171
234
rename_func(tmp_name, new)
173
# Default is to just use the python builtins
174
abspath = os.path.abspath
175
realpath = os.path.realpath
237
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
238
# choke on a Unicode string containing a relative path if
239
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
241
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
242
def _posix_abspath(path):
243
# jam 20060426 rather than encoding to fsencoding
244
# copy posixpath.abspath, but use os.getcwdu instead
245
if not posixpath.isabs(path):
246
path = posixpath.join(getcwd(), path)
247
return posixpath.normpath(path)
250
def _posix_realpath(path):
251
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
254
def _win32_fixdrive(path):
255
"""Force drive letters to be consistent.
257
win32 is inconsistent whether it returns lower or upper case
258
and even if it was consistent the user might type the other
259
so we force it to uppercase
260
running python.exe under cmd.exe return capital C:\\
261
running win32 python inside a cygwin shell returns lowercase c:\\
263
drive, path = _nt_splitdrive(path)
264
return drive.upper() + path
267
def _win32_abspath(path):
268
# Real _nt_abspath doesn't have a problem with a unicode cwd
269
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
272
def _win98_abspath(path):
273
"""Return the absolute version of a path.
274
Windows 98 safe implementation (python reimplementation
275
of Win32 API function GetFullPathNameW)
280
# \\HOST\path => //HOST/path
281
# //HOST/path => //HOST/path
282
# path => C:/cwd/path
285
# check for absolute path
286
drive = _nt_splitdrive(path)[0]
287
if drive == '' and path[:2] not in('//','\\\\'):
289
# we cannot simply os.path.join cwd and path
290
# because os.path.join('C:','/path') produce '/path'
291
# and this is incorrect
292
if path[:1] in ('/','\\'):
293
cwd = _nt_splitdrive(cwd)[0]
295
path = cwd + '\\' + path
296
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
298
if win32utils.winver == 'Windows 98':
299
_win32_abspath = _win98_abspath
302
def _win32_realpath(path):
303
# Real _nt_realpath doesn't have a problem with a unicode cwd
304
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
307
def _win32_pathjoin(*args):
308
return _nt_join(*args).replace('\\', '/')
311
def _win32_normpath(path):
312
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
316
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
319
def _win32_mkdtemp(*args, **kwargs):
320
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
323
def _win32_rename(old, new):
324
"""We expect to be able to atomically replace 'new' with old.
326
On win32, if new exists, it must be moved out of the way first,
330
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
332
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
333
# If we try to rename a non-existant file onto cwd, we get
334
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
335
# if the old path doesn't exist, sometimes we get EACCES
336
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
342
return unicodedata.normalize('NFKC', os.getcwdu())
345
# Default is to just use the python builtins, but these can be rebound on
346
# particular platforms.
347
abspath = _posix_abspath
348
realpath = _posix_realpath
176
349
pathjoin = os.path.join
177
350
normpath = os.path.normpath
178
351
getcwd = os.getcwdu
179
mkdtemp = tempfile.mkdtemp
180
352
rename = os.rename
181
353
dirname = os.path.dirname
182
354
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)
355
split = os.path.split
356
splitext = os.path.splitext
357
# These were already imported into local scope
358
# mkdtemp = tempfile.mkdtemp
359
# rmtree = shutil.rmtree
361
MIN_ABS_PATHLENGTH = 1
196
364
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)
365
abspath = _win32_abspath
366
realpath = _win32_realpath
367
pathjoin = _win32_pathjoin
368
normpath = _win32_normpath
369
getcwd = _win32_getcwd
370
mkdtemp = _win32_mkdtemp
371
rename = _win32_rename
373
MIN_ABS_PATHLENGTH = 3
375
def _win32_delete_readonly(function, path, excinfo):
376
"""Error handler for shutil.rmtree function [for win32]
377
Helps to remove files and dirs marked as read-only.
379
exception = excinfo[1]
380
if function in (os.remove, os.rmdir) \
381
and isinstance(exception, OSError) \
382
and exception.errno == errno.EACCES:
388
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
389
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
390
return shutil.rmtree(path, ignore_errors, onerror)
391
elif sys.platform == 'darwin':
395
def get_terminal_encoding():
396
"""Find the best encoding for printing to the screen.
398
This attempts to check both sys.stdout and sys.stdin to see
399
what encoding they are in, and if that fails it falls back to
400
bzrlib.user_encoding.
401
The problem is that on Windows, locale.getpreferredencoding()
402
is not the same encoding as that used by the console:
403
http://mail.python.org/pipermail/python-list/2003-May/162357.html
405
On my standard US Windows XP, the preferred encoding is
406
cp1252, but the console is cp437
408
output_encoding = getattr(sys.stdout, 'encoding', None)
409
if not output_encoding:
410
input_encoding = getattr(sys.stdin, 'encoding', None)
411
if not input_encoding:
412
output_encoding = bzrlib.user_encoding
413
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
415
output_encoding = input_encoding
416
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
418
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
419
if output_encoding == 'cp0':
420
# invalid encoding (cp0 means 'no codepage' on Windows)
421
output_encoding = bzrlib.user_encoding
422
mutter('cp0 is invalid encoding.'
423
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
426
codecs.lookup(output_encoding)
428
sys.stderr.write('bzr: warning:'
429
' unknown terminal encoding %s.\n'
430
' Using encoding %s instead.\n'
431
% (output_encoding, bzrlib.user_encoding)
433
output_encoding = bzrlib.user_encoding
435
return output_encoding
221
438
def normalizepath(f):
222
if hasattr(os.path, 'realpath'):
439
if getattr(os.path, 'realpath', None) is not None:
636
896
return unicode_or_utf8_string.decode('utf8')
637
897
except UnicodeDecodeError:
638
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
898
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
901
def safe_utf8(unicode_or_utf8_string):
902
"""Coerce unicode_or_utf8_string to a utf8 string.
904
If it is a str, it is returned.
905
If it is Unicode, it is encoded into a utf-8 string.
907
if isinstance(unicode_or_utf8_string, str):
908
# TODO: jam 20070209 This is overkill, and probably has an impact on
909
# performance if we are dealing with lots of apis that want a
912
# Make sure it is a valid utf-8 string
913
unicode_or_utf8_string.decode('utf-8')
914
except UnicodeDecodeError:
915
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
916
return unicode_or_utf8_string
917
return unicode_or_utf8_string.encode('utf-8')
920
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
921
' Revision id generators should be creating utf8'
925
def safe_revision_id(unicode_or_utf8_string, warn=True):
926
"""Revision ids should now be utf8, but at one point they were unicode.
928
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
930
:param warn: Functions that are sanitizing user data can set warn=False
931
:return: None or a utf8 revision id.
933
if (unicode_or_utf8_string is None
934
or unicode_or_utf8_string.__class__ == str):
935
return unicode_or_utf8_string
937
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
939
return cache_utf8.encode(unicode_or_utf8_string)
942
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
943
' generators should be creating utf8 file ids.')
946
def safe_file_id(unicode_or_utf8_string, warn=True):
947
"""File ids should now be utf8, but at one point they were unicode.
949
This is the same as safe_utf8, except it uses the cached encode functions
950
to save a little bit of performance.
952
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
954
:param warn: Functions that are sanitizing user data can set warn=False
955
:return: None or a utf8 file id.
957
if (unicode_or_utf8_string is None
958
or unicode_or_utf8_string.__class__ == str):
959
return unicode_or_utf8_string
961
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
963
return cache_utf8.encode(unicode_or_utf8_string)
966
_platform_normalizes_filenames = False
967
if sys.platform == 'darwin':
968
_platform_normalizes_filenames = True
971
def normalizes_filenames():
972
"""Return True if this platform normalizes unicode filenames.
974
Mac OSX does, Windows/Linux do not.
976
return _platform_normalizes_filenames
979
def _accessible_normalized_filename(path):
980
"""Get the unicode normalized path, and if you can access the file.
982
On platforms where the system normalizes filenames (Mac OSX),
983
you can access a file by any path which will normalize correctly.
984
On platforms where the system does not normalize filenames
985
(Windows, Linux), you have to access a file by its exact path.
987
Internally, bzr only supports NFC/NFKC normalization, since that is
988
the standard for XML documents.
990
So return the normalized path, and a flag indicating if the file
991
can be accessed by that path.
994
return unicodedata.normalize('NFKC', unicode(path)), True
997
def _inaccessible_normalized_filename(path):
998
__doc__ = _accessible_normalized_filename.__doc__
1000
normalized = unicodedata.normalize('NFKC', unicode(path))
1001
return normalized, normalized == path
1004
if _platform_normalizes_filenames:
1005
normalized_filename = _accessible_normalized_filename
1007
normalized_filename = _inaccessible_normalized_filename
641
1010
def terminal_width():
642
1011
"""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
1012
if sys.platform == 'win32':
1013
return win32utils.get_console_size()[0]
650
return int(os.environ['COLUMNS'])
651
except (IndexError, KeyError, ValueError):
1016
import struct, fcntl, termios
1017
s = struct.pack('HHHH', 0, 0, 0, 0)
1018
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1019
width = struct.unpack('HHHH', x)[1]
1024
width = int(os.environ['COLUMNS'])
654
1033
def supports_executable():
655
1034
return sys.platform != "win32"
1037
def supports_posix_readonly():
1038
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1040
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1041
directory controls creation/deletion, etc.
1043
And under win32, readonly means that the directory itself cannot be
1044
deleted. The contents of a readonly directory can be changed, unlike POSIX
1045
where files in readonly directories cannot be added, deleted or renamed.
1047
return sys.platform != "win32"
1050
def set_or_unset_env(env_variable, value):
1051
"""Modify the environment, setting or removing the env_variable.
1053
:param env_variable: The environment variable in question
1054
:param value: The value to set the environment to. If None, then
1055
the variable will be removed.
1056
:return: The original value of the environment variable.
1058
orig_val = os.environ.get(env_variable)
1060
if orig_val is not None:
1061
del os.environ[env_variable]
1063
if isinstance(value, unicode):
1064
value = value.encode(bzrlib.user_encoding)
1065
os.environ[env_variable] = value
1069
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1072
def check_legal_path(path):
1073
"""Check whether the supplied path is legal.
1074
This is only required on Windows, so we don't test on other platforms
1077
if sys.platform != "win32":
1079
if _validWin32PathRE.match(path) is None:
1080
raise errors.IllegalPath(path)
1083
def walkdirs(top, prefix=""):
1084
"""Yield data about all the directories in a tree.
1086
This yields all the data about the contents of a directory at a time.
1087
After each directory has been yielded, if the caller has mutated the list
1088
to exclude some directories, they are then not descended into.
1090
The data yielded is of the form:
1091
((directory-relpath, directory-path-from-top),
1092
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1093
- directory-relpath is the relative path of the directory being returned
1094
with respect to top. prefix is prepended to this.
1095
- directory-path-from-root is the path including top for this directory.
1096
It is suitable for use with os functions.
1097
- relpath is the relative path within the subtree being walked.
1098
- basename is the basename of the path
1099
- kind is the kind of the file now. If unknown then the file is not
1100
present within the tree - but it may be recorded as versioned. See
1102
- lstat is the stat data *if* the file was statted.
1103
- planned, not implemented:
1104
path_from_tree_root is the path from the root of the tree.
1106
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1107
allows one to walk a subtree but get paths that are relative to a tree
1109
:return: an iterator over the dirs.
1111
#TODO there is a bit of a smell where the results of the directory-
1112
# summary in this, and the path from the root, may not agree
1113
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1114
# potentially confusing output. We should make this more robust - but
1115
# not at a speed cost. RBC 20060731
1117
_directory = _directory_kind
1118
_listdir = os.listdir
1119
_kind_from_mode = _formats.get
1120
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1122
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1123
relroot, _, _, _, top = pending.pop()
1125
relprefix = relroot + u'/'
1128
top_slash = top + u'/'
1131
append = dirblock.append
1132
for name in sorted(_listdir(top)):
1133
abspath = top_slash + name
1134
statvalue = _lstat(abspath)
1135
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1136
append((relprefix + name, name, kind, statvalue, abspath))
1137
yield (relroot, top), dirblock
1139
# push the user specified dirs from dirblock
1140
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1143
def _walkdirs_utf8(top, prefix=""):
1144
"""Yield data about all the directories in a tree.
1146
This yields the same information as walkdirs() only each entry is yielded
1147
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1148
are returned as exact byte-strings.
1150
:return: yields a tuple of (dir_info, [file_info])
1151
dir_info is (utf8_relpath, path-from-top)
1152
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1153
if top is an absolute path, path-from-top is also an absolute path.
1154
path-from-top might be unicode or utf8, but it is the correct path to
1155
pass to os functions to affect the file in question. (such as os.lstat)
1157
fs_encoding = _fs_enc.upper()
1158
if (sys.platform == 'win32' or
1159
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1160
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1162
return _walkdirs_fs_utf8(top, prefix=prefix)
1165
def _walkdirs_fs_utf8(top, prefix=""):
1166
"""See _walkdirs_utf8.
1168
This sub-function is called when we know the filesystem is already in utf8
1169
encoding. So we don't need to transcode filenames.
1172
_directory = _directory_kind
1173
_listdir = os.listdir
1174
_kind_from_mode = _formats.get
1176
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1177
# But we don't actually uses 1-3 in pending, so set them to None
1178
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1180
relroot, _, _, _, top = pending.pop()
1182
relprefix = relroot + '/'
1185
top_slash = top + '/'
1188
append = dirblock.append
1189
for name in sorted(_listdir(top)):
1190
abspath = top_slash + name
1191
statvalue = _lstat(abspath)
1192
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1193
append((relprefix + name, name, kind, statvalue, abspath))
1194
yield (relroot, top), dirblock
1196
# push the user specified dirs from dirblock
1197
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1200
def _walkdirs_unicode_to_utf8(top, prefix=""):
1201
"""See _walkdirs_utf8
1203
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1205
This is currently the fallback code path when the filesystem encoding is
1206
not UTF-8. It may be better to implement an alternative so that we can
1207
safely handle paths that are not properly decodable in the current
1210
_utf8_encode = codecs.getencoder('utf8')
1212
_directory = _directory_kind
1213
_listdir = os.listdir
1214
_kind_from_mode = _formats.get
1216
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1218
relroot, _, _, _, top = pending.pop()
1220
relprefix = relroot + '/'
1223
top_slash = top + u'/'
1226
append = dirblock.append
1227
for name in sorted(_listdir(top)):
1228
name_utf8 = _utf8_encode(name)[0]
1229
abspath = top_slash + name
1230
statvalue = _lstat(abspath)
1231
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1232
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1233
yield (relroot, top), dirblock
1235
# push the user specified dirs from dirblock
1236
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1239
def copy_tree(from_path, to_path, handlers={}):
1240
"""Copy all of the entries in from_path into to_path.
1242
:param from_path: The base directory to copy.
1243
:param to_path: The target directory. If it does not exist, it will
1245
:param handlers: A dictionary of functions, which takes a source and
1246
destinations for files, directories, etc.
1247
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1248
'file', 'directory', and 'symlink' should always exist.
1249
If they are missing, they will be replaced with 'os.mkdir()',
1250
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1252
# Now, just copy the existing cached tree to the new location
1253
# We use a cheap trick here.
1254
# Absolute paths are prefixed with the first parameter
1255
# relative paths are prefixed with the second.
1256
# So we can get both the source and target returned
1257
# without any extra work.
1259
def copy_dir(source, dest):
1262
def copy_link(source, dest):
1263
"""Copy the contents of a symlink"""
1264
link_to = os.readlink(source)
1265
os.symlink(link_to, dest)
1267
real_handlers = {'file':shutil.copy2,
1268
'symlink':copy_link,
1269
'directory':copy_dir,
1271
real_handlers.update(handlers)
1273
if not os.path.exists(to_path):
1274
real_handlers['directory'](from_path, to_path)
1276
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1277
for relpath, name, kind, st, abspath in entries:
1278
real_handlers[kind](abspath, relpath)
1281
def path_prefix_key(path):
1282
"""Generate a prefix-order path key for path.
1284
This can be used to sort paths in the same way that walkdirs does.
1286
return (dirname(path) , path)
1289
def compare_paths_prefix_order(path_a, path_b):
1290
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1291
key_a = path_prefix_key(path_a)
1292
key_b = path_prefix_key(path_b)
1293
return cmp(key_a, key_b)
1296
_cached_user_encoding = None
1299
def get_user_encoding(use_cache=True):
1300
"""Find out what the preferred user encoding is.
1302
This is generally the encoding that is used for command line parameters
1303
and file contents. This may be different from the terminal encoding
1304
or the filesystem encoding.
1306
:param use_cache: Enable cache for detected encoding.
1307
(This parameter is turned on by default,
1308
and required only for selftesting)
1310
:return: A string defining the preferred user encoding
1312
global _cached_user_encoding
1313
if _cached_user_encoding is not None and use_cache:
1314
return _cached_user_encoding
1316
if sys.platform == 'darwin':
1317
# work around egregious python 2.4 bug
1318
sys.platform = 'posix'
1322
sys.platform = 'darwin'
1327
user_encoding = locale.getpreferredencoding()
1328
except locale.Error, e:
1329
sys.stderr.write('bzr: warning: %s\n'
1330
' Could not determine what text encoding to use.\n'
1331
' This error usually means your Python interpreter\n'
1332
' doesn\'t support the locale set by $LANG (%s)\n'
1333
" Continuing with ascii encoding.\n"
1334
% (e, os.environ.get('LANG')))
1335
user_encoding = 'ascii'
1337
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1338
# treat that as ASCII, and not support printing unicode characters to the
1340
if user_encoding in (None, 'cp0'):
1341
user_encoding = 'ascii'
1345
codecs.lookup(user_encoding)
1347
sys.stderr.write('bzr: warning:'
1348
' unknown encoding %s.'
1349
' Continuing with ascii encoding.\n'
1352
user_encoding = 'ascii'
1355
_cached_user_encoding = user_encoding
1357
return user_encoding
1360
def recv_all(socket, bytes):
1361
"""Receive an exact number of bytes.
1363
Regular Socket.recv() may return less than the requested number of bytes,
1364
dependning on what's in the OS buffer. MSG_WAITALL is not available
1365
on all platforms, but this should work everywhere. This will return
1366
less than the requested amount if the remote end closes.
1368
This isn't optimized and is intended mostly for use in testing.
1371
while len(b) < bytes:
1372
new = socket.recv(bytes - len(b))
1378
def dereference_path(path):
1379
"""Determine the real path to a file.
1381
All parent elements are dereferenced. But the file itself is not
1383
:param path: The original path. May be absolute or relative.
1384
:return: the real path *to* the file
1386
parent, base = os.path.split(path)
1387
# The pathjoin for '.' is a workaround for Python bug #1213894.
1388
# (initial path components aren't dereferenced)
1389
return pathjoin(realpath(pathjoin('.', parent)), base)
added
removed
bzrlib/osutils.py
