171
226
rename_func(tmp_name, new)
173
# Default is to just use the python builtins
174
abspath = os.path.abspath
175
realpath = os.path.realpath
229
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
230
# choke on a Unicode string containing a relative path if
231
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
233
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
234
def _posix_abspath(path):
235
# jam 20060426 rather than encoding to fsencoding
236
# copy posixpath.abspath, but use os.getcwdu instead
237
if not posixpath.isabs(path):
238
path = posixpath.join(getcwd(), path)
239
return posixpath.normpath(path)
242
def _posix_realpath(path):
243
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
246
def _win32_fixdrive(path):
247
"""Force drive letters to be consistent.
249
win32 is inconsistent whether it returns lower or upper case
250
and even if it was consistent the user might type the other
251
so we force it to uppercase
252
running python.exe under cmd.exe return capital C:\\
253
running win32 python inside a cygwin shell returns lowercase c:\\
255
drive, path = _nt_splitdrive(path)
256
return drive.upper() + path
259
def _win32_abspath(path):
260
# Real _nt_abspath doesn't have a problem with a unicode cwd
261
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
264
def _win98_abspath(path):
265
"""Return the absolute version of a path.
266
Windows 98 safe implementation (python reimplementation
267
of Win32 API function GetFullPathNameW)
272
# \\HOST\path => //HOST/path
273
# //HOST/path => //HOST/path
274
# path => C:/cwd/path
277
# check for absolute path
278
drive = _nt_splitdrive(path)[0]
279
if drive == '' and path[:2] not in('//','\\\\'):
281
# we cannot simply os.path.join cwd and path
282
# because os.path.join('C:','/path') produce '/path'
283
# and this is incorrect
284
if path[:1] in ('/','\\'):
285
cwd = _nt_splitdrive(cwd)[0]
287
path = cwd + '\\' + path
288
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
290
if win32utils.winver == 'Windows 98':
291
_win32_abspath = _win98_abspath
294
def _win32_realpath(path):
295
# Real _nt_realpath doesn't have a problem with a unicode cwd
296
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
299
def _win32_pathjoin(*args):
300
return _nt_join(*args).replace('\\', '/')
303
def _win32_normpath(path):
304
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
308
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
311
def _win32_mkdtemp(*args, **kwargs):
312
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
315
def _win32_rename(old, new):
316
"""We expect to be able to atomically replace 'new' with old.
318
On win32, if new exists, it must be moved out of the way first,
322
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
324
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
325
# If we try to rename a non-existant file onto cwd, we get
326
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
327
# if the old path doesn't exist, sometimes we get EACCES
328
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
334
return unicodedata.normalize('NFKC', os.getcwdu())
337
# Default is to just use the python builtins, but these can be rebound on
338
# particular platforms.
339
abspath = _posix_abspath
340
realpath = _posix_realpath
176
341
pathjoin = os.path.join
177
342
normpath = os.path.normpath
178
343
getcwd = os.getcwdu
179
mkdtemp = tempfile.mkdtemp
180
344
rename = os.rename
181
345
dirname = os.path.dirname
182
346
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)
347
split = os.path.split
348
splitext = os.path.splitext
349
# These were already imported into local scope
350
# mkdtemp = tempfile.mkdtemp
351
# rmtree = shutil.rmtree
353
MIN_ABS_PATHLENGTH = 1
196
356
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)
357
abspath = _win32_abspath
358
realpath = _win32_realpath
359
pathjoin = _win32_pathjoin
360
normpath = _win32_normpath
361
getcwd = _win32_getcwd
362
mkdtemp = _win32_mkdtemp
363
rename = _win32_rename
365
MIN_ABS_PATHLENGTH = 3
367
def _win32_delete_readonly(function, path, excinfo):
368
"""Error handler for shutil.rmtree function [for win32]
369
Helps to remove files and dirs marked as read-only.
371
exception = excinfo[1]
372
if function in (os.remove, os.rmdir) \
373
and isinstance(exception, OSError) \
374
and exception.errno == errno.EACCES:
380
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
381
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
382
return shutil.rmtree(path, ignore_errors, onerror)
383
elif sys.platform == 'darwin':
387
def get_terminal_encoding():
388
"""Find the best encoding for printing to the screen.
390
This attempts to check both sys.stdout and sys.stdin to see
391
what encoding they are in, and if that fails it falls back to
392
bzrlib.user_encoding.
393
The problem is that on Windows, locale.getpreferredencoding()
394
is not the same encoding as that used by the console:
395
http://mail.python.org/pipermail/python-list/2003-May/162357.html
397
On my standard US Windows XP, the preferred encoding is
398
cp1252, but the console is cp437
400
output_encoding = getattr(sys.stdout, 'encoding', None)
401
if not output_encoding:
402
input_encoding = getattr(sys.stdin, 'encoding', None)
403
if not input_encoding:
404
output_encoding = bzrlib.user_encoding
405
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
407
output_encoding = input_encoding
408
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
410
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
411
if output_encoding == 'cp0':
412
# invalid encoding (cp0 means 'no codepage' on Windows)
413
output_encoding = bzrlib.user_encoding
414
mutter('cp0 is invalid encoding.'
415
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
418
codecs.lookup(output_encoding)
420
sys.stderr.write('bzr: warning:'
421
' unknown terminal encoding %s.\n'
422
' Using encoding %s instead.\n'
423
% (output_encoding, bzrlib.user_encoding)
425
output_encoding = bzrlib.user_encoding
427
return output_encoding
221
430
def normalizepath(f):
222
if hasattr(os.path, 'realpath'):
431
if getattr(os.path, 'realpath', None) is not None:
636
882
return unicode_or_utf8_string.decode('utf8')
637
883
except UnicodeDecodeError:
638
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
884
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
887
def safe_utf8(unicode_or_utf8_string):
888
"""Coerce unicode_or_utf8_string to a utf8 string.
890
If it is a str, it is returned.
891
If it is Unicode, it is encoded into a utf-8 string.
893
if isinstance(unicode_or_utf8_string, str):
894
# TODO: jam 20070209 This is overkill, and probably has an impact on
895
# performance if we are dealing with lots of apis that want a
898
# Make sure it is a valid utf-8 string
899
unicode_or_utf8_string.decode('utf-8')
900
except UnicodeDecodeError:
901
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
902
return unicode_or_utf8_string
903
return unicode_or_utf8_string.encode('utf-8')
906
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
907
' Revision id generators should be creating utf8'
911
def safe_revision_id(unicode_or_utf8_string, warn=True):
912
"""Revision ids should now be utf8, but at one point they were unicode.
914
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
916
:param warn: Functions that are sanitizing user data can set warn=False
917
:return: None or a utf8 revision id.
919
if (unicode_or_utf8_string is None
920
or unicode_or_utf8_string.__class__ == str):
921
return unicode_or_utf8_string
923
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
925
return cache_utf8.encode(unicode_or_utf8_string)
928
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
929
' generators should be creating utf8 file ids.')
932
def safe_file_id(unicode_or_utf8_string, warn=True):
933
"""File ids should now be utf8, but at one point they were unicode.
935
This is the same as safe_utf8, except it uses the cached encode functions
936
to save a little bit of performance.
938
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
940
:param warn: Functions that are sanitizing user data can set warn=False
941
:return: None or a utf8 file id.
943
if (unicode_or_utf8_string is None
944
or unicode_or_utf8_string.__class__ == str):
945
return unicode_or_utf8_string
947
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
949
return cache_utf8.encode(unicode_or_utf8_string)
952
_platform_normalizes_filenames = False
953
if sys.platform == 'darwin':
954
_platform_normalizes_filenames = True
957
def normalizes_filenames():
958
"""Return True if this platform normalizes unicode filenames.
960
Mac OSX does, Windows/Linux do not.
962
return _platform_normalizes_filenames
965
def _accessible_normalized_filename(path):
966
"""Get the unicode normalized path, and if you can access the file.
968
On platforms where the system normalizes filenames (Mac OSX),
969
you can access a file by any path which will normalize correctly.
970
On platforms where the system does not normalize filenames
971
(Windows, Linux), you have to access a file by its exact path.
973
Internally, bzr only supports NFC/NFKC normalization, since that is
974
the standard for XML documents.
976
So return the normalized path, and a flag indicating if the file
977
can be accessed by that path.
980
return unicodedata.normalize('NFKC', unicode(path)), True
983
def _inaccessible_normalized_filename(path):
984
__doc__ = _accessible_normalized_filename.__doc__
986
normalized = unicodedata.normalize('NFKC', unicode(path))
987
return normalized, normalized == path
990
if _platform_normalizes_filenames:
991
normalized_filename = _accessible_normalized_filename
993
normalized_filename = _inaccessible_normalized_filename
641
996
def terminal_width():
642
997
"""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
998
if sys.platform == 'win32':
999
return win32utils.get_console_size()[0]
650
return int(os.environ['COLUMNS'])
651
except (IndexError, KeyError, ValueError):
1002
import struct, fcntl, termios
1003
s = struct.pack('HHHH', 0, 0, 0, 0)
1004
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1005
width = struct.unpack('HHHH', x)[1]
1010
width = int(os.environ['COLUMNS'])
654
1019
def supports_executable():
655
1020
return sys.platform != "win32"
1023
def supports_posix_readonly():
1024
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1026
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1027
directory controls creation/deletion, etc.
1029
And under win32, readonly means that the directory itself cannot be
1030
deleted. The contents of a readonly directory can be changed, unlike POSIX
1031
where files in readonly directories cannot be added, deleted or renamed.
1033
return sys.platform != "win32"
1036
def set_or_unset_env(env_variable, value):
1037
"""Modify the environment, setting or removing the env_variable.
1039
:param env_variable: The environment variable in question
1040
:param value: The value to set the environment to. If None, then
1041
the variable will be removed.
1042
:return: The original value of the environment variable.
1044
orig_val = os.environ.get(env_variable)
1046
if orig_val is not None:
1047
del os.environ[env_variable]
1049
if isinstance(value, unicode):
1050
value = value.encode(bzrlib.user_encoding)
1051
os.environ[env_variable] = value
1055
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1058
def check_legal_path(path):
1059
"""Check whether the supplied path is legal.
1060
This is only required on Windows, so we don't test on other platforms
1063
if sys.platform != "win32":
1065
if _validWin32PathRE.match(path) is None:
1066
raise errors.IllegalPath(path)
1069
def walkdirs(top, prefix=""):
1070
"""Yield data about all the directories in a tree.
1072
This yields all the data about the contents of a directory at a time.
1073
After each directory has been yielded, if the caller has mutated the list
1074
to exclude some directories, they are then not descended into.
1076
The data yielded is of the form:
1077
((directory-relpath, directory-path-from-top),
1078
[(directory-relpath, basename, kind, lstat, path-from-top), ...]),
1079
- directory-relpath is the relative path of the directory being returned
1080
with respect to top. prefix is prepended to this.
1081
- directory-path-from-root is the path including top for this directory.
1082
It is suitable for use with os functions.
1083
- relpath is the relative path within the subtree being walked.
1084
- basename is the basename of the path
1085
- kind is the kind of the file now. If unknown then the file is not
1086
present within the tree - but it may be recorded as versioned. See
1088
- lstat is the stat data *if* the file was statted.
1089
- planned, not implemented:
1090
path_from_tree_root is the path from the root of the tree.
1092
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1093
allows one to walk a subtree but get paths that are relative to a tree
1095
:return: an iterator over the dirs.
1097
#TODO there is a bit of a smell where the results of the directory-
1098
# summary in this, and the path from the root, may not agree
1099
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1100
# potentially confusing output. We should make this more robust - but
1101
# not at a speed cost. RBC 20060731
1103
_directory = _directory_kind
1104
_listdir = os.listdir
1105
_kind_from_mode = _formats.get
1106
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1108
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1109
relroot, _, _, _, top = pending.pop()
1111
relprefix = relroot + u'/'
1114
top_slash = top + u'/'
1117
append = dirblock.append
1118
for name in sorted(_listdir(top)):
1119
abspath = top_slash + name
1120
statvalue = _lstat(abspath)
1121
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1122
append((relprefix + name, name, kind, statvalue, abspath))
1123
yield (relroot, top), dirblock
1125
# push the user specified dirs from dirblock
1126
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1129
def _walkdirs_utf8(top, prefix=""):
1130
"""Yield data about all the directories in a tree.
1132
This yields the same information as walkdirs() only each entry is yielded
1133
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1134
are returned as exact byte-strings.
1136
:return: yields a tuple of (dir_info, [file_info])
1137
dir_info is (utf8_relpath, path-from-top)
1138
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1139
if top is an absolute path, path-from-top is also an absolute path.
1140
path-from-top might be unicode or utf8, but it is the correct path to
1141
pass to os functions to affect the file in question. (such as os.lstat)
1143
fs_encoding = sys.getfilesystemencoding()
1144
if (sys.platform == 'win32' or
1145
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1146
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1148
return _walkdirs_fs_utf8(top, prefix=prefix)
1151
def _walkdirs_fs_utf8(top, prefix=""):
1152
"""See _walkdirs_utf8.
1154
This sub-function is called when we know the filesystem is already in utf8
1155
encoding. So we don't need to transcode filenames.
1158
_directory = _directory_kind
1159
_listdir = os.listdir
1160
_kind_from_mode = _formats.get
1162
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1163
# But we don't actually uses 1-3 in pending, so set them to None
1164
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1166
relroot, _, _, _, top = pending.pop()
1168
relprefix = relroot + '/'
1171
top_slash = top + '/'
1174
append = dirblock.append
1175
for name in sorted(_listdir(top)):
1176
abspath = top_slash + name
1177
statvalue = _lstat(abspath)
1178
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1179
append((relprefix + name, name, kind, statvalue, abspath))
1180
yield (relroot, top), dirblock
1182
# push the user specified dirs from dirblock
1183
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1186
def _walkdirs_unicode_to_utf8(top, prefix=""):
1187
"""See _walkdirs_utf8
1189
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1191
This is currently the fallback code path when the filesystem encoding is
1192
not UTF-8. It may be better to implement an alternative so that we can
1193
safely handle paths that are not properly decodable in the current
1196
_utf8_encode = codecs.getencoder('utf8')
1198
_directory = _directory_kind
1199
_listdir = os.listdir
1200
_kind_from_mode = _formats.get
1202
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1204
relroot, _, _, _, top = pending.pop()
1206
relprefix = relroot + '/'
1209
top_slash = top + u'/'
1212
append = dirblock.append
1213
for name in sorted(_listdir(top)):
1214
name_utf8 = _utf8_encode(name)[0]
1215
abspath = top_slash + name
1216
statvalue = _lstat(abspath)
1217
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1218
append((relprefix + name_utf8, name_utf8, 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 copy_tree(from_path, to_path, handlers={}):
1226
"""Copy all of the entries in from_path into to_path.
1228
:param from_path: The base directory to copy.
1229
:param to_path: The target directory. If it does not exist, it will
1231
:param handlers: A dictionary of functions, which takes a source and
1232
destinations for files, directories, etc.
1233
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1234
'file', 'directory', and 'symlink' should always exist.
1235
If they are missing, they will be replaced with 'os.mkdir()',
1236
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1238
# Now, just copy the existing cached tree to the new location
1239
# We use a cheap trick here.
1240
# Absolute paths are prefixed with the first parameter
1241
# relative paths are prefixed with the second.
1242
# So we can get both the source and target returned
1243
# without any extra work.
1245
def copy_dir(source, dest):
1248
def copy_link(source, dest):
1249
"""Copy the contents of a symlink"""
1250
link_to = os.readlink(source)
1251
os.symlink(link_to, dest)
1253
real_handlers = {'file':shutil.copy2,
1254
'symlink':copy_link,
1255
'directory':copy_dir,
1257
real_handlers.update(handlers)
1259
if not os.path.exists(to_path):
1260
real_handlers['directory'](from_path, to_path)
1262
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1263
for relpath, name, kind, st, abspath in entries:
1264
real_handlers[kind](abspath, relpath)
1267
def path_prefix_key(path):
1268
"""Generate a prefix-order path key for path.
1270
This can be used to sort paths in the same way that walkdirs does.
1272
return (dirname(path) , path)
1275
def compare_paths_prefix_order(path_a, path_b):
1276
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1277
key_a = path_prefix_key(path_a)
1278
key_b = path_prefix_key(path_b)
1279
return cmp(key_a, key_b)
1282
_cached_user_encoding = None
1285
def get_user_encoding(use_cache=True):
1286
"""Find out what the preferred user encoding is.
1288
This is generally the encoding that is used for command line parameters
1289
and file contents. This may be different from the terminal encoding
1290
or the filesystem encoding.
1292
:param use_cache: Enable cache for detected encoding.
1293
(This parameter is turned on by default,
1294
and required only for selftesting)
1296
:return: A string defining the preferred user encoding
1298
global _cached_user_encoding
1299
if _cached_user_encoding is not None and use_cache:
1300
return _cached_user_encoding
1302
if sys.platform == 'darwin':
1303
# work around egregious python 2.4 bug
1304
sys.platform = 'posix'
1308
sys.platform = 'darwin'
1313
user_encoding = locale.getpreferredencoding()
1314
except locale.Error, e:
1315
sys.stderr.write('bzr: warning: %s\n'
1316
' Could not determine what text encoding to use.\n'
1317
' This error usually means your Python interpreter\n'
1318
' doesn\'t support the locale set by $LANG (%s)\n'
1319
" Continuing with ascii encoding.\n"
1320
% (e, os.environ.get('LANG')))
1321
user_encoding = 'ascii'
1323
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1324
# treat that as ASCII, and not support printing unicode characters to the
1326
if user_encoding in (None, 'cp0'):
1327
user_encoding = 'ascii'
1331
codecs.lookup(user_encoding)
1333
sys.stderr.write('bzr: warning:'
1334
' unknown encoding %s.'
1335
' Continuing with ascii encoding.\n'
1338
user_encoding = 'ascii'
1341
_cached_user_encoding = user_encoding
1343
return user_encoding
1346
def recv_all(socket, bytes):
1347
"""Receive an exact number of bytes.
1349
Regular Socket.recv() may return less than the requested number of bytes,
1350
dependning on what's in the OS buffer. MSG_WAITALL is not available
1351
on all platforms, but this should work everywhere. This will return
1352
less than the requested amount if the remote end closes.
1354
This isn't optimized and is intended mostly for use in testing.
1357
while len(b) < bytes:
1358
new = socket.recv(bytes - len(b))
1364
def dereference_path(path):
1365
"""Determine the real path to a file.
1367
All parent elements are dereferenced. But the file itself is not
1369
:param path: The original path. May be absolute or relative.
1370
:return: the real path *to* the file
1372
parent, base = os.path.split(path)
1373
# The pathjoin for '.' is a workaround for Python bug #1213894.
1374
# (initial path components aren't dereferenced)
1375
return pathjoin(realpath(pathjoin('.', parent)), base)