1
# Copyright (C) 2005-2010 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25
from bzrlib.lazy_import import lazy_import
26
lazy_import(globals(), """
27
from datetime import datetime
31
# We need to import both shutil and rmtree as we export the later on posix
32
# and need the former on windows
34
from shutil import rmtree
37
# We need to import both tempfile and mkdtemp as we export the later on posix
38
# and need the former on windows
40
from tempfile import mkdtemp
51
from bzrlib.symbol_versioning import (
56
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
58
if sys.version_info < (2, 5):
59
import md5 as _mod_md5
61
import sha as _mod_sha
71
from bzrlib import symbol_versioning
74
# Cross platform wall-clock time functionality with decent resolution.
75
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
76
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
77
# synchronized with ``time.time()``, this is only meant to be used to find
78
# delta times by subtracting from another call to this function.
79
timer_func = time.time
80
if sys.platform == 'win32':
81
timer_func = time.clock
83
# On win32, O_BINARY is used to indicate the file should
84
# be opened in binary mode, rather than text mode.
85
# On other platforms, O_BINARY doesn't exist, because
86
# they always open in binary mode, so it is okay to
87
# OR with 0 on those platforms.
88
# O_NOINHERIT and O_TEXT exists only on win32 too.
89
O_BINARY = getattr(os, 'O_BINARY', 0)
90
O_TEXT = getattr(os, 'O_TEXT', 0)
91
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
94
def get_unicode_argv():
96
user_encoding = get_user_encoding()
97
return [a.decode(user_encoding) for a in sys.argv[1:]]
98
except UnicodeDecodeError:
99
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
103
def make_readonly(filename):
104
"""Make a filename read-only."""
105
mod = os.lstat(filename).st_mode
106
if not stat.S_ISLNK(mod):
108
os.chmod(filename, mod)
111
def make_writable(filename):
112
mod = os.lstat(filename).st_mode
113
if not stat.S_ISLNK(mod):
115
os.chmod(filename, mod)
118
def minimum_path_selection(paths):
119
"""Return the smallset subset of paths which are outside paths.
121
:param paths: A container (and hence not None) of paths.
122
:return: A set of paths sufficient to include everything in paths via
123
is_inside, drawn from the paths parameter.
129
return path.split('/')
130
sorted_paths = sorted(list(paths), key=sort_key)
132
search_paths = [sorted_paths[0]]
133
for path in sorted_paths[1:]:
134
if not is_inside(search_paths[-1], path):
135
# This path is unique, add it
136
search_paths.append(path)
138
return set(search_paths)
145
"""Return a quoted filename filename
147
This previously used backslash quoting, but that works poorly on
149
# TODO: I'm not really sure this is the best format either.x
151
if _QUOTE_RE is None:
152
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
154
if _QUOTE_RE.search(f):
160
_directory_kind = 'directory'
163
"""Return the current umask"""
164
# Assume that people aren't messing with the umask while running
165
# XXX: This is not thread safe, but there is no way to get the
166
# umask without setting it
174
_directory_kind: "/",
176
'tree-reference': '+',
180
def kind_marker(kind):
182
return _kind_marker_map[kind]
184
# Slightly faster than using .get(, '') when the common case is that
189
lexists = getattr(os.path, 'lexists', None)
193
stat = getattr(os, 'lstat', os.stat)
197
if e.errno == errno.ENOENT:
200
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
203
def fancy_rename(old, new, rename_func, unlink_func):
204
"""A fancy rename, when you don't have atomic rename.
206
:param old: The old path, to rename from
207
:param new: The new path, to rename to
208
:param rename_func: The potentially non-atomic rename function
209
:param unlink_func: A way to delete the target file if the full rename
212
# sftp rename doesn't allow overwriting, so play tricks:
213
base = os.path.basename(new)
214
dirname = os.path.dirname(new)
215
# callers use different encodings for the paths so the following MUST
216
# respect that. We rely on python upcasting to unicode if new is unicode
217
# and keeping a str if not.
218
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
219
os.getpid(), rand_chars(10))
220
tmp_name = pathjoin(dirname, tmp_name)
222
# Rename the file out of the way, but keep track if it didn't exist
223
# We don't want to grab just any exception
224
# something like EACCES should prevent us from continuing
225
# The downside is that the rename_func has to throw an exception
226
# with an errno = ENOENT, or NoSuchFile
229
rename_func(new, tmp_name)
230
except (errors.NoSuchFile,), e:
233
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
234
# function raises an IOError with errno is None when a rename fails.
235
# This then gets caught here.
236
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
239
if (getattr(e, 'errno', None) is None
240
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
249
# This may throw an exception, in which case success will
251
rename_func(old, new)
253
except (IOError, OSError), e:
254
# source and target may be aliases of each other (e.g. on a
255
# case-insensitive filesystem), so we may have accidentally renamed
256
# source by when we tried to rename target
257
failure_exc = sys.exc_info()
258
if (file_existed and e.errno in (None, errno.ENOENT)
259
and old.lower() == new.lower()):
260
# source and target are the same file on a case-insensitive
261
# filesystem, so we don't generate an exception
265
# If the file used to exist, rename it back into place
266
# otherwise just delete it from the tmp location
268
unlink_func(tmp_name)
270
rename_func(tmp_name, new)
271
if failure_exc is not None:
272
raise failure_exc[0], failure_exc[1], failure_exc[2]
275
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
276
# choke on a Unicode string containing a relative path if
277
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
279
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
280
def _posix_abspath(path):
281
# jam 20060426 rather than encoding to fsencoding
282
# copy posixpath.abspath, but use os.getcwdu instead
283
if not posixpath.isabs(path):
284
path = posixpath.join(getcwd(), path)
285
return posixpath.normpath(path)
288
def _posix_realpath(path):
289
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
292
def _win32_fixdrive(path):
293
"""Force drive letters to be consistent.
295
win32 is inconsistent whether it returns lower or upper case
296
and even if it was consistent the user might type the other
297
so we force it to uppercase
298
running python.exe under cmd.exe return capital C:\\
299
running win32 python inside a cygwin shell returns lowercase c:\\
301
drive, path = ntpath.splitdrive(path)
302
return drive.upper() + path
305
def _win32_abspath(path):
306
# Real ntpath.abspath doesn't have a problem with a unicode cwd
307
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
310
def _win98_abspath(path):
311
"""Return the absolute version of a path.
312
Windows 98 safe implementation (python reimplementation
313
of Win32 API function GetFullPathNameW)
318
# \\HOST\path => //HOST/path
319
# //HOST/path => //HOST/path
320
# path => C:/cwd/path
323
# check for absolute path
324
drive = ntpath.splitdrive(path)[0]
325
if drive == '' and path[:2] not in('//','\\\\'):
327
# we cannot simply os.path.join cwd and path
328
# because os.path.join('C:','/path') produce '/path'
329
# and this is incorrect
330
if path[:1] in ('/','\\'):
331
cwd = ntpath.splitdrive(cwd)[0]
333
path = cwd + '\\' + path
334
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
337
def _win32_realpath(path):
338
# Real ntpath.realpath doesn't have a problem with a unicode cwd
339
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
342
def _win32_pathjoin(*args):
343
return ntpath.join(*args).replace('\\', '/')
346
def _win32_normpath(path):
347
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
351
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
354
def _win32_mkdtemp(*args, **kwargs):
355
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
358
def _win32_rename(old, new):
359
"""We expect to be able to atomically replace 'new' with old.
361
On win32, if new exists, it must be moved out of the way first,
365
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
367
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
368
# If we try to rename a non-existant file onto cwd, we get
369
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
370
# if the old path doesn't exist, sometimes we get EACCES
371
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
377
return unicodedata.normalize('NFC', os.getcwdu())
380
# Default is to just use the python builtins, but these can be rebound on
381
# particular platforms.
382
abspath = _posix_abspath
383
realpath = _posix_realpath
384
pathjoin = os.path.join
385
normpath = os.path.normpath
388
dirname = os.path.dirname
389
basename = os.path.basename
390
split = os.path.split
391
splitext = os.path.splitext
392
# These were already lazily imported into local scope
393
# mkdtemp = tempfile.mkdtemp
394
# rmtree = shutil.rmtree
396
MIN_ABS_PATHLENGTH = 1
399
if sys.platform == 'win32':
400
if win32utils.winver == 'Windows 98':
401
abspath = _win98_abspath
403
abspath = _win32_abspath
404
realpath = _win32_realpath
405
pathjoin = _win32_pathjoin
406
normpath = _win32_normpath
407
getcwd = _win32_getcwd
408
mkdtemp = _win32_mkdtemp
409
rename = _win32_rename
411
MIN_ABS_PATHLENGTH = 3
413
def _win32_delete_readonly(function, path, excinfo):
414
"""Error handler for shutil.rmtree function [for win32]
415
Helps to remove files and dirs marked as read-only.
417
exception = excinfo[1]
418
if function in (os.remove, os.rmdir) \
419
and isinstance(exception, OSError) \
420
and exception.errno == errno.EACCES:
426
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
427
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
428
return shutil.rmtree(path, ignore_errors, onerror)
430
f = win32utils.get_unicode_argv # special function or None
434
elif sys.platform == 'darwin':
438
def get_terminal_encoding(trace=False):
439
"""Find the best encoding for printing to the screen.
441
This attempts to check both sys.stdout and sys.stdin to see
442
what encoding they are in, and if that fails it falls back to
443
osutils.get_user_encoding().
444
The problem is that on Windows, locale.getpreferredencoding()
445
is not the same encoding as that used by the console:
446
http://mail.python.org/pipermail/python-list/2003-May/162357.html
448
On my standard US Windows XP, the preferred encoding is
449
cp1252, but the console is cp437
451
:param trace: If True trace the selected encoding via mutter().
453
from bzrlib.trace import mutter
454
output_encoding = getattr(sys.stdout, 'encoding', None)
455
if not output_encoding:
456
input_encoding = getattr(sys.stdin, 'encoding', None)
457
if not input_encoding:
458
output_encoding = get_user_encoding()
460
mutter('encoding stdout as osutils.get_user_encoding() %r',
463
output_encoding = input_encoding
465
mutter('encoding stdout as sys.stdin encoding %r',
469
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
470
if output_encoding == 'cp0':
471
# invalid encoding (cp0 means 'no codepage' on Windows)
472
output_encoding = get_user_encoding()
474
mutter('cp0 is invalid encoding.'
475
' encoding stdout as osutils.get_user_encoding() %r',
479
codecs.lookup(output_encoding)
481
sys.stderr.write('bzr: warning:'
482
' unknown terminal encoding %s.\n'
483
' Using encoding %s instead.\n'
484
% (output_encoding, get_user_encoding())
486
output_encoding = get_user_encoding()
488
return output_encoding
491
def normalizepath(f):
492
if getattr(os.path, 'realpath', None) is not None:
496
[p,e] = os.path.split(f)
497
if e == "" or e == "." or e == "..":
500
return pathjoin(F(p), e)
504
"""True if f is an accessible directory."""
506
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
512
"""True if f is a regular file."""
514
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
519
"""True if f is a symlink."""
521
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
525
def is_inside(dir, fname):
526
"""True if fname is inside dir.
528
The parameters should typically be passed to osutils.normpath first, so
529
that . and .. and repeated slashes are eliminated, and the separators
530
are canonical for the platform.
532
The empty string as a dir name is taken as top-of-tree and matches
535
# XXX: Most callers of this can actually do something smarter by
536
# looking at the inventory
546
return fname.startswith(dir)
549
def is_inside_any(dir_list, fname):
550
"""True if fname is inside any of given dirs."""
551
for dirname in dir_list:
552
if is_inside(dirname, fname):
557
def is_inside_or_parent_of_any(dir_list, fname):
558
"""True if fname is a child or a parent of any of the given files."""
559
for dirname in dir_list:
560
if is_inside(dirname, fname) or is_inside(fname, dirname):
565
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
566
report_activity=None, direction='read'):
567
"""Copy contents of one file to another.
569
The read_length can either be -1 to read to end-of-file (EOF) or
570
it can specify the maximum number of bytes to read.
572
The buff_size represents the maximum size for each read operation
573
performed on from_file.
575
:param report_activity: Call this as bytes are read, see
576
Transport._report_activity
577
:param direction: Will be passed to report_activity
579
:return: The number of bytes copied.
583
# read specified number of bytes
585
while read_length > 0:
586
num_bytes_to_read = min(read_length, buff_size)
588
block = from_file.read(num_bytes_to_read)
592
if report_activity is not None:
593
report_activity(len(block), direction)
596
actual_bytes_read = len(block)
597
read_length -= actual_bytes_read
598
length += actual_bytes_read
602
block = from_file.read(buff_size)
606
if report_activity is not None:
607
report_activity(len(block), direction)
613
def pump_string_file(bytes, file_handle, segment_size=None):
614
"""Write bytes to file_handle in many smaller writes.
616
:param bytes: The string to write.
617
:param file_handle: The file to write to.
619
# Write data in chunks rather than all at once, because very large
620
# writes fail on some platforms (e.g. Windows with SMB mounted
623
segment_size = 5242880 # 5MB
624
segments = range(len(bytes) / segment_size + 1)
625
write = file_handle.write
626
for segment_index in segments:
627
segment = buffer(bytes, segment_index * segment_size, segment_size)
631
def file_iterator(input_file, readsize=32768):
633
b = input_file.read(readsize)
640
"""Calculate the hexdigest of an open file.
642
The file cursor should be already at the start.
654
def size_sha_file(f):
655
"""Calculate the size and hexdigest of an open file.
657
The file cursor should be already at the start and
658
the caller is responsible for closing the file afterwards.
669
return size, s.hexdigest()
672
def sha_file_by_name(fname):
673
"""Calculate the SHA1 of a file by reading the full text"""
675
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
678
b = os.read(f, 1<<16)
686
def sha_strings(strings, _factory=sha):
687
"""Return the sha-1 of concatenation of strings"""
689
map(s.update, strings)
693
def sha_string(f, _factory=sha):
694
return _factory(f).hexdigest()
697
def fingerprint_file(f):
699
return {'size': len(b),
700
'sha1': sha(b).hexdigest()}
703
def compare_files(a, b):
704
"""Returns true if equal in contents"""
715
def local_time_offset(t=None):
716
"""Return offset of local zone from GMT, either at present or at time t."""
719
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
720
return offset.days * 86400 + offset.seconds
722
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
723
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
726
def format_date(t, offset=0, timezone='original', date_fmt=None,
728
"""Return a formatted date string.
730
:param t: Seconds since the epoch.
731
:param offset: Timezone offset in seconds east of utc.
732
:param timezone: How to display the time: 'utc', 'original' for the
733
timezone specified by offset, or 'local' for the process's current
735
:param date_fmt: strftime format.
736
:param show_offset: Whether to append the timezone.
738
(date_fmt, tt, offset_str) = \
739
_format_date(t, offset, timezone, date_fmt, show_offset)
740
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
741
date_str = time.strftime(date_fmt, tt)
742
return date_str + offset_str
745
# Cache of formatted offset strings
749
def format_date_with_offset_in_original_timezone(t, offset=0,
750
_cache=_offset_cache):
751
"""Return a formatted date string in the original timezone.
753
This routine may be faster then format_date.
755
:param t: Seconds since the epoch.
756
:param offset: Timezone offset in seconds east of utc.
760
tt = time.gmtime(t + offset)
761
date_fmt = _default_format_by_weekday_num[tt[6]]
762
date_str = time.strftime(date_fmt, tt)
763
offset_str = _cache.get(offset, None)
764
if offset_str is None:
765
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
766
_cache[offset] = offset_str
767
return date_str + offset_str
770
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
772
"""Return an unicode date string formatted according to the current locale.
774
:param t: Seconds since the epoch.
775
:param offset: Timezone offset in seconds east of utc.
776
:param timezone: How to display the time: 'utc', 'original' for the
777
timezone specified by offset, or 'local' for the process's current
779
:param date_fmt: strftime format.
780
:param show_offset: Whether to append the timezone.
782
(date_fmt, tt, offset_str) = \
783
_format_date(t, offset, timezone, date_fmt, show_offset)
784
date_str = time.strftime(date_fmt, tt)
785
if not isinstance(date_str, unicode):
786
date_str = date_str.decode(get_user_encoding(), 'replace')
787
return date_str + offset_str
790
def _format_date(t, offset, timezone, date_fmt, show_offset):
791
if timezone == 'utc':
794
elif timezone == 'original':
797
tt = time.gmtime(t + offset)
798
elif timezone == 'local':
799
tt = time.localtime(t)
800
offset = local_time_offset(t)
802
raise errors.UnsupportedTimezoneFormat(timezone)
804
date_fmt = "%a %Y-%m-%d %H:%M:%S"
806
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
809
return (date_fmt, tt, offset_str)
812
def compact_date(when):
813
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
816
def format_delta(delta):
817
"""Get a nice looking string for a time delta.
819
:param delta: The time difference in seconds, can be positive or negative.
820
positive indicates time in the past, negative indicates time in the
821
future. (usually time.time() - stored_time)
822
:return: String formatted to show approximate resolution
828
direction = 'in the future'
832
if seconds < 90: # print seconds up to 90 seconds
834
return '%d second %s' % (seconds, direction,)
836
return '%d seconds %s' % (seconds, direction)
838
minutes = int(seconds / 60)
839
seconds -= 60 * minutes
844
if minutes < 90: # print minutes, seconds up to 90 minutes
846
return '%d minute, %d second%s %s' % (
847
minutes, seconds, plural_seconds, direction)
849
return '%d minutes, %d second%s %s' % (
850
minutes, seconds, plural_seconds, direction)
852
hours = int(minutes / 60)
853
minutes -= 60 * hours
860
return '%d hour, %d minute%s %s' % (hours, minutes,
861
plural_minutes, direction)
862
return '%d hours, %d minute%s %s' % (hours, minutes,
863
plural_minutes, direction)
866
"""Return size of given open file."""
867
return os.fstat(f.fileno())[stat.ST_SIZE]
870
# Define rand_bytes based on platform.
872
# Python 2.4 and later have os.urandom,
873
# but it doesn't work on some arches
875
rand_bytes = os.urandom
876
except (NotImplementedError, AttributeError):
877
# If python doesn't have os.urandom, or it doesn't work,
878
# then try to first pull random data from /dev/urandom
880
rand_bytes = file('/dev/urandom', 'rb').read
881
# Otherwise, use this hack as a last resort
882
except (IOError, OSError):
883
# not well seeded, but better than nothing
888
s += chr(random.randint(0, 255))
893
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
895
"""Return a random string of num alphanumeric characters
897
The result only contains lowercase chars because it may be used on
898
case-insensitive filesystems.
901
for raw_byte in rand_bytes(num):
902
s += ALNUM[ord(raw_byte) % 36]
906
## TODO: We could later have path objects that remember their list
907
## decomposition (might be too tricksy though.)
910
"""Turn string into list of parts."""
911
# split on either delimiter because people might use either on
913
ps = re.split(r'[\\/]', p)
918
raise errors.BzrError("sorry, %r not allowed in path" % f)
919
elif (f == '.') or (f == ''):
928
if (f == '..') or (f is None) or (f == ''):
929
raise errors.BzrError("sorry, %r not allowed in path" % f)
933
def parent_directories(filename):
934
"""Return the list of parent directories, deepest first.
936
For example, parent_directories("a/b/c") -> ["a/b", "a"].
939
parts = splitpath(dirname(filename))
941
parents.append(joinpath(parts))
946
_extension_load_failures = []
949
def failed_to_load_extension(exception):
950
"""Handle failing to load a binary extension.
952
This should be called from the ImportError block guarding the attempt to
953
import the native extension. If this function returns, the pure-Python
954
implementation should be loaded instead::
957
>>> import bzrlib._fictional_extension_pyx
958
>>> except ImportError, e:
959
>>> bzrlib.osutils.failed_to_load_extension(e)
960
>>> import bzrlib._fictional_extension_py
962
# NB: This docstring is just an example, not a doctest, because doctest
963
# currently can't cope with the use of lazy imports in this namespace --
966
# This currently doesn't report the failure at the time it occurs, because
967
# they tend to happen very early in startup when we can't check config
968
# files etc, and also we want to report all failures but not spam the user
970
from bzrlib import trace
971
exception_str = str(exception)
972
if exception_str not in _extension_load_failures:
973
trace.mutter("failed to load compiled extension: %s" % exception_str)
974
_extension_load_failures.append(exception_str)
977
def report_extension_load_failures():
978
if not _extension_load_failures:
980
from bzrlib.config import GlobalConfig
981
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
983
# the warnings framework should by default show this only once
984
from bzrlib.trace import warning
986
"bzr: warning: some compiled extensions could not be loaded; "
987
"see <https://answers.launchpad.net/bzr/+faq/703>")
988
# we no longer show the specific missing extensions here, because it makes
989
# the message too long and scary - see
990
# https://bugs.launchpad.net/bzr/+bug/430529
994
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
995
except ImportError, e:
996
failed_to_load_extension(e)
997
from bzrlib._chunks_to_lines_py import chunks_to_lines
1001
"""Split s into lines, but without removing the newline characters."""
1002
# Trivially convert a fulltext into a 'chunked' representation, and let
1003
# chunks_to_lines do the heavy lifting.
1004
if isinstance(s, str):
1005
# chunks_to_lines only supports 8-bit strings
1006
return chunks_to_lines([s])
1008
return _split_lines(s)
1011
def _split_lines(s):
1012
"""Split s into lines, but without removing the newline characters.
1014
This supports Unicode or plain string objects.
1016
lines = s.split('\n')
1017
result = [line + '\n' for line in lines[:-1]]
1019
result.append(lines[-1])
1023
def hardlinks_good():
1024
return sys.platform not in ('win32', 'cygwin', 'darwin')
1027
def link_or_copy(src, dest):
1028
"""Hardlink a file, or copy it if it can't be hardlinked."""
1029
if not hardlinks_good():
1030
shutil.copyfile(src, dest)
1034
except (OSError, IOError), e:
1035
if e.errno != errno.EXDEV:
1037
shutil.copyfile(src, dest)
1040
def delete_any(path):
1041
"""Delete a file, symlink or directory.
1043
Will delete even if readonly.
1046
_delete_file_or_dir(path)
1047
except (OSError, IOError), e:
1048
if e.errno in (errno.EPERM, errno.EACCES):
1049
# make writable and try again
1052
except (OSError, IOError):
1054
_delete_file_or_dir(path)
1059
def _delete_file_or_dir(path):
1060
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1061
# Forgiveness than Permission (EAFP) because:
1062
# - root can damage a solaris file system by using unlink,
1063
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1064
# EACCES, OSX: EPERM) when invoked on a directory.
1065
if isdir(path): # Takes care of symlinks
1072
if getattr(os, 'symlink', None) is not None:
1078
def has_hardlinks():
1079
if getattr(os, 'link', None) is not None:
1085
def host_os_dereferences_symlinks():
1086
return (has_symlinks()
1087
and sys.platform not in ('cygwin', 'win32'))
1090
def readlink(abspath):
1091
"""Return a string representing the path to which the symbolic link points.
1093
:param abspath: The link absolute unicode path.
1095
This his guaranteed to return the symbolic link in unicode in all python
1098
link = abspath.encode(_fs_enc)
1099
target = os.readlink(link)
1100
target = target.decode(_fs_enc)
1104
def contains_whitespace(s):
1105
"""True if there are any whitespace characters in s."""
1106
# string.whitespace can include '\xa0' in certain locales, because it is
1107
# considered "non-breaking-space" as part of ISO-8859-1. But it
1108
# 1) Isn't a breaking whitespace
1109
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1111
# 3) '\xa0' isn't unicode safe since it is >128.
1113
# This should *not* be a unicode set of characters in case the source
1114
# string is not a Unicode string. We can auto-up-cast the characters since
1115
# they are ascii, but we don't want to auto-up-cast the string in case it
1117
for ch in ' \t\n\r\v\f':
1124
def contains_linebreaks(s):
1125
"""True if there is any vertical whitespace in s."""
1133
def relpath(base, path):
1134
"""Return path relative to base, or raise PathNotChild exception.
1136
The path may be either an absolute path or a path relative to the
1137
current working directory.
1139
os.path.commonprefix (python2.4) has a bad bug that it works just
1140
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1141
avoids that problem.
1143
NOTE: `base` should not have a trailing slash otherwise you'll get
1144
PathNotChild exceptions regardless of `path`.
1147
if len(base) < MIN_ABS_PATHLENGTH:
1148
# must have space for e.g. a drive letter
1149
raise ValueError('%r is too short to calculate a relative path'
1157
if len(head) <= len(base) and head != base:
1158
raise errors.PathNotChild(rp, base)
1161
head, tail = split(head)
1166
return pathjoin(*reversed(s))
1171
def _cicp_canonical_relpath(base, path):
1172
"""Return the canonical path relative to base.
1174
Like relpath, but on case-insensitive-case-preserving file-systems, this
1175
will return the relpath as stored on the file-system rather than in the
1176
case specified in the input string, for all existing portions of the path.
1178
This will cause O(N) behaviour if called for every path in a tree; if you
1179
have a number of paths to convert, you should use canonical_relpaths().
1181
# TODO: it should be possible to optimize this for Windows by using the
1182
# win32 API FindFiles function to look for the specified name - but using
1183
# os.listdir() still gives us the correct, platform agnostic semantics in
1186
rel = relpath(base, path)
1187
# '.' will have been turned into ''
1191
abs_base = abspath(base)
1193
_listdir = os.listdir
1195
# use an explicit iterator so we can easily consume the rest on early exit.
1196
bit_iter = iter(rel.split('/'))
1197
for bit in bit_iter:
1200
next_entries = _listdir(current)
1201
except OSError: # enoent, eperm, etc
1202
# We can't find this in the filesystem, so just append the
1204
current = pathjoin(current, bit, *list(bit_iter))
1206
for look in next_entries:
1207
if lbit == look.lower():
1208
current = pathjoin(current, look)
1211
# got to the end, nothing matched, so we just return the
1212
# non-existing bits as they were specified (the filename may be
1213
# the target of a move, for example).
1214
current = pathjoin(current, bit, *list(bit_iter))
1216
return current[len(abs_base):].lstrip('/')
1218
# XXX - TODO - we need better detection/integration of case-insensitive
1219
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1220
# filesystems), for example, so could probably benefit from the same basic
1221
# support there. For now though, only Windows and OSX get that support, and
1222
# they get it for *all* file-systems!
1223
if sys.platform in ('win32', 'darwin'):
1224
canonical_relpath = _cicp_canonical_relpath
1226
canonical_relpath = relpath
1228
def canonical_relpaths(base, paths):
1229
"""Create an iterable to canonicalize a sequence of relative paths.
1231
The intent is for this implementation to use a cache, vastly speeding
1232
up multiple transformations in the same directory.
1234
# but for now, we haven't optimized...
1235
return [canonical_relpath(base, p) for p in paths]
1238
def decode_filename(filename):
1239
"""Decode the filename using the filesystem encoding
1241
If it is unicode, it is returned.
1242
Otherwise it is decoded from the the filesystem's encoding. If decoding
1243
fails, a errors.BadFilenameEncoding exception is raised.
1245
if type(filename) is unicode:
1248
return filename.decode(_fs_enc)
1249
except UnicodeDecodeError:
1250
raise errors.BadFilenameEncoding(filename, _fs_enc)
1253
def safe_unicode(unicode_or_utf8_string):
1254
"""Coerce unicode_or_utf8_string into unicode.
1256
If it is unicode, it is returned.
1257
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1258
wrapped in a BzrBadParameterNotUnicode exception.
1260
if isinstance(unicode_or_utf8_string, unicode):
1261
return unicode_or_utf8_string
1263
return unicode_or_utf8_string.decode('utf8')
1264
except UnicodeDecodeError:
1265
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1268
def safe_utf8(unicode_or_utf8_string):
1269
"""Coerce unicode_or_utf8_string to a utf8 string.
1271
If it is a str, it is returned.
1272
If it is Unicode, it is encoded into a utf-8 string.
1274
if isinstance(unicode_or_utf8_string, str):
1275
# TODO: jam 20070209 This is overkill, and probably has an impact on
1276
# performance if we are dealing with lots of apis that want a
1279
# Make sure it is a valid utf-8 string
1280
unicode_or_utf8_string.decode('utf-8')
1281
except UnicodeDecodeError:
1282
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1283
return unicode_or_utf8_string
1284
return unicode_or_utf8_string.encode('utf-8')
1287
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1288
' Revision id generators should be creating utf8'
1292
def safe_revision_id(unicode_or_utf8_string, warn=True):
1293
"""Revision ids should now be utf8, but at one point they were unicode.
1295
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1297
:param warn: Functions that are sanitizing user data can set warn=False
1298
:return: None or a utf8 revision id.
1300
if (unicode_or_utf8_string is None
1301
or unicode_or_utf8_string.__class__ == str):
1302
return unicode_or_utf8_string
1304
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1306
return cache_utf8.encode(unicode_or_utf8_string)
1309
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1310
' generators should be creating utf8 file ids.')
1313
def safe_file_id(unicode_or_utf8_string, warn=True):
1314
"""File ids should now be utf8, but at one point they were unicode.
1316
This is the same as safe_utf8, except it uses the cached encode functions
1317
to save a little bit of performance.
1319
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1321
:param warn: Functions that are sanitizing user data can set warn=False
1322
:return: None or a utf8 file id.
1324
if (unicode_or_utf8_string is None
1325
or unicode_or_utf8_string.__class__ == str):
1326
return unicode_or_utf8_string
1328
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1330
return cache_utf8.encode(unicode_or_utf8_string)
1333
_platform_normalizes_filenames = False
1334
if sys.platform == 'darwin':
1335
_platform_normalizes_filenames = True
1338
def normalizes_filenames():
1339
"""Return True if this platform normalizes unicode filenames.
1343
return _platform_normalizes_filenames
1346
def _accessible_normalized_filename(path):
1347
"""Get the unicode normalized path, and if you can access the file.
1349
On platforms where the system normalizes filenames (Mac OSX),
1350
you can access a file by any path which will normalize correctly.
1351
On platforms where the system does not normalize filenames
1352
(everything else), you have to access a file by its exact path.
1354
Internally, bzr only supports NFC normalization, since that is
1355
the standard for XML documents.
1357
So return the normalized path, and a flag indicating if the file
1358
can be accessed by that path.
1361
return unicodedata.normalize('NFC', unicode(path)), True
1364
def _inaccessible_normalized_filename(path):
1365
__doc__ = _accessible_normalized_filename.__doc__
1367
normalized = unicodedata.normalize('NFC', unicode(path))
1368
return normalized, normalized == path
1371
if _platform_normalizes_filenames:
1372
normalized_filename = _accessible_normalized_filename
1374
normalized_filename = _inaccessible_normalized_filename
1377
def set_signal_handler(signum, handler, restart_syscall=True):
1378
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1379
on platforms that support that.
1381
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1382
automatically restart (by calling `signal.siginterrupt(signum,
1383
False)`). May be ignored if the feature is not available on this
1384
platform or Python version.
1388
siginterrupt = signal.siginterrupt
1390
# This python implementation doesn't provide signal support, hence no
1393
except AttributeError:
1394
# siginterrupt doesn't exist on this platform, or for this version
1396
siginterrupt = lambda signum, flag: None
1398
def sig_handler(*args):
1399
# Python resets the siginterrupt flag when a signal is
1400
# received. <http://bugs.python.org/issue8354>
1401
# As a workaround for some cases, set it back the way we want it.
1402
siginterrupt(signum, False)
1403
# Now run the handler function passed to set_signal_handler.
1406
sig_handler = handler
1407
old_handler = signal.signal(signum, sig_handler)
1409
siginterrupt(signum, False)
1413
default_terminal_width = 80
1414
"""The default terminal width for ttys.
1416
This is defined so that higher levels can share a common fallback value when
1417
terminal_width() returns None.
1420
# Keep some state so that terminal_width can detect if _terminal_size has
1421
# returned a different size since the process started. See docstring and
1422
# comments of terminal_width for details.
1423
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1424
_terminal_size_state = 'no_data'
1425
_first_terminal_size = None
1427
def terminal_width():
1428
"""Return terminal width.
1430
None is returned if the width can't established precisely.
1433
- if BZR_COLUMNS is set, returns its value
1434
- if there is no controlling terminal, returns None
1435
- query the OS, if the queried size has changed since the last query,
1437
- if COLUMNS is set, returns its value,
1438
- if the OS has a value (even though it's never changed), return its value.
1440
From there, we need to query the OS to get the size of the controlling
1443
On Unices we query the OS by:
1444
- get termios.TIOCGWINSZ
1445
- if an error occurs or a negative value is obtained, returns None
1447
On Windows we query the OS by:
1448
- win32utils.get_console_size() decides,
1449
- returns None on error (provided default value)
1451
# Note to implementors: if changing the rules for determining the width,
1452
# make sure you've considered the behaviour in these cases:
1453
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1454
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1456
# - (add more interesting cases here, if you find any)
1457
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1458
# but we don't want to register a signal handler because it is impossible
1459
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1460
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1461
# time so we can notice if the reported size has changed, which should have
1464
# If BZR_COLUMNS is set, take it, user is always right
1466
return int(os.environ['BZR_COLUMNS'])
1467
except (KeyError, ValueError):
1470
isatty = getattr(sys.stdout, 'isatty', None)
1471
if isatty is None or not isatty():
1472
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1476
width, height = os_size = _terminal_size(None, None)
1477
global _first_terminal_size, _terminal_size_state
1478
if _terminal_size_state == 'no_data':
1479
_first_terminal_size = os_size
1480
_terminal_size_state = 'unchanged'
1481
elif (_terminal_size_state == 'unchanged' and
1482
_first_terminal_size != os_size):
1483
_terminal_size_state = 'changed'
1485
# If the OS claims to know how wide the terminal is, and this value has
1486
# ever changed, use that.
1487
if _terminal_size_state == 'changed':
1488
if width is not None and width > 0:
1491
# If COLUMNS is set, use it.
1493
return int(os.environ['COLUMNS'])
1494
except (KeyError, ValueError):
1497
# Finally, use an unchanged size from the OS, if we have one.
1498
if _terminal_size_state == 'unchanged':
1499
if width is not None and width > 0:
1502
# The width could not be determined.
1506
def _win32_terminal_size(width, height):
1507
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1508
return width, height
1511
def _ioctl_terminal_size(width, height):
1513
import struct, fcntl, termios
1514
s = struct.pack('HHHH', 0, 0, 0, 0)
1515
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1516
height, width = struct.unpack('HHHH', x)[0:2]
1517
except (IOError, AttributeError):
1519
return width, height
1521
_terminal_size = None
1522
"""Returns the terminal size as (width, height).
1524
:param width: Default value for width.
1525
:param height: Default value for height.
1527
This is defined specifically for each OS and query the size of the controlling
1528
terminal. If any error occurs, the provided default values should be returned.
1530
if sys.platform == 'win32':
1531
_terminal_size = _win32_terminal_size
1533
_terminal_size = _ioctl_terminal_size
1536
def supports_executable():
1537
return sys.platform != "win32"
1540
def supports_posix_readonly():
1541
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1543
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1544
directory controls creation/deletion, etc.
1546
And under win32, readonly means that the directory itself cannot be
1547
deleted. The contents of a readonly directory can be changed, unlike POSIX
1548
where files in readonly directories cannot be added, deleted or renamed.
1550
return sys.platform != "win32"
1553
def set_or_unset_env(env_variable, value):
1554
"""Modify the environment, setting or removing the env_variable.
1556
:param env_variable: The environment variable in question
1557
:param value: The value to set the environment to. If None, then
1558
the variable will be removed.
1559
:return: The original value of the environment variable.
1561
orig_val = os.environ.get(env_variable)
1563
if orig_val is not None:
1564
del os.environ[env_variable]
1566
if isinstance(value, unicode):
1567
value = value.encode(get_user_encoding())
1568
os.environ[env_variable] = value
1572
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1575
def check_legal_path(path):
1576
"""Check whether the supplied path is legal.
1577
This is only required on Windows, so we don't test on other platforms
1580
if sys.platform != "win32":
1582
if _validWin32PathRE.match(path) is None:
1583
raise errors.IllegalPath(path)
1586
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1588
def _is_error_enotdir(e):
1589
"""Check if this exception represents ENOTDIR.
1591
Unfortunately, python is very inconsistent about the exception
1592
here. The cases are:
1593
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1594
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1595
which is the windows error code.
1596
3) Windows, Python2.5 uses errno == EINVAL and
1597
winerror == ERROR_DIRECTORY
1599
:param e: An Exception object (expected to be OSError with an errno
1600
attribute, but we should be able to cope with anything)
1601
:return: True if this represents an ENOTDIR error. False otherwise.
1603
en = getattr(e, 'errno', None)
1604
if (en == errno.ENOTDIR
1605
or (sys.platform == 'win32'
1606
and (en == _WIN32_ERROR_DIRECTORY
1607
or (en == errno.EINVAL
1608
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1614
def walkdirs(top, prefix=""):
1615
"""Yield data about all the directories in a tree.
1617
This yields all the data about the contents of a directory at a time.
1618
After each directory has been yielded, if the caller has mutated the list
1619
to exclude some directories, they are then not descended into.
1621
The data yielded is of the form:
1622
((directory-relpath, directory-path-from-top),
1623
[(relpath, basename, kind, lstat, path-from-top), ...]),
1624
- directory-relpath is the relative path of the directory being returned
1625
with respect to top. prefix is prepended to this.
1626
- directory-path-from-root is the path including top for this directory.
1627
It is suitable for use with os functions.
1628
- relpath is the relative path within the subtree being walked.
1629
- basename is the basename of the path
1630
- kind is the kind of the file now. If unknown then the file is not
1631
present within the tree - but it may be recorded as versioned. See
1633
- lstat is the stat data *if* the file was statted.
1634
- planned, not implemented:
1635
path_from_tree_root is the path from the root of the tree.
1637
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1638
allows one to walk a subtree but get paths that are relative to a tree
1640
:return: an iterator over the dirs.
1642
#TODO there is a bit of a smell where the results of the directory-
1643
# summary in this, and the path from the root, may not agree
1644
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1645
# potentially confusing output. We should make this more robust - but
1646
# not at a speed cost. RBC 20060731
1648
_directory = _directory_kind
1649
_listdir = os.listdir
1650
_kind_from_mode = file_kind_from_stat_mode
1651
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1653
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1654
relroot, _, _, _, top = pending.pop()
1656
relprefix = relroot + u'/'
1659
top_slash = top + u'/'
1662
append = dirblock.append
1664
names = sorted(map(decode_filename, _listdir(top)))
1666
if not _is_error_enotdir(e):
1670
abspath = top_slash + name
1671
statvalue = _lstat(abspath)
1672
kind = _kind_from_mode(statvalue.st_mode)
1673
append((relprefix + name, name, kind, statvalue, abspath))
1674
yield (relroot, top), dirblock
1676
# push the user specified dirs from dirblock
1677
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1680
class DirReader(object):
1681
"""An interface for reading directories."""
1683
def top_prefix_to_starting_dir(self, top, prefix=""):
1684
"""Converts top and prefix to a starting dir entry
1686
:param top: A utf8 path
1687
:param prefix: An optional utf8 path to prefix output relative paths
1689
:return: A tuple starting with prefix, and ending with the native
1692
raise NotImplementedError(self.top_prefix_to_starting_dir)
1694
def read_dir(self, prefix, top):
1695
"""Read a specific dir.
1697
:param prefix: A utf8 prefix to be preprended to the path basenames.
1698
:param top: A natively encoded path to read.
1699
:return: A list of the directories contents. Each item contains:
1700
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1702
raise NotImplementedError(self.read_dir)
1705
_selected_dir_reader = None
1708
def _walkdirs_utf8(top, prefix=""):
1709
"""Yield data about all the directories in a tree.
1711
This yields the same information as walkdirs() only each entry is yielded
1712
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1713
are returned as exact byte-strings.
1715
:return: yields a tuple of (dir_info, [file_info])
1716
dir_info is (utf8_relpath, path-from-top)
1717
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1718
if top is an absolute path, path-from-top is also an absolute path.
1719
path-from-top might be unicode or utf8, but it is the correct path to
1720
pass to os functions to affect the file in question. (such as os.lstat)
1722
global _selected_dir_reader
1723
if _selected_dir_reader is None:
1724
fs_encoding = _fs_enc.upper()
1725
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1726
# Win98 doesn't have unicode apis like FindFirstFileW
1727
# TODO: We possibly could support Win98 by falling back to the
1728
# original FindFirstFile, and using TCHAR instead of WCHAR,
1729
# but that gets a bit tricky, and requires custom compiling
1732
from bzrlib._walkdirs_win32 import Win32ReadDir
1733
_selected_dir_reader = Win32ReadDir()
1736
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1737
# ANSI_X3.4-1968 is a form of ASCII
1739
from bzrlib._readdir_pyx import UTF8DirReader
1740
_selected_dir_reader = UTF8DirReader()
1741
except ImportError, e:
1742
failed_to_load_extension(e)
1745
if _selected_dir_reader is None:
1746
# Fallback to the python version
1747
_selected_dir_reader = UnicodeDirReader()
1749
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1750
# But we don't actually uses 1-3 in pending, so set them to None
1751
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1752
read_dir = _selected_dir_reader.read_dir
1753
_directory = _directory_kind
1755
relroot, _, _, _, top = pending[-1].pop()
1758
dirblock = sorted(read_dir(relroot, top))
1759
yield (relroot, top), dirblock
1760
# push the user specified dirs from dirblock
1761
next = [d for d in reversed(dirblock) if d[2] == _directory]
1763
pending.append(next)
1766
class UnicodeDirReader(DirReader):
1767
"""A dir reader for non-utf8 file systems, which transcodes."""
1769
__slots__ = ['_utf8_encode']
1772
self._utf8_encode = codecs.getencoder('utf8')
1774
def top_prefix_to_starting_dir(self, top, prefix=""):
1775
"""See DirReader.top_prefix_to_starting_dir."""
1776
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1778
def read_dir(self, prefix, top):
1779
"""Read a single directory from a non-utf8 file system.
1781
top, and the abspath element in the output are unicode, all other paths
1782
are utf8. Local disk IO is done via unicode calls to listdir etc.
1784
This is currently the fallback code path when the filesystem encoding is
1785
not UTF-8. It may be better to implement an alternative so that we can
1786
safely handle paths that are not properly decodable in the current
1789
See DirReader.read_dir for details.
1791
_utf8_encode = self._utf8_encode
1793
_listdir = os.listdir
1794
_kind_from_mode = file_kind_from_stat_mode
1797
relprefix = prefix + '/'
1800
top_slash = top + u'/'
1803
append = dirblock.append
1804
for name in sorted(_listdir(top)):
1806
name_utf8 = _utf8_encode(name)[0]
1807
except UnicodeDecodeError:
1808
raise errors.BadFilenameEncoding(
1809
_utf8_encode(relprefix)[0] + name, _fs_enc)
1810
abspath = top_slash + name
1811
statvalue = _lstat(abspath)
1812
kind = _kind_from_mode(statvalue.st_mode)
1813
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1817
def copy_tree(from_path, to_path, handlers={}):
1818
"""Copy all of the entries in from_path into to_path.
1820
:param from_path: The base directory to copy.
1821
:param to_path: The target directory. If it does not exist, it will
1823
:param handlers: A dictionary of functions, which takes a source and
1824
destinations for files, directories, etc.
1825
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1826
'file', 'directory', and 'symlink' should always exist.
1827
If they are missing, they will be replaced with 'os.mkdir()',
1828
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1830
# Now, just copy the existing cached tree to the new location
1831
# We use a cheap trick here.
1832
# Absolute paths are prefixed with the first parameter
1833
# relative paths are prefixed with the second.
1834
# So we can get both the source and target returned
1835
# without any extra work.
1837
def copy_dir(source, dest):
1840
def copy_link(source, dest):
1841
"""Copy the contents of a symlink"""
1842
link_to = os.readlink(source)
1843
os.symlink(link_to, dest)
1845
real_handlers = {'file':shutil.copy2,
1846
'symlink':copy_link,
1847
'directory':copy_dir,
1849
real_handlers.update(handlers)
1851
if not os.path.exists(to_path):
1852
real_handlers['directory'](from_path, to_path)
1854
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1855
for relpath, name, kind, st, abspath in entries:
1856
real_handlers[kind](abspath, relpath)
1859
def copy_ownership_from_path(dst, src=None):
1860
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1862
If src is None, the containing directory is used as source. If chown
1863
fails, the error is ignored and a warning is printed.
1865
chown = getattr(os, 'chown', None)
1870
src = os.path.dirname(dst)
1876
chown(dst, s.st_uid, s.st_gid)
1878
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1881
def path_prefix_key(path):
1882
"""Generate a prefix-order path key for path.
1884
This can be used to sort paths in the same way that walkdirs does.
1886
return (dirname(path) , path)
1889
def compare_paths_prefix_order(path_a, path_b):
1890
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1891
key_a = path_prefix_key(path_a)
1892
key_b = path_prefix_key(path_b)
1893
return cmp(key_a, key_b)
1896
_cached_user_encoding = None
1899
def get_user_encoding(use_cache=True):
1900
"""Find out what the preferred user encoding is.
1902
This is generally the encoding that is used for command line parameters
1903
and file contents. This may be different from the terminal encoding
1904
or the filesystem encoding.
1906
:param use_cache: Enable cache for detected encoding.
1907
(This parameter is turned on by default,
1908
and required only for selftesting)
1910
:return: A string defining the preferred user encoding
1912
global _cached_user_encoding
1913
if _cached_user_encoding is not None and use_cache:
1914
return _cached_user_encoding
1916
if sys.platform == 'darwin':
1917
# python locale.getpreferredencoding() always return
1918
# 'mac-roman' on darwin. That's a lie.
1919
sys.platform = 'posix'
1921
if os.environ.get('LANG', None) is None:
1922
# If LANG is not set, we end up with 'ascii', which is bad
1923
# ('mac-roman' is more than ascii), so we set a default which
1924
# will give us UTF-8 (which appears to work in all cases on
1925
# OSX). Users are still free to override LANG of course, as
1926
# long as it give us something meaningful. This work-around
1927
# *may* not be needed with python 3k and/or OSX 10.5, but will
1928
# work with them too -- vila 20080908
1929
os.environ['LANG'] = 'en_US.UTF-8'
1932
sys.platform = 'darwin'
1937
user_encoding = locale.getpreferredencoding()
1938
except locale.Error, e:
1939
sys.stderr.write('bzr: warning: %s\n'
1940
' Could not determine what text encoding to use.\n'
1941
' This error usually means your Python interpreter\n'
1942
' doesn\'t support the locale set by $LANG (%s)\n'
1943
" Continuing with ascii encoding.\n"
1944
% (e, os.environ.get('LANG')))
1945
user_encoding = 'ascii'
1947
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1948
# treat that as ASCII, and not support printing unicode characters to the
1951
# For python scripts run under vim, we get '', so also treat that as ASCII
1952
if user_encoding in (None, 'cp0', ''):
1953
user_encoding = 'ascii'
1957
codecs.lookup(user_encoding)
1959
sys.stderr.write('bzr: warning:'
1960
' unknown encoding %s.'
1961
' Continuing with ascii encoding.\n'
1964
user_encoding = 'ascii'
1967
_cached_user_encoding = user_encoding
1969
return user_encoding
1972
def get_diff_header_encoding():
1973
return get_terminal_encoding()
1976
def get_host_name():
1977
"""Return the current unicode host name.
1979
This is meant to be used in place of socket.gethostname() because that
1980
behaves inconsistently on different platforms.
1982
if sys.platform == "win32":
1984
return win32utils.get_host_name()
1987
return socket.gethostname().decode(get_user_encoding())
1990
# We must not read/write any more than 64k at a time from/to a socket so we
1991
# don't risk "no buffer space available" errors on some platforms. Windows in
1992
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1994
MAX_SOCKET_CHUNK = 64 * 1024
1996
def read_bytes_from_socket(sock, report_activity=None,
1997
max_read_size=MAX_SOCKET_CHUNK):
1998
"""Read up to max_read_size of bytes from sock and notify of progress.
2000
Translates "Connection reset by peer" into file-like EOF (return an
2001
empty string rather than raise an error), and repeats the recv if
2002
interrupted by a signal.
2006
bytes = sock.recv(max_read_size)
2007
except socket.error, e:
2009
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
2010
# The connection was closed by the other side. Callers expect
2011
# an empty string to signal end-of-stream.
2013
elif eno == errno.EINTR:
2014
# Retry the interrupted recv.
2018
if report_activity is not None:
2019
report_activity(len(bytes), 'read')
2023
def recv_all(socket, count):
2024
"""Receive an exact number of bytes.
2026
Regular Socket.recv() may return less than the requested number of bytes,
2027
depending on what's in the OS buffer. MSG_WAITALL is not available
2028
on all platforms, but this should work everywhere. This will return
2029
less than the requested amount if the remote end closes.
2031
This isn't optimized and is intended mostly for use in testing.
2034
while len(b) < count:
2035
new = read_bytes_from_socket(socket, None, count - len(b))
2042
def send_all(sock, bytes, report_activity=None):
2043
"""Send all bytes on a socket.
2045
Breaks large blocks in smaller chunks to avoid buffering limitations on
2046
some platforms, and catches EINTR which may be thrown if the send is
2047
interrupted by a signal.
2049
This is preferred to socket.sendall(), because it avoids portability bugs
2050
and provides activity reporting.
2052
:param report_activity: Call this as bytes are read, see
2053
Transport._report_activity
2056
byte_count = len(bytes)
2057
while sent_total < byte_count:
2059
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2060
except socket.error, e:
2061
if e.args[0] != errno.EINTR:
2065
report_activity(sent, 'write')
2068
def connect_socket(address):
2069
# Slight variation of the socket.create_connection() function (provided by
2070
# python-2.6) that can fail if getaddrinfo returns an empty list. We also
2071
# provide it for previous python versions. Also, we don't use the timeout
2072
# parameter (provided by the python implementation) so we don't implement
2074
err = socket.error('getaddrinfo returns an empty list')
2075
host, port = address
2076
for res in socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM):
2077
af, socktype, proto, canonname, sa = res
2080
sock = socket.socket(af, socktype, proto)
2084
except socket.error, err:
2085
# 'err' is now the most recent error
2086
if sock is not None:
2091
def dereference_path(path):
2092
"""Determine the real path to a file.
2094
All parent elements are dereferenced. But the file itself is not
2096
:param path: The original path. May be absolute or relative.
2097
:return: the real path *to* the file
2099
parent, base = os.path.split(path)
2100
# The pathjoin for '.' is a workaround for Python bug #1213894.
2101
# (initial path components aren't dereferenced)
2102
return pathjoin(realpath(pathjoin('.', parent)), base)
2105
def supports_mapi():
2106
"""Return True if we can use MAPI to launch a mail client."""
2107
return sys.platform == "win32"
2110
def resource_string(package, resource_name):
2111
"""Load a resource from a package and return it as a string.
2113
Note: Only packages that start with bzrlib are currently supported.
2115
This is designed to be a lightweight implementation of resource
2116
loading in a way which is API compatible with the same API from
2118
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2119
If and when pkg_resources becomes a standard library, this routine
2122
# Check package name is within bzrlib
2123
if package == "bzrlib":
2124
resource_relpath = resource_name
2125
elif package.startswith("bzrlib."):
2126
package = package[len("bzrlib."):].replace('.', os.sep)
2127
resource_relpath = pathjoin(package, resource_name)
2129
raise errors.BzrError('resource package %s not in bzrlib' % package)
2131
# Map the resource to a file and read its contents
2132
base = dirname(bzrlib.__file__)
2133
if getattr(sys, 'frozen', None): # bzr.exe
2134
base = abspath(pathjoin(base, '..', '..'))
2135
f = file(pathjoin(base, resource_relpath), "rU")
2141
def file_kind_from_stat_mode_thunk(mode):
2142
global file_kind_from_stat_mode
2143
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2145
from bzrlib._readdir_pyx import UTF8DirReader
2146
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2147
except ImportError, e:
2148
# This is one time where we won't warn that an extension failed to
2149
# load. The extension is never available on Windows anyway.
2150
from bzrlib._readdir_py import (
2151
_kind_from_mode as file_kind_from_stat_mode
2153
return file_kind_from_stat_mode(mode)
2154
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2157
def file_kind(f, _lstat=os.lstat):
2159
return file_kind_from_stat_mode(_lstat(f).st_mode)
2161
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2162
raise errors.NoSuchFile(f)
2166
def until_no_eintr(f, *a, **kw):
2167
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2169
WARNING: you must be certain that it is safe to retry the call repeatedly
2170
if EINTR does occur. This is typically only true for low-level operations
2171
like os.read. If in any doubt, don't use this.
2173
Keep in mind that this is not a complete solution to EINTR. There is
2174
probably code in the Python standard library and other dependencies that
2175
may encounter EINTR if a signal arrives (and there is signal handler for
2176
that signal). So this function can reduce the impact for IO that bzrlib
2177
directly controls, but it is not a complete solution.
2179
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2183
except (IOError, OSError), e:
2184
if e.errno == errno.EINTR:
2189
@deprecated_function(deprecated_in((2, 2, 0)))
2190
def re_compile_checked(re_string, flags=0, where=""):
2191
"""Return a compiled re, or raise a sensible error.
2193
This should only be used when compiling user-supplied REs.
2195
:param re_string: Text form of regular expression.
2196
:param flags: eg re.IGNORECASE
2197
:param where: Message explaining to the user the context where
2198
it occurred, eg 'log search filter'.
2200
# from https://bugs.launchpad.net/bzr/+bug/251352
2202
re_obj = re.compile(re_string, flags)
2205
except errors.InvalidPattern, e:
2207
where = ' in ' + where
2208
# despite the name 'error' is a type
2209
raise errors.BzrCommandError('Invalid regular expression%s: %s'
2213
if sys.platform == "win32":
2216
return msvcrt.getch()
2221
fd = sys.stdin.fileno()
2222
settings = termios.tcgetattr(fd)
2225
ch = sys.stdin.read(1)
2227
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2231
if sys.platform == 'linux2':
2232
def _local_concurrency():
2234
prefix = 'processor'
2235
for line in file('/proc/cpuinfo', 'rb'):
2236
if line.startswith(prefix):
2237
concurrency = int(line[line.find(':')+1:]) + 1
2239
elif sys.platform == 'darwin':
2240
def _local_concurrency():
2241
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2242
stdout=subprocess.PIPE).communicate()[0]
2243
elif sys.platform[0:7] == 'freebsd':
2244
def _local_concurrency():
2245
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2246
stdout=subprocess.PIPE).communicate()[0]
2247
elif sys.platform == 'sunos5':
2248
def _local_concurrency():
2249
return subprocess.Popen(['psrinfo', '-p',],
2250
stdout=subprocess.PIPE).communicate()[0]
2251
elif sys.platform == "win32":
2252
def _local_concurrency():
2253
# This appears to return the number of cores.
2254
return os.environ.get('NUMBER_OF_PROCESSORS')
2256
def _local_concurrency():
2261
_cached_local_concurrency = None
2263
def local_concurrency(use_cache=True):
2264
"""Return how many processes can be run concurrently.
2266
Rely on platform specific implementations and default to 1 (one) if
2267
anything goes wrong.
2269
global _cached_local_concurrency
2271
if _cached_local_concurrency is not None and use_cache:
2272
return _cached_local_concurrency
2274
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2275
if concurrency is None:
2277
concurrency = _local_concurrency()
2278
except (OSError, IOError):
2281
concurrency = int(concurrency)
2282
except (TypeError, ValueError):
2285
_cached_concurrency = concurrency
2289
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2290
"""A stream writer that doesn't decode str arguments."""
2292
def __init__(self, encode, stream, errors='strict'):
2293
codecs.StreamWriter.__init__(self, stream, errors)
2294
self.encode = encode
2296
def write(self, object):
2297
if type(object) is str:
2298
self.stream.write(object)
2300
data, _ = self.encode(object, self.errors)
2301
self.stream.write(data)
2303
if sys.platform == 'win32':
2304
def open_file(filename, mode='r', bufsize=-1):
2305
"""This function is used to override the ``open`` builtin.
2307
But it uses O_NOINHERIT flag so the file handle is not inherited by
2308
child processes. Deleting or renaming a closed file opened with this
2309
function is not blocking child processes.
2311
writing = 'w' in mode
2312
appending = 'a' in mode
2313
updating = '+' in mode
2314
binary = 'b' in mode
2317
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2318
# for flags for each modes.
2328
flags |= os.O_WRONLY
2329
flags |= os.O_CREAT | os.O_TRUNC
2334
flags |= os.O_WRONLY
2335
flags |= os.O_CREAT | os.O_APPEND
2340
flags |= os.O_RDONLY
2342
return os.fdopen(os.open(filename, flags), mode, bufsize)
2347
def getuser_unicode():
2348
"""Return the username as unicode.
2351
user_encoding = get_user_encoding()
2352
username = getpass.getuser().decode(user_encoding)
2353
except UnicodeDecodeError:
2354
raise errors.BzrError("Can't decode username as %s." % \
added
removed
bzrlib/osutils.py
