247
171
rename_func(tmp_name, new)
250
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
251
# choke on a Unicode string containing a relative path if
252
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
254
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
255
def _posix_abspath(path):
256
# jam 20060426 rather than encoding to fsencoding
257
# copy posixpath.abspath, but use os.getcwdu instead
258
if not posixpath.isabs(path):
259
path = posixpath.join(getcwd(), path)
260
return posixpath.normpath(path)
263
def _posix_realpath(path):
264
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
267
def _win32_fixdrive(path):
268
"""Force drive letters to be consistent.
270
win32 is inconsistent whether it returns lower or upper case
271
and even if it was consistent the user might type the other
272
so we force it to uppercase
273
running python.exe under cmd.exe return capital C:\\
274
running win32 python inside a cygwin shell returns lowercase c:\\
276
drive, path = _nt_splitdrive(path)
277
return drive.upper() + path
280
def _win32_abspath(path):
281
# Real _nt_abspath doesn't have a problem with a unicode cwd
282
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
285
def _win98_abspath(path):
286
"""Return the absolute version of a path.
287
Windows 98 safe implementation (python reimplementation
288
of Win32 API function GetFullPathNameW)
293
# \\HOST\path => //HOST/path
294
# //HOST/path => //HOST/path
295
# path => C:/cwd/path
298
# check for absolute path
299
drive = _nt_splitdrive(path)[0]
300
if drive == '' and path[:2] not in('//','\\\\'):
302
# we cannot simply os.path.join cwd and path
303
# because os.path.join('C:','/path') produce '/path'
304
# and this is incorrect
305
if path[:1] in ('/','\\'):
306
cwd = _nt_splitdrive(cwd)[0]
308
path = cwd + '\\' + path
309
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
311
if win32utils.winver == 'Windows 98':
312
_win32_abspath = _win98_abspath
315
def _win32_realpath(path):
316
# Real _nt_realpath doesn't have a problem with a unicode cwd
317
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
320
def _win32_pathjoin(*args):
321
return _nt_join(*args).replace('\\', '/')
324
def _win32_normpath(path):
325
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
329
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
332
def _win32_mkdtemp(*args, **kwargs):
333
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
336
def _win32_rename(old, new):
337
"""We expect to be able to atomically replace 'new' with old.
339
On win32, if new exists, it must be moved out of the way first,
343
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
345
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
346
# If we try to rename a non-existant file onto cwd, we get
347
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
348
# if the old path doesn't exist, sometimes we get EACCES
349
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
355
return unicodedata.normalize('NFKC', os.getcwdu())
358
# Default is to just use the python builtins, but these can be rebound on
359
# particular platforms.
360
abspath = _posix_abspath
361
realpath = _posix_realpath
173
# Default is to just use the python builtins
174
abspath = os.path.abspath
175
realpath = os.path.realpath
362
176
pathjoin = os.path.join
363
177
normpath = os.path.normpath
364
178
getcwd = os.getcwdu
179
mkdtemp = tempfile.mkdtemp
365
180
rename = os.rename
366
181
dirname = os.path.dirname
367
182
basename = os.path.basename
368
split = os.path.split
369
splitext = os.path.splitext
370
# These were already imported into local scope
371
# mkdtemp = tempfile.mkdtemp
372
# rmtree = shutil.rmtree
374
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)
377
196
if sys.platform == 'win32':
378
abspath = _win32_abspath
379
realpath = _win32_realpath
380
pathjoin = _win32_pathjoin
381
normpath = _win32_normpath
382
getcwd = _win32_getcwd
383
mkdtemp = _win32_mkdtemp
384
rename = _win32_rename
386
MIN_ABS_PATHLENGTH = 3
388
def _win32_delete_readonly(function, path, excinfo):
389
"""Error handler for shutil.rmtree function [for win32]
390
Helps to remove files and dirs marked as read-only.
392
exception = excinfo[1]
393
if function in (os.remove, os.rmdir) \
394
and isinstance(exception, OSError) \
395
and exception.errno == errno.EACCES:
401
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
402
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
403
return shutil.rmtree(path, ignore_errors, onerror)
404
elif sys.platform == 'darwin':
408
def get_terminal_encoding():
409
"""Find the best encoding for printing to the screen.
411
This attempts to check both sys.stdout and sys.stdin to see
412
what encoding they are in, and if that fails it falls back to
413
bzrlib.user_encoding.
414
The problem is that on Windows, locale.getpreferredencoding()
415
is not the same encoding as that used by the console:
416
http://mail.python.org/pipermail/python-list/2003-May/162357.html
418
On my standard US Windows XP, the preferred encoding is
419
cp1252, but the console is cp437
421
output_encoding = getattr(sys.stdout, 'encoding', None)
422
if not output_encoding:
423
input_encoding = getattr(sys.stdin, 'encoding', None)
424
if not input_encoding:
425
output_encoding = bzrlib.user_encoding
426
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
428
output_encoding = input_encoding
429
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
431
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
432
if output_encoding == 'cp0':
433
# invalid encoding (cp0 means 'no codepage' on Windows)
434
output_encoding = bzrlib.user_encoding
435
mutter('cp0 is invalid encoding.'
436
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
439
codecs.lookup(output_encoding)
441
sys.stderr.write('bzr: warning:'
442
' unknown terminal encoding %s.\n'
443
' Using encoding %s instead.\n'
444
% (output_encoding, bzrlib.user_encoding)
446
output_encoding = bzrlib.user_encoding
448
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)
451
221
def normalizepath(f):
452
if getattr(os.path, 'realpath', None) is not None:
222
if hasattr(os.path, 'realpath'):
918
604
return unicode_or_utf8_string.decode('utf8')
919
605
except UnicodeDecodeError:
920
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
923
def safe_utf8(unicode_or_utf8_string):
924
"""Coerce unicode_or_utf8_string to a utf8 string.
926
If it is a str, it is returned.
927
If it is Unicode, it is encoded into a utf-8 string.
929
if isinstance(unicode_or_utf8_string, str):
930
# TODO: jam 20070209 This is overkill, and probably has an impact on
931
# performance if we are dealing with lots of apis that want a
934
# Make sure it is a valid utf-8 string
935
unicode_or_utf8_string.decode('utf-8')
936
except UnicodeDecodeError:
937
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
938
return unicode_or_utf8_string
939
return unicode_or_utf8_string.encode('utf-8')
942
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
943
' Revision id generators should be creating utf8'
947
def safe_revision_id(unicode_or_utf8_string, warn=True):
948
"""Revision ids should now be utf8, but at one point they were unicode.
950
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
952
:param warn: Functions that are sanitizing user data can set warn=False
953
:return: None or a utf8 revision id.
955
if (unicode_or_utf8_string is None
956
or unicode_or_utf8_string.__class__ == str):
957
return unicode_or_utf8_string
959
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
961
return cache_utf8.encode(unicode_or_utf8_string)
964
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
965
' generators should be creating utf8 file ids.')
968
def safe_file_id(unicode_or_utf8_string, warn=True):
969
"""File ids should now be utf8, but at one point they were unicode.
971
This is the same as safe_utf8, except it uses the cached encode functions
972
to save a little bit of performance.
974
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
976
:param warn: Functions that are sanitizing user data can set warn=False
977
:return: None or a utf8 file id.
979
if (unicode_or_utf8_string is None
980
or unicode_or_utf8_string.__class__ == str):
981
return unicode_or_utf8_string
983
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
985
return cache_utf8.encode(unicode_or_utf8_string)
988
_platform_normalizes_filenames = False
989
if sys.platform == 'darwin':
990
_platform_normalizes_filenames = True
993
def normalizes_filenames():
994
"""Return True if this platform normalizes unicode filenames.
996
Mac OSX does, Windows/Linux do not.
998
return _platform_normalizes_filenames
1001
def _accessible_normalized_filename(path):
1002
"""Get the unicode normalized path, and if you can access the file.
1004
On platforms where the system normalizes filenames (Mac OSX),
1005
you can access a file by any path which will normalize correctly.
1006
On platforms where the system does not normalize filenames
1007
(Windows, Linux), you have to access a file by its exact path.
1009
Internally, bzr only supports NFC/NFKC normalization, since that is
1010
the standard for XML documents.
1012
So return the normalized path, and a flag indicating if the file
1013
can be accessed by that path.
1016
return unicodedata.normalize('NFKC', unicode(path)), True
1019
def _inaccessible_normalized_filename(path):
1020
__doc__ = _accessible_normalized_filename.__doc__
1022
normalized = unicodedata.normalize('NFKC', unicode(path))
1023
return normalized, normalized == path
1026
if _platform_normalizes_filenames:
1027
normalized_filename = _accessible_normalized_filename
1029
normalized_filename = _inaccessible_normalized_filename
606
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
1032
609
def terminal_width():
1033
610
"""Return estimated terminal width."""
1034
if sys.platform == 'win32':
1035
return win32utils.get_console_size()[0]
1038
import struct, fcntl, termios
1039
s = struct.pack('HHHH', 0, 0, 0, 0)
1040
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1041
width = struct.unpack('HHHH', x)[1]
1046
width = int(os.environ['COLUMNS'])
1055
def supports_executable():
1056
return sys.platform != "win32"
1059
def supports_posix_readonly():
1060
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1062
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1063
directory controls creation/deletion, etc.
1065
And under win32, readonly means that the directory itself cannot be
1066
deleted. The contents of a readonly directory can be changed, unlike POSIX
1067
where files in readonly directories cannot be added, deleted or renamed.
1069
return sys.platform != "win32"
1072
def set_or_unset_env(env_variable, value):
1073
"""Modify the environment, setting or removing the env_variable.
1075
:param env_variable: The environment variable in question
1076
:param value: The value to set the environment to. If None, then
1077
the variable will be removed.
1078
:return: The original value of the environment variable.
1080
orig_val = os.environ.get(env_variable)
1082
if orig_val is not None:
1083
del os.environ[env_variable]
1085
if isinstance(value, unicode):
1086
value = value.encode(bzrlib.user_encoding)
1087
os.environ[env_variable] = value
1091
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1094
def check_legal_path(path):
1095
"""Check whether the supplied path is legal.
1096
This is only required on Windows, so we don't test on other platforms
1099
if sys.platform != "win32":
1101
if _validWin32PathRE.match(path) is None:
1102
raise errors.IllegalPath(path)
1105
def walkdirs(top, prefix=""):
1106
"""Yield data about all the directories in a tree.
1108
This yields all the data about the contents of a directory at a time.
1109
After each directory has been yielded, if the caller has mutated the list
1110
to exclude some directories, they are then not descended into.
1112
The data yielded is of the form:
1113
((directory-relpath, directory-path-from-top),
1114
[(relpath, basename, kind, lstat, path-from-top), ...]),
1115
- directory-relpath is the relative path of the directory being returned
1116
with respect to top. prefix is prepended to this.
1117
- directory-path-from-root is the path including top for this directory.
1118
It is suitable for use with os functions.
1119
- relpath is the relative path within the subtree being walked.
1120
- basename is the basename of the path
1121
- kind is the kind of the file now. If unknown then the file is not
1122
present within the tree - but it may be recorded as versioned. See
1124
- lstat is the stat data *if* the file was statted.
1125
- planned, not implemented:
1126
path_from_tree_root is the path from the root of the tree.
1128
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1129
allows one to walk a subtree but get paths that are relative to a tree
1131
:return: an iterator over the dirs.
1133
#TODO there is a bit of a smell where the results of the directory-
1134
# summary in this, and the path from the root, may not agree
1135
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1136
# potentially confusing output. We should make this more robust - but
1137
# not at a speed cost. RBC 20060731
1139
_directory = _directory_kind
1140
_listdir = os.listdir
1141
_kind_from_mode = _formats.get
1142
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1144
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1145
relroot, _, _, _, top = pending.pop()
1147
relprefix = relroot + u'/'
1150
top_slash = top + u'/'
1153
append = dirblock.append
1154
for name in sorted(_listdir(top)):
1155
abspath = top_slash + name
1156
statvalue = _lstat(abspath)
1157
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1158
append((relprefix + name, name, kind, statvalue, abspath))
1159
yield (relroot, top), dirblock
1161
# push the user specified dirs from dirblock
1162
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1165
def _walkdirs_utf8(top, prefix=""):
1166
"""Yield data about all the directories in a tree.
1168
This yields the same information as walkdirs() only each entry is yielded
1169
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1170
are returned as exact byte-strings.
1172
:return: yields a tuple of (dir_info, [file_info])
1173
dir_info is (utf8_relpath, path-from-top)
1174
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1175
if top is an absolute path, path-from-top is also an absolute path.
1176
path-from-top might be unicode or utf8, but it is the correct path to
1177
pass to os functions to affect the file in question. (such as os.lstat)
1179
fs_encoding = _fs_enc.upper()
1180
if (sys.platform == 'win32' or
1181
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1182
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1184
return _walkdirs_fs_utf8(top, prefix=prefix)
1187
def _walkdirs_fs_utf8(top, prefix=""):
1188
"""See _walkdirs_utf8.
1190
This sub-function is called when we know the filesystem is already in utf8
1191
encoding. So we don't need to transcode filenames.
1194
_directory = _directory_kind
1195
_listdir = os.listdir
1196
_kind_from_mode = _formats.get
1198
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1199
# But we don't actually uses 1-3 in pending, so set them to None
1200
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1202
relroot, _, _, _, top = pending.pop()
1204
relprefix = relroot + '/'
1207
top_slash = top + '/'
1210
append = dirblock.append
1211
for name in sorted(_listdir(top)):
1212
abspath = top_slash + name
1213
statvalue = _lstat(abspath)
1214
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1215
append((relprefix + name, name, kind, statvalue, abspath))
1216
yield (relroot, top), dirblock
1218
# push the user specified dirs from dirblock
1219
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1222
def _walkdirs_unicode_to_utf8(top, prefix=""):
1223
"""See _walkdirs_utf8
1225
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1227
This is currently the fallback code path when the filesystem encoding is
1228
not UTF-8. It may be better to implement an alternative so that we can
1229
safely handle paths that are not properly decodable in the current
1232
_utf8_encode = codecs.getencoder('utf8')
1234
_directory = _directory_kind
1235
_listdir = os.listdir
1236
_kind_from_mode = _formats.get
1238
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1240
relroot, _, _, _, top = pending.pop()
1242
relprefix = relroot + '/'
1245
top_slash = top + u'/'
1248
append = dirblock.append
1249
for name in sorted(_listdir(top)):
1250
name_utf8 = _utf8_encode(name)[0]
1251
abspath = top_slash + name
1252
statvalue = _lstat(abspath)
1253
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1254
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1255
yield (relroot, top), dirblock
1257
# push the user specified dirs from dirblock
1258
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1261
def copy_tree(from_path, to_path, handlers={}):
1262
"""Copy all of the entries in from_path into to_path.
1264
:param from_path: The base directory to copy.
1265
:param to_path: The target directory. If it does not exist, it will
1267
:param handlers: A dictionary of functions, which takes a source and
1268
destinations for files, directories, etc.
1269
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1270
'file', 'directory', and 'symlink' should always exist.
1271
If they are missing, they will be replaced with 'os.mkdir()',
1272
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1274
# Now, just copy the existing cached tree to the new location
1275
# We use a cheap trick here.
1276
# Absolute paths are prefixed with the first parameter
1277
# relative paths are prefixed with the second.
1278
# So we can get both the source and target returned
1279
# without any extra work.
1281
def copy_dir(source, dest):
1284
def copy_link(source, dest):
1285
"""Copy the contents of a symlink"""
1286
link_to = os.readlink(source)
1287
os.symlink(link_to, dest)
1289
real_handlers = {'file':shutil.copy2,
1290
'symlink':copy_link,
1291
'directory':copy_dir,
1293
real_handlers.update(handlers)
1295
if not os.path.exists(to_path):
1296
real_handlers['directory'](from_path, to_path)
1298
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1299
for relpath, name, kind, st, abspath in entries:
1300
real_handlers[kind](abspath, relpath)
1303
def path_prefix_key(path):
1304
"""Generate a prefix-order path key for path.
1306
This can be used to sort paths in the same way that walkdirs does.
1308
return (dirname(path) , path)
1311
def compare_paths_prefix_order(path_a, path_b):
1312
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1313
key_a = path_prefix_key(path_a)
1314
key_b = path_prefix_key(path_b)
1315
return cmp(key_a, key_b)
1318
_cached_user_encoding = None
1321
def get_user_encoding(use_cache=True):
1322
"""Find out what the preferred user encoding is.
1324
This is generally the encoding that is used for command line parameters
1325
and file contents. This may be different from the terminal encoding
1326
or the filesystem encoding.
1328
:param use_cache: Enable cache for detected encoding.
1329
(This parameter is turned on by default,
1330
and required only for selftesting)
1332
:return: A string defining the preferred user encoding
1334
global _cached_user_encoding
1335
if _cached_user_encoding is not None and use_cache:
1336
return _cached_user_encoding
1338
if sys.platform == 'darwin':
1339
# work around egregious python 2.4 bug
1340
sys.platform = 'posix'
1344
sys.platform = 'darwin'
1349
user_encoding = locale.getpreferredencoding()
1350
except locale.Error, e:
1351
sys.stderr.write('bzr: warning: %s\n'
1352
' Could not determine what text encoding to use.\n'
1353
' This error usually means your Python interpreter\n'
1354
' doesn\'t support the locale set by $LANG (%s)\n'
1355
" Continuing with ascii encoding.\n"
1356
% (e, os.environ.get('LANG')))
1357
user_encoding = 'ascii'
1359
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1360
# treat that as ASCII, and not support printing unicode characters to the
1362
if user_encoding in (None, 'cp0'):
1363
user_encoding = 'ascii'
1367
codecs.lookup(user_encoding)
1369
sys.stderr.write('bzr: warning:'
1370
' unknown encoding %s.'
1371
' Continuing with ascii encoding.\n'
1374
user_encoding = 'ascii'
1377
_cached_user_encoding = user_encoding
1379
return user_encoding
1382
def recv_all(socket, bytes):
1383
"""Receive an exact number of bytes.
1385
Regular Socket.recv() may return less than the requested number of bytes,
1386
dependning on what's in the OS buffer. MSG_WAITALL is not available
1387
on all platforms, but this should work everywhere. This will return
1388
less than the requested amount if the remote end closes.
1390
This isn't optimized and is intended mostly for use in testing.
1393
while len(b) < bytes:
1394
new = socket.recv(bytes - len(b))
1400
def dereference_path(path):
1401
"""Determine the real path to a file.
1403
All parent elements are dereferenced. But the file itself is not
1405
:param path: The original path. May be absolute or relative.
1406
:return: the real path *to* the file
1408
parent, base = os.path.split(path)
1409
# The pathjoin for '.' is a workaround for Python bug #1213894.
1410
# (initial path components aren't dereferenced)
1411
return pathjoin(realpath(pathjoin('.', parent)), base)
1414
def supports_mapi():
1415
"""Return True if we can use MAPI to launch a mail client."""
1416
return sys.platform == "win32"
612
# TODO: Do something smart on Windows?
614
# TODO: Is there anything that gets a better update when the window
615
# is resized while the program is running? We could use the Python termcap
618
return int(os.environ['COLUMNS'])
619
except (IndexError, KeyError, ValueError):
added
removed
bzrlib/osutils.py
