47
133
This previously used backslash quoting, but that works poorly on
49
135
# TODO: I'm not really sure this is the best format either.x
137
if _QUOTE_RE is None:
138
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
50
140
if _QUOTE_RE.search(f):
51
141
return '"' + f + '"'
57
mode = os.lstat(f)[ST_MODE]
65
raise BzrError("can't handle file kind with mode %o of %r" % (mode, f))
146
_directory_kind = 'directory'
149
"""Return the current umask"""
150
# Assume that people aren't messing with the umask while running
151
# XXX: This is not thread safe, but there is no way to get the
152
# umask without setting it
160
_directory_kind: "/",
162
'tree-reference': '+',
68
166
def kind_marker(kind):
71
elif kind == 'directory':
73
elif kind == 'symlink':
76
raise BzrError('invalid file kind %r' % kind)
81
"""Copy a file to a backup.
83
Backups are named in GNU-style, with a ~ suffix.
85
If the file is already a backup, it's not copied.
98
outf = file(bfn, 'wb')
104
def rename(path_from, path_to):
105
"""Basically the same as os.rename() just special for win32"""
106
if sys.platform == 'win32':
168
return _kind_marker_map[kind]
170
raise errors.BzrError('invalid file kind %r' % kind)
173
lexists = getattr(os.path, 'lexists', None)
177
stat = getattr(os, 'lstat', os.stat)
109
180
except OSError, e:
110
if e.errno != e.ENOENT:
181
if e.errno == errno.ENOENT:
184
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
187
def fancy_rename(old, new, rename_func, unlink_func):
188
"""A fancy rename, when you don't have atomic rename.
190
:param old: The old path, to rename from
191
:param new: The new path, to rename to
192
:param rename_func: The potentially non-atomic rename function
193
:param unlink_func: A way to delete the target file if the full rename succeeds
196
# sftp rename doesn't allow overwriting, so play tricks:
197
base = os.path.basename(new)
198
dirname = os.path.dirname(new)
199
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
200
tmp_name = pathjoin(dirname, tmp_name)
202
# Rename the file out of the way, but keep track if it didn't exist
203
# We don't want to grab just any exception
204
# something like EACCES should prevent us from continuing
205
# The downside is that the rename_func has to throw an exception
206
# with an errno = ENOENT, or NoSuchFile
209
rename_func(new, tmp_name)
210
except (errors.NoSuchFile,), e:
213
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
214
# function raises an IOError with errno is None when a rename fails.
215
# This then gets caught here.
216
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
219
if (getattr(e, 'errno', None) is None
220
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
228
# This may throw an exception, in which case success will
230
rename_func(old, new)
232
except (IOError, OSError), e:
233
# source and target may be aliases of each other (e.g. on a
234
# case-insensitive filesystem), so we may have accidentally renamed
235
# source by when we tried to rename target
236
if not (file_existed and e.errno in (None, errno.ENOENT)):
112
os.rename(path_from, path_to)
240
# If the file used to exist, rename it back into place
241
# otherwise just delete it from the tmp location
243
unlink_func(tmp_name)
245
rename_func(tmp_name, new)
248
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
249
# choke on a Unicode string containing a relative path if
250
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
252
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
253
def _posix_abspath(path):
254
# jam 20060426 rather than encoding to fsencoding
255
# copy posixpath.abspath, but use os.getcwdu instead
256
if not posixpath.isabs(path):
257
path = posixpath.join(getcwd(), path)
258
return posixpath.normpath(path)
261
def _posix_realpath(path):
262
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
265
def _win32_fixdrive(path):
266
"""Force drive letters to be consistent.
268
win32 is inconsistent whether it returns lower or upper case
269
and even if it was consistent the user might type the other
270
so we force it to uppercase
271
running python.exe under cmd.exe return capital C:\\
272
running win32 python inside a cygwin shell returns lowercase c:\\
274
drive, path = _nt_splitdrive(path)
275
return drive.upper() + path
278
def _win32_abspath(path):
279
# Real _nt_abspath doesn't have a problem with a unicode cwd
280
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
283
def _win98_abspath(path):
284
"""Return the absolute version of a path.
285
Windows 98 safe implementation (python reimplementation
286
of Win32 API function GetFullPathNameW)
291
# \\HOST\path => //HOST/path
292
# //HOST/path => //HOST/path
293
# path => C:/cwd/path
296
# check for absolute path
297
drive = _nt_splitdrive(path)[0]
298
if drive == '' and path[:2] not in('//','\\\\'):
300
# we cannot simply os.path.join cwd and path
301
# because os.path.join('C:','/path') produce '/path'
302
# and this is incorrect
303
if path[:1] in ('/','\\'):
304
cwd = _nt_splitdrive(cwd)[0]
306
path = cwd + '\\' + path
307
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
310
def _win32_realpath(path):
311
# Real _nt_realpath doesn't have a problem with a unicode cwd
312
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
315
def _win32_pathjoin(*args):
316
return _nt_join(*args).replace('\\', '/')
319
def _win32_normpath(path):
320
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
324
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
327
def _win32_mkdtemp(*args, **kwargs):
328
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
331
def _win32_rename(old, new):
332
"""We expect to be able to atomically replace 'new' with old.
334
On win32, if new exists, it must be moved out of the way first,
338
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
340
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
341
# If we try to rename a non-existant file onto cwd, we get
342
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
343
# if the old path doesn't exist, sometimes we get EACCES
344
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
350
return unicodedata.normalize('NFC', os.getcwdu())
353
# Default is to just use the python builtins, but these can be rebound on
354
# particular platforms.
355
abspath = _posix_abspath
356
realpath = _posix_realpath
357
pathjoin = os.path.join
358
normpath = os.path.normpath
361
dirname = os.path.dirname
362
basename = os.path.basename
363
split = os.path.split
364
splitext = os.path.splitext
365
# These were already imported into local scope
366
# mkdtemp = tempfile.mkdtemp
367
# rmtree = shutil.rmtree
369
MIN_ABS_PATHLENGTH = 1
372
if sys.platform == 'win32':
373
if win32utils.winver == 'Windows 98':
374
abspath = _win98_abspath
376
abspath = _win32_abspath
377
realpath = _win32_realpath
378
pathjoin = _win32_pathjoin
379
normpath = _win32_normpath
380
getcwd = _win32_getcwd
381
mkdtemp = _win32_mkdtemp
382
rename = _win32_rename
384
MIN_ABS_PATHLENGTH = 3
386
def _win32_delete_readonly(function, path, excinfo):
387
"""Error handler for shutil.rmtree function [for win32]
388
Helps to remove files and dirs marked as read-only.
390
exception = excinfo[1]
391
if function in (os.remove, os.rmdir) \
392
and isinstance(exception, OSError) \
393
and exception.errno == errno.EACCES:
399
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
400
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
401
return shutil.rmtree(path, ignore_errors, onerror)
403
f = win32utils.get_unicode_argv # special function or None
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
osutils.get_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
from bzrlib.trace import mutter
425
output_encoding = getattr(sys.stdout, 'encoding', None)
426
if not output_encoding:
427
input_encoding = getattr(sys.stdin, 'encoding', None)
428
if not input_encoding:
429
output_encoding = get_user_encoding()
430
mutter('encoding stdout as osutils.get_user_encoding() %r',
433
output_encoding = input_encoding
434
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
436
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
437
if output_encoding == 'cp0':
438
# invalid encoding (cp0 means 'no codepage' on Windows)
439
output_encoding = get_user_encoding()
440
mutter('cp0 is invalid encoding.'
441
' encoding stdout as osutils.get_user_encoding() %r',
445
codecs.lookup(output_encoding)
447
sys.stderr.write('bzr: warning:'
448
' unknown terminal encoding %s.\n'
449
' Using encoding %s instead.\n'
450
% (output_encoding, get_user_encoding())
452
output_encoding = get_user_encoding()
454
return output_encoding
457
def normalizepath(f):
458
if getattr(os.path, 'realpath', None) is not None:
462
[p,e] = os.path.split(f)
463
if e == "" or e == "." or e == "..":
466
return pathjoin(F(p), e)
342
681
def local_time_offset(t=None):
343
682
"""Return offset of local zone from GMT, either at present or at time t."""
344
# python2.3 localtime() can't take None
348
if time.localtime(t).tm_isdst and time.daylight:
351
return -time.timezone
354
def format_date(t, offset=0, timezone='original'):
355
## TODO: Perhaps a global option to use either universal or local time?
356
## Or perhaps just let people set $TZ?
357
assert isinstance(t, float)
685
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
686
return offset.days * 86400 + offset.seconds
688
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
690
def format_date(t, offset=0, timezone='original', date_fmt=None,
692
"""Return a formatted date string.
694
:param t: Seconds since the epoch.
695
:param offset: Timezone offset in seconds east of utc.
696
:param timezone: How to display the time: 'utc', 'original' for the
697
timezone specified by offset, or 'local' for the process's current
699
:param date_fmt: strftime format.
700
:param show_offset: Whether to append the timezone.
702
(date_fmt, tt, offset_str) = \
703
_format_date(t, offset, timezone, date_fmt, show_offset)
704
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
705
date_str = time.strftime(date_fmt, tt)
706
return date_str + offset_str
708
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
710
"""Return an unicode date string formatted according to the current locale.
712
:param t: Seconds since the epoch.
713
:param offset: Timezone offset in seconds east of utc.
714
:param timezone: How to display the time: 'utc', 'original' for the
715
timezone specified by offset, or 'local' for the process's current
717
:param date_fmt: strftime format.
718
:param show_offset: Whether to append the timezone.
720
(date_fmt, tt, offset_str) = \
721
_format_date(t, offset, timezone, date_fmt, show_offset)
722
date_str = time.strftime(date_fmt, tt)
723
if not isinstance(date_str, unicode):
724
date_str = date_str.decode(bzrlib.user_encoding, 'replace')
725
return date_str + offset_str
727
def _format_date(t, offset, timezone, date_fmt, show_offset):
359
728
if timezone == 'utc':
360
729
tt = time.gmtime(t)
362
731
elif timezone == 'original':
365
734
tt = time.gmtime(t + offset)
366
735
elif timezone == 'local':
367
736
tt = time.localtime(t)
368
737
offset = local_time_offset(t)
370
raise BzrError("unsupported timezone format %r" % timezone,
371
['options are "utc", "original", "local"'])
373
return (time.strftime("%a %Y-%m-%d %H:%M:%S", tt)
374
+ ' %+03d%02d' % (offset / 3600, (offset / 60) % 60))
739
raise errors.UnsupportedTimezoneFormat(timezone)
741
date_fmt = "%a %Y-%m-%d %H:%M:%S"
743
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
746
return (date_fmt, tt, offset_str)
377
749
def compact_date(when):
378
750
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
753
def format_delta(delta):
754
"""Get a nice looking string for a time delta.
756
:param delta: The time difference in seconds, can be positive or negative.
757
positive indicates time in the past, negative indicates time in the
758
future. (usually time.time() - stored_time)
759
:return: String formatted to show approximate resolution
765
direction = 'in the future'
769
if seconds < 90: # print seconds up to 90 seconds
771
return '%d second %s' % (seconds, direction,)
773
return '%d seconds %s' % (seconds, direction)
775
minutes = int(seconds / 60)
776
seconds -= 60 * minutes
781
if minutes < 90: # print minutes, seconds up to 90 minutes
783
return '%d minute, %d second%s %s' % (
784
minutes, seconds, plural_seconds, direction)
786
return '%d minutes, %d second%s %s' % (
787
minutes, seconds, plural_seconds, direction)
789
hours = int(minutes / 60)
790
minutes -= 60 * hours
797
return '%d hour, %d minute%s %s' % (hours, minutes,
798
plural_minutes, direction)
799
return '%d hours, %d minute%s %s' % (hours, minutes,
800
plural_minutes, direction)
383
803
"""Return size of given open file."""
384
804
return os.fstat(f.fileno())[ST_SIZE]
387
if hasattr(os, 'urandom'): # python 2.4 and later
807
# Define rand_bytes based on platform.
809
# Python 2.4 and later have os.urandom,
810
# but it doesn't work on some arches
388
812
rand_bytes = os.urandom
389
elif sys.platform == 'linux2':
390
rand_bytes = file('/dev/urandom', 'rb').read
392
# not well seeded, but better than nothing
397
s += chr(random.randint(0, 255))
813
except (NotImplementedError, AttributeError):
814
# If python doesn't have os.urandom, or it doesn't work,
815
# then try to first pull random data from /dev/urandom
817
rand_bytes = file('/dev/urandom', 'rb').read
818
# Otherwise, use this hack as a last resort
819
except (IOError, OSError):
820
# not well seeded, but better than nothing
825
s += chr(random.randint(0, 255))
830
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
832
"""Return a random string of num alphanumeric characters
834
The result only contains lowercase chars because it may be used on
835
case-insensitive filesystems.
838
for raw_byte in rand_bytes(num):
839
s += ALNUM[ord(raw_byte) % 36]
402
843
## TODO: We could later have path objects that remember their list
403
844
## decomposition (might be too tricksy though.)
405
846
def splitpath(p):
406
"""Turn string into list of parts.
412
>>> splitpath('a/./b')
414
>>> splitpath('a/.b')
416
>>> splitpath('a/../b')
417
Traceback (most recent call last):
419
BzrError: sorry, '..' not allowed in path
421
assert isinstance(p, types.StringTypes)
847
"""Turn string into list of parts."""
423
848
# split on either delimiter because people might use either on
425
850
ps = re.split(r'[\\/]', p)
430
raise BzrError("sorry, %r not allowed in path" % f)
855
raise errors.BzrError("sorry, %r not allowed in path" % f)
431
856
elif (f == '.') or (f == ''):
438
assert isinstance(p, list)
440
if (f == '..') or (f == None) or (f == ''):
441
raise BzrError("sorry, %r not allowed in path" % f)
442
return os.path.join(*p)
445
def appendpath(p1, p2):
449
return os.path.join(p1, p2)
865
if (f == '..') or (f is None) or (f == ''):
866
raise errors.BzrError("sorry, %r not allowed in path" % f)
870
def parent_directories(filename):
871
"""Return the list of parent directories, deepest first.
452
def extern_command(cmd, ignore_errors = False):
453
mutter('external command: %s' % `cmd`)
455
if not ignore_errors:
456
raise BzrError('command failed')
459
def _read_config_value(name):
460
"""Read a config value from the file ~/.bzr.conf/<name>
461
Return None if the file does not exist"""
463
f = file(os.path.join(config_dir(), name), "r")
464
return f.read().decode(bzrlib.user_encoding).rstrip("\r\n")
466
if e.errno == errno.ENOENT:
873
For example, parent_directories("a/b/c") -> ["a/b", "a"].
876
parts = splitpath(dirname(filename))
878
parents.append(joinpath(parts))
884
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
886
from bzrlib._chunks_to_lines_py import chunks_to_lines
890
"""Split s into lines, but without removing the newline characters."""
891
# Trivially convert a fulltext into a 'chunked' representation, and let
892
# chunks_to_lines do the heavy lifting.
893
if isinstance(s, str):
894
# chunks_to_lines only supports 8-bit strings
895
return chunks_to_lines([s])
897
return _split_lines(s)
901
"""Split s into lines, but without removing the newline characters.
903
This supports Unicode or plain string objects.
905
lines = s.split('\n')
906
result = [line + '\n' for line in lines[:-1]]
908
result.append(lines[-1])
912
def hardlinks_good():
913
return sys.platform not in ('win32', 'cygwin', 'darwin')
916
def link_or_copy(src, dest):
917
"""Hardlink a file, or copy it if it can't be hardlinked."""
918
if not hardlinks_good():
919
shutil.copyfile(src, dest)
923
except (OSError, IOError), e:
924
if e.errno != errno.EXDEV:
926
shutil.copyfile(src, dest)
929
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
930
# Forgiveness than Permission (EAFP) because:
931
# - root can damage a solaris file system by using unlink,
932
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
933
# EACCES, OSX: EPERM) when invoked on a directory.
934
def delete_any(path):
935
"""Delete a file or directory."""
936
if isdir(path): # Takes care of symlinks
943
if getattr(os, 'symlink', None) is not None:
950
if getattr(os, 'link', None) is not None:
956
def host_os_dereferences_symlinks():
957
return (has_symlinks()
958
and sys.platform not in ('cygwin', 'win32'))
961
def readlink(abspath):
962
"""Return a string representing the path to which the symbolic link points.
964
:param abspath: The link absolute unicode path.
966
This his guaranteed to return the symbolic link in unicode in all python
969
link = abspath.encode(_fs_enc)
970
target = os.readlink(link)
971
target = target.decode(_fs_enc)
975
def contains_whitespace(s):
976
"""True if there are any whitespace characters in s."""
977
# string.whitespace can include '\xa0' in certain locales, because it is
978
# considered "non-breaking-space" as part of ISO-8859-1. But it
979
# 1) Isn't a breaking whitespace
980
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
982
# 3) '\xa0' isn't unicode safe since it is >128.
984
# This should *not* be a unicode set of characters in case the source
985
# string is not a Unicode string. We can auto-up-cast the characters since
986
# they are ascii, but we don't want to auto-up-cast the string in case it
988
for ch in ' \t\n\r\v\f':
995
def contains_linebreaks(s):
996
"""True if there is any vertical whitespace in s."""
1004
def relpath(base, path):
1005
"""Return path relative to base, or raise exception.
1007
The path may be either an absolute path or a path relative to the
1008
current working directory.
1010
os.path.commonprefix (python2.4) has a bad bug that it works just
1011
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1012
avoids that problem.
1015
if len(base) < MIN_ABS_PATHLENGTH:
1016
# must have space for e.g. a drive letter
1017
raise ValueError('%r is too short to calculate a relative path'
1024
while len(head) >= len(base):
1027
head, tail = os.path.split(head)
1031
raise errors.PathNotChild(rp, base)
1039
def _cicp_canonical_relpath(base, path):
1040
"""Return the canonical path relative to base.
1042
Like relpath, but on case-insensitive-case-preserving file-systems, this
1043
will return the relpath as stored on the file-system rather than in the
1044
case specified in the input string, for all existing portions of the path.
1046
This will cause O(N) behaviour if called for every path in a tree; if you
1047
have a number of paths to convert, you should use canonical_relpaths().
1049
# TODO: it should be possible to optimize this for Windows by using the
1050
# win32 API FindFiles function to look for the specified name - but using
1051
# os.listdir() still gives us the correct, platform agnostic semantics in
1054
rel = relpath(base, path)
1055
# '.' will have been turned into ''
1059
abs_base = abspath(base)
1061
_listdir = os.listdir
1063
# use an explicit iterator so we can easily consume the rest on early exit.
1064
bit_iter = iter(rel.split('/'))
1065
for bit in bit_iter:
1067
for look in _listdir(current):
1068
if lbit == look.lower():
1069
current = pathjoin(current, look)
1072
# got to the end, nothing matched, so we just return the
1073
# non-existing bits as they were specified (the filename may be
1074
# the target of a move, for example).
1075
current = pathjoin(current, bit, *list(bit_iter))
1077
return current[len(abs_base)+1:]
1079
# XXX - TODO - we need better detection/integration of case-insensitive
1080
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1081
# filesystems), for example, so could probably benefit from the same basic
1082
# support there. For now though, only Windows and OSX get that support, and
1083
# they get it for *all* file-systems!
1084
if sys.platform in ('win32', 'darwin'):
1085
canonical_relpath = _cicp_canonical_relpath
1087
canonical_relpath = relpath
1089
def canonical_relpaths(base, paths):
1090
"""Create an iterable to canonicalize a sequence of relative paths.
1092
The intent is for this implementation to use a cache, vastly speeding
1093
up multiple transformations in the same directory.
1095
# but for now, we haven't optimized...
1096
return [canonical_relpath(base, p) for p in paths]
1098
def safe_unicode(unicode_or_utf8_string):
1099
"""Coerce unicode_or_utf8_string into unicode.
1101
If it is unicode, it is returned.
1102
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1103
wrapped in a BzrBadParameterNotUnicode exception.
1105
if isinstance(unicode_or_utf8_string, unicode):
1106
return unicode_or_utf8_string
1108
return unicode_or_utf8_string.decode('utf8')
1109
except UnicodeDecodeError:
1110
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1113
def safe_utf8(unicode_or_utf8_string):
1114
"""Coerce unicode_or_utf8_string to a utf8 string.
1116
If it is a str, it is returned.
1117
If it is Unicode, it is encoded into a utf-8 string.
1119
if isinstance(unicode_or_utf8_string, str):
1120
# TODO: jam 20070209 This is overkill, and probably has an impact on
1121
# performance if we are dealing with lots of apis that want a
1124
# Make sure it is a valid utf-8 string
1125
unicode_or_utf8_string.decode('utf-8')
1126
except UnicodeDecodeError:
1127
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1128
return unicode_or_utf8_string
1129
return unicode_or_utf8_string.encode('utf-8')
1132
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1133
' Revision id generators should be creating utf8'
1137
def safe_revision_id(unicode_or_utf8_string, warn=True):
1138
"""Revision ids should now be utf8, but at one point they were unicode.
1140
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1142
:param warn: Functions that are sanitizing user data can set warn=False
1143
:return: None or a utf8 revision id.
1145
if (unicode_or_utf8_string is None
1146
or unicode_or_utf8_string.__class__ == str):
1147
return unicode_or_utf8_string
1149
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1151
return cache_utf8.encode(unicode_or_utf8_string)
1154
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1155
' generators should be creating utf8 file ids.')
1158
def safe_file_id(unicode_or_utf8_string, warn=True):
1159
"""File ids should now be utf8, but at one point they were unicode.
1161
This is the same as safe_utf8, except it uses the cached encode functions
1162
to save a little bit of performance.
1164
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1166
:param warn: Functions that are sanitizing user data can set warn=False
1167
:return: None or a utf8 file id.
1169
if (unicode_or_utf8_string is None
1170
or unicode_or_utf8_string.__class__ == str):
1171
return unicode_or_utf8_string
1173
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1175
return cache_utf8.encode(unicode_or_utf8_string)
1178
_platform_normalizes_filenames = False
1179
if sys.platform == 'darwin':
1180
_platform_normalizes_filenames = True
1183
def normalizes_filenames():
1184
"""Return True if this platform normalizes unicode filenames.
1186
Mac OSX does, Windows/Linux do not.
1188
return _platform_normalizes_filenames
1191
def _accessible_normalized_filename(path):
1192
"""Get the unicode normalized path, and if you can access the file.
1194
On platforms where the system normalizes filenames (Mac OSX),
1195
you can access a file by any path which will normalize correctly.
1196
On platforms where the system does not normalize filenames
1197
(Windows, Linux), you have to access a file by its exact path.
1199
Internally, bzr only supports NFC normalization, since that is
1200
the standard for XML documents.
1202
So return the normalized path, and a flag indicating if the file
1203
can be accessed by that path.
1206
return unicodedata.normalize('NFC', unicode(path)), True
1209
def _inaccessible_normalized_filename(path):
1210
__doc__ = _accessible_normalized_filename.__doc__
1212
normalized = unicodedata.normalize('NFC', unicode(path))
1213
return normalized, normalized == path
1216
if _platform_normalizes_filenames:
1217
normalized_filename = _accessible_normalized_filename
1219
normalized_filename = _inaccessible_normalized_filename
1222
def terminal_width():
1223
"""Return estimated terminal width."""
1224
if sys.platform == 'win32':
1225
return win32utils.get_console_size()[0]
1228
import struct, fcntl, termios
1229
s = struct.pack('HHHH', 0, 0, 0, 0)
1230
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1231
width = struct.unpack('HHHH', x)[1]
1236
width = int(os.environ['COLUMNS'])
1245
def supports_executable():
1246
return sys.platform != "win32"
1249
def supports_posix_readonly():
1250
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1252
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1253
directory controls creation/deletion, etc.
1255
And under win32, readonly means that the directory itself cannot be
1256
deleted. The contents of a readonly directory can be changed, unlike POSIX
1257
where files in readonly directories cannot be added, deleted or renamed.
1259
return sys.platform != "win32"
1262
def set_or_unset_env(env_variable, value):
1263
"""Modify the environment, setting or removing the env_variable.
1265
:param env_variable: The environment variable in question
1266
:param value: The value to set the environment to. If None, then
1267
the variable will be removed.
1268
:return: The original value of the environment variable.
1270
orig_val = os.environ.get(env_variable)
1272
if orig_val is not None:
1273
del os.environ[env_variable]
1275
if isinstance(value, unicode):
1276
value = value.encode(get_user_encoding())
1277
os.environ[env_variable] = value
1281
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1284
def check_legal_path(path):
1285
"""Check whether the supplied path is legal.
1286
This is only required on Windows, so we don't test on other platforms
1289
if sys.platform != "win32":
1291
if _validWin32PathRE.match(path) is None:
1292
raise errors.IllegalPath(path)
1295
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1297
def _is_error_enotdir(e):
1298
"""Check if this exception represents ENOTDIR.
1300
Unfortunately, python is very inconsistent about the exception
1301
here. The cases are:
1302
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1303
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1304
which is the windows error code.
1305
3) Windows, Python2.5 uses errno == EINVAL and
1306
winerror == ERROR_DIRECTORY
1308
:param e: An Exception object (expected to be OSError with an errno
1309
attribute, but we should be able to cope with anything)
1310
:return: True if this represents an ENOTDIR error. False otherwise.
1312
en = getattr(e, 'errno', None)
1313
if (en == errno.ENOTDIR
1314
or (sys.platform == 'win32'
1315
and (en == _WIN32_ERROR_DIRECTORY
1316
or (en == errno.EINVAL
1317
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1323
def walkdirs(top, prefix=""):
1324
"""Yield data about all the directories in a tree.
1326
This yields all the data about the contents of a directory at a time.
1327
After each directory has been yielded, if the caller has mutated the list
1328
to exclude some directories, they are then not descended into.
1330
The data yielded is of the form:
1331
((directory-relpath, directory-path-from-top),
1332
[(relpath, basename, kind, lstat, path-from-top), ...]),
1333
- directory-relpath is the relative path of the directory being returned
1334
with respect to top. prefix is prepended to this.
1335
- directory-path-from-root is the path including top for this directory.
1336
It is suitable for use with os functions.
1337
- relpath is the relative path within the subtree being walked.
1338
- basename is the basename of the path
1339
- kind is the kind of the file now. If unknown then the file is not
1340
present within the tree - but it may be recorded as versioned. See
1342
- lstat is the stat data *if* the file was statted.
1343
- planned, not implemented:
1344
path_from_tree_root is the path from the root of the tree.
1346
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1347
allows one to walk a subtree but get paths that are relative to a tree
1349
:return: an iterator over the dirs.
1351
#TODO there is a bit of a smell where the results of the directory-
1352
# summary in this, and the path from the root, may not agree
1353
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1354
# potentially confusing output. We should make this more robust - but
1355
# not at a speed cost. RBC 20060731
1357
_directory = _directory_kind
1358
_listdir = os.listdir
1359
_kind_from_mode = file_kind_from_stat_mode
1360
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1362
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1363
relroot, _, _, _, top = pending.pop()
1365
relprefix = relroot + u'/'
1368
top_slash = top + u'/'
1371
append = dirblock.append
1373
names = sorted(_listdir(top))
1375
if not _is_error_enotdir(e):
1379
abspath = top_slash + name
1380
statvalue = _lstat(abspath)
1381
kind = _kind_from_mode(statvalue.st_mode)
1382
append((relprefix + name, name, kind, statvalue, abspath))
1383
yield (relroot, top), dirblock
1385
# push the user specified dirs from dirblock
1386
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1389
class DirReader(object):
1390
"""An interface for reading directories."""
1392
def top_prefix_to_starting_dir(self, top, prefix=""):
1393
"""Converts top and prefix to a starting dir entry
1395
:param top: A utf8 path
1396
:param prefix: An optional utf8 path to prefix output relative paths
1398
:return: A tuple starting with prefix, and ending with the native
1401
raise NotImplementedError(self.top_prefix_to_starting_dir)
1403
def read_dir(self, prefix, top):
1404
"""Read a specific dir.
1406
:param prefix: A utf8 prefix to be preprended to the path basenames.
1407
:param top: A natively encoded path to read.
1408
:return: A list of the directories contents. Each item contains:
1409
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1411
raise NotImplementedError(self.read_dir)
1414
_selected_dir_reader = None
1417
def _walkdirs_utf8(top, prefix=""):
1418
"""Yield data about all the directories in a tree.
1420
This yields the same information as walkdirs() only each entry is yielded
1421
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1422
are returned as exact byte-strings.
1424
:return: yields a tuple of (dir_info, [file_info])
1425
dir_info is (utf8_relpath, path-from-top)
1426
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1427
if top is an absolute path, path-from-top is also an absolute path.
1428
path-from-top might be unicode or utf8, but it is the correct path to
1429
pass to os functions to affect the file in question. (such as os.lstat)
1431
global _selected_dir_reader
1432
if _selected_dir_reader is None:
1433
fs_encoding = _fs_enc.upper()
1434
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1435
# Win98 doesn't have unicode apis like FindFirstFileW
1436
# TODO: We possibly could support Win98 by falling back to the
1437
# original FindFirstFile, and using TCHAR instead of WCHAR,
1438
# but that gets a bit tricky, and requires custom compiling
1441
from bzrlib._walkdirs_win32 import Win32ReadDir
1442
_selected_dir_reader = Win32ReadDir()
1445
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1446
# ANSI_X3.4-1968 is a form of ASCII
1448
from bzrlib._readdir_pyx import UTF8DirReader
1449
_selected_dir_reader = UTF8DirReader()
1453
if _selected_dir_reader is None:
1454
# Fallback to the python version
1455
_selected_dir_reader = UnicodeDirReader()
1457
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1458
# But we don't actually uses 1-3 in pending, so set them to None
1459
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1460
read_dir = _selected_dir_reader.read_dir
1461
_directory = _directory_kind
1463
relroot, _, _, _, top = pending[-1].pop()
1466
dirblock = sorted(read_dir(relroot, top))
1467
yield (relroot, top), dirblock
1468
# push the user specified dirs from dirblock
1469
next = [d for d in reversed(dirblock) if d[2] == _directory]
1471
pending.append(next)
1474
class UnicodeDirReader(DirReader):
1475
"""A dir reader for non-utf8 file systems, which transcodes."""
1477
__slots__ = ['_utf8_encode']
1480
self._utf8_encode = codecs.getencoder('utf8')
1482
def top_prefix_to_starting_dir(self, top, prefix=""):
1483
"""See DirReader.top_prefix_to_starting_dir."""
1484
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1486
def read_dir(self, prefix, top):
1487
"""Read a single directory from a non-utf8 file system.
1489
top, and the abspath element in the output are unicode, all other paths
1490
are utf8. Local disk IO is done via unicode calls to listdir etc.
1492
This is currently the fallback code path when the filesystem encoding is
1493
not UTF-8. It may be better to implement an alternative so that we can
1494
safely handle paths that are not properly decodable in the current
1497
See DirReader.read_dir for details.
1499
_utf8_encode = self._utf8_encode
1501
_listdir = os.listdir
1502
_kind_from_mode = file_kind_from_stat_mode
1505
relprefix = prefix + '/'
1508
top_slash = top + u'/'
1511
append = dirblock.append
1512
for name in sorted(_listdir(top)):
1514
name_utf8 = _utf8_encode(name)[0]
1515
except UnicodeDecodeError:
1516
raise errors.BadFilenameEncoding(
1517
_utf8_encode(relprefix)[0] + name, _fs_enc)
1518
abspath = top_slash + name
1519
statvalue = _lstat(abspath)
1520
kind = _kind_from_mode(statvalue.st_mode)
1521
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1525
def copy_tree(from_path, to_path, handlers={}):
1526
"""Copy all of the entries in from_path into to_path.
1528
:param from_path: The base directory to copy.
1529
:param to_path: The target directory. If it does not exist, it will
1531
:param handlers: A dictionary of functions, which takes a source and
1532
destinations for files, directories, etc.
1533
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1534
'file', 'directory', and 'symlink' should always exist.
1535
If they are missing, they will be replaced with 'os.mkdir()',
1536
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1538
# Now, just copy the existing cached tree to the new location
1539
# We use a cheap trick here.
1540
# Absolute paths are prefixed with the first parameter
1541
# relative paths are prefixed with the second.
1542
# So we can get both the source and target returned
1543
# without any extra work.
1545
def copy_dir(source, dest):
1548
def copy_link(source, dest):
1549
"""Copy the contents of a symlink"""
1550
link_to = os.readlink(source)
1551
os.symlink(link_to, dest)
1553
real_handlers = {'file':shutil.copy2,
1554
'symlink':copy_link,
1555
'directory':copy_dir,
1557
real_handlers.update(handlers)
1559
if not os.path.exists(to_path):
1560
real_handlers['directory'](from_path, to_path)
1562
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1563
for relpath, name, kind, st, abspath in entries:
1564
real_handlers[kind](abspath, relpath)
1567
def path_prefix_key(path):
1568
"""Generate a prefix-order path key for path.
1570
This can be used to sort paths in the same way that walkdirs does.
1572
return (dirname(path) , path)
1575
def compare_paths_prefix_order(path_a, path_b):
1576
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1577
key_a = path_prefix_key(path_a)
1578
key_b = path_prefix_key(path_b)
1579
return cmp(key_a, key_b)
1582
_cached_user_encoding = None
1585
def get_user_encoding(use_cache=True):
1586
"""Find out what the preferred user encoding is.
1588
This is generally the encoding that is used for command line parameters
1589
and file contents. This may be different from the terminal encoding
1590
or the filesystem encoding.
1592
:param use_cache: Enable cache for detected encoding.
1593
(This parameter is turned on by default,
1594
and required only for selftesting)
1596
:return: A string defining the preferred user encoding
1598
global _cached_user_encoding
1599
if _cached_user_encoding is not None and use_cache:
1600
return _cached_user_encoding
1602
if sys.platform == 'darwin':
1603
# python locale.getpreferredencoding() always return
1604
# 'mac-roman' on darwin. That's a lie.
1605
sys.platform = 'posix'
1607
if os.environ.get('LANG', None) is None:
1608
# If LANG is not set, we end up with 'ascii', which is bad
1609
# ('mac-roman' is more than ascii), so we set a default which
1610
# will give us UTF-8 (which appears to work in all cases on
1611
# OSX). Users are still free to override LANG of course, as
1612
# long as it give us something meaningful. This work-around
1613
# *may* not be needed with python 3k and/or OSX 10.5, but will
1614
# work with them too -- vila 20080908
1615
os.environ['LANG'] = 'en_US.UTF-8'
1618
sys.platform = 'darwin'
1623
user_encoding = locale.getpreferredencoding()
1624
except locale.Error, e:
1625
sys.stderr.write('bzr: warning: %s\n'
1626
' Could not determine what text encoding to use.\n'
1627
' This error usually means your Python interpreter\n'
1628
' doesn\'t support the locale set by $LANG (%s)\n'
1629
" Continuing with ascii encoding.\n"
1630
% (e, os.environ.get('LANG')))
1631
user_encoding = 'ascii'
1633
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1634
# treat that as ASCII, and not support printing unicode characters to the
1637
# For python scripts run under vim, we get '', so also treat that as ASCII
1638
if user_encoding in (None, 'cp0', ''):
1639
user_encoding = 'ascii'
1643
codecs.lookup(user_encoding)
1645
sys.stderr.write('bzr: warning:'
1646
' unknown encoding %s.'
1647
' Continuing with ascii encoding.\n'
1650
user_encoding = 'ascii'
1653
_cached_user_encoding = user_encoding
1655
return user_encoding
1658
def get_host_name():
1659
"""Return the current unicode host name.
1661
This is meant to be used in place of socket.gethostname() because that
1662
behaves inconsistently on different platforms.
1664
if sys.platform == "win32":
1666
return win32utils.get_host_name()
1669
return socket.gethostname().decode(get_user_encoding())
1672
def recv_all(socket, bytes):
1673
"""Receive an exact number of bytes.
1675
Regular Socket.recv() may return less than the requested number of bytes,
1676
dependning on what's in the OS buffer. MSG_WAITALL is not available
1677
on all platforms, but this should work everywhere. This will return
1678
less than the requested amount if the remote end closes.
1680
This isn't optimized and is intended mostly for use in testing.
1683
while len(b) < bytes:
1684
new = until_no_eintr(socket.recv, bytes - len(b))
1691
def send_all(socket, bytes, report_activity=None):
1692
"""Send all bytes on a socket.
1694
Regular socket.sendall() can give socket error 10053 on Windows. This
1695
implementation sends no more than 64k at a time, which avoids this problem.
1697
:param report_activity: Call this as bytes are read, see
1698
Transport._report_activity
1701
for pos in xrange(0, len(bytes), chunk_size):
1702
block = bytes[pos:pos+chunk_size]
1703
if report_activity is not None:
1704
report_activity(len(block), 'write')
1705
until_no_eintr(socket.sendall, block)
1708
def dereference_path(path):
1709
"""Determine the real path to a file.
1711
All parent elements are dereferenced. But the file itself is not
1713
:param path: The original path. May be absolute or relative.
1714
:return: the real path *to* the file
1716
parent, base = os.path.split(path)
1717
# The pathjoin for '.' is a workaround for Python bug #1213894.
1718
# (initial path components aren't dereferenced)
1719
return pathjoin(realpath(pathjoin('.', parent)), base)
1722
def supports_mapi():
1723
"""Return True if we can use MAPI to launch a mail client."""
1724
return sys.platform == "win32"
1727
def resource_string(package, resource_name):
1728
"""Load a resource from a package and return it as a string.
1730
Note: Only packages that start with bzrlib are currently supported.
1732
This is designed to be a lightweight implementation of resource
1733
loading in a way which is API compatible with the same API from
1735
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1736
If and when pkg_resources becomes a standard library, this routine
1739
# Check package name is within bzrlib
1740
if package == "bzrlib":
1741
resource_relpath = resource_name
1742
elif package.startswith("bzrlib."):
1743
package = package[len("bzrlib."):].replace('.', os.sep)
1744
resource_relpath = pathjoin(package, resource_name)
1746
raise errors.BzrError('resource package %s not in bzrlib' % package)
1748
# Map the resource to a file and read its contents
1749
base = dirname(bzrlib.__file__)
1750
if getattr(sys, 'frozen', None): # bzr.exe
1751
base = abspath(pathjoin(base, '..', '..'))
1752
filename = pathjoin(base, resource_relpath)
1753
return open(filename, 'rU').read()
1756
def file_kind_from_stat_mode_thunk(mode):
1757
global file_kind_from_stat_mode
1758
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1760
from bzrlib._readdir_pyx import UTF8DirReader
1761
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1763
from bzrlib._readdir_py import (
1764
_kind_from_mode as file_kind_from_stat_mode
1766
return file_kind_from_stat_mode(mode)
1767
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1770
def file_kind(f, _lstat=os.lstat):
1772
return file_kind_from_stat_mode(_lstat(f).st_mode)
1774
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1775
raise errors.NoSuchFile(f)
472
"""Return a sequence of possible editor binaries for the current platform"""
473
e = _read_config_value("editor")
477
if os.name == "windows":
479
elif os.name == "posix":
1779
def until_no_eintr(f, *a, **kw):
1780
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1781
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
481
yield os.environ["EDITOR"]
486
def _run_editor(filename):
487
"""Try to execute an editor to edit the commit message. Returns True on success,
489
for e in _get_editor():
490
x = os.spawnvp(os.P_WAIT, e, (e, filename))
497
raise BzrError("Could not start any editor. Please specify $EDITOR or use ~/.bzr.conf/editor")
501
def get_text_message(infotext, ignoreline = "default"):
504
if ignoreline == "default":
505
ignoreline = "-- This line and the following will be ignored --"
1785
except (IOError, OSError), e:
1786
if e.errno == errno.EINTR:
1790
def re_compile_checked(re_string, flags=0, where=""):
1791
"""Return a compiled re, or raise a sensible error.
1793
This should only be used when compiling user-supplied REs.
1795
:param re_string: Text form of regular expression.
1796
:param flags: eg re.IGNORECASE
1797
:param where: Message explaining to the user the context where
1798
it occurred, eg 'log search filter'.
1800
# from https://bugs.launchpad.net/bzr/+bug/251352
508
tmp_fileno, msgfilename = tempfile.mkstemp()
509
msgfile = os.close(tmp_fileno)
510
if infotext is not None and infotext != "":
512
msgfile = file(msgfilename, "w")
513
msgfile.write("\n\n%s\n\n%s" % (ignoreline, infotext))
518
if not _run_editor(msgfilename):
523
lastline, nlines = 0, 0
524
for line in file(msgfilename, "r"):
525
stripped_line = line.strip()
526
# strip empty line before the log message starts
528
if stripped_line != "":
532
# check for the ignore line only if there
533
# is additional information at the end
534
if hasinfo and stripped_line == ignoreline:
537
# keep track of the last line that had some content
538
if stripped_line != "":
544
# delete empty lines at the end
546
# add a newline at the end, if needed
547
if not msg[-1].endswith("\n"):
548
return "%s%s" % ("".join(msg), "\n")
552
# delete the msg file in any case
553
try: os.unlink(msgfilename)
1802
re_obj = re.compile(re_string, flags)
1807
where = ' in ' + where
1808
# despite the name 'error' is a type
1809
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
1810
% (where, re_string, e))
1813
if sys.platform == "win32":
1816
return msvcrt.getch()
1821
fd = sys.stdin.fileno()
1822
settings = termios.tcgetattr(fd)
1825
ch = sys.stdin.read(1)
1827
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
added
removed
bzrlib/osutils.py
