58
136
# TODO: I'm not really sure this is the best format either.x
138
if _QUOTE_RE is None:
61
139
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
63
141
if _QUOTE_RE.search(f):
64
142
return '"' + f + '"'
70
mode = os.lstat(f)[ST_MODE]
147
_directory_kind = 'directory'
150
"""Return the current umask"""
151
# Assume that people aren't messing with the umask while running
152
# XXX: This is not thread safe, but there is no way to get the
153
# umask without setting it
161
_directory_kind: "/",
163
'tree-reference': '+',
89
167
def kind_marker(kind):
92
elif kind == 'directory':
94
elif kind == 'symlink':
97
raise BzrError('invalid file kind %r' % kind)
101
if hasattr(os, 'lstat'):
107
if e.errno == errno.ENOENT:
110
raise BzrError("lstat/stat of (%r): %r" % (f, e))
169
return _kind_marker_map[kind]
171
raise errors.BzrError('invalid file kind %r' % kind)
174
lexists = getattr(os.path, 'lexists', None)
178
stat = getattr(os, 'lstat', os.stat)
182
if e.errno == errno.ENOENT:
185
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
188
def fancy_rename(old, new, rename_func, unlink_func):
189
"""A fancy rename, when you don't have atomic rename.
191
:param old: The old path, to rename from
192
:param new: The new path, to rename to
193
:param rename_func: The potentially non-atomic rename function
194
:param unlink_func: A way to delete the target file if the full rename succeeds
197
# sftp rename doesn't allow overwriting, so play tricks:
198
base = os.path.basename(new)
199
dirname = os.path.dirname(new)
200
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
201
tmp_name = pathjoin(dirname, tmp_name)
203
# Rename the file out of the way, but keep track if it didn't exist
204
# We don't want to grab just any exception
205
# something like EACCES should prevent us from continuing
206
# The downside is that the rename_func has to throw an exception
207
# with an errno = ENOENT, or NoSuchFile
210
rename_func(new, tmp_name)
211
except (errors.NoSuchFile,), e:
214
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
215
# function raises an IOError with errno is None when a rename fails.
216
# This then gets caught here.
217
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
220
if (getattr(e, 'errno', None) is None
221
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
229
# This may throw an exception, in which case success will
231
rename_func(old, new)
233
except (IOError, OSError), e:
234
# source and target may be aliases of each other (e.g. on a
235
# case-insensitive filesystem), so we may have accidentally renamed
236
# source by when we tried to rename target
237
if not (file_existed and e.errno in (None, errno.ENOENT)):
241
# If the file used to exist, rename it back into place
242
# otherwise just delete it from the tmp location
244
unlink_func(tmp_name)
246
rename_func(tmp_name, new)
249
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
250
# choke on a Unicode string containing a relative path if
251
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
253
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
254
def _posix_abspath(path):
255
# jam 20060426 rather than encoding to fsencoding
256
# copy posixpath.abspath, but use os.getcwdu instead
257
if not posixpath.isabs(path):
258
path = posixpath.join(getcwd(), path)
259
return posixpath.normpath(path)
262
def _posix_realpath(path):
263
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
266
def _win32_fixdrive(path):
267
"""Force drive letters to be consistent.
269
win32 is inconsistent whether it returns lower or upper case
270
and even if it was consistent the user might type the other
271
so we force it to uppercase
272
running python.exe under cmd.exe return capital C:\\
273
running win32 python inside a cygwin shell returns lowercase c:\\
275
drive, path = _nt_splitdrive(path)
276
return drive.upper() + path
279
def _win32_abspath(path):
280
# Real _nt_abspath doesn't have a problem with a unicode cwd
281
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
284
def _win98_abspath(path):
285
"""Return the absolute version of a path.
286
Windows 98 safe implementation (python reimplementation
287
of Win32 API function GetFullPathNameW)
292
# \\HOST\path => //HOST/path
293
# //HOST/path => //HOST/path
294
# path => C:/cwd/path
297
# check for absolute path
298
drive = _nt_splitdrive(path)[0]
299
if drive == '' and path[:2] not in('//','\\\\'):
301
# we cannot simply os.path.join cwd and path
302
# because os.path.join('C:','/path') produce '/path'
303
# and this is incorrect
304
if path[:1] in ('/','\\'):
305
cwd = _nt_splitdrive(cwd)[0]
307
path = cwd + '\\' + path
308
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
311
def _win32_realpath(path):
312
# Real _nt_realpath doesn't have a problem with a unicode cwd
313
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
316
def _win32_pathjoin(*args):
317
return _nt_join(*args).replace('\\', '/')
320
def _win32_normpath(path):
321
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
325
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
328
def _win32_mkdtemp(*args, **kwargs):
329
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
332
def _win32_rename(old, new):
333
"""We expect to be able to atomically replace 'new' with old.
335
On win32, if new exists, it must be moved out of the way first,
339
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
341
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
342
# If we try to rename a non-existant file onto cwd, we get
343
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
344
# if the old path doesn't exist, sometimes we get EACCES
345
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
351
return unicodedata.normalize('NFC', os.getcwdu())
354
# Default is to just use the python builtins, but these can be rebound on
355
# particular platforms.
356
abspath = _posix_abspath
357
realpath = _posix_realpath
358
pathjoin = os.path.join
359
normpath = os.path.normpath
362
dirname = os.path.dirname
363
basename = os.path.basename
364
split = os.path.split
365
splitext = os.path.splitext
366
# These were already imported into local scope
367
# mkdtemp = tempfile.mkdtemp
368
# rmtree = shutil.rmtree
370
MIN_ABS_PATHLENGTH = 1
373
if sys.platform == 'win32':
374
if win32utils.winver == 'Windows 98':
375
abspath = _win98_abspath
377
abspath = _win32_abspath
378
realpath = _win32_realpath
379
pathjoin = _win32_pathjoin
380
normpath = _win32_normpath
381
getcwd = _win32_getcwd
382
mkdtemp = _win32_mkdtemp
383
rename = _win32_rename
385
MIN_ABS_PATHLENGTH = 3
387
def _win32_delete_readonly(function, path, excinfo):
388
"""Error handler for shutil.rmtree function [for win32]
389
Helps to remove files and dirs marked as read-only.
391
exception = excinfo[1]
392
if function in (os.remove, os.rmdir) \
393
and isinstance(exception, OSError) \
394
and exception.errno == errno.EACCES:
400
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
401
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
402
return shutil.rmtree(path, ignore_errors, onerror)
404
f = win32utils.get_unicode_argv # special function or None
408
elif sys.platform == 'darwin':
412
def get_terminal_encoding():
413
"""Find the best encoding for printing to the screen.
415
This attempts to check both sys.stdout and sys.stdin to see
416
what encoding they are in, and if that fails it falls back to
417
osutils.get_user_encoding().
418
The problem is that on Windows, locale.getpreferredencoding()
419
is not the same encoding as that used by the console:
420
http://mail.python.org/pipermail/python-list/2003-May/162357.html
422
On my standard US Windows XP, the preferred encoding is
423
cp1252, but the console is cp437
425
from bzrlib.trace import mutter
426
output_encoding = getattr(sys.stdout, 'encoding', None)
427
if not output_encoding:
428
input_encoding = getattr(sys.stdin, 'encoding', None)
429
if not input_encoding:
430
output_encoding = get_user_encoding()
431
mutter('encoding stdout as osutils.get_user_encoding() %r',
434
output_encoding = input_encoding
435
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
437
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
438
if output_encoding == 'cp0':
439
# invalid encoding (cp0 means 'no codepage' on Windows)
440
output_encoding = get_user_encoding()
441
mutter('cp0 is invalid encoding.'
442
' encoding stdout as osutils.get_user_encoding() %r',
446
codecs.lookup(output_encoding)
448
sys.stderr.write('bzr: warning:'
449
' unknown terminal encoding %s.\n'
450
' Using encoding %s instead.\n'
451
% (output_encoding, get_user_encoding())
453
output_encoding = get_user_encoding()
455
return output_encoding
112
458
def normalizepath(f):
113
if hasattr(os.path, 'realpath'):
459
if getattr(os.path, 'realpath', None) is not None:
117
463
[p,e] = os.path.split(f)
118
464
if e == "" or e == "." or e == "..":
121
return os.path.join(F(p), e)
123
if os.name == "posix":
124
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
125
# choke on a Unicode string containing a relative path if
126
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
128
_fs_enc = sys.getfilesystemencoding()
130
return os.path.abspath(path.encode(_fs_enc)).decode(_fs_enc)
132
return os.path.realpath(path.encode(_fs_enc)).decode(_fs_enc)
134
# We need to use the Unicode-aware os.path.abspath and
135
# os.path.realpath on Windows systems.
136
abspath = os.path.abspath
137
realpath = os.path.realpath
140
"""Copy a file to a backup.
142
Backups are named in GNU-style, with a ~ suffix.
144
If the file is already a backup, it's not copied.
150
if has_symlinks() and os.path.islink(fn):
151
target = os.readlink(fn)
152
os.symlink(target, bfn)
160
outf = file(bfn, 'wb')
467
return pathjoin(F(p), e)
468
1029
os.path.commonprefix (python2.4) has a bad bug that it works just
469
1030
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
470
avoids that problem."""
1031
avoids that problem.
1034
if len(base) < MIN_ABS_PATHLENGTH:
1035
# must have space for e.g. a drive letter
1036
raise ValueError('%r is too short to calculate a relative path'
471
1039
rp = abspath(path)
475
while len(head) >= len(base):
1044
if len(head) <= len(base) and head != base:
1045
raise errors.PathNotChild(rp, base)
476
1046
if head == base:
478
head, tail = os.path.split(head)
1048
head, tail = split(head)
482
# XXX This should raise a NotChildPath exception, as its not tied
484
raise NotBranchError("path %r is not within branch %r" % (rp, base))
486
return os.sep.join(s)
1053
return pathjoin(*reversed(s))
1058
def _cicp_canonical_relpath(base, path):
1059
"""Return the canonical path relative to base.
1061
Like relpath, but on case-insensitive-case-preserving file-systems, this
1062
will return the relpath as stored on the file-system rather than in the
1063
case specified in the input string, for all existing portions of the path.
1065
This will cause O(N) behaviour if called for every path in a tree; if you
1066
have a number of paths to convert, you should use canonical_relpaths().
1068
# TODO: it should be possible to optimize this for Windows by using the
1069
# win32 API FindFiles function to look for the specified name - but using
1070
# os.listdir() still gives us the correct, platform agnostic semantics in
1073
rel = relpath(base, path)
1074
# '.' will have been turned into ''
1078
abs_base = abspath(base)
1080
_listdir = os.listdir
1082
# use an explicit iterator so we can easily consume the rest on early exit.
1083
bit_iter = iter(rel.split('/'))
1084
for bit in bit_iter:
1087
next_entries = _listdir(current)
1088
except OSError: # enoent, eperm, etc
1089
# We can't find this in the filesystem, so just append the
1091
current = pathjoin(current, bit, *list(bit_iter))
1093
for look in next_entries:
1094
if lbit == look.lower():
1095
current = pathjoin(current, look)
1098
# got to the end, nothing matched, so we just return the
1099
# non-existing bits as they were specified (the filename may be
1100
# the target of a move, for example).
1101
current = pathjoin(current, bit, *list(bit_iter))
1103
return current[len(abs_base):].lstrip('/')
1105
# XXX - TODO - we need better detection/integration of case-insensitive
1106
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1107
# filesystems), for example, so could probably benefit from the same basic
1108
# support there. For now though, only Windows and OSX get that support, and
1109
# they get it for *all* file-systems!
1110
if sys.platform in ('win32', 'darwin'):
1111
canonical_relpath = _cicp_canonical_relpath
1113
canonical_relpath = relpath
1115
def canonical_relpaths(base, paths):
1116
"""Create an iterable to canonicalize a sequence of relative paths.
1118
The intent is for this implementation to use a cache, vastly speeding
1119
up multiple transformations in the same directory.
1121
# but for now, we haven't optimized...
1122
return [canonical_relpath(base, p) for p in paths]
1124
def safe_unicode(unicode_or_utf8_string):
1125
"""Coerce unicode_or_utf8_string into unicode.
1127
If it is unicode, it is returned.
1128
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1129
wrapped in a BzrBadParameterNotUnicode exception.
1131
if isinstance(unicode_or_utf8_string, unicode):
1132
return unicode_or_utf8_string
1134
return unicode_or_utf8_string.decode('utf8')
1135
except UnicodeDecodeError:
1136
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1139
def safe_utf8(unicode_or_utf8_string):
1140
"""Coerce unicode_or_utf8_string to a utf8 string.
1142
If it is a str, it is returned.
1143
If it is Unicode, it is encoded into a utf-8 string.
1145
if isinstance(unicode_or_utf8_string, str):
1146
# TODO: jam 20070209 This is overkill, and probably has an impact on
1147
# performance if we are dealing with lots of apis that want a
1150
# Make sure it is a valid utf-8 string
1151
unicode_or_utf8_string.decode('utf-8')
1152
except UnicodeDecodeError:
1153
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1154
return unicode_or_utf8_string
1155
return unicode_or_utf8_string.encode('utf-8')
1158
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1159
' Revision id generators should be creating utf8'
1163
def safe_revision_id(unicode_or_utf8_string, warn=True):
1164
"""Revision ids should now be utf8, but at one point they were unicode.
1166
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1168
:param warn: Functions that are sanitizing user data can set warn=False
1169
:return: None or a utf8 revision id.
1171
if (unicode_or_utf8_string is None
1172
or unicode_or_utf8_string.__class__ == str):
1173
return unicode_or_utf8_string
1175
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1177
return cache_utf8.encode(unicode_or_utf8_string)
1180
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1181
' generators should be creating utf8 file ids.')
1184
def safe_file_id(unicode_or_utf8_string, warn=True):
1185
"""File ids should now be utf8, but at one point they were unicode.
1187
This is the same as safe_utf8, except it uses the cached encode functions
1188
to save a little bit of performance.
1190
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1192
:param warn: Functions that are sanitizing user data can set warn=False
1193
:return: None or a utf8 file id.
1195
if (unicode_or_utf8_string is None
1196
or unicode_or_utf8_string.__class__ == str):
1197
return unicode_or_utf8_string
1199
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1201
return cache_utf8.encode(unicode_or_utf8_string)
1204
_platform_normalizes_filenames = False
1205
if sys.platform == 'darwin':
1206
_platform_normalizes_filenames = True
1209
def normalizes_filenames():
1210
"""Return True if this platform normalizes unicode filenames.
1212
Mac OSX does, Windows/Linux do not.
1214
return _platform_normalizes_filenames
1217
def _accessible_normalized_filename(path):
1218
"""Get the unicode normalized path, and if you can access the file.
1220
On platforms where the system normalizes filenames (Mac OSX),
1221
you can access a file by any path which will normalize correctly.
1222
On platforms where the system does not normalize filenames
1223
(Windows, Linux), you have to access a file by its exact path.
1225
Internally, bzr only supports NFC normalization, since that is
1226
the standard for XML documents.
1228
So return the normalized path, and a flag indicating if the file
1229
can be accessed by that path.
1232
return unicodedata.normalize('NFC', unicode(path)), True
1235
def _inaccessible_normalized_filename(path):
1236
__doc__ = _accessible_normalized_filename.__doc__
1238
normalized = unicodedata.normalize('NFC', unicode(path))
1239
return normalized, normalized == path
1242
if _platform_normalizes_filenames:
1243
normalized_filename = _accessible_normalized_filename
1245
normalized_filename = _inaccessible_normalized_filename
1248
def terminal_width():
1249
"""Return estimated terminal width."""
1250
if sys.platform == 'win32':
1251
return win32utils.get_console_size()[0]
1254
import struct, fcntl, termios
1255
s = struct.pack('HHHH', 0, 0, 0, 0)
1256
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1257
width = struct.unpack('HHHH', x)[1]
1262
width = int(os.environ['COLUMNS'])
1271
def supports_executable():
1272
return sys.platform != "win32"
1275
def supports_posix_readonly():
1276
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1278
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1279
directory controls creation/deletion, etc.
1281
And under win32, readonly means that the directory itself cannot be
1282
deleted. The contents of a readonly directory can be changed, unlike POSIX
1283
where files in readonly directories cannot be added, deleted or renamed.
1285
return sys.platform != "win32"
1288
def set_or_unset_env(env_variable, value):
1289
"""Modify the environment, setting or removing the env_variable.
1291
:param env_variable: The environment variable in question
1292
:param value: The value to set the environment to. If None, then
1293
the variable will be removed.
1294
:return: The original value of the environment variable.
1296
orig_val = os.environ.get(env_variable)
1298
if orig_val is not None:
1299
del os.environ[env_variable]
1301
if isinstance(value, unicode):
1302
value = value.encode(get_user_encoding())
1303
os.environ[env_variable] = value
1307
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1310
def check_legal_path(path):
1311
"""Check whether the supplied path is legal.
1312
This is only required on Windows, so we don't test on other platforms
1315
if sys.platform != "win32":
1317
if _validWin32PathRE.match(path) is None:
1318
raise errors.IllegalPath(path)
1321
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1323
def _is_error_enotdir(e):
1324
"""Check if this exception represents ENOTDIR.
1326
Unfortunately, python is very inconsistent about the exception
1327
here. The cases are:
1328
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1329
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1330
which is the windows error code.
1331
3) Windows, Python2.5 uses errno == EINVAL and
1332
winerror == ERROR_DIRECTORY
1334
:param e: An Exception object (expected to be OSError with an errno
1335
attribute, but we should be able to cope with anything)
1336
:return: True if this represents an ENOTDIR error. False otherwise.
1338
en = getattr(e, 'errno', None)
1339
if (en == errno.ENOTDIR
1340
or (sys.platform == 'win32'
1341
and (en == _WIN32_ERROR_DIRECTORY
1342
or (en == errno.EINVAL
1343
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1349
def walkdirs(top, prefix=""):
1350
"""Yield data about all the directories in a tree.
1352
This yields all the data about the contents of a directory at a time.
1353
After each directory has been yielded, if the caller has mutated the list
1354
to exclude some directories, they are then not descended into.
1356
The data yielded is of the form:
1357
((directory-relpath, directory-path-from-top),
1358
[(relpath, basename, kind, lstat, path-from-top), ...]),
1359
- directory-relpath is the relative path of the directory being returned
1360
with respect to top. prefix is prepended to this.
1361
- directory-path-from-root is the path including top for this directory.
1362
It is suitable for use with os functions.
1363
- relpath is the relative path within the subtree being walked.
1364
- basename is the basename of the path
1365
- kind is the kind of the file now. If unknown then the file is not
1366
present within the tree - but it may be recorded as versioned. See
1368
- lstat is the stat data *if* the file was statted.
1369
- planned, not implemented:
1370
path_from_tree_root is the path from the root of the tree.
1372
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1373
allows one to walk a subtree but get paths that are relative to a tree
1375
:return: an iterator over the dirs.
1377
#TODO there is a bit of a smell where the results of the directory-
1378
# summary in this, and the path from the root, may not agree
1379
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1380
# potentially confusing output. We should make this more robust - but
1381
# not at a speed cost. RBC 20060731
1383
_directory = _directory_kind
1384
_listdir = os.listdir
1385
_kind_from_mode = file_kind_from_stat_mode
1386
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1388
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1389
relroot, _, _, _, top = pending.pop()
1391
relprefix = relroot + u'/'
1394
top_slash = top + u'/'
1397
append = dirblock.append
1399
names = sorted(_listdir(top))
1401
if not _is_error_enotdir(e):
1405
abspath = top_slash + name
1406
statvalue = _lstat(abspath)
1407
kind = _kind_from_mode(statvalue.st_mode)
1408
append((relprefix + name, name, kind, statvalue, abspath))
1409
yield (relroot, top), dirblock
1411
# push the user specified dirs from dirblock
1412
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1415
class DirReader(object):
1416
"""An interface for reading directories."""
1418
def top_prefix_to_starting_dir(self, top, prefix=""):
1419
"""Converts top and prefix to a starting dir entry
1421
:param top: A utf8 path
1422
:param prefix: An optional utf8 path to prefix output relative paths
1424
:return: A tuple starting with prefix, and ending with the native
1427
raise NotImplementedError(self.top_prefix_to_starting_dir)
1429
def read_dir(self, prefix, top):
1430
"""Read a specific dir.
1432
:param prefix: A utf8 prefix to be preprended to the path basenames.
1433
:param top: A natively encoded path to read.
1434
:return: A list of the directories contents. Each item contains:
1435
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1437
raise NotImplementedError(self.read_dir)
1440
_selected_dir_reader = None
1443
def _walkdirs_utf8(top, prefix=""):
1444
"""Yield data about all the directories in a tree.
1446
This yields the same information as walkdirs() only each entry is yielded
1447
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1448
are returned as exact byte-strings.
1450
:return: yields a tuple of (dir_info, [file_info])
1451
dir_info is (utf8_relpath, path-from-top)
1452
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1453
if top is an absolute path, path-from-top is also an absolute path.
1454
path-from-top might be unicode or utf8, but it is the correct path to
1455
pass to os functions to affect the file in question. (such as os.lstat)
1457
global _selected_dir_reader
1458
if _selected_dir_reader is None:
1459
fs_encoding = _fs_enc.upper()
1460
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1461
# Win98 doesn't have unicode apis like FindFirstFileW
1462
# TODO: We possibly could support Win98 by falling back to the
1463
# original FindFirstFile, and using TCHAR instead of WCHAR,
1464
# but that gets a bit tricky, and requires custom compiling
1467
from bzrlib._walkdirs_win32 import Win32ReadDir
1468
_selected_dir_reader = Win32ReadDir()
1471
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1472
# ANSI_X3.4-1968 is a form of ASCII
1474
from bzrlib._readdir_pyx import UTF8DirReader
1475
_selected_dir_reader = UTF8DirReader()
1479
if _selected_dir_reader is None:
1480
# Fallback to the python version
1481
_selected_dir_reader = UnicodeDirReader()
1483
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1484
# But we don't actually uses 1-3 in pending, so set them to None
1485
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1486
read_dir = _selected_dir_reader.read_dir
1487
_directory = _directory_kind
1489
relroot, _, _, _, top = pending[-1].pop()
1492
dirblock = sorted(read_dir(relroot, top))
1493
yield (relroot, top), dirblock
1494
# push the user specified dirs from dirblock
1495
next = [d for d in reversed(dirblock) if d[2] == _directory]
1497
pending.append(next)
1500
class UnicodeDirReader(DirReader):
1501
"""A dir reader for non-utf8 file systems, which transcodes."""
1503
__slots__ = ['_utf8_encode']
1506
self._utf8_encode = codecs.getencoder('utf8')
1508
def top_prefix_to_starting_dir(self, top, prefix=""):
1509
"""See DirReader.top_prefix_to_starting_dir."""
1510
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1512
def read_dir(self, prefix, top):
1513
"""Read a single directory from a non-utf8 file system.
1515
top, and the abspath element in the output are unicode, all other paths
1516
are utf8. Local disk IO is done via unicode calls to listdir etc.
1518
This is currently the fallback code path when the filesystem encoding is
1519
not UTF-8. It may be better to implement an alternative so that we can
1520
safely handle paths that are not properly decodable in the current
1523
See DirReader.read_dir for details.
1525
_utf8_encode = self._utf8_encode
1527
_listdir = os.listdir
1528
_kind_from_mode = file_kind_from_stat_mode
1531
relprefix = prefix + '/'
1534
top_slash = top + u'/'
1537
append = dirblock.append
1538
for name in sorted(_listdir(top)):
1540
name_utf8 = _utf8_encode(name)[0]
1541
except UnicodeDecodeError:
1542
raise errors.BadFilenameEncoding(
1543
_utf8_encode(relprefix)[0] + name, _fs_enc)
1544
abspath = top_slash + name
1545
statvalue = _lstat(abspath)
1546
kind = _kind_from_mode(statvalue.st_mode)
1547
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1551
def copy_tree(from_path, to_path, handlers={}):
1552
"""Copy all of the entries in from_path into to_path.
1554
:param from_path: The base directory to copy.
1555
:param to_path: The target directory. If it does not exist, it will
1557
:param handlers: A dictionary of functions, which takes a source and
1558
destinations for files, directories, etc.
1559
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1560
'file', 'directory', and 'symlink' should always exist.
1561
If they are missing, they will be replaced with 'os.mkdir()',
1562
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1564
# Now, just copy the existing cached tree to the new location
1565
# We use a cheap trick here.
1566
# Absolute paths are prefixed with the first parameter
1567
# relative paths are prefixed with the second.
1568
# So we can get both the source and target returned
1569
# without any extra work.
1571
def copy_dir(source, dest):
1574
def copy_link(source, dest):
1575
"""Copy the contents of a symlink"""
1576
link_to = os.readlink(source)
1577
os.symlink(link_to, dest)
1579
real_handlers = {'file':shutil.copy2,
1580
'symlink':copy_link,
1581
'directory':copy_dir,
1583
real_handlers.update(handlers)
1585
if not os.path.exists(to_path):
1586
real_handlers['directory'](from_path, to_path)
1588
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1589
for relpath, name, kind, st, abspath in entries:
1590
real_handlers[kind](abspath, relpath)
1593
def path_prefix_key(path):
1594
"""Generate a prefix-order path key for path.
1596
This can be used to sort paths in the same way that walkdirs does.
1598
return (dirname(path) , path)
1601
def compare_paths_prefix_order(path_a, path_b):
1602
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1603
key_a = path_prefix_key(path_a)
1604
key_b = path_prefix_key(path_b)
1605
return cmp(key_a, key_b)
1608
_cached_user_encoding = None
1611
def get_user_encoding(use_cache=True):
1612
"""Find out what the preferred user encoding is.
1614
This is generally the encoding that is used for command line parameters
1615
and file contents. This may be different from the terminal encoding
1616
or the filesystem encoding.
1618
:param use_cache: Enable cache for detected encoding.
1619
(This parameter is turned on by default,
1620
and required only for selftesting)
1622
:return: A string defining the preferred user encoding
1624
global _cached_user_encoding
1625
if _cached_user_encoding is not None and use_cache:
1626
return _cached_user_encoding
1628
if sys.platform == 'darwin':
1629
# python locale.getpreferredencoding() always return
1630
# 'mac-roman' on darwin. That's a lie.
1631
sys.platform = 'posix'
1633
if os.environ.get('LANG', None) is None:
1634
# If LANG is not set, we end up with 'ascii', which is bad
1635
# ('mac-roman' is more than ascii), so we set a default which
1636
# will give us UTF-8 (which appears to work in all cases on
1637
# OSX). Users are still free to override LANG of course, as
1638
# long as it give us something meaningful. This work-around
1639
# *may* not be needed with python 3k and/or OSX 10.5, but will
1640
# work with them too -- vila 20080908
1641
os.environ['LANG'] = 'en_US.UTF-8'
1644
sys.platform = 'darwin'
1649
user_encoding = locale.getpreferredencoding()
1650
except locale.Error, e:
1651
sys.stderr.write('bzr: warning: %s\n'
1652
' Could not determine what text encoding to use.\n'
1653
' This error usually means your Python interpreter\n'
1654
' doesn\'t support the locale set by $LANG (%s)\n'
1655
" Continuing with ascii encoding.\n"
1656
% (e, os.environ.get('LANG')))
1657
user_encoding = 'ascii'
1659
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1660
# treat that as ASCII, and not support printing unicode characters to the
1663
# For python scripts run under vim, we get '', so also treat that as ASCII
1664
if user_encoding in (None, 'cp0', ''):
1665
user_encoding = 'ascii'
1669
codecs.lookup(user_encoding)
1671
sys.stderr.write('bzr: warning:'
1672
' unknown encoding %s.'
1673
' Continuing with ascii encoding.\n'
1676
user_encoding = 'ascii'
1679
_cached_user_encoding = user_encoding
1681
return user_encoding
1684
def get_host_name():
1685
"""Return the current unicode host name.
1687
This is meant to be used in place of socket.gethostname() because that
1688
behaves inconsistently on different platforms.
1690
if sys.platform == "win32":
1692
return win32utils.get_host_name()
1695
return socket.gethostname().decode(get_user_encoding())
1698
def recv_all(socket, bytes):
1699
"""Receive an exact number of bytes.
1701
Regular Socket.recv() may return less than the requested number of bytes,
1702
dependning on what's in the OS buffer. MSG_WAITALL is not available
1703
on all platforms, but this should work everywhere. This will return
1704
less than the requested amount if the remote end closes.
1706
This isn't optimized and is intended mostly for use in testing.
1709
while len(b) < bytes:
1710
new = until_no_eintr(socket.recv, bytes - len(b))
1717
def send_all(socket, bytes, report_activity=None):
1718
"""Send all bytes on a socket.
1720
Regular socket.sendall() can give socket error 10053 on Windows. This
1721
implementation sends no more than 64k at a time, which avoids this problem.
1723
:param report_activity: Call this as bytes are read, see
1724
Transport._report_activity
1727
for pos in xrange(0, len(bytes), chunk_size):
1728
block = bytes[pos:pos+chunk_size]
1729
if report_activity is not None:
1730
report_activity(len(block), 'write')
1731
until_no_eintr(socket.sendall, block)
1734
def dereference_path(path):
1735
"""Determine the real path to a file.
1737
All parent elements are dereferenced. But the file itself is not
1739
:param path: The original path. May be absolute or relative.
1740
:return: the real path *to* the file
1742
parent, base = os.path.split(path)
1743
# The pathjoin for '.' is a workaround for Python bug #1213894.
1744
# (initial path components aren't dereferenced)
1745
return pathjoin(realpath(pathjoin('.', parent)), base)
1748
def supports_mapi():
1749
"""Return True if we can use MAPI to launch a mail client."""
1750
return sys.platform == "win32"
1753
def resource_string(package, resource_name):
1754
"""Load a resource from a package and return it as a string.
1756
Note: Only packages that start with bzrlib are currently supported.
1758
This is designed to be a lightweight implementation of resource
1759
loading in a way which is API compatible with the same API from
1761
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1762
If and when pkg_resources becomes a standard library, this routine
1765
# Check package name is within bzrlib
1766
if package == "bzrlib":
1767
resource_relpath = resource_name
1768
elif package.startswith("bzrlib."):
1769
package = package[len("bzrlib."):].replace('.', os.sep)
1770
resource_relpath = pathjoin(package, resource_name)
1772
raise errors.BzrError('resource package %s not in bzrlib' % package)
1774
# Map the resource to a file and read its contents
1775
base = dirname(bzrlib.__file__)
1776
if getattr(sys, 'frozen', None): # bzr.exe
1777
base = abspath(pathjoin(base, '..', '..'))
1778
filename = pathjoin(base, resource_relpath)
1779
return open(filename, 'rU').read()
1782
def file_kind_from_stat_mode_thunk(mode):
1783
global file_kind_from_stat_mode
1784
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1786
from bzrlib._readdir_pyx import UTF8DirReader
1787
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1789
from bzrlib._readdir_py import (
1790
_kind_from_mode as file_kind_from_stat_mode
1792
return file_kind_from_stat_mode(mode)
1793
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1796
def file_kind(f, _lstat=os.lstat):
1798
return file_kind_from_stat_mode(_lstat(f).st_mode)
1800
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1801
raise errors.NoSuchFile(f)
1805
def until_no_eintr(f, *a, **kw):
1806
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1807
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1811
except (IOError, OSError), e:
1812
if e.errno == errno.EINTR:
1816
def re_compile_checked(re_string, flags=0, where=""):
1817
"""Return a compiled re, or raise a sensible error.
1819
This should only be used when compiling user-supplied REs.
1821
:param re_string: Text form of regular expression.
1822
:param flags: eg re.IGNORECASE
1823
:param where: Message explaining to the user the context where
1824
it occurred, eg 'log search filter'.
1826
# from https://bugs.launchpad.net/bzr/+bug/251352
1828
re_obj = re.compile(re_string, flags)
1833
where = ' in ' + where
1834
# despite the name 'error' is a type
1835
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
1836
% (where, re_string, e))
1839
if sys.platform == "win32":
1842
return msvcrt.getch()
1847
fd = sys.stdin.fileno()
1848
settings = termios.tcgetattr(fd)
1851
ch = sys.stdin.read(1)
1853
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
1857
if sys.platform == 'linux2':
1858
def _local_concurrency():
1860
prefix = 'processor'
1861
for line in file('/proc/cpuinfo', 'rb'):
1862
if line.startswith(prefix):
1863
concurrency = int(line[line.find(':')+1:]) + 1
1865
elif sys.platform == 'darwin':
1866
def _local_concurrency():
1867
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
1868
stdout=subprocess.PIPE).communicate()[0]
1869
elif sys.platform[0:7] == 'freebsd':
1870
def _local_concurrency():
1871
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
1872
stdout=subprocess.PIPE).communicate()[0]
1873
elif sys.platform == 'sunos5':
1874
def _local_concurrency():
1875
return subprocess.Popen(['psrinfo', '-p',],
1876
stdout=subprocess.PIPE).communicate()[0]
1877
elif sys.platform == "win32":
1878
def _local_concurrency():
1879
# This appears to return the number of cores.
1880
return os.environ.get('NUMBER_OF_PROCESSORS')
1882
def _local_concurrency():
1887
_cached_local_concurrency = None
1889
def local_concurrency(use_cache=True):
1890
"""Return how many processes can be run concurrently.
1892
Rely on platform specific implementations and default to 1 (one) if
1893
anything goes wrong.
1895
global _cached_local_concurrency
1896
if _cached_local_concurrency is not None and use_cache:
1897
return _cached_local_concurrency
1900
concurrency = _local_concurrency()
1901
except (OSError, IOError):
1904
concurrency = int(concurrency)
1905
except (TypeError, ValueError):
1908
_cached_concurrency = concurrency
added
removed
bzrlib/osutils.py
