171
250
rename_func(tmp_name, new)
173
# Default is to just use the python builtins
174
abspath = os.path.abspath
175
realpath = os.path.realpath
253
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
254
# choke on a Unicode string containing a relative path if
255
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
257
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
258
def _posix_abspath(path):
259
# jam 20060426 rather than encoding to fsencoding
260
# copy posixpath.abspath, but use os.getcwdu instead
261
if not posixpath.isabs(path):
262
path = posixpath.join(getcwd(), path)
263
return posixpath.normpath(path)
266
def _posix_realpath(path):
267
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
270
def _win32_fixdrive(path):
271
"""Force drive letters to be consistent.
273
win32 is inconsistent whether it returns lower or upper case
274
and even if it was consistent the user might type the other
275
so we force it to uppercase
276
running python.exe under cmd.exe return capital C:\\
277
running win32 python inside a cygwin shell returns lowercase c:\\
279
drive, path = _nt_splitdrive(path)
280
return drive.upper() + path
283
def _win32_abspath(path):
284
# Real _nt_abspath doesn't have a problem with a unicode cwd
285
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
288
def _win98_abspath(path):
289
"""Return the absolute version of a path.
290
Windows 98 safe implementation (python reimplementation
291
of Win32 API function GetFullPathNameW)
296
# \\HOST\path => //HOST/path
297
# //HOST/path => //HOST/path
298
# path => C:/cwd/path
301
# check for absolute path
302
drive = _nt_splitdrive(path)[0]
303
if drive == '' and path[:2] not in('//','\\\\'):
305
# we cannot simply os.path.join cwd and path
306
# because os.path.join('C:','/path') produce '/path'
307
# and this is incorrect
308
if path[:1] in ('/','\\'):
309
cwd = _nt_splitdrive(cwd)[0]
311
path = cwd + '\\' + path
312
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
314
if win32utils.winver == 'Windows 98':
315
_win32_abspath = _win98_abspath
318
def _win32_realpath(path):
319
# Real _nt_realpath doesn't have a problem with a unicode cwd
320
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
323
def _win32_pathjoin(*args):
324
return _nt_join(*args).replace('\\', '/')
327
def _win32_normpath(path):
328
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
332
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
335
def _win32_mkdtemp(*args, **kwargs):
336
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
339
def _win32_rename(old, new):
340
"""We expect to be able to atomically replace 'new' with old.
342
On win32, if new exists, it must be moved out of the way first,
346
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
348
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
349
# If we try to rename a non-existant file onto cwd, we get
350
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
351
# if the old path doesn't exist, sometimes we get EACCES
352
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
358
return unicodedata.normalize('NFKC', os.getcwdu())
361
# Default is to just use the python builtins, but these can be rebound on
362
# particular platforms.
363
abspath = _posix_abspath
364
realpath = _posix_realpath
176
365
pathjoin = os.path.join
177
366
normpath = os.path.normpath
178
367
getcwd = os.getcwdu
179
mkdtemp = tempfile.mkdtemp
180
368
rename = os.rename
181
369
dirname = os.path.dirname
182
370
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)
371
split = os.path.split
372
splitext = os.path.splitext
373
# These were already imported into local scope
374
# mkdtemp = tempfile.mkdtemp
375
# rmtree = shutil.rmtree
377
MIN_ABS_PATHLENGTH = 1
196
380
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)
381
abspath = _win32_abspath
382
realpath = _win32_realpath
383
pathjoin = _win32_pathjoin
384
normpath = _win32_normpath
385
getcwd = _win32_getcwd
386
mkdtemp = _win32_mkdtemp
387
rename = _win32_rename
389
MIN_ABS_PATHLENGTH = 3
391
def _win32_delete_readonly(function, path, excinfo):
392
"""Error handler for shutil.rmtree function [for win32]
393
Helps to remove files and dirs marked as read-only.
395
exception = excinfo[1]
396
if function in (os.remove, os.rmdir) \
397
and isinstance(exception, OSError) \
398
and exception.errno == errno.EACCES:
404
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
405
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
406
return shutil.rmtree(path, ignore_errors, onerror)
407
elif sys.platform == 'darwin':
411
def get_terminal_encoding():
412
"""Find the best encoding for printing to the screen.
414
This attempts to check both sys.stdout and sys.stdin to see
415
what encoding they are in, and if that fails it falls back to
416
bzrlib.user_encoding.
417
The problem is that on Windows, locale.getpreferredencoding()
418
is not the same encoding as that used by the console:
419
http://mail.python.org/pipermail/python-list/2003-May/162357.html
421
On my standard US Windows XP, the preferred encoding is
422
cp1252, but the console is cp437
424
output_encoding = getattr(sys.stdout, 'encoding', None)
425
if not output_encoding:
426
input_encoding = getattr(sys.stdin, 'encoding', None)
427
if not input_encoding:
428
output_encoding = bzrlib.user_encoding
429
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
431
output_encoding = input_encoding
432
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
434
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
435
if output_encoding == 'cp0':
436
# invalid encoding (cp0 means 'no codepage' on Windows)
437
output_encoding = bzrlib.user_encoding
438
mutter('cp0 is invalid encoding.'
439
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
442
codecs.lookup(output_encoding)
444
sys.stderr.write('bzr: warning:'
445
' unknown terminal encoding %s.\n'
446
' Using encoding %s instead.\n'
447
% (output_encoding, bzrlib.user_encoding)
449
output_encoding = bzrlib.user_encoding
451
return output_encoding
221
454
def normalizepath(f):
222
if hasattr(os.path, 'realpath'):
455
if getattr(os.path, 'realpath', None) is not None:
636
921
return unicode_or_utf8_string.decode('utf8')
637
922
except UnicodeDecodeError:
638
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
923
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
926
def safe_utf8(unicode_or_utf8_string):
927
"""Coerce unicode_or_utf8_string to a utf8 string.
929
If it is a str, it is returned.
930
If it is Unicode, it is encoded into a utf-8 string.
932
if isinstance(unicode_or_utf8_string, str):
933
# TODO: jam 20070209 This is overkill, and probably has an impact on
934
# performance if we are dealing with lots of apis that want a
937
# Make sure it is a valid utf-8 string
938
unicode_or_utf8_string.decode('utf-8')
939
except UnicodeDecodeError:
940
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
941
return unicode_or_utf8_string
942
return unicode_or_utf8_string.encode('utf-8')
945
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
946
' Revision id generators should be creating utf8'
950
def safe_revision_id(unicode_or_utf8_string, warn=True):
951
"""Revision ids should now be utf8, but at one point they were unicode.
953
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
955
:param warn: Functions that are sanitizing user data can set warn=False
956
:return: None or a utf8 revision id.
958
if (unicode_or_utf8_string is None
959
or unicode_or_utf8_string.__class__ == str):
960
return unicode_or_utf8_string
962
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
964
return cache_utf8.encode(unicode_or_utf8_string)
967
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
968
' generators should be creating utf8 file ids.')
971
def safe_file_id(unicode_or_utf8_string, warn=True):
972
"""File ids should now be utf8, but at one point they were unicode.
974
This is the same as safe_utf8, except it uses the cached encode functions
975
to save a little bit of performance.
977
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
979
:param warn: Functions that are sanitizing user data can set warn=False
980
:return: None or a utf8 file id.
982
if (unicode_or_utf8_string is None
983
or unicode_or_utf8_string.__class__ == str):
984
return unicode_or_utf8_string
986
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
988
return cache_utf8.encode(unicode_or_utf8_string)
991
_platform_normalizes_filenames = False
992
if sys.platform == 'darwin':
993
_platform_normalizes_filenames = True
996
def normalizes_filenames():
997
"""Return True if this platform normalizes unicode filenames.
999
Mac OSX does, Windows/Linux do not.
1001
return _platform_normalizes_filenames
1004
def _accessible_normalized_filename(path):
1005
"""Get the unicode normalized path, and if you can access the file.
1007
On platforms where the system normalizes filenames (Mac OSX),
1008
you can access a file by any path which will normalize correctly.
1009
On platforms where the system does not normalize filenames
1010
(Windows, Linux), you have to access a file by its exact path.
1012
Internally, bzr only supports NFC/NFKC normalization, since that is
1013
the standard for XML documents.
1015
So return the normalized path, and a flag indicating if the file
1016
can be accessed by that path.
1019
return unicodedata.normalize('NFKC', unicode(path)), True
1022
def _inaccessible_normalized_filename(path):
1023
__doc__ = _accessible_normalized_filename.__doc__
1025
normalized = unicodedata.normalize('NFKC', unicode(path))
1026
return normalized, normalized == path
1029
if _platform_normalizes_filenames:
1030
normalized_filename = _accessible_normalized_filename
1032
normalized_filename = _inaccessible_normalized_filename
641
1035
def terminal_width():
642
1036
"""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
1037
if sys.platform == 'win32':
1038
return win32utils.get_console_size()[0]
650
return int(os.environ['COLUMNS'])
651
except (IndexError, KeyError, ValueError):
1041
import struct, fcntl, termios
1042
s = struct.pack('HHHH', 0, 0, 0, 0)
1043
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1044
width = struct.unpack('HHHH', x)[1]
1049
width = int(os.environ['COLUMNS'])
654
1058
def supports_executable():
655
1059
return sys.platform != "win32"
1062
def supports_posix_readonly():
1063
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1065
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1066
directory controls creation/deletion, etc.
1068
And under win32, readonly means that the directory itself cannot be
1069
deleted. The contents of a readonly directory can be changed, unlike POSIX
1070
where files in readonly directories cannot be added, deleted or renamed.
1072
return sys.platform != "win32"
1075
def set_or_unset_env(env_variable, value):
1076
"""Modify the environment, setting or removing the env_variable.
1078
:param env_variable: The environment variable in question
1079
:param value: The value to set the environment to. If None, then
1080
the variable will be removed.
1081
:return: The original value of the environment variable.
1083
orig_val = os.environ.get(env_variable)
1085
if orig_val is not None:
1086
del os.environ[env_variable]
1088
if isinstance(value, unicode):
1089
value = value.encode(bzrlib.user_encoding)
1090
os.environ[env_variable] = value
1094
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1097
def check_legal_path(path):
1098
"""Check whether the supplied path is legal.
1099
This is only required on Windows, so we don't test on other platforms
1102
if sys.platform != "win32":
1104
if _validWin32PathRE.match(path) is None:
1105
raise errors.IllegalPath(path)
1108
def walkdirs(top, prefix=""):
1109
"""Yield data about all the directories in a tree.
1111
This yields all the data about the contents of a directory at a time.
1112
After each directory has been yielded, if the caller has mutated the list
1113
to exclude some directories, they are then not descended into.
1115
The data yielded is of the form:
1116
((directory-relpath, directory-path-from-top),
1117
[(relpath, basename, kind, lstat, path-from-top), ...]),
1118
- directory-relpath is the relative path of the directory being returned
1119
with respect to top. prefix is prepended to this.
1120
- directory-path-from-root is the path including top for this directory.
1121
It is suitable for use with os functions.
1122
- relpath is the relative path within the subtree being walked.
1123
- basename is the basename of the path
1124
- kind is the kind of the file now. If unknown then the file is not
1125
present within the tree - but it may be recorded as versioned. See
1127
- lstat is the stat data *if* the file was statted.
1128
- planned, not implemented:
1129
path_from_tree_root is the path from the root of the tree.
1131
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1132
allows one to walk a subtree but get paths that are relative to a tree
1134
:return: an iterator over the dirs.
1136
#TODO there is a bit of a smell where the results of the directory-
1137
# summary in this, and the path from the root, may not agree
1138
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1139
# potentially confusing output. We should make this more robust - but
1140
# not at a speed cost. RBC 20060731
1142
_directory = _directory_kind
1143
_listdir = os.listdir
1144
_kind_from_mode = _formats.get
1145
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1147
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1148
relroot, _, _, _, top = pending.pop()
1150
relprefix = relroot + u'/'
1153
top_slash = top + u'/'
1156
append = dirblock.append
1157
for name in sorted(_listdir(top)):
1158
abspath = top_slash + name
1159
statvalue = _lstat(abspath)
1160
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1161
append((relprefix + name, name, kind, statvalue, abspath))
1162
yield (relroot, top), dirblock
1164
# push the user specified dirs from dirblock
1165
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1168
def _walkdirs_utf8(top, prefix=""):
1169
"""Yield data about all the directories in a tree.
1171
This yields the same information as walkdirs() only each entry is yielded
1172
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1173
are returned as exact byte-strings.
1175
:return: yields a tuple of (dir_info, [file_info])
1176
dir_info is (utf8_relpath, path-from-top)
1177
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1178
if top is an absolute path, path-from-top is also an absolute path.
1179
path-from-top might be unicode or utf8, but it is the correct path to
1180
pass to os functions to affect the file in question. (such as os.lstat)
1182
fs_encoding = _fs_enc.upper()
1183
if (sys.platform == 'win32' or
1184
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1185
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1187
return _walkdirs_fs_utf8(top, prefix=prefix)
1190
def _walkdirs_fs_utf8(top, prefix=""):
1191
"""See _walkdirs_utf8.
1193
This sub-function is called when we know the filesystem is already in utf8
1194
encoding. So we don't need to transcode filenames.
1197
_directory = _directory_kind
1198
_listdir = os.listdir
1199
_kind_from_mode = _formats.get
1201
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1202
# But we don't actually uses 1-3 in pending, so set them to None
1203
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1205
relroot, _, _, _, top = pending.pop()
1207
relprefix = relroot + '/'
1210
top_slash = top + '/'
1213
append = dirblock.append
1214
for name in sorted(_listdir(top)):
1215
abspath = top_slash + name
1216
statvalue = _lstat(abspath)
1217
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1218
append((relprefix + name, name, kind, statvalue, abspath))
1219
yield (relroot, top), dirblock
1221
# push the user specified dirs from dirblock
1222
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1225
def _walkdirs_unicode_to_utf8(top, prefix=""):
1226
"""See _walkdirs_utf8
1228
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1230
This is currently the fallback code path when the filesystem encoding is
1231
not UTF-8. It may be better to implement an alternative so that we can
1232
safely handle paths that are not properly decodable in the current
1235
_utf8_encode = codecs.getencoder('utf8')
1237
_directory = _directory_kind
1238
_listdir = os.listdir
1239
_kind_from_mode = _formats.get
1241
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1243
relroot, _, _, _, top = pending.pop()
1245
relprefix = relroot + '/'
1248
top_slash = top + u'/'
1251
append = dirblock.append
1252
for name in sorted(_listdir(top)):
1253
name_utf8 = _utf8_encode(name)[0]
1254
abspath = top_slash + name
1255
statvalue = _lstat(abspath)
1256
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1257
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1258
yield (relroot, top), dirblock
1260
# push the user specified dirs from dirblock
1261
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1264
def copy_tree(from_path, to_path, handlers={}):
1265
"""Copy all of the entries in from_path into to_path.
1267
:param from_path: The base directory to copy.
1268
:param to_path: The target directory. If it does not exist, it will
1270
:param handlers: A dictionary of functions, which takes a source and
1271
destinations for files, directories, etc.
1272
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1273
'file', 'directory', and 'symlink' should always exist.
1274
If they are missing, they will be replaced with 'os.mkdir()',
1275
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1277
# Now, just copy the existing cached tree to the new location
1278
# We use a cheap trick here.
1279
# Absolute paths are prefixed with the first parameter
1280
# relative paths are prefixed with the second.
1281
# So we can get both the source and target returned
1282
# without any extra work.
1284
def copy_dir(source, dest):
1287
def copy_link(source, dest):
1288
"""Copy the contents of a symlink"""
1289
link_to = os.readlink(source)
1290
os.symlink(link_to, dest)
1292
real_handlers = {'file':shutil.copy2,
1293
'symlink':copy_link,
1294
'directory':copy_dir,
1296
real_handlers.update(handlers)
1298
if not os.path.exists(to_path):
1299
real_handlers['directory'](from_path, to_path)
1301
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1302
for relpath, name, kind, st, abspath in entries:
1303
real_handlers[kind](abspath, relpath)
1306
def path_prefix_key(path):
1307
"""Generate a prefix-order path key for path.
1309
This can be used to sort paths in the same way that walkdirs does.
1311
return (dirname(path) , path)
1314
def compare_paths_prefix_order(path_a, path_b):
1315
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1316
key_a = path_prefix_key(path_a)
1317
key_b = path_prefix_key(path_b)
1318
return cmp(key_a, key_b)
1321
_cached_user_encoding = None
1324
def get_user_encoding(use_cache=True):
1325
"""Find out what the preferred user encoding is.
1327
This is generally the encoding that is used for command line parameters
1328
and file contents. This may be different from the terminal encoding
1329
or the filesystem encoding.
1331
:param use_cache: Enable cache for detected encoding.
1332
(This parameter is turned on by default,
1333
and required only for selftesting)
1335
:return: A string defining the preferred user encoding
1337
global _cached_user_encoding
1338
if _cached_user_encoding is not None and use_cache:
1339
return _cached_user_encoding
1341
if sys.platform == 'darwin':
1342
# work around egregious python 2.4 bug
1343
sys.platform = 'posix'
1347
sys.platform = 'darwin'
1352
user_encoding = locale.getpreferredencoding()
1353
except locale.Error, e:
1354
sys.stderr.write('bzr: warning: %s\n'
1355
' Could not determine what text encoding to use.\n'
1356
' This error usually means your Python interpreter\n'
1357
' doesn\'t support the locale set by $LANG (%s)\n'
1358
" Continuing with ascii encoding.\n"
1359
% (e, os.environ.get('LANG')))
1360
user_encoding = 'ascii'
1362
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1363
# treat that as ASCII, and not support printing unicode characters to the
1365
if user_encoding in (None, 'cp0'):
1366
user_encoding = 'ascii'
1370
codecs.lookup(user_encoding)
1372
sys.stderr.write('bzr: warning:'
1373
' unknown encoding %s.'
1374
' Continuing with ascii encoding.\n'
1377
user_encoding = 'ascii'
1380
_cached_user_encoding = user_encoding
1382
return user_encoding
1385
def recv_all(socket, bytes):
1386
"""Receive an exact number of bytes.
1388
Regular Socket.recv() may return less than the requested number of bytes,
1389
dependning on what's in the OS buffer. MSG_WAITALL is not available
1390
on all platforms, but this should work everywhere. This will return
1391
less than the requested amount if the remote end closes.
1393
This isn't optimized and is intended mostly for use in testing.
1396
while len(b) < bytes:
1397
new = socket.recv(bytes - len(b))
1403
def dereference_path(path):
1404
"""Determine the real path to a file.
1406
All parent elements are dereferenced. But the file itself is not
1408
:param path: The original path. May be absolute or relative.
1409
:return: the real path *to* the file
1411
parent, base = os.path.split(path)
1412
# The pathjoin for '.' is a workaround for Python bug #1213894.
1413
# (initial path components aren't dereferenced)
1414
return pathjoin(realpath(pathjoin('.', parent)), base)
1417
def supports_mapi():
1418
"""Return True if we can use MAPI to launch a mail client."""
1419
return sys.platform == "win32"