225
171
rename_func(tmp_name, new)
228
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
229
# choke on a Unicode string containing a relative path if
230
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
232
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
233
def _posix_abspath(path):
234
# jam 20060426 rather than encoding to fsencoding
235
# copy posixpath.abspath, but use os.getcwdu instead
236
if not posixpath.isabs(path):
237
path = posixpath.join(getcwd(), path)
238
return posixpath.normpath(path)
241
def _posix_realpath(path):
242
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
245
def _win32_fixdrive(path):
246
"""Force drive letters to be consistent.
248
win32 is inconsistent whether it returns lower or upper case
249
and even if it was consistent the user might type the other
250
so we force it to uppercase
251
running python.exe under cmd.exe return capital C:\\
252
running win32 python inside a cygwin shell returns lowercase c:\\
254
drive, path = _nt_splitdrive(path)
255
return drive.upper() + path
258
def _win32_abspath(path):
259
# Real _nt_abspath doesn't have a problem with a unicode cwd
260
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
263
def _win98_abspath(path):
264
"""Return the absolute version of a path.
265
Windows 98 safe implementation (python reimplementation
266
of Win32 API function GetFullPathNameW)
271
# \\HOST\path => //HOST/path
272
# //HOST/path => //HOST/path
273
# path => C:/cwd/path
276
# check for absolute path
277
drive = _nt_splitdrive(path)[0]
278
if drive == '' and path[:2] not in('//','\\\\'):
280
# we cannot simply os.path.join cwd and path
281
# because os.path.join('C:','/path') produce '/path'
282
# and this is incorrect
283
if path[:1] in ('/','\\'):
284
cwd = _nt_splitdrive(cwd)[0]
286
path = cwd + '\\' + path
287
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
289
if win32utils.winver == 'Windows 98':
290
_win32_abspath = _win98_abspath
293
def _win32_realpath(path):
294
# Real _nt_realpath doesn't have a problem with a unicode cwd
295
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
298
def _win32_pathjoin(*args):
299
return _nt_join(*args).replace('\\', '/')
302
def _win32_normpath(path):
303
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
307
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
310
def _win32_mkdtemp(*args, **kwargs):
311
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
314
def _win32_rename(old, new):
315
"""We expect to be able to atomically replace 'new' with old.
317
On win32, if new exists, it must be moved out of the way first,
321
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
323
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
324
# If we try to rename a non-existant file onto cwd, we get
325
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
326
# if the old path doesn't exist, sometimes we get EACCES
327
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
333
return unicodedata.normalize('NFKC', os.getcwdu())
336
# Default is to just use the python builtins, but these can be rebound on
337
# particular platforms.
338
abspath = _posix_abspath
339
realpath = _posix_realpath
173
# Default is to just use the python builtins
174
abspath = os.path.abspath
175
realpath = os.path.realpath
340
176
pathjoin = os.path.join
341
177
normpath = os.path.normpath
342
178
getcwd = os.getcwdu
179
mkdtemp = tempfile.mkdtemp
343
180
rename = os.rename
344
181
dirname = os.path.dirname
345
182
basename = os.path.basename
346
split = os.path.split
347
splitext = os.path.splitext
348
# These were already imported into local scope
349
# mkdtemp = tempfile.mkdtemp
350
# rmtree = shutil.rmtree
352
MIN_ABS_PATHLENGTH = 1
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
196
if sys.platform == 'win32':
356
abspath = _win32_abspath
357
realpath = _win32_realpath
358
pathjoin = _win32_pathjoin
359
normpath = _win32_normpath
360
getcwd = _win32_getcwd
361
mkdtemp = _win32_mkdtemp
362
rename = _win32_rename
364
MIN_ABS_PATHLENGTH = 3
366
def _win32_delete_readonly(function, path, excinfo):
367
"""Error handler for shutil.rmtree function [for win32]
368
Helps to remove files and dirs marked as read-only.
370
exception = excinfo[1]
371
if function in (os.remove, os.rmdir) \
372
and isinstance(exception, OSError) \
373
and exception.errno == errno.EACCES:
379
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
380
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
381
return shutil.rmtree(path, ignore_errors, onerror)
382
elif sys.platform == 'darwin':
386
def get_terminal_encoding():
387
"""Find the best encoding for printing to the screen.
389
This attempts to check both sys.stdout and sys.stdin to see
390
what encoding they are in, and if that fails it falls back to
391
bzrlib.user_encoding.
392
The problem is that on Windows, locale.getpreferredencoding()
393
is not the same encoding as that used by the console:
394
http://mail.python.org/pipermail/python-list/2003-May/162357.html
396
On my standard US Windows XP, the preferred encoding is
397
cp1252, but the console is cp437
399
output_encoding = getattr(sys.stdout, 'encoding', None)
400
if not output_encoding:
401
input_encoding = getattr(sys.stdin, 'encoding', None)
402
if not input_encoding:
403
output_encoding = bzrlib.user_encoding
404
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
406
output_encoding = input_encoding
407
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
409
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
410
if output_encoding == 'cp0':
411
# invalid encoding (cp0 means 'no codepage' on Windows)
412
output_encoding = bzrlib.user_encoding
413
mutter('cp0 is invalid encoding.'
414
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
417
codecs.lookup(output_encoding)
419
sys.stderr.write('bzr: warning:'
420
' unknown terminal encoding %s.\n'
421
' Using encoding %s instead.\n'
422
% (output_encoding, bzrlib.user_encoding)
424
output_encoding = bzrlib.user_encoding
426
return output_encoding
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)
429
221
def normalizepath(f):
430
if getattr(os.path, 'realpath', None) is not None:
222
if hasattr(os.path, 'realpath'):
883
619
return unicode_or_utf8_string.decode('utf8')
884
620
except UnicodeDecodeError:
885
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
888
def safe_utf8(unicode_or_utf8_string):
889
"""Coerce unicode_or_utf8_string to a utf8 string.
891
If it is a str, it is returned.
892
If it is Unicode, it is encoded into a utf-8 string.
894
if isinstance(unicode_or_utf8_string, str):
895
# TODO: jam 20070209 This is overkill, and probably has an impact on
896
# performance if we are dealing with lots of apis that want a
899
# Make sure it is a valid utf-8 string
900
unicode_or_utf8_string.decode('utf-8')
901
except UnicodeDecodeError:
902
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
903
return unicode_or_utf8_string
904
return unicode_or_utf8_string.encode('utf-8')
907
def safe_revision_id(unicode_or_utf8_string):
908
"""Revision ids should now be utf8, but at one point they were unicode.
910
This is the same as safe_utf8, except it uses the cached encode functions
911
to save a little bit of performance.
913
if unicode_or_utf8_string is None:
915
if isinstance(unicode_or_utf8_string, str):
916
# TODO: jam 20070209 Eventually just remove this check.
918
utf8_str = cache_utf8.get_cached_utf8(unicode_or_utf8_string)
919
except UnicodeDecodeError:
920
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
922
return cache_utf8.encode(unicode_or_utf8_string)
925
# TODO: jam 20070217 We start by just re-using safe_revision_id, but ultimately
926
# we want to use a different dictionary cache, because trapping file ids
927
# and revision ids in the same dict seemed to have a noticable effect on
929
safe_file_id = safe_revision_id
932
_platform_normalizes_filenames = False
933
if sys.platform == 'darwin':
934
_platform_normalizes_filenames = True
937
def normalizes_filenames():
938
"""Return True if this platform normalizes unicode filenames.
940
Mac OSX does, Windows/Linux do not.
942
return _platform_normalizes_filenames
945
def _accessible_normalized_filename(path):
946
"""Get the unicode normalized path, and if you can access the file.
948
On platforms where the system normalizes filenames (Mac OSX),
949
you can access a file by any path which will normalize correctly.
950
On platforms where the system does not normalize filenames
951
(Windows, Linux), you have to access a file by its exact path.
953
Internally, bzr only supports NFC/NFKC normalization, since that is
954
the standard for XML documents.
956
So return the normalized path, and a flag indicating if the file
957
can be accessed by that path.
960
return unicodedata.normalize('NFKC', unicode(path)), True
963
def _inaccessible_normalized_filename(path):
964
__doc__ = _accessible_normalized_filename.__doc__
966
normalized = unicodedata.normalize('NFKC', unicode(path))
967
return normalized, normalized == path
970
if _platform_normalizes_filenames:
971
normalized_filename = _accessible_normalized_filename
973
normalized_filename = _inaccessible_normalized_filename
621
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
976
624
def terminal_width():
977
625
"""Return estimated terminal width."""
978
if sys.platform == 'win32':
979
return win32utils.get_console_size()[0]
627
# TODO: Do something smart on Windows?
629
# TODO: Is there anything that gets a better update when the window
630
# is resized while the program is running? We could use the Python termcap
982
import struct, fcntl, termios
983
s = struct.pack('HHHH', 0, 0, 0, 0)
984
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
985
width = struct.unpack('HHHH', x)[1]
990
width = int(os.environ['COLUMNS'])
633
return int(os.environ['COLUMNS'])
634
except (IndexError, KeyError, ValueError):
999
637
def supports_executable():
1000
638
return sys.platform != "win32"
1003
def supports_posix_readonly():
1004
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1006
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1007
directory controls creation/deletion, etc.
1009
And under win32, readonly means that the directory itself cannot be
1010
deleted. The contents of a readonly directory can be changed, unlike POSIX
1011
where files in readonly directories cannot be added, deleted or renamed.
1013
return sys.platform != "win32"
1016
def set_or_unset_env(env_variable, value):
1017
"""Modify the environment, setting or removing the env_variable.
1019
:param env_variable: The environment variable in question
1020
:param value: The value to set the environment to. If None, then
1021
the variable will be removed.
1022
:return: The original value of the environment variable.
1024
orig_val = os.environ.get(env_variable)
1026
if orig_val is not None:
1027
del os.environ[env_variable]
1029
if isinstance(value, unicode):
1030
value = value.encode(bzrlib.user_encoding)
1031
os.environ[env_variable] = value
1035
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1038
def check_legal_path(path):
1039
"""Check whether the supplied path is legal.
1040
This is only required on Windows, so we don't test on other platforms
1043
if sys.platform != "win32":
1045
if _validWin32PathRE.match(path) is None:
1046
raise errors.IllegalPath(path)
1049
def walkdirs(top, prefix=""):
1050
"""Yield data about all the directories in a tree.
1052
This yields all the data about the contents of a directory at a time.
1053
After each directory has been yielded, if the caller has mutated the list
1054
to exclude some directories, they are then not descended into.
1056
The data yielded is of the form:
1057
((directory-relpath, directory-path-from-top),
1058
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1059
- directory-relpath is the relative path of the directory being returned
1060
with respect to top. prefix is prepended to this.
1061
- directory-path-from-root is the path including top for this directory.
1062
It is suitable for use with os functions.
1063
- relpath is the relative path within the subtree being walked.
1064
- basename is the basename of the path
1065
- kind is the kind of the file now. If unknown then the file is not
1066
present within the tree - but it may be recorded as versioned. See
1068
- lstat is the stat data *if* the file was statted.
1069
- planned, not implemented:
1070
path_from_tree_root is the path from the root of the tree.
1072
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1073
allows one to walk a subtree but get paths that are relative to a tree
1075
:return: an iterator over the dirs.
1077
#TODO there is a bit of a smell where the results of the directory-
1078
# summary in this, and the path from the root, may not agree
1079
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1080
# potentially confusing output. We should make this more robust - but
1081
# not at a speed cost. RBC 20060731
1083
_directory = _directory_kind
1084
_listdir = os.listdir
1085
_kind_from_mode = _formats.get
1086
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1088
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1089
relroot, _, _, _, top = pending.pop()
1091
relprefix = relroot + u'/'
1094
top_slash = top + u'/'
1097
append = dirblock.append
1098
for name in sorted(_listdir(top)):
1099
abspath = top_slash + name
1100
statvalue = _lstat(abspath)
1101
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1102
append((relprefix + name, name, kind, statvalue, abspath))
1103
yield (relroot, top), dirblock
1105
# push the user specified dirs from dirblock
1106
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1109
def _walkdirs_utf8(top, prefix=""):
1110
"""Yield data about all the directories in a tree.
1112
This yields the same information as walkdirs() only each entry is yielded
1113
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1114
are returned as exact byte-strings.
1116
:return: yields a tuple of (dir_info, [file_info])
1117
dir_info is (utf8_relpath, path-from-top)
1118
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1119
if top is an absolute path, path-from-top is also an absolute path.
1120
path-from-top might be unicode or utf8, but it is the correct path to
1121
pass to os functions to affect the file in question. (such as os.lstat)
1123
fs_encoding = sys.getfilesystemencoding()
1124
if (sys.platform == 'win32' or
1125
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1126
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1128
return _walkdirs_fs_utf8(top, prefix=prefix)
1131
def _walkdirs_fs_utf8(top, prefix=""):
1132
"""See _walkdirs_utf8.
1134
This sub-function is called when we know the filesystem is already in utf8
1135
encoding. So we don't need to transcode filenames.
1138
_directory = _directory_kind
1139
_listdir = os.listdir
1140
_kind_from_mode = _formats.get
1142
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1143
# But we don't actually uses 1-3 in pending, so set them to None
1144
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1146
relroot, _, _, _, top = pending.pop()
1148
relprefix = relroot + '/'
1151
top_slash = top + '/'
1154
append = dirblock.append
1155
for name in sorted(_listdir(top)):
1156
abspath = top_slash + name
1157
statvalue = _lstat(abspath)
1158
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1159
append((relprefix + name, name, kind, statvalue, abspath))
1160
yield (relroot, top), dirblock
1162
# push the user specified dirs from dirblock
1163
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1166
def _walkdirs_unicode_to_utf8(top, prefix=""):
1167
"""See _walkdirs_utf8
1169
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1171
This is currently the fallback code path when the filesystem encoding is
1172
not UTF-8. It may be better to implement an alternative so that we can
1173
safely handle paths that are not properly decodable in the current
1176
_utf8_encode = codecs.getencoder('utf8')
1178
_directory = _directory_kind
1179
_listdir = os.listdir
1180
_kind_from_mode = _formats.get
1182
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1184
relroot, _, _, _, top = pending.pop()
1186
relprefix = relroot + '/'
1189
top_slash = top + u'/'
1192
append = dirblock.append
1193
for name in sorted(_listdir(top)):
1194
name_utf8 = _utf8_encode(name)[0]
1195
abspath = top_slash + name
1196
statvalue = _lstat(abspath)
1197
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1198
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1199
yield (relroot, top), dirblock
1201
# push the user specified dirs from dirblock
1202
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1205
def copy_tree(from_path, to_path, handlers={}):
1206
"""Copy all of the entries in from_path into to_path.
1208
:param from_path: The base directory to copy.
1209
:param to_path: The target directory. If it does not exist, it will
1211
:param handlers: A dictionary of functions, which takes a source and
1212
destinations for files, directories, etc.
1213
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1214
'file', 'directory', and 'symlink' should always exist.
1215
If they are missing, they will be replaced with 'os.mkdir()',
1216
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1218
# Now, just copy the existing cached tree to the new location
1219
# We use a cheap trick here.
1220
# Absolute paths are prefixed with the first parameter
1221
# relative paths are prefixed with the second.
1222
# So we can get both the source and target returned
1223
# without any extra work.
1225
def copy_dir(source, dest):
1228
def copy_link(source, dest):
1229
"""Copy the contents of a symlink"""
1230
link_to = os.readlink(source)
1231
os.symlink(link_to, dest)
1233
real_handlers = {'file':shutil.copy2,
1234
'symlink':copy_link,
1235
'directory':copy_dir,
1237
real_handlers.update(handlers)
1239
if not os.path.exists(to_path):
1240
real_handlers['directory'](from_path, to_path)
1242
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1243
for relpath, name, kind, st, abspath in entries:
1244
real_handlers[kind](abspath, relpath)
1247
def path_prefix_key(path):
1248
"""Generate a prefix-order path key for path.
1250
This can be used to sort paths in the same way that walkdirs does.
1252
return (dirname(path) , path)
1255
def compare_paths_prefix_order(path_a, path_b):
1256
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1257
key_a = path_prefix_key(path_a)
1258
key_b = path_prefix_key(path_b)
1259
return cmp(key_a, key_b)
1262
_cached_user_encoding = None
1265
def get_user_encoding(use_cache=True):
1266
"""Find out what the preferred user encoding is.
1268
This is generally the encoding that is used for command line parameters
1269
and file contents. This may be different from the terminal encoding
1270
or the filesystem encoding.
1272
:param use_cache: Enable cache for detected encoding.
1273
(This parameter is turned on by default,
1274
and required only for selftesting)
1276
:return: A string defining the preferred user encoding
1278
global _cached_user_encoding
1279
if _cached_user_encoding is not None and use_cache:
1280
return _cached_user_encoding
1282
if sys.platform == 'darwin':
1283
# work around egregious python 2.4 bug
1284
sys.platform = 'posix'
1288
sys.platform = 'darwin'
1293
user_encoding = locale.getpreferredencoding()
1294
except locale.Error, e:
1295
sys.stderr.write('bzr: warning: %s\n'
1296
' Could not determine what text encoding to use.\n'
1297
' This error usually means your Python interpreter\n'
1298
' doesn\'t support the locale set by $LANG (%s)\n'
1299
" Continuing with ascii encoding.\n"
1300
% (e, os.environ.get('LANG')))
1301
user_encoding = 'ascii'
1303
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1304
# treat that as ASCII, and not support printing unicode characters to the
1306
if user_encoding in (None, 'cp0'):
1307
user_encoding = 'ascii'
1311
codecs.lookup(user_encoding)
1313
sys.stderr.write('bzr: warning:'
1314
' unknown encoding %s.'
1315
' Continuing with ascii encoding.\n'
1318
user_encoding = 'ascii'
1321
_cached_user_encoding = user_encoding
1323
return user_encoding
1326
def recv_all(socket, bytes):
1327
"""Receive an exact number of bytes.
1329
Regular Socket.recv() may return less than the requested number of bytes,
1330
dependning on what's in the OS buffer. MSG_WAITALL is not available
1331
on all platforms, but this should work everywhere. This will return
1332
less than the requested amount if the remote end closes.
1334
This isn't optimized and is intended mostly for use in testing.
1337
while len(b) < bytes:
1338
new = socket.recv(bytes - len(b))
1344
def dereference_path(path):
1345
"""Determine the real path to a file.
1347
All parent elements are dereferenced. But the file itself is not
1349
:param path: The original path. May be absolute or relative.
1350
:return: the real path *to* the file
1352
parent, base = os.path.split(path)
1353
# The pathjoin for '.' is a workaround for Python bug #1213894.
1354
# (initial path components aren't dereferenced)
1355
return pathjoin(realpath(pathjoin('.', parent)), base)
added
removed
bzrlib/osutils.py
