1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
21
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
43
from tempfile import (
57
from bzrlib import symbol_versioning
60
# On win32, O_BINARY is used to indicate the file should
61
# be opened in binary mode, rather than text mode.
62
# On other platforms, O_BINARY doesn't exist, because
63
# they always open in binary mode, so it is okay to
64
# OR with 0 on those platforms
65
O_BINARY = getattr(os, 'O_BINARY', 0)
68
def make_readonly(filename):
69
"""Make a filename read-only."""
70
mod = os.lstat(filename).st_mode
71
if not stat.S_ISLNK(mod):
73
os.chmod(filename, mod)
76
def make_writable(filename):
77
mod = os.lstat(filename).st_mode
78
if not stat.S_ISLNK(mod):
80
os.chmod(filename, mod)
83
def minimum_path_selection(paths):
84
"""Return the smallset subset of paths which are outside paths.
86
:param paths: A container (and hence not None) of paths.
87
:return: A set of paths sufficient to include everything in paths via
88
is_inside_any, drawn from the paths parameter.
93
other_paths = paths.difference([path])
94
if not is_inside_any(other_paths, path):
95
# this is a top level path, we must check it.
96
search_paths.add(path)
104
"""Return a quoted filename filename
106
This previously used backslash quoting, but that works poorly on
108
# TODO: I'm not really sure this is the best format either.x
110
if _QUOTE_RE is None:
111
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
113
if _QUOTE_RE.search(f):
119
_directory_kind = 'directory'
122
stat.S_IFDIR:_directory_kind,
123
stat.S_IFCHR:'chardev',
124
stat.S_IFBLK:'block',
127
stat.S_IFLNK:'symlink',
128
stat.S_IFSOCK:'socket',
132
def file_kind_from_stat_mode(stat_mode, _formats=_formats, _unknown='unknown'):
133
"""Generate a file kind from a stat mode. This is used in walkdirs.
135
Its performance is critical: Do not mutate without careful benchmarking.
138
return _formats[stat_mode & 0170000]
143
def file_kind(f, _lstat=os.lstat, _mapper=file_kind_from_stat_mode):
145
return _mapper(_lstat(f).st_mode)
147
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
148
raise errors.NoSuchFile(f)
153
"""Return the current umask"""
154
# Assume that people aren't messing with the umask while running
155
# XXX: This is not thread safe, but there is no way to get the
156
# umask without setting it
164
_directory_kind: "/",
166
'tree-reference': '+',
170
def kind_marker(kind):
172
return _kind_marker_map[kind]
174
raise errors.BzrError('invalid file kind %r' % kind)
177
lexists = getattr(os.path, 'lexists', None)
181
stat = getattr(os, 'lstat', os.stat)
185
if e.errno == errno.ENOENT:
188
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
191
def fancy_rename(old, new, rename_func, unlink_func):
192
"""A fancy rename, when you don't have atomic rename.
194
:param old: The old path, to rename from
195
:param new: The new path, to rename to
196
:param rename_func: The potentially non-atomic rename function
197
:param unlink_func: A way to delete the target file if the full rename succeeds
200
# sftp rename doesn't allow overwriting, so play tricks:
201
base = os.path.basename(new)
202
dirname = os.path.dirname(new)
203
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
204
tmp_name = pathjoin(dirname, tmp_name)
206
# Rename the file out of the way, but keep track if it didn't exist
207
# We don't want to grab just any exception
208
# something like EACCES should prevent us from continuing
209
# The downside is that the rename_func has to throw an exception
210
# with an errno = ENOENT, or NoSuchFile
213
rename_func(new, tmp_name)
214
except (errors.NoSuchFile,), e:
217
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
218
# function raises an IOError with errno is None when a rename fails.
219
# This then gets caught here.
220
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
223
if (getattr(e, 'errno', None) is None
224
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
232
# This may throw an exception, in which case success will
234
rename_func(old, new)
236
except (IOError, OSError), e:
237
# source and target may be aliases of each other (e.g. on a
238
# case-insensitive filesystem), so we may have accidentally renamed
239
# source by when we tried to rename target
240
if not (file_existed and e.errno in (None, errno.ENOENT)):
244
# If the file used to exist, rename it back into place
245
# otherwise just delete it from the tmp location
247
unlink_func(tmp_name)
249
rename_func(tmp_name, new)
252
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
253
# choke on a Unicode string containing a relative path if
254
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
256
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
257
def _posix_abspath(path):
258
# jam 20060426 rather than encoding to fsencoding
259
# copy posixpath.abspath, but use os.getcwdu instead
260
if not posixpath.isabs(path):
261
path = posixpath.join(getcwd(), path)
262
return posixpath.normpath(path)
265
def _posix_realpath(path):
266
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
269
def _win32_fixdrive(path):
270
"""Force drive letters to be consistent.
272
win32 is inconsistent whether it returns lower or upper case
273
and even if it was consistent the user might type the other
274
so we force it to uppercase
275
running python.exe under cmd.exe return capital C:\\
276
running win32 python inside a cygwin shell returns lowercase c:\\
278
drive, path = _nt_splitdrive(path)
279
return drive.upper() + path
282
def _win32_abspath(path):
283
global _win32_abspath
284
if win32utils.winver == 'Windows 98':
285
_win32_abspath = _win98_abspath
287
_win32_abspath = _real_win32_abspath
288
return _win32_abspath(path)
291
def _real_win32_abspath(path):
292
# Real _nt_abspath doesn't have a problem with a unicode cwd
293
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
296
def _win98_abspath(path):
297
"""Return the absolute version of a path.
298
Windows 98 safe implementation (python reimplementation
299
of Win32 API function GetFullPathNameW)
304
# \\HOST\path => //HOST/path
305
# //HOST/path => //HOST/path
306
# path => C:/cwd/path
309
# check for absolute path
310
drive = _nt_splitdrive(path)[0]
311
if drive == '' and path[:2] not in('//','\\\\'):
313
# we cannot simply os.path.join cwd and path
314
# because os.path.join('C:','/path') produce '/path'
315
# and this is incorrect
316
if path[:1] in ('/','\\'):
317
cwd = _nt_splitdrive(cwd)[0]
319
path = cwd + '\\' + path
320
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
323
def _win32_realpath(path):
324
# Real _nt_realpath doesn't have a problem with a unicode cwd
325
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
328
def _win32_pathjoin(*args):
329
return _nt_join(*args).replace('\\', '/')
332
def _win32_normpath(path):
333
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
337
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
340
def _win32_mkdtemp(*args, **kwargs):
341
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
344
def _win32_rename(old, new):
345
"""We expect to be able to atomically replace 'new' with old.
347
On win32, if new exists, it must be moved out of the way first,
351
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
353
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
354
# If we try to rename a non-existant file onto cwd, we get
355
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
356
# if the old path doesn't exist, sometimes we get EACCES
357
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
363
return unicodedata.normalize('NFC', os.getcwdu())
366
# Default is to just use the python builtins, but these can be rebound on
367
# particular platforms.
368
abspath = _posix_abspath
369
realpath = _posix_realpath
370
pathjoin = os.path.join
371
normpath = os.path.normpath
374
dirname = os.path.dirname
375
basename = os.path.basename
376
split = os.path.split
377
splitext = os.path.splitext
378
# These were already imported into local scope
379
# mkdtemp = tempfile.mkdtemp
380
# rmtree = shutil.rmtree
382
MIN_ABS_PATHLENGTH = 1
385
if sys.platform == 'win32':
386
abspath = _win32_abspath
387
realpath = _win32_realpath
388
pathjoin = _win32_pathjoin
389
normpath = _win32_normpath
390
getcwd = _win32_getcwd
391
mkdtemp = _win32_mkdtemp
392
rename = _win32_rename
394
MIN_ABS_PATHLENGTH = 3
396
def _win32_delete_readonly(function, path, excinfo):
397
"""Error handler for shutil.rmtree function [for win32]
398
Helps to remove files and dirs marked as read-only.
400
exception = excinfo[1]
401
if function in (os.remove, os.rmdir) \
402
and isinstance(exception, OSError) \
403
and exception.errno == errno.EACCES:
409
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
410
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
411
return shutil.rmtree(path, ignore_errors, onerror)
412
elif sys.platform == 'darwin':
416
def get_terminal_encoding():
417
"""Find the best encoding for printing to the screen.
419
This attempts to check both sys.stdout and sys.stdin to see
420
what encoding they are in, and if that fails it falls back to
421
osutils.get_user_encoding().
422
The problem is that on Windows, locale.getpreferredencoding()
423
is not the same encoding as that used by the console:
424
http://mail.python.org/pipermail/python-list/2003-May/162357.html
426
On my standard US Windows XP, the preferred encoding is
427
cp1252, but the console is cp437
429
from bzrlib.trace import mutter
430
output_encoding = getattr(sys.stdout, 'encoding', None)
431
if not output_encoding:
432
input_encoding = getattr(sys.stdin, 'encoding', None)
433
if not input_encoding:
434
output_encoding = get_user_encoding()
435
mutter('encoding stdout as osutils.get_user_encoding() %r',
438
output_encoding = input_encoding
439
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
441
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
442
if output_encoding == 'cp0':
443
# invalid encoding (cp0 means 'no codepage' on Windows)
444
output_encoding = get_user_encoding()
445
mutter('cp0 is invalid encoding.'
446
' encoding stdout as osutils.get_user_encoding() %r',
450
codecs.lookup(output_encoding)
452
sys.stderr.write('bzr: warning:'
453
' unknown terminal encoding %s.\n'
454
' Using encoding %s instead.\n'
455
% (output_encoding, get_user_encoding())
457
output_encoding = get_user_encoding()
459
return output_encoding
462
def normalizepath(f):
463
if getattr(os.path, 'realpath', None) is not None:
467
[p,e] = os.path.split(f)
468
if e == "" or e == "." or e == "..":
471
return pathjoin(F(p), e)
475
"""True if f is an accessible directory."""
477
return S_ISDIR(os.lstat(f)[ST_MODE])
483
"""True if f is a regular file."""
485
return S_ISREG(os.lstat(f)[ST_MODE])
490
"""True if f is a symlink."""
492
return S_ISLNK(os.lstat(f)[ST_MODE])
496
def is_inside(dir, fname):
497
"""True if fname is inside dir.
499
The parameters should typically be passed to osutils.normpath first, so
500
that . and .. and repeated slashes are eliminated, and the separators
501
are canonical for the platform.
503
The empty string as a dir name is taken as top-of-tree and matches
506
# XXX: Most callers of this can actually do something smarter by
507
# looking at the inventory
517
return fname.startswith(dir)
520
def is_inside_any(dir_list, fname):
521
"""True if fname is inside any of given dirs."""
522
for dirname in dir_list:
523
if is_inside(dirname, fname):
528
def is_inside_or_parent_of_any(dir_list, fname):
529
"""True if fname is a child or a parent of any of the given files."""
530
for dirname in dir_list:
531
if is_inside(dirname, fname) or is_inside(fname, dirname):
536
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768):
537
"""Copy contents of one file to another.
539
The read_length can either be -1 to read to end-of-file (EOF) or
540
it can specify the maximum number of bytes to read.
542
The buff_size represents the maximum size for each read operation
543
performed on from_file.
545
:return: The number of bytes copied.
549
# read specified number of bytes
551
while read_length > 0:
552
num_bytes_to_read = min(read_length, buff_size)
554
block = from_file.read(num_bytes_to_read)
560
actual_bytes_read = len(block)
561
read_length -= actual_bytes_read
562
length += actual_bytes_read
566
block = from_file.read(buff_size)
575
def pump_string_file(bytes, file_handle, segment_size=None):
576
"""Write bytes to file_handle in many smaller writes.
578
:param bytes: The string to write.
579
:param file_handle: The file to write to.
581
# Write data in chunks rather than all at once, because very large
582
# writes fail on some platforms (e.g. Windows with SMB mounted
585
segment_size = 5242880 # 5MB
586
segments = range(len(bytes) / segment_size + 1)
587
write = file_handle.write
588
for segment_index in segments:
589
segment = buffer(bytes, segment_index * segment_size, segment_size)
593
def file_iterator(input_file, readsize=32768):
595
b = input_file.read(readsize)
602
"""Calculate the hexdigest of an open file.
604
The file cursor should be already at the start.
616
def sha_file_by_name(fname):
617
"""Calculate the SHA1 of a file by reading the full text"""
619
f = os.open(fname, os.O_RDONLY | O_BINARY)
622
b = os.read(f, 1<<16)
630
def sha_strings(strings, _factory=sha.new):
631
"""Return the sha-1 of concatenation of strings"""
633
map(s.update, strings)
637
def sha_string(f, _factory=sha.new):
638
return _factory(f).hexdigest()
641
def fingerprint_file(f):
643
return {'size': len(b),
644
'sha1': sha.new(b).hexdigest()}
647
def compare_files(a, b):
648
"""Returns true if equal in contents"""
659
def local_time_offset(t=None):
660
"""Return offset of local zone from GMT, either at present or at time t."""
663
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
664
return offset.days * 86400 + offset.seconds
666
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
668
def format_date(t, offset=0, timezone='original', date_fmt=None,
670
"""Return a formatted date string.
672
:param t: Seconds since the epoch.
673
:param offset: Timezone offset in seconds east of utc.
674
:param timezone: How to display the time: 'utc', 'original' for the
675
timezone specified by offset, or 'local' for the process's current
677
:param show_offset: Whether to append the timezone.
678
:param date_fmt: strftime format.
680
if timezone == 'utc':
683
elif timezone == 'original':
686
tt = time.gmtime(t + offset)
687
elif timezone == 'local':
688
tt = time.localtime(t)
689
offset = local_time_offset(t)
691
raise errors.UnsupportedTimezoneFormat(timezone)
693
date_fmt = "%a %Y-%m-%d %H:%M:%S"
695
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
698
# day of week depends on locale, so we do this ourself
699
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
700
return (time.strftime(date_fmt, tt) + offset_str)
703
def compact_date(when):
704
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
707
def format_delta(delta):
708
"""Get a nice looking string for a time delta.
710
:param delta: The time difference in seconds, can be positive or negative.
711
positive indicates time in the past, negative indicates time in the
712
future. (usually time.time() - stored_time)
713
:return: String formatted to show approximate resolution
719
direction = 'in the future'
723
if seconds < 90: # print seconds up to 90 seconds
725
return '%d second %s' % (seconds, direction,)
727
return '%d seconds %s' % (seconds, direction)
729
minutes = int(seconds / 60)
730
seconds -= 60 * minutes
735
if minutes < 90: # print minutes, seconds up to 90 minutes
737
return '%d minute, %d second%s %s' % (
738
minutes, seconds, plural_seconds, direction)
740
return '%d minutes, %d second%s %s' % (
741
minutes, seconds, plural_seconds, direction)
743
hours = int(minutes / 60)
744
minutes -= 60 * hours
751
return '%d hour, %d minute%s %s' % (hours, minutes,
752
plural_minutes, direction)
753
return '%d hours, %d minute%s %s' % (hours, minutes,
754
plural_minutes, direction)
757
"""Return size of given open file."""
758
return os.fstat(f.fileno())[ST_SIZE]
761
# Define rand_bytes based on platform.
763
# Python 2.4 and later have os.urandom,
764
# but it doesn't work on some arches
766
rand_bytes = os.urandom
767
except (NotImplementedError, AttributeError):
768
# If python doesn't have os.urandom, or it doesn't work,
769
# then try to first pull random data from /dev/urandom
771
rand_bytes = file('/dev/urandom', 'rb').read
772
# Otherwise, use this hack as a last resort
773
except (IOError, OSError):
774
# not well seeded, but better than nothing
779
s += chr(random.randint(0, 255))
784
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
786
"""Return a random string of num alphanumeric characters
788
The result only contains lowercase chars because it may be used on
789
case-insensitive filesystems.
792
for raw_byte in rand_bytes(num):
793
s += ALNUM[ord(raw_byte) % 36]
797
## TODO: We could later have path objects that remember their list
798
## decomposition (might be too tricksy though.)
801
"""Turn string into list of parts."""
802
# split on either delimiter because people might use either on
804
ps = re.split(r'[\\/]', p)
809
raise errors.BzrError("sorry, %r not allowed in path" % f)
810
elif (f == '.') or (f == ''):
818
if (f == '..') or (f is None) or (f == ''):
819
raise errors.BzrError("sorry, %r not allowed in path" % f)
824
"""Split s into lines, but without removing the newline characters."""
825
lines = s.split('\n')
826
result = [line + '\n' for line in lines[:-1]]
828
result.append(lines[-1])
832
def hardlinks_good():
833
return sys.platform not in ('win32', 'cygwin', 'darwin')
836
def link_or_copy(src, dest):
837
"""Hardlink a file, or copy it if it can't be hardlinked."""
838
if not hardlinks_good():
839
shutil.copyfile(src, dest)
843
except (OSError, IOError), e:
844
if e.errno != errno.EXDEV:
846
shutil.copyfile(src, dest)
849
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
850
# Forgiveness than Permission (EAFP) because:
851
# - root can damage a solaris file system by using unlink,
852
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
853
# EACCES, OSX: EPERM) when invoked on a directory.
854
def delete_any(path):
855
"""Delete a file or directory."""
856
if isdir(path): # Takes care of symlinks
863
if getattr(os, 'symlink', None) is not None:
870
if getattr(os, 'link', None) is not None:
876
def host_os_dereferences_symlinks():
877
return (has_symlinks()
878
and sys.platform not in ('cygwin', 'win32'))
881
def contains_whitespace(s):
882
"""True if there are any whitespace characters in s."""
883
# string.whitespace can include '\xa0' in certain locales, because it is
884
# considered "non-breaking-space" as part of ISO-8859-1. But it
885
# 1) Isn't a breaking whitespace
886
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
888
# 3) '\xa0' isn't unicode safe since it is >128.
890
# This should *not* be a unicode set of characters in case the source
891
# string is not a Unicode string. We can auto-up-cast the characters since
892
# they are ascii, but we don't want to auto-up-cast the string in case it
894
for ch in ' \t\n\r\v\f':
901
def contains_linebreaks(s):
902
"""True if there is any vertical whitespace in s."""
910
def relpath(base, path):
911
"""Return path relative to base, or raise exception.
913
The path may be either an absolute path or a path relative to the
914
current working directory.
916
os.path.commonprefix (python2.4) has a bad bug that it works just
917
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
921
if len(base) < MIN_ABS_PATHLENGTH:
922
# must have space for e.g. a drive letter
923
raise ValueError('%r is too short to calculate a relative path'
930
while len(head) >= len(base):
933
head, tail = os.path.split(head)
937
raise errors.PathNotChild(rp, base)
945
def safe_unicode(unicode_or_utf8_string):
946
"""Coerce unicode_or_utf8_string into unicode.
948
If it is unicode, it is returned.
949
Otherwise it is decoded from utf-8. If a decoding error
950
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
951
as a BzrBadParameter exception.
953
if isinstance(unicode_or_utf8_string, unicode):
954
return unicode_or_utf8_string
956
return unicode_or_utf8_string.decode('utf8')
957
except UnicodeDecodeError:
958
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
961
def safe_utf8(unicode_or_utf8_string):
962
"""Coerce unicode_or_utf8_string to a utf8 string.
964
If it is a str, it is returned.
965
If it is Unicode, it is encoded into a utf-8 string.
967
if isinstance(unicode_or_utf8_string, str):
968
# TODO: jam 20070209 This is overkill, and probably has an impact on
969
# performance if we are dealing with lots of apis that want a
972
# Make sure it is a valid utf-8 string
973
unicode_or_utf8_string.decode('utf-8')
974
except UnicodeDecodeError:
975
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
976
return unicode_or_utf8_string
977
return unicode_or_utf8_string.encode('utf-8')
980
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
981
' Revision id generators should be creating utf8'
985
def safe_revision_id(unicode_or_utf8_string, warn=True):
986
"""Revision ids should now be utf8, but at one point they were unicode.
988
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
990
:param warn: Functions that are sanitizing user data can set warn=False
991
:return: None or a utf8 revision id.
993
if (unicode_or_utf8_string is None
994
or unicode_or_utf8_string.__class__ == str):
995
return unicode_or_utf8_string
997
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
999
return cache_utf8.encode(unicode_or_utf8_string)
1002
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1003
' generators should be creating utf8 file ids.')
1006
def safe_file_id(unicode_or_utf8_string, warn=True):
1007
"""File ids should now be utf8, but at one point they were unicode.
1009
This is the same as safe_utf8, except it uses the cached encode functions
1010
to save a little bit of performance.
1012
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1014
:param warn: Functions that are sanitizing user data can set warn=False
1015
:return: None or a utf8 file id.
1017
if (unicode_or_utf8_string is None
1018
or unicode_or_utf8_string.__class__ == str):
1019
return unicode_or_utf8_string
1021
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1023
return cache_utf8.encode(unicode_or_utf8_string)
1026
_platform_normalizes_filenames = False
1027
if sys.platform == 'darwin':
1028
_platform_normalizes_filenames = True
1031
def normalizes_filenames():
1032
"""Return True if this platform normalizes unicode filenames.
1034
Mac OSX does, Windows/Linux do not.
1036
return _platform_normalizes_filenames
1039
def _accessible_normalized_filename(path):
1040
"""Get the unicode normalized path, and if you can access the file.
1042
On platforms where the system normalizes filenames (Mac OSX),
1043
you can access a file by any path which will normalize correctly.
1044
On platforms where the system does not normalize filenames
1045
(Windows, Linux), you have to access a file by its exact path.
1047
Internally, bzr only supports NFC normalization, since that is
1048
the standard for XML documents.
1050
So return the normalized path, and a flag indicating if the file
1051
can be accessed by that path.
1054
return unicodedata.normalize('NFC', unicode(path)), True
1057
def _inaccessible_normalized_filename(path):
1058
__doc__ = _accessible_normalized_filename.__doc__
1060
normalized = unicodedata.normalize('NFC', unicode(path))
1061
return normalized, normalized == path
1064
if _platform_normalizes_filenames:
1065
normalized_filename = _accessible_normalized_filename
1067
normalized_filename = _inaccessible_normalized_filename
1070
def terminal_width():
1071
"""Return estimated terminal width."""
1072
if sys.platform == 'win32':
1073
return win32utils.get_console_size()[0]
1076
import struct, fcntl, termios
1077
s = struct.pack('HHHH', 0, 0, 0, 0)
1078
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1079
width = struct.unpack('HHHH', x)[1]
1084
width = int(os.environ['COLUMNS'])
1093
def supports_executable():
1094
return sys.platform != "win32"
1097
def supports_posix_readonly():
1098
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1100
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1101
directory controls creation/deletion, etc.
1103
And under win32, readonly means that the directory itself cannot be
1104
deleted. The contents of a readonly directory can be changed, unlike POSIX
1105
where files in readonly directories cannot be added, deleted or renamed.
1107
return sys.platform != "win32"
1110
def set_or_unset_env(env_variable, value):
1111
"""Modify the environment, setting or removing the env_variable.
1113
:param env_variable: The environment variable in question
1114
:param value: The value to set the environment to. If None, then
1115
the variable will be removed.
1116
:return: The original value of the environment variable.
1118
orig_val = os.environ.get(env_variable)
1120
if orig_val is not None:
1121
del os.environ[env_variable]
1123
if isinstance(value, unicode):
1124
value = value.encode(get_user_encoding())
1125
os.environ[env_variable] = value
1129
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1132
def check_legal_path(path):
1133
"""Check whether the supplied path is legal.
1134
This is only required on Windows, so we don't test on other platforms
1137
if sys.platform != "win32":
1139
if _validWin32PathRE.match(path) is None:
1140
raise errors.IllegalPath(path)
1143
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1145
def _is_error_enotdir(e):
1146
"""Check if this exception represents ENOTDIR.
1148
Unfortunately, python is very inconsistent about the exception
1149
here. The cases are:
1150
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1151
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1152
which is the windows error code.
1153
3) Windows, Python2.5 uses errno == EINVAL and
1154
winerror == ERROR_DIRECTORY
1156
:param e: An Exception object (expected to be OSError with an errno
1157
attribute, but we should be able to cope with anything)
1158
:return: True if this represents an ENOTDIR error. False otherwise.
1160
en = getattr(e, 'errno', None)
1161
if (en == errno.ENOTDIR
1162
or (sys.platform == 'win32'
1163
and (en == _WIN32_ERROR_DIRECTORY
1164
or (en == errno.EINVAL
1165
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1171
def walkdirs(top, prefix=""):
1172
"""Yield data about all the directories in a tree.
1174
This yields all the data about the contents of a directory at a time.
1175
After each directory has been yielded, if the caller has mutated the list
1176
to exclude some directories, they are then not descended into.
1178
The data yielded is of the form:
1179
((directory-relpath, directory-path-from-top),
1180
[(relpath, basename, kind, lstat, path-from-top), ...]),
1181
- directory-relpath is the relative path of the directory being returned
1182
with respect to top. prefix is prepended to this.
1183
- directory-path-from-root is the path including top for this directory.
1184
It is suitable for use with os functions.
1185
- relpath is the relative path within the subtree being walked.
1186
- basename is the basename of the path
1187
- kind is the kind of the file now. If unknown then the file is not
1188
present within the tree - but it may be recorded as versioned. See
1190
- lstat is the stat data *if* the file was statted.
1191
- planned, not implemented:
1192
path_from_tree_root is the path from the root of the tree.
1194
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1195
allows one to walk a subtree but get paths that are relative to a tree
1197
:return: an iterator over the dirs.
1199
#TODO there is a bit of a smell where the results of the directory-
1200
# summary in this, and the path from the root, may not agree
1201
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1202
# potentially confusing output. We should make this more robust - but
1203
# not at a speed cost. RBC 20060731
1205
_directory = _directory_kind
1206
_listdir = os.listdir
1207
_kind_from_mode = _formats.get
1208
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1210
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1211
relroot, _, _, _, top = pending.pop()
1213
relprefix = relroot + u'/'
1216
top_slash = top + u'/'
1219
append = dirblock.append
1221
names = sorted(_listdir(top))
1223
if not _is_error_enotdir(e):
1227
abspath = top_slash + name
1228
statvalue = _lstat(abspath)
1229
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1230
append((relprefix + name, name, kind, statvalue, abspath))
1231
yield (relroot, top), dirblock
1233
# push the user specified dirs from dirblock
1234
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1237
_real_walkdirs_utf8 = None
1239
def _walkdirs_utf8(top, prefix=""):
1240
"""Yield data about all the directories in a tree.
1242
This yields the same information as walkdirs() only each entry is yielded
1243
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1244
are returned as exact byte-strings.
1246
:return: yields a tuple of (dir_info, [file_info])
1247
dir_info is (utf8_relpath, path-from-top)
1248
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1249
if top is an absolute path, path-from-top is also an absolute path.
1250
path-from-top might be unicode or utf8, but it is the correct path to
1251
pass to os functions to affect the file in question. (such as os.lstat)
1253
global _real_walkdirs_utf8
1254
if _real_walkdirs_utf8 is None:
1255
fs_encoding = _fs_enc.upper()
1256
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1257
# Win98 doesn't have unicode apis like FindFirstFileW
1258
# TODO: We possibly could support Win98 by falling back to the
1259
# original FindFirstFile, and using TCHAR instead of WCHAR,
1260
# but that gets a bit tricky, and requires custom compiling
1263
from bzrlib._walkdirs_win32 import _walkdirs_utf8_win32_find_file
1265
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1267
_real_walkdirs_utf8 = _walkdirs_utf8_win32_find_file
1268
elif fs_encoding not in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1269
# ANSI_X3.4-1968 is a form of ASCII
1270
_real_walkdirs_utf8 = _walkdirs_unicode_to_utf8
1272
_real_walkdirs_utf8 = _walkdirs_fs_utf8
1273
return _real_walkdirs_utf8(top, prefix=prefix)
1276
def _walkdirs_fs_utf8(top, prefix=""):
1277
"""See _walkdirs_utf8.
1279
This sub-function is called when we know the filesystem is already in utf8
1280
encoding. So we don't need to transcode filenames.
1283
_directory = _directory_kind
1284
# Use C accelerated directory listing.
1285
_listdir = _read_dir
1286
_kind_from_mode = _formats.get
1288
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1289
# But we don't actually uses 1-3 in pending, so set them to None
1290
pending = [(safe_utf8(prefix), None, None, None, safe_utf8(top))]
1292
relroot, _, _, _, top = pending.pop()
1294
relprefix = relroot + '/'
1297
top_slash = top + '/'
1300
append = dirblock.append
1301
# read_dir supplies in should-stat order.
1302
for _, name in sorted(_listdir(top)):
1303
abspath = top_slash + name
1304
statvalue = _lstat(abspath)
1305
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1306
append((relprefix + name, name, kind, statvalue, abspath))
1308
yield (relroot, top), dirblock
1310
# push the user specified dirs from dirblock
1311
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1314
def _walkdirs_unicode_to_utf8(top, prefix=""):
1315
"""See _walkdirs_utf8
1317
Because Win32 has a Unicode api, all of the 'path-from-top' entries will be
1319
This is currently the fallback code path when the filesystem encoding is
1320
not UTF-8. It may be better to implement an alternative so that we can
1321
safely handle paths that are not properly decodable in the current
1324
_utf8_encode = codecs.getencoder('utf8')
1326
_directory = _directory_kind
1327
_listdir = os.listdir
1328
_kind_from_mode = _formats.get
1330
pending = [(safe_utf8(prefix), None, None, None, safe_unicode(top))]
1332
relroot, _, _, _, top = pending.pop()
1334
relprefix = relroot + '/'
1337
top_slash = top + u'/'
1340
append = dirblock.append
1341
for name in sorted(_listdir(top)):
1342
name_utf8 = _utf8_encode(name)[0]
1343
abspath = top_slash + name
1344
statvalue = _lstat(abspath)
1345
kind = _kind_from_mode(statvalue.st_mode & 0170000, 'unknown')
1346
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1347
yield (relroot, top), dirblock
1349
# push the user specified dirs from dirblock
1350
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1353
def copy_tree(from_path, to_path, handlers={}):
1354
"""Copy all of the entries in from_path into to_path.
1356
:param from_path: The base directory to copy.
1357
:param to_path: The target directory. If it does not exist, it will
1359
:param handlers: A dictionary of functions, which takes a source and
1360
destinations for files, directories, etc.
1361
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1362
'file', 'directory', and 'symlink' should always exist.
1363
If they are missing, they will be replaced with 'os.mkdir()',
1364
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1366
# Now, just copy the existing cached tree to the new location
1367
# We use a cheap trick here.
1368
# Absolute paths are prefixed with the first parameter
1369
# relative paths are prefixed with the second.
1370
# So we can get both the source and target returned
1371
# without any extra work.
1373
def copy_dir(source, dest):
1376
def copy_link(source, dest):
1377
"""Copy the contents of a symlink"""
1378
link_to = os.readlink(source)
1379
os.symlink(link_to, dest)
1381
real_handlers = {'file':shutil.copy2,
1382
'symlink':copy_link,
1383
'directory':copy_dir,
1385
real_handlers.update(handlers)
1387
if not os.path.exists(to_path):
1388
real_handlers['directory'](from_path, to_path)
1390
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1391
for relpath, name, kind, st, abspath in entries:
1392
real_handlers[kind](abspath, relpath)
1395
def path_prefix_key(path):
1396
"""Generate a prefix-order path key for path.
1398
This can be used to sort paths in the same way that walkdirs does.
1400
return (dirname(path) , path)
1403
def compare_paths_prefix_order(path_a, path_b):
1404
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1405
key_a = path_prefix_key(path_a)
1406
key_b = path_prefix_key(path_b)
1407
return cmp(key_a, key_b)
1410
_cached_user_encoding = None
1413
def get_user_encoding(use_cache=True):
1414
"""Find out what the preferred user encoding is.
1416
This is generally the encoding that is used for command line parameters
1417
and file contents. This may be different from the terminal encoding
1418
or the filesystem encoding.
1420
:param use_cache: Enable cache for detected encoding.
1421
(This parameter is turned on by default,
1422
and required only for selftesting)
1424
:return: A string defining the preferred user encoding
1426
global _cached_user_encoding
1427
if _cached_user_encoding is not None and use_cache:
1428
return _cached_user_encoding
1430
if sys.platform == 'darwin':
1431
# work around egregious python 2.4 bug
1432
sys.platform = 'posix'
1436
sys.platform = 'darwin'
1441
user_encoding = locale.getpreferredencoding()
1442
except locale.Error, e:
1443
sys.stderr.write('bzr: warning: %s\n'
1444
' Could not determine what text encoding to use.\n'
1445
' This error usually means your Python interpreter\n'
1446
' doesn\'t support the locale set by $LANG (%s)\n'
1447
" Continuing with ascii encoding.\n"
1448
% (e, os.environ.get('LANG')))
1449
user_encoding = 'ascii'
1451
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1452
# treat that as ASCII, and not support printing unicode characters to the
1455
# For python scripts run under vim, we get '', so also treat that as ASCII
1456
if user_encoding in (None, 'cp0', ''):
1457
user_encoding = 'ascii'
1461
codecs.lookup(user_encoding)
1463
sys.stderr.write('bzr: warning:'
1464
' unknown encoding %s.'
1465
' Continuing with ascii encoding.\n'
1468
user_encoding = 'ascii'
1471
_cached_user_encoding = user_encoding
1473
return user_encoding
1476
def get_host_name():
1477
"""Return the current unicode host name.
1479
This is meant to be used in place of socket.gethostname() because that
1480
behaves inconsistently on different platforms.
1482
if sys.platform == "win32":
1484
return win32utils.get_host_name()
1487
return socket.gethostname().decode(get_user_encoding())
1490
def recv_all(socket, bytes):
1491
"""Receive an exact number of bytes.
1493
Regular Socket.recv() may return less than the requested number of bytes,
1494
dependning on what's in the OS buffer. MSG_WAITALL is not available
1495
on all platforms, but this should work everywhere. This will return
1496
less than the requested amount if the remote end closes.
1498
This isn't optimized and is intended mostly for use in testing.
1501
while len(b) < bytes:
1502
new = socket.recv(bytes - len(b))
1509
def send_all(socket, bytes):
1510
"""Send all bytes on a socket.
1512
Regular socket.sendall() can give socket error 10053 on Windows. This
1513
implementation sends no more than 64k at a time, which avoids this problem.
1516
for pos in xrange(0, len(bytes), chunk_size):
1517
socket.sendall(bytes[pos:pos+chunk_size])
1520
def dereference_path(path):
1521
"""Determine the real path to a file.
1523
All parent elements are dereferenced. But the file itself is not
1525
:param path: The original path. May be absolute or relative.
1526
:return: the real path *to* the file
1528
parent, base = os.path.split(path)
1529
# The pathjoin for '.' is a workaround for Python bug #1213894.
1530
# (initial path components aren't dereferenced)
1531
return pathjoin(realpath(pathjoin('.', parent)), base)
1534
def supports_mapi():
1535
"""Return True if we can use MAPI to launch a mail client."""
1536
return sys.platform == "win32"
1539
def resource_string(package, resource_name):
1540
"""Load a resource from a package and return it as a string.
1542
Note: Only packages that start with bzrlib are currently supported.
1544
This is designed to be a lightweight implementation of resource
1545
loading in a way which is API compatible with the same API from
1547
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1548
If and when pkg_resources becomes a standard library, this routine
1551
# Check package name is within bzrlib
1552
if package == "bzrlib":
1553
resource_relpath = resource_name
1554
elif package.startswith("bzrlib."):
1555
package = package[len("bzrlib."):].replace('.', os.sep)
1556
resource_relpath = pathjoin(package, resource_name)
1558
raise errors.BzrError('resource package %s not in bzrlib' % package)
1560
# Map the resource to a file and read its contents
1561
base = dirname(bzrlib.__file__)
1562
if getattr(sys, 'frozen', None): # bzr.exe
1563
base = abspath(pathjoin(base, '..', '..'))
1564
filename = pathjoin(base, resource_relpath)
1565
return open(filename, 'rU').read()
1569
from bzrlib._readdir_pyx import read_dir as _read_dir
1571
from bzrlib._readdir_py import read_dir as _read_dir
added
removed
bzrlib/osutils.py
