot + u'/'
1113
top_slash = top + u'/'
1116
append = dirblock.append
1117
for name in sorted(_listdir(top)):
1118
abspath = top_slash + name
1119
statvalue = _lstat(abspath)
1120
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1121
append((relprefix + name, name, kind, statvalue, abspath))
1122
yield (relroot, top), dirblock
1124
# push the user specified dirs from dirblock
1125
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1128
def _walkdirs_utf8(top, prefix=""):
1129
"""Yield data about all the directories in a tree.
1131
This yields the same information as walkdirs() only each entry is yielded
1132
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1133
are returned as exact byte-strings.
1135
:return: yields a tuple of (dir_info, [file_info])
1136
dir_info is (utf8_relpath, path-from-top)
1137
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1138
if top is an absolute path, path-from-top is also an absolute path.
1139
path-from-top might be unicode or utf8, but it is the correct path to
1140
pass to os functions to affect the file in question. (such as os.lstat)
1142
fs_encoding = _fs_enc.upper()
1143
if (sys.platform == 'win32' or
1144
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1145
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1147
return _walkdirs_fs_utf8(top, prefix=prefix)
1150
def _walkdirs_fs_utf8(top, prefix=""):
1151
"""See _walkdirs_utf8.
1153
This sub-function is called when we know the filesystem is already in utf8
1154
encoding. So we don't need to transcode filenames.
1157
_directory = _directory_kind
1158
_listdir = os.listdir
1159
_kind_from_mode = _formats.get
1161
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1162
# But we don't actually uses 1-3 in pending, so set them to None
1163
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1165
relroot, _, _, _, top = pending.pop()
1167
relprefix = relroot + '/'
1170
top_slash = top + '/'
1173
append = dirblock.append
1174
for name in sorted(_listdir(top)):
1175
abspath = top_slash + name
1176
statvalue = _lstat(abspath)
1177
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1178
append((relprefix + name, name, kind, statvalue, abspath))
1179
yield (relroot, top), dirblock
1181
# push the user specified dirs from dirblock
1182
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1185
def _walkdirs_unicode_to_utf8(top, prefix=""):
1186
"""See _walkdirs_utf8
1188
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1190
This is currently the fallback code path when the filesystem encoding is
1191
not UTF-8. It may be better to implement an alternative so that we can
1192
safely handle paths that are not properly decodable in the current
1195
_utf8_encode = codecs.getencoder('utf8')
1197
_directory = _directory_kind
1198
_listdir = os.listdir
1199
_kind_from_mode = _formats.get
1201
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1203
relroot, _, _, _, top = pending.pop()
1205
relprefix = relroot + '/'
1208
top_slash = top + u'/'
1211
append = dirblock.append
1212
for name in sorted(_listdir(top)):
1213
name_utf8 = _utf8_encode(name)[0]
1214
abspath = top_slash + name
1215
statvalue = _lstat(abspath)
1216
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1217
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1218
yield (relroot, top), dirblock
1220
# push the user specified dirs from dirblock
1221
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1224
def copy_tree(from_path, to_path, handlers={}):
1225
"""Copy all of the entries in from_path into to_path.
1227
:param from_path: The base directory to copy.
1228
:param to_path: The target directory. If it does not exist, it will
1230
:param handlers: A dictionary of functions, which takes a source and
1231
destinations for files, directories, etc.
1232
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1233
'file', 'directory', and 'symlink' should always exist.
1234
If they are missing, they will be replaced with 'os.mkdir()',
1235
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1237
# Now, just copy the existing cached tree to the new location
1238
# We use a cheap trick here.
1239
# Absolute paths are prefixed with the first parameter
1240
# relative paths are prefixed with the second.
1241
# So we can get both the source and target returned
1242
# without any extra work.
1244
def copy_dir(source, dest):
1247
def copy_link(source, dest):
1248
"""Copy the contents of a symlink"""
1249
link_to = os.readlink(source)
1250
os.symlink(link_to, dest)
1252
real_handlers = {'file':shutil.copy2,
1253
'symlink':copy_link,
1254
'directory':copy_dir,
1256
real_handlers.update(handlers)
1258
if not os.path.exists(to_path):
1259
real_handlers['directory'](from_path, to_path)
1261
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1262
for relpath, name, kind, st, abspath in entries:
1263
real_handlers[kind](abspath, relpath)
1266
def path_prefix_key(path):
1267
"""Generate a prefix-order path key for path.
1269
This can be used to sort paths in the same way that walkdirs does.
1271
return (dirname(path) , path)
1274
def compare_paths_prefix_order(path_a, path_b):
1275
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1276
key_a = path_prefix_key(path_a)
1277
key_b = path_prefix_key(path_b)
1278
return cmp(key_a, key_b)
1281
_cached_user_encoding = None
1284
def get_user_encoding(use_cache=True):
1285
"""Find out what the preferred user encoding is.
1287
This is generally the encoding that is used for command line parameters
1288
and file contents. This may be different from the terminal encoding
1289
or the filesystem encoding.
1291
:param use_cache: Enable cache for detected encoding.
1292
(This parameter is turned on by default,
1293
and required only for selftesting)
1295
:return: A string defining the preferred user encoding
1297
global _cached_user_encoding
1298
if _cached_user_encoding is not None and use_cache:
1299
return _cached_user_encoding
1301
if sys.platform == 'darwin':
1302
# work around egregious python 2.4 bug
1303
sys.platform = 'posix'
1307
sys.platform = 'darwin'
1312
user_encoding = locale.getpreferredencoding()
1313
except locale.Error, e:
1314
sys.stderr.write('bzr: warning: %s\n'
1315
' Could not determine what text encoding to use.\n'
1316
' This error usually means your Python interpreter\n'
1317
' doesn\'t support the locale set by $LANG (%s)\n'
1318
" Continuing with ascii encoding.\n"
1319
% (e, os.environ.get('LANG')))
1320
user_encoding = 'ascii'
1322
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1323
# treat that as ASCII, and not support printing unicode characters to the
1325
if user_encoding in (None, 'cp0'):
1326
user_encoding = 'ascii'
1330
codecs.lookup(user_encoding)
1332
sys.stderr.write('bzr: warning:'
1333
' unknown encoding %s.'
1334
' Continuing with ascii encoding.\n'
1337
user_encoding = 'ascii'
1340
_cached_user_encoding = user_encoding
1342
return user_encoding
1345
def recv_all(socket, bytes):
1346
"""Receive an exact number of bytes.
1348
Regular Socket.recv() may return less than the requested number of bytes,
1349
dependning on what's in the OS buffer. MSG_WAITALL is not available
1350
on all platforms, but this should work everywhere. This will return
1351
less than the requested amount if the remote end closes.
1353
This isn't optimized and is intended mostly for use in testing.
1356
while len(b) < bytes:
1357
new = socket.recv(bytes - len(b))
1363
def dereference_path(path):
1364
"""Determine the real path to a file.
1366
All parent elements are dereferenced. But the file itself is not
1368
:param path: The original path. May be absolute or relative.
1369
:return: the real path *to* the file
1371
parent, base = os.path.split(path)
1372
# The pathjoin for '.' is a workaround for Python bug #1213894.
1373
# (initial path components aren't dereferenced)
1374
return pathjoin(realpath(pathjoin('.', parent)), base)
1377
def supports_mapi():
1378
"""Return True if we can use MAPI to launch a mail client."""
1379
return sys.platform == "win32"
added
removed
bzrlib/osutils.py
