171
248
rename_func(tmp_name, new)
173
# Default is to just use the python builtins
174
abspath = os.path.abspath
175
realpath = os.path.realpath
251
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
252
# choke on a Unicode string containing a relative path if
253
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
255
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
256
def _posix_abspath(path):
257
# jam 20060426 rather than encoding to fsencoding
258
# copy posixpath.abspath, but use os.getcwdu instead
259
if not posixpath.isabs(path):
260
path = posixpath.join(getcwd(), path)
261
return posixpath.normpath(path)
264
def _posix_realpath(path):
265
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
268
def _win32_fixdrive(path):
269
"""Force drive letters to be consistent.
271
win32 is inconsistent whether it returns lower or upper case
272
and even if it was consistent the user might type the other
273
so we force it to uppercase
274
running python.exe under cmd.exe return capital C:\\
275
running win32 python inside a cygwin shell returns lowercase c:\\
277
drive, path = _nt_splitdrive(path)
278
return drive.upper() + path
281
def _win32_abspath(path):
282
# Real _nt_abspath doesn't have a problem with a unicode cwd
283
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
286
def _win98_abspath(path):
287
"""Return the absolute version of a path.
288
Windows 98 safe implementation (python reimplementation
289
of Win32 API function GetFullPathNameW)
294
# \\HOST\path => //HOST/path
295
# //HOST/path => //HOST/path
296
# path => C:/cwd/path
299
# check for absolute path
300
drive = _nt_splitdrive(path)[0]
301
if drive == '' and path[:2] not in('//','\\\\'):
303
# we cannot simply os.path.join cwd and path
304
# because os.path.join('C:','/path') produce '/path'
305
# and this is incorrect
306
if path[:1] in ('/','\\'):
307
cwd = _nt_splitdrive(cwd)[0]
309
path = cwd + '\\' + path
310
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
312
if win32utils.winver == 'Windows 98':
313
_win32_abspath = _win98_abspath
316
def _win32_realpath(path):
317
# Real _nt_realpath doesn't have a problem with a unicode cwd
318
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
321
def _win32_pathjoin(*args):
322
return _nt_join(*args).replace('\\', '/')
325
def _win32_normpath(path):
326
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
330
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
333
def _win32_mkdtemp(*args, **kwargs):
334
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
337
def _win32_rename(old, new):
338
"""We expect to be able to atomically replace 'new' with old.
340
On win32, if new exists, it must be moved out of the way first,
344
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
346
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
347
# If we try to rename a non-existant file onto cwd, we get
348
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
349
# if the old path doesn't exist, sometimes we get EACCES
350
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
356
return unicodedata.normalize('NFKC', os.getcwdu())
359
# Default is to just use the python builtins, but these can be rebound on
360
# particular platforms.
361
abspath = _posix_abspath
362
realpath = _posix_realpath
176
363
pathjoin = os.path.join
177
364
normpath = os.path.normpath
178
365
getcwd = os.getcwdu
179
mkdtemp = tempfile.mkdtemp
180
366
rename = os.rename
181
367
dirname = os.path.dirname
182
368
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)
369
split = os.path.split
370
splitext = os.path.splitext
371
# These were already imported into local scope
372
# mkdtemp = tempfile.mkdtemp
373
# rmtree = shutil.rmtree
375
MIN_ABS_PATHLENGTH = 1
196
378
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)
379
abspath = _win32_abspath
380
realpath = _win32_realpath
381
pathjoin = _win32_pathjoin
382
normpath = _win32_normpath
383
getcwd = _win32_getcwd
384
mkdtemp = _win32_mkdtemp
385
rename = _win32_rename
387
MIN_ABS_PATHLENGTH = 3
389
def _win32_delete_readonly(function, path, excinfo):
390
"""Error handler for shutil.rmtree function [for win32]
391
Helps to remove files and dirs marked as read-only.
393
exception = excinfo[1]
394
if function in (os.remove, os.rmdir) \
395
and isinstance(exception, OSError) \
396
and exception.errno == errno.EACCES:
402
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
403
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
404
return shutil.rmtree(path, ignore_errors, onerror)
405
elif sys.platform == 'darwin':
409
def get_terminal_encoding():
410
"""Find the best encoding for printing to the screen.
412
This attempts to check both sys.stdout and sys.stdin to see
413
what encoding they are in, and if that fails it falls back to
414
bzrlib.user_encoding.
415
The problem is that on Windows, locale.getpreferredencoding()
416
is not the same encoding as that used by the console:
417
http://mail.python.org/pipermail/python-list/2003-May/162357.html
419
On my standard US Windows XP, the preferred encoding is
420
cp1252, but the console is cp437
422
output_encoding = getattr(sys.stdout, 'encoding', None)
423
if not output_encoding:
424
input_encoding = getattr(sys.stdin, 'encoding', None)
425
if not input_encoding:
426
output_encoding = bzrlib.user_encoding
427
mutter('encoding stdout as bzrlib.user_encoding %r', output_encoding)
429
output_encoding = input_encoding
430
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
432
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
433
if output_encoding == 'cp0':
434
# invalid encoding (cp0 means 'no codepage' on Windows)
435
output_encoding = bzrlib.user_encoding
436
mutter('cp0 is invalid encoding.'
437
' encoding stdout as bzrlib.user_encoding %r', output_encoding)
440
codecs.lookup(output_encoding)
442
sys.stderr.write('bzr: warning:'
443
' unknown terminal encoding %s.\n'
444
' Using encoding %s instead.\n'
445
% (output_encoding, bzrlib.user_encoding)
447
output_encoding = bzrlib.user_encoding
449
return output_encoding
221
452
def normalizepath(f):
222
if hasattr(os.path, 'realpath'):
453
if getattr(os.path, 'realpath', None) is not None:
596
920
return unicode_or_utf8_string.decode('utf8')
597
921
except UnicodeDecodeError:
598
raise BzrBadParameter(unicode_or_utf8_string)
922
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
925
def safe_utf8(unicode_or_utf8_string):
926
"""Coerce unicode_or_utf8_string to a utf8 string.
928
If it is a str, it is returned.
929
If it is Unicode, it is encoded into a utf-8 string.
931
if isinstance(unicode_or_utf8_string, str):
932
# TODO: jam 20070209 This is overkill, and probably has an impact on
933
# performance if we are dealing with lots of apis that want a
936
# Make sure it is a valid utf-8 string
937
unicode_or_utf8_string.decode('utf-8')
938
except UnicodeDecodeError:
939
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
940
return unicode_or_utf8_string
941
return unicode_or_utf8_string.encode('utf-8')
944
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
945
' Revision id generators should be creating utf8'
949
def safe_revision_id(unicode_or_utf8_string, warn=True):
950
"""Revision ids should now be utf8, but at one point they were unicode.
952
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
954
:param warn: Functions that are sanitizing user data can set warn=False
955
:return: None or a utf8 revision id.
957
if (unicode_or_utf8_string is None
958
or unicode_or_utf8_string.__class__ == str):
959
return unicode_or_utf8_string
961
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
963
return cache_utf8.encode(unicode_or_utf8_string)
966
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
967
' generators should be creating utf8 file ids.')
970
def safe_file_id(unicode_or_utf8_string, warn=True):
971
"""File ids should now be utf8, but at one point they were unicode.
973
This is the same as safe_utf8, except it uses the cached encode functions
974
to save a little bit of performance.
976
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
978
:param warn: Functions that are sanitizing user data can set warn=False
979
:return: None or a utf8 file id.
981
if (unicode_or_utf8_string is None
982
or unicode_or_utf8_string.__class__ == str):
983
return unicode_or_utf8_string
985
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
987
return cache_utf8.encode(unicode_or_utf8_string)
990
_platform_normalizes_filenames = False
991
if sys.platform == 'darwin':
992
_platform_normalizes_filenames = True
995
def normalizes_filenames():
996
"""Return True if this platform normalizes unicode filenames.
998
Mac OSX does, Windows/Linux do not.
1000
return _platform_normalizes_filenames
1003
def _accessible_normalized_filename(path):
1004
"""Get the unicode normalized path, and if you can access the file.
1006
On platforms where the system normalizes filenames (Mac OSX),
1007
you can access a file by any path which will normalize correctly.
1008
On platforms where the system does not normalize filenames
1009
(Windows, Linux), you have to access a file by its exact path.
1011
Internally, bzr only supports NFC/NFKC normalization, since that is
1012
the standard for XML documents.
1014
So return the normalized path, and a flag indicating if the file
1015
can be accessed by that path.
1018
return unicodedata.normalize('NFKC', unicode(path)), True
1021
def _inaccessible_normalized_filename(path):
1022
__doc__ = _accessible_normalized_filename.__doc__
1024
normalized = unicodedata.normalize('NFKC', unicode(path))
1025
return normalized, normalized == path
1028
if _platform_normalizes_filenames:
1029
normalized_filename = _accessible_normalized_filename
1031
normalized_filename = _inaccessible_normalized_filename
601
1034
def terminal_width():
602
1035
"""Return estimated terminal width."""
604
# TODO: Do something smart on Windows?
606
# TODO: Is there anything that gets a better update when the window
607
# is resized while the program is running? We could use the Python termcap
610
return int(os.environ['COLUMNS'])
611
except (IndexError, KeyError, ValueError):
1036
if sys.platform == 'win32':
1037
return win32utils.get_console_size()[0]
1040
import struct, fcntl, termios
1041
s = struct.pack('HHHH', 0, 0, 0, 0)
1042
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1043
width = struct.unpack('HHHH', x)[1]
1048
width = int(os.environ['COLUMNS'])
1057
def supports_executable():
1058
return sys.platform != "win32"
1061
def supports_posix_readonly():
1062
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1064
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1065
directory controls creation/deletion, etc.
1067
And under win32, readonly means that the directory itself cannot be
1068
deleted. The contents of a readonly directory can be changed, unlike POSIX
1069
where files in readonly directories cannot be added, deleted or renamed.
1071
return sys.platform != "win32"
1074
def set_or_unset_env(env_variable, value):
1075
"""Modify the environment, setting or removing the env_variable.
1077
:param env_variable: The environment variable in question
1078
:param value: The value to set the environment to. If None, then
1079
the variable will be removed.
1080
:return: The original value of the environment variable.
1082
orig_val = os.environ.get(env_variable)
1084
if orig_val is not None:
1085
del os.environ[env_variable]
1087
if isinstance(value, unicode):
1088
value = value.encode(bzrlib.user_encoding)
1089
os.environ[env_variable] = value
1093
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1096
def check_legal_path(path):
1097
"""Check whether the supplied path is legal.
1098
This is only required on Windows, so we don't test on other platforms
1101
if sys.platform != "win32":
1103
if _validWin32PathRE.match(path) is None:
1104
raise errors.IllegalPath(path)
1107
def walkdirs(top, prefix=""):
1108
"""Yield data about all the directories in a tree.
1110
This yields all the data about the contents of a directory at a time.
1111
After each directory has been yielded, if the caller has mutated the list
1112
to exclude some directories, they are then not descended into.
1114
The data yielded is of the form:
1115
((directory-relpath, directory-path-from-top),
1116
[(relpath, basename, kind, lstat, path-from-top), ...]),
1117
- directory-relpath is the relative path of the directory being returned
1118
with respect to top. prefix is prepended to this.
1119
- directory-path-from-root is the path including top for this directory.
1120
It is suitable for use with os functions.
1121
- relpath is the relative path within the subtree being walked.
1122
- basename is the basename of the path
1123
- kind is the kind of the file now. If unknown then the file is not
1124
present within the tree - but it may be recorded as versioned. See
1126
- lstat is the stat data *if* the file was statted.
1127
- planned, not implemented:
1128
path_from_tree_root is the path from the root of the tree.
1130
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1131
allows one to walk a subtree but get paths that are relative to a tree
1133
:return: an iterator over the dirs.
1135
#TODO there is a bit of a smell where the results of the directory-
1136
# summary in this, and the path from the root, may not agree
1137
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1138
# potentially confusing output. We should make this more robust - but
1139
# not at a speed cost. RBC 20060731
1141
_directory = _directory_kind
1142
_listdir = os.listdir
1143
_kind_from_mode = _formats.get
1144
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1146
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1147
relroot, _, _, _, top = pending.pop()
1149
relprefix = relroot + u'/'
1152
top_slash = top + u'/'
1155
append = dirblock.append
1156
for name in sorted(_listdir(top)):
1157
abspath = top_slash + name
1158
statvalue = _lstat(abspath)
1159
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1160
append((relprefix + name, name, kind, statvalue, abspath))
1161
yield (relroot, top), dirblock
1163
# push the user specified dirs from dirblock
1164
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1167
def _walkdirs_utf8(top, prefix=""):
1168
"""Yield data about all the directories in a tree.
1170
This yields the same information as walkdirs() only each entry is yielded
1171
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1172
are returned as exact byte-strings.
1174
:return: yields a tuple of (dir_info, [file_info])
1175
dir_info is (utf8_relpath, path-from-top)
1176
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1177
if top is an absolute path, path-from-top is also an absolute path.
1178
path-from-top might be unicode or utf8, but it is the correct path to
1179
pass to os functions to affect the file in question. (such as os.lstat)
1181
fs_encoding = _fs_enc.upper()
1182
if (sys.platform == 'win32' or
1183
fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968')): # ascii
1184
return _walkdirs_unicode_to_utf8(top, prefix=prefix)
1186
return _walkdirs_fs_utf8(top, prefix=prefix)
1189
def _walkdirs_fs_utf8(top, prefix=""):
1190
"""See _walkdirs_utf8.
1192
This sub-function is called when we know the filesystem is already in utf8
1193
encoding. So we don't need to transcode filenames.
1196
_directory = _directory_kind
1197
_listdir = os.listdir
1198
_kind_from_mode = _formats.get
1200
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1201
# But we don't actually uses 1-3 in pending, so set them to None
1202
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1204
relroot, _, _, _, top = pending.pop()
1206
relprefix = relroot + '/'
1209
top_slash = top + '/'
1212
append = dirblock.append
1213
for name in sorted(_listdir(top)):
1214
abspath = top_slash + name
1215
statvalue = _lstat(abspath)
1216
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1217
append((relprefix + name, name, kind, statvalue, abspath))
1218
yield (relroot, top), dirblock
1220
# push the user specified dirs from dirblock
1221
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1224
def _walkdirs_unicode_to_utf8(top, prefix=""):
1225
"""See _walkdirs_utf8
1227
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1229
This is currently the fallback code path when the filesystem encoding is
1230
not UTF-8. It may be better to implement an alternative so that we can
1231
safely handle paths that are not properly decodable in the current
1234
_utf8_encode = codecs.getencoder('utf8')
1236
_directory = _directory_kind
1237
_listdir = os.listdir
1238
_kind_from_mode = _formats.get
1240
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1242
relroot, _, _, _, top = pending.pop()
1244
relprefix = relroot + '/'
1247
top_slash = top + u'/'
1250
append = dirblock.append
1251
for name in sorted(_listdir(top)):
1252
name_utf8 = _utf8_encode(name)[0]
1253
abspath = top_slash + name
1254
statvalue = _lstat(abspath)
1255
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1256
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1257
yield (relroot, top), dirblock
1259
# push the user specified dirs from dirblock
1260
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1263
def copy_tree(from_path, to_path, handlers={}):
1264
"""Copy all of the entries in from_path into to_path.
1266
:param from_path: The base directory to copy.
1267
:param to_path: The target directory. If it does not exist, it will
1269
:param handlers: A dictionary of functions, which takes a source and
1270
destinations for files, directories, etc.
1271
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1272
'file', 'directory', and 'symlink' should always exist.
1273
If they are missing, they will be replaced with 'os.mkdir()',
1274
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1276
# Now, just copy the existing cached tree to the new location
1277
# We use a cheap trick here.
1278
# Absolute paths are prefixed with the first parameter
1279
# relative paths are prefixed with the second.
1280
# So we can get both the source and target returned
1281
# without any extra work.
1283
def copy_dir(source, dest):
1286
def copy_link(source, dest):
1287
"""Copy the contents of a symlink"""
1288
link_to = os.readlink(source)
1289
os.symlink(link_to, dest)
1291
real_handlers = {'file':shutil.copy2,
1292
'symlink':copy_link,
1293
'directory':copy_dir,
1295
real_handlers.update(handlers)
1297
if not os.path.exists(to_path):
1298
real_handlers['directory'](from_path, to_path)
1300
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1301
for relpath, name, kind, st, abspath in entries:
1302
real_handlers[kind](abspath, relpath)
1305
def path_prefix_key(path):
1306
"""Generate a prefix-order path key for path.
1308
This can be used to sort paths in the same way that walkdirs does.
1310
return (dirname(path) , path)
1313
def compare_paths_prefix_order(path_a, path_b):
1314
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1315
key_a = path_prefix_key(path_a)
1316
key_b = path_prefix_key(path_b)
1317
return cmp(key_a, key_b)
1320
_cached_user_encoding = None
1323
def get_user_encoding(use_cache=True):
1324
"""Find out what the preferred user encoding is.
1326
This is generally the encoding that is used for command line parameters
1327
and file contents. This may be different from the terminal encoding
1328
or the filesystem encoding.
1330
:param use_cache: Enable cache for detected encoding.
1331
(This parameter is turned on by default,
1332
and required only for selftesting)
1334
:return: A string defining the preferred user encoding
1336
global _cached_user_encoding
1337
if _cached_user_encoding is not None and use_cache:
1338
return _cached_user_encoding
1340
if sys.platform == 'darwin':
1341
# work around egregious python 2.4 bug
1342
sys.platform = 'posix'
1346
sys.platform = 'darwin'
1351
user_encoding = locale.getpreferredencoding()
1352
except locale.Error, e:
1353
sys.stderr.write('bzr: warning: %s\n'
1354
' Could not determine what text encoding to use.\n'
1355
' This error usually means your Python interpreter\n'
1356
' doesn\'t support the locale set by $LANG (%s)\n'
1357
" Continuing with ascii encoding.\n"
1358
% (e, os.environ.get('LANG')))
1359
user_encoding = 'ascii'
1361
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1362
# treat that as ASCII, and not support printing unicode characters to the
1364
if user_encoding in (None, 'cp0'):
1365
user_encoding = 'ascii'
1369
codecs.lookup(user_encoding)
1371
sys.stderr.write('bzr: warning:'
1372
' unknown encoding %s.'
1373
' Continuing with ascii encoding.\n'
1376
user_encoding = 'ascii'
1379
_cached_user_encoding = user_encoding
1381
return user_encoding
1384
def recv_all(socket, bytes):
1385
"""Receive an exact number of bytes.
1387
Regular Socket.recv() may return less than the requested number of bytes,
1388
dependning on what's in the OS buffer. MSG_WAITALL is not available
1389
on all platforms, but this should work everywhere. This will return
1390
less than the requested amount if the remote end closes.
1392
This isn't optimized and is intended mostly for use in testing.
1395
while len(b) < bytes:
1396
new = socket.recv(bytes - len(b))
1402
def dereference_path(path):
1403
"""Determine the real path to a file.
1405
All parent elements are dereferenced. But the file itself is not
1407
:param path: The original path. May be absolute or relative.
1408
:return: the real path *to* the file
1410
parent, base = os.path.split(path)
1411
# The pathjoin for '.' is a workaround for Python bug #1213894.
1412
# (initial path components aren't dereferenced)
1413
return pathjoin(realpath(pathjoin('.', parent)), base)
1416
def supports_mapi():
1417
"""Return True if we can use MAPI to launch a mail client."""
1418
return sys.platform == "win32"