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
45
from bzrlib.symbol_versioning import (
50
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
52
if sys.version_info < (2, 5):
53
import md5 as _mod_md5
55
import sha as _mod_sha
65
from bzrlib import symbol_versioning
68
# Cross platform wall-clock time functionality with decent resolution.
69
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
70
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
71
# synchronized with ``time.time()``, this is only meant to be used to find
72
# delta times by subtracting from another call to this function.
73
timer_func = time.time
74
if sys.platform == 'win32':
75
timer_func = time.clock
77
# On win32, O_BINARY is used to indicate the file should
78
# be opened in binary mode, rather than text mode.
79
# On other platforms, O_BINARY doesn't exist, because
80
# they always open in binary mode, so it is okay to
81
# OR with 0 on those platforms.
82
# O_NOINHERIT and O_TEXT exists only on win32 too.
83
O_BINARY = getattr(os, 'O_BINARY', 0)
84
O_TEXT = getattr(os, 'O_TEXT', 0)
85
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
88
def get_unicode_argv():
90
user_encoding = get_user_encoding()
91
return [a.decode(user_encoding) for a in sys.argv[1:]]
92
except UnicodeDecodeError:
93
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
97
def make_readonly(filename):
98
"""Make a filename read-only."""
99
mod = os.lstat(filename).st_mode
100
if not stat.S_ISLNK(mod):
102
os.chmod(filename, mod)
105
def make_writable(filename):
106
mod = os.lstat(filename).st_mode
107
if not stat.S_ISLNK(mod):
109
os.chmod(filename, mod)
112
def minimum_path_selection(paths):
113
"""Return the smallset subset of paths which are outside paths.
115
:param paths: A container (and hence not None) of paths.
116
:return: A set of paths sufficient to include everything in paths via
117
is_inside, drawn from the paths parameter.
123
return path.split('/')
124
sorted_paths = sorted(list(paths), key=sort_key)
126
search_paths = [sorted_paths[0]]
127
for path in sorted_paths[1:]:
128
if not is_inside(search_paths[-1], path):
129
# This path is unique, add it
130
search_paths.append(path)
132
return set(search_paths)
139
"""Return a quoted filename filename
141
This previously used backslash quoting, but that works poorly on
143
# TODO: I'm not really sure this is the best format either.x
145
if _QUOTE_RE is None:
146
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
148
if _QUOTE_RE.search(f):
154
_directory_kind = 'directory'
157
"""Return the current umask"""
158
# Assume that people aren't messing with the umask while running
159
# XXX: This is not thread safe, but there is no way to get the
160
# umask without setting it
168
_directory_kind: "/",
170
'tree-reference': '+',
174
def kind_marker(kind):
176
return _kind_marker_map[kind]
178
# Slightly faster than using .get(, '') when the common case is that
183
lexists = getattr(os.path, 'lexists', None)
187
stat = getattr(os, 'lstat', os.stat)
191
if e.errno == errno.ENOENT:
194
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
197
def fancy_rename(old, new, rename_func, unlink_func):
198
"""A fancy rename, when you don't have atomic rename.
200
:param old: The old path, to rename from
201
:param new: The new path, to rename to
202
:param rename_func: The potentially non-atomic rename function
203
:param unlink_func: A way to delete the target file if the full rename
206
# sftp rename doesn't allow overwriting, so play tricks:
207
base = os.path.basename(new)
208
dirname = os.path.dirname(new)
209
# callers use different encodings for the paths so the following MUST
210
# respect that. We rely on python upcasting to unicode if new is unicode
211
# and keeping a str if not.
212
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
213
os.getpid(), rand_chars(10))
214
tmp_name = pathjoin(dirname, tmp_name)
216
# Rename the file out of the way, but keep track if it didn't exist
217
# We don't want to grab just any exception
218
# something like EACCES should prevent us from continuing
219
# The downside is that the rename_func has to throw an exception
220
# with an errno = ENOENT, or NoSuchFile
223
rename_func(new, tmp_name)
224
except (errors.NoSuchFile,), e:
227
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
228
# function raises an IOError with errno is None when a rename fails.
229
# This then gets caught here.
230
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
233
if (getattr(e, 'errno', None) is None
234
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
243
# This may throw an exception, in which case success will
245
rename_func(old, new)
247
except (IOError, OSError), e:
248
# source and target may be aliases of each other (e.g. on a
249
# case-insensitive filesystem), so we may have accidentally renamed
250
# source by when we tried to rename target
251
failure_exc = sys.exc_info()
252
if (file_existed and e.errno in (None, errno.ENOENT)
253
and old.lower() == new.lower()):
254
# source and target are the same file on a case-insensitive
255
# filesystem, so we don't generate an exception
259
# If the file used to exist, rename it back into place
260
# otherwise just delete it from the tmp location
262
unlink_func(tmp_name)
264
rename_func(tmp_name, new)
265
if failure_exc is not None:
266
raise failure_exc[0], failure_exc[1], failure_exc[2]
269
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
270
# choke on a Unicode string containing a relative path if
271
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
273
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
274
def _posix_abspath(path):
275
# jam 20060426 rather than encoding to fsencoding
276
# copy posixpath.abspath, but use os.getcwdu instead
277
if not posixpath.isabs(path):
278
path = posixpath.join(getcwd(), path)
279
return posixpath.normpath(path)
282
def _posix_realpath(path):
283
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
286
def _win32_fixdrive(path):
287
"""Force drive letters to be consistent.
289
win32 is inconsistent whether it returns lower or upper case
290
and even if it was consistent the user might type the other
291
so we force it to uppercase
292
running python.exe under cmd.exe return capital C:\\
293
running win32 python inside a cygwin shell returns lowercase c:\\
295
drive, path = ntpath.splitdrive(path)
296
return drive.upper() + path
299
def _win32_abspath(path):
300
# Real ntpath.abspath doesn't have a problem with a unicode cwd
301
return _win32_fixdrive(ntpath.abspath(unicode(path)).replace('\\', '/'))
304
def _win98_abspath(path):
305
"""Return the absolute version of a path.
306
Windows 98 safe implementation (python reimplementation
307
of Win32 API function GetFullPathNameW)
312
# \\HOST\path => //HOST/path
313
# //HOST/path => //HOST/path
314
# path => C:/cwd/path
317
# check for absolute path
318
drive = ntpath.splitdrive(path)[0]
319
if drive == '' and path[:2] not in('//','\\\\'):
321
# we cannot simply os.path.join cwd and path
322
# because os.path.join('C:','/path') produce '/path'
323
# and this is incorrect
324
if path[:1] in ('/','\\'):
325
cwd = ntpath.splitdrive(cwd)[0]
327
path = cwd + '\\' + path
328
return _win32_fixdrive(ntpath.normpath(path).replace('\\', '/'))
331
def _win32_realpath(path):
332
# Real ntpath.realpath doesn't have a problem with a unicode cwd
333
return _win32_fixdrive(ntpath.realpath(unicode(path)).replace('\\', '/'))
336
def _win32_pathjoin(*args):
337
return ntpath.join(*args).replace('\\', '/')
340
def _win32_normpath(path):
341
return _win32_fixdrive(ntpath.normpath(unicode(path)).replace('\\', '/'))
345
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
348
def _win32_mkdtemp(*args, **kwargs):
349
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
352
def _win32_rename(old, new):
353
"""We expect to be able to atomically replace 'new' with old.
355
On win32, if new exists, it must be moved out of the way first,
359
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
361
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
362
# If we try to rename a non-existant file onto cwd, we get
363
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
364
# if the old path doesn't exist, sometimes we get EACCES
365
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
371
return unicodedata.normalize('NFC', os.getcwdu())
374
# Default is to just use the python builtins, but these can be rebound on
375
# particular platforms.
376
abspath = _posix_abspath
377
realpath = _posix_realpath
378
pathjoin = os.path.join
379
normpath = os.path.normpath
382
dirname = os.path.dirname
383
basename = os.path.basename
384
split = os.path.split
385
splitext = os.path.splitext
386
mkdtemp = tempfile.mkdtemp
387
rmtree = shutil.rmtree
389
MIN_ABS_PATHLENGTH = 1
392
if sys.platform == 'win32':
393
if win32utils.winver == 'Windows 98':
394
abspath = _win98_abspath
396
abspath = _win32_abspath
397
realpath = _win32_realpath
398
pathjoin = _win32_pathjoin
399
normpath = _win32_normpath
400
getcwd = _win32_getcwd
401
mkdtemp = _win32_mkdtemp
402
rename = _win32_rename
404
MIN_ABS_PATHLENGTH = 3
406
def _win32_delete_readonly(function, path, excinfo):
407
"""Error handler for shutil.rmtree function [for win32]
408
Helps to remove files and dirs marked as read-only.
410
exception = excinfo[1]
411
if function in (os.remove, os.rmdir) \
412
and isinstance(exception, OSError) \
413
and exception.errno == errno.EACCES:
419
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
420
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
421
return shutil.rmtree(path, ignore_errors, onerror)
423
f = win32utils.get_unicode_argv # special function or None
427
elif sys.platform == 'darwin':
431
def get_terminal_encoding():
432
"""Find the best encoding for printing to the screen.
434
This attempts to check both sys.stdout and sys.stdin to see
435
what encoding they are in, and if that fails it falls back to
436
osutils.get_user_encoding().
437
The problem is that on Windows, locale.getpreferredencoding()
438
is not the same encoding as that used by the console:
439
http://mail.python.org/pipermail/python-list/2003-May/162357.html
441
On my standard US Windows XP, the preferred encoding is
442
cp1252, but the console is cp437
444
from bzrlib.trace import mutter
445
output_encoding = getattr(sys.stdout, 'encoding', None)
446
if not output_encoding:
447
input_encoding = getattr(sys.stdin, 'encoding', None)
448
if not input_encoding:
449
output_encoding = get_user_encoding()
450
mutter('encoding stdout as osutils.get_user_encoding() %r',
453
output_encoding = input_encoding
454
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
456
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
457
if output_encoding == 'cp0':
458
# invalid encoding (cp0 means 'no codepage' on Windows)
459
output_encoding = get_user_encoding()
460
mutter('cp0 is invalid encoding.'
461
' encoding stdout as osutils.get_user_encoding() %r',
465
codecs.lookup(output_encoding)
467
sys.stderr.write('bzr: warning:'
468
' unknown terminal encoding %s.\n'
469
' Using encoding %s instead.\n'
470
% (output_encoding, get_user_encoding())
472
output_encoding = get_user_encoding()
474
return output_encoding
477
def normalizepath(f):
478
if getattr(os.path, 'realpath', None) is not None:
482
[p,e] = os.path.split(f)
483
if e == "" or e == "." or e == "..":
486
return pathjoin(F(p), e)
490
"""True if f is an accessible directory."""
492
return stat.S_ISDIR(os.lstat(f)[stat.ST_MODE])
498
"""True if f is a regular file."""
500
return stat.S_ISREG(os.lstat(f)[stat.ST_MODE])
505
"""True if f is a symlink."""
507
return stat.S_ISLNK(os.lstat(f)[stat.ST_MODE])
511
def is_inside(dir, fname):
512
"""True if fname is inside dir.
514
The parameters should typically be passed to osutils.normpath first, so
515
that . and .. and repeated slashes are eliminated, and the separators
516
are canonical for the platform.
518
The empty string as a dir name is taken as top-of-tree and matches
521
# XXX: Most callers of this can actually do something smarter by
522
# looking at the inventory
532
return fname.startswith(dir)
535
def is_inside_any(dir_list, fname):
536
"""True if fname is inside any of given dirs."""
537
for dirname in dir_list:
538
if is_inside(dirname, fname):
543
def is_inside_or_parent_of_any(dir_list, fname):
544
"""True if fname is a child or a parent of any of the given files."""
545
for dirname in dir_list:
546
if is_inside(dirname, fname) or is_inside(fname, dirname):
551
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
552
report_activity=None, direction='read'):
553
"""Copy contents of one file to another.
555
The read_length can either be -1 to read to end-of-file (EOF) or
556
it can specify the maximum number of bytes to read.
558
The buff_size represents the maximum size for each read operation
559
performed on from_file.
561
:param report_activity: Call this as bytes are read, see
562
Transport._report_activity
563
:param direction: Will be passed to report_activity
565
:return: The number of bytes copied.
569
# read specified number of bytes
571
while read_length > 0:
572
num_bytes_to_read = min(read_length, buff_size)
574
block = from_file.read(num_bytes_to_read)
578
if report_activity is not None:
579
report_activity(len(block), direction)
582
actual_bytes_read = len(block)
583
read_length -= actual_bytes_read
584
length += actual_bytes_read
588
block = from_file.read(buff_size)
592
if report_activity is not None:
593
report_activity(len(block), direction)
599
def pump_string_file(bytes, file_handle, segment_size=None):
600
"""Write bytes to file_handle in many smaller writes.
602
:param bytes: The string to write.
603
:param file_handle: The file to write to.
605
# Write data in chunks rather than all at once, because very large
606
# writes fail on some platforms (e.g. Windows with SMB mounted
609
segment_size = 5242880 # 5MB
610
segments = range(len(bytes) / segment_size + 1)
611
write = file_handle.write
612
for segment_index in segments:
613
segment = buffer(bytes, segment_index * segment_size, segment_size)
617
def file_iterator(input_file, readsize=32768):
619
b = input_file.read(readsize)
626
"""Calculate the hexdigest of an open file.
628
The file cursor should be already at the start.
640
def size_sha_file(f):
641
"""Calculate the size and hexdigest of an open file.
643
The file cursor should be already at the start and
644
the caller is responsible for closing the file afterwards.
655
return size, s.hexdigest()
658
def sha_file_by_name(fname):
659
"""Calculate the SHA1 of a file by reading the full text"""
661
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
664
b = os.read(f, 1<<16)
672
def sha_strings(strings, _factory=sha):
673
"""Return the sha-1 of concatenation of strings"""
675
map(s.update, strings)
679
def sha_string(f, _factory=sha):
680
return _factory(f).hexdigest()
683
def fingerprint_file(f):
685
return {'size': len(b),
686
'sha1': sha(b).hexdigest()}
689
def compare_files(a, b):
690
"""Returns true if equal in contents"""
701
def local_time_offset(t=None):
702
"""Return offset of local zone from GMT, either at present or at time t."""
705
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
706
return offset.days * 86400 + offset.seconds
708
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
709
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
712
def format_date(t, offset=0, timezone='original', date_fmt=None,
714
"""Return a formatted date string.
716
:param t: Seconds since the epoch.
717
:param offset: Timezone offset in seconds east of utc.
718
:param timezone: How to display the time: 'utc', 'original' for the
719
timezone specified by offset, or 'local' for the process's current
721
:param date_fmt: strftime format.
722
:param show_offset: Whether to append the timezone.
724
(date_fmt, tt, offset_str) = \
725
_format_date(t, offset, timezone, date_fmt, show_offset)
726
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
727
date_str = time.strftime(date_fmt, tt)
728
return date_str + offset_str
731
# Cache of formatted offset strings
735
def format_date_with_offset_in_original_timezone(t, offset=0,
736
_cache=_offset_cache):
737
"""Return a formatted date string in the original timezone.
739
This routine may be faster then format_date.
741
:param t: Seconds since the epoch.
742
:param offset: Timezone offset in seconds east of utc.
746
tt = time.gmtime(t + offset)
747
date_fmt = _default_format_by_weekday_num[tt[6]]
748
date_str = time.strftime(date_fmt, tt)
749
offset_str = _cache.get(offset, None)
750
if offset_str is None:
751
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
752
_cache[offset] = offset_str
753
return date_str + offset_str
756
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
758
"""Return an unicode date string formatted according to the current locale.
760
:param t: Seconds since the epoch.
761
:param offset: Timezone offset in seconds east of utc.
762
:param timezone: How to display the time: 'utc', 'original' for the
763
timezone specified by offset, or 'local' for the process's current
765
:param date_fmt: strftime format.
766
:param show_offset: Whether to append the timezone.
768
(date_fmt, tt, offset_str) = \
769
_format_date(t, offset, timezone, date_fmt, show_offset)
770
date_str = time.strftime(date_fmt, tt)
771
if not isinstance(date_str, unicode):
772
date_str = date_str.decode(get_user_encoding(), 'replace')
773
return date_str + offset_str
776
def _format_date(t, offset, timezone, date_fmt, show_offset):
777
if timezone == 'utc':
780
elif timezone == 'original':
783
tt = time.gmtime(t + offset)
784
elif timezone == 'local':
785
tt = time.localtime(t)
786
offset = local_time_offset(t)
788
raise errors.UnsupportedTimezoneFormat(timezone)
790
date_fmt = "%a %Y-%m-%d %H:%M:%S"
792
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
795
return (date_fmt, tt, offset_str)
798
def compact_date(when):
799
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
802
def format_delta(delta):
803
"""Get a nice looking string for a time delta.
805
:param delta: The time difference in seconds, can be positive or negative.
806
positive indicates time in the past, negative indicates time in the
807
future. (usually time.time() - stored_time)
808
:return: String formatted to show approximate resolution
814
direction = 'in the future'
818
if seconds < 90: # print seconds up to 90 seconds
820
return '%d second %s' % (seconds, direction,)
822
return '%d seconds %s' % (seconds, direction)
824
minutes = int(seconds / 60)
825
seconds -= 60 * minutes
830
if minutes < 90: # print minutes, seconds up to 90 minutes
832
return '%d minute, %d second%s %s' % (
833
minutes, seconds, plural_seconds, direction)
835
return '%d minutes, %d second%s %s' % (
836
minutes, seconds, plural_seconds, direction)
838
hours = int(minutes / 60)
839
minutes -= 60 * hours
846
return '%d hour, %d minute%s %s' % (hours, minutes,
847
plural_minutes, direction)
848
return '%d hours, %d minute%s %s' % (hours, minutes,
849
plural_minutes, direction)
852
"""Return size of given open file."""
853
return os.fstat(f.fileno())[stat.ST_SIZE]
856
# Define rand_bytes based on platform.
858
# Python 2.4 and later have os.urandom,
859
# but it doesn't work on some arches
861
rand_bytes = os.urandom
862
except (NotImplementedError, AttributeError):
863
# If python doesn't have os.urandom, or it doesn't work,
864
# then try to first pull random data from /dev/urandom
866
rand_bytes = file('/dev/urandom', 'rb').read
867
# Otherwise, use this hack as a last resort
868
except (IOError, OSError):
869
# not well seeded, but better than nothing
874
s += chr(random.randint(0, 255))
879
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
881
"""Return a random string of num alphanumeric characters
883
The result only contains lowercase chars because it may be used on
884
case-insensitive filesystems.
887
for raw_byte in rand_bytes(num):
888
s += ALNUM[ord(raw_byte) % 36]
892
## TODO: We could later have path objects that remember their list
893
## decomposition (might be too tricksy though.)
896
"""Turn string into list of parts."""
897
# split on either delimiter because people might use either on
899
ps = re.split(r'[\\/]', p)
904
raise errors.BzrError("sorry, %r not allowed in path" % f)
905
elif (f == '.') or (f == ''):
914
if (f == '..') or (f is None) or (f == ''):
915
raise errors.BzrError("sorry, %r not allowed in path" % f)
919
def parent_directories(filename):
920
"""Return the list of parent directories, deepest first.
922
For example, parent_directories("a/b/c") -> ["a/b", "a"].
925
parts = splitpath(dirname(filename))
927
parents.append(joinpath(parts))
932
_extension_load_failures = []
935
def failed_to_load_extension(exception):
936
"""Handle failing to load a binary extension.
938
This should be called from the ImportError block guarding the attempt to
939
import the native extension. If this function returns, the pure-Python
940
implementation should be loaded instead::
943
>>> import bzrlib._fictional_extension_pyx
944
>>> except ImportError, e:
945
>>> bzrlib.osutils.failed_to_load_extension(e)
946
>>> import bzrlib._fictional_extension_py
948
# NB: This docstring is just an example, not a doctest, because doctest
949
# currently can't cope with the use of lazy imports in this namespace --
952
# This currently doesn't report the failure at the time it occurs, because
953
# they tend to happen very early in startup when we can't check config
954
# files etc, and also we want to report all failures but not spam the user
956
from bzrlib import trace
957
exception_str = str(exception)
958
if exception_str not in _extension_load_failures:
959
trace.mutter("failed to load compiled extension: %s" % exception_str)
960
_extension_load_failures.append(exception_str)
963
def report_extension_load_failures():
964
if not _extension_load_failures:
966
from bzrlib.config import GlobalConfig
967
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
969
# the warnings framework should by default show this only once
970
from bzrlib.trace import warning
972
"bzr: warning: some compiled extensions could not be loaded; "
973
"see <https://answers.launchpad.net/bzr/+faq/703>")
974
# we no longer show the specific missing extensions here, because it makes
975
# the message too long and scary - see
976
# https://bugs.launchpad.net/bzr/+bug/430529
980
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
981
except ImportError, e:
982
failed_to_load_extension(e)
983
from bzrlib._chunks_to_lines_py import chunks_to_lines
987
"""Split s into lines, but without removing the newline characters."""
988
# Trivially convert a fulltext into a 'chunked' representation, and let
989
# chunks_to_lines do the heavy lifting.
990
if isinstance(s, str):
991
# chunks_to_lines only supports 8-bit strings
992
return chunks_to_lines([s])
994
return _split_lines(s)
998
"""Split s into lines, but without removing the newline characters.
1000
This supports Unicode or plain string objects.
1002
lines = s.split('\n')
1003
result = [line + '\n' for line in lines[:-1]]
1005
result.append(lines[-1])
1009
def hardlinks_good():
1010
return sys.platform not in ('win32', 'cygwin', 'darwin')
1013
def link_or_copy(src, dest):
1014
"""Hardlink a file, or copy it if it can't be hardlinked."""
1015
if not hardlinks_good():
1016
shutil.copyfile(src, dest)
1020
except (OSError, IOError), e:
1021
if e.errno != errno.EXDEV:
1023
shutil.copyfile(src, dest)
1026
def delete_any(path):
1027
"""Delete a file, symlink or directory.
1029
Will delete even if readonly.
1032
_delete_file_or_dir(path)
1033
except (OSError, IOError), e:
1034
if e.errno in (errno.EPERM, errno.EACCES):
1035
# make writable and try again
1038
except (OSError, IOError):
1040
_delete_file_or_dir(path)
1045
def _delete_file_or_dir(path):
1046
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1047
# Forgiveness than Permission (EAFP) because:
1048
# - root can damage a solaris file system by using unlink,
1049
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1050
# EACCES, OSX: EPERM) when invoked on a directory.
1051
if isdir(path): # Takes care of symlinks
1058
if getattr(os, 'symlink', None) is not None:
1064
def has_hardlinks():
1065
if getattr(os, 'link', None) is not None:
1071
def host_os_dereferences_symlinks():
1072
return (has_symlinks()
1073
and sys.platform not in ('cygwin', 'win32'))
1076
def readlink(abspath):
1077
"""Return a string representing the path to which the symbolic link points.
1079
:param abspath: The link absolute unicode path.
1081
This his guaranteed to return the symbolic link in unicode in all python
1084
link = abspath.encode(_fs_enc)
1085
target = os.readlink(link)
1086
target = target.decode(_fs_enc)
1090
def contains_whitespace(s):
1091
"""True if there are any whitespace characters in s."""
1092
# string.whitespace can include '\xa0' in certain locales, because it is
1093
# considered "non-breaking-space" as part of ISO-8859-1. But it
1094
# 1) Isn't a breaking whitespace
1095
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1097
# 3) '\xa0' isn't unicode safe since it is >128.
1099
# This should *not* be a unicode set of characters in case the source
1100
# string is not a Unicode string. We can auto-up-cast the characters since
1101
# they are ascii, but we don't want to auto-up-cast the string in case it
1103
for ch in ' \t\n\r\v\f':
1110
def contains_linebreaks(s):
1111
"""True if there is any vertical whitespace in s."""
1119
def relpath(base, path):
1120
"""Return path relative to base, or raise PathNotChild exception.
1122
The path may be either an absolute path or a path relative to the
1123
current working directory.
1125
os.path.commonprefix (python2.4) has a bad bug that it works just
1126
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1127
avoids that problem.
1129
NOTE: `base` should not have a trailing slash otherwise you'll get
1130
PathNotChild exceptions regardless of `path`.
1133
if len(base) < MIN_ABS_PATHLENGTH:
1134
# must have space for e.g. a drive letter
1135
raise ValueError('%r is too short to calculate a relative path'
1143
if len(head) <= len(base) and head != base:
1144
raise errors.PathNotChild(rp, base)
1147
head, tail = split(head)
1152
return pathjoin(*reversed(s))
1157
def _cicp_canonical_relpath(base, path):
1158
"""Return the canonical path relative to base.
1160
Like relpath, but on case-insensitive-case-preserving file-systems, this
1161
will return the relpath as stored on the file-system rather than in the
1162
case specified in the input string, for all existing portions of the path.
1164
This will cause O(N) behaviour if called for every path in a tree; if you
1165
have a number of paths to convert, you should use canonical_relpaths().
1167
# TODO: it should be possible to optimize this for Windows by using the
1168
# win32 API FindFiles function to look for the specified name - but using
1169
# os.listdir() still gives us the correct, platform agnostic semantics in
1172
rel = relpath(base, path)
1173
# '.' will have been turned into ''
1177
abs_base = abspath(base)
1179
_listdir = os.listdir
1181
# use an explicit iterator so we can easily consume the rest on early exit.
1182
bit_iter = iter(rel.split('/'))
1183
for bit in bit_iter:
1186
next_entries = _listdir(current)
1187
except OSError: # enoent, eperm, etc
1188
# We can't find this in the filesystem, so just append the
1190
current = pathjoin(current, bit, *list(bit_iter))
1192
for look in next_entries:
1193
if lbit == look.lower():
1194
current = pathjoin(current, look)
1197
# got to the end, nothing matched, so we just return the
1198
# non-existing bits as they were specified (the filename may be
1199
# the target of a move, for example).
1200
current = pathjoin(current, bit, *list(bit_iter))
1202
return current[len(abs_base):].lstrip('/')
1204
# XXX - TODO - we need better detection/integration of case-insensitive
1205
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1206
# filesystems), for example, so could probably benefit from the same basic
1207
# support there. For now though, only Windows and OSX get that support, and
1208
# they get it for *all* file-systems!
1209
if sys.platform in ('win32', 'darwin'):
1210
canonical_relpath = _cicp_canonical_relpath
1212
canonical_relpath = relpath
1214
def canonical_relpaths(base, paths):
1215
"""Create an iterable to canonicalize a sequence of relative paths.
1217
The intent is for this implementation to use a cache, vastly speeding
1218
up multiple transformations in the same directory.
1220
# but for now, we haven't optimized...
1221
return [canonical_relpath(base, p) for p in paths]
1224
def decode_filename(filename):
1225
"""Decode the filename using the filesystem encoding
1227
If it is unicode, it is returned.
1228
Otherwise it is decoded from the the filesystem's encoding. If decoding
1229
fails, a errors.BadFilenameEncoding exception is raised.
1231
if type(filename) is unicode:
1234
return filename.decode(_fs_enc)
1235
except UnicodeDecodeError:
1236
raise errors.BadFilenameEncoding(filename, _fs_enc)
1239
def safe_unicode(unicode_or_utf8_string):
1240
"""Coerce unicode_or_utf8_string into unicode.
1242
If it is unicode, it is returned.
1243
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1244
wrapped in a BzrBadParameterNotUnicode exception.
1246
if isinstance(unicode_or_utf8_string, unicode):
1247
return unicode_or_utf8_string
1249
return unicode_or_utf8_string.decode('utf8')
1250
except UnicodeDecodeError:
1251
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1254
def safe_utf8(unicode_or_utf8_string):
1255
"""Coerce unicode_or_utf8_string to a utf8 string.
1257
If it is a str, it is returned.
1258
If it is Unicode, it is encoded into a utf-8 string.
1260
if isinstance(unicode_or_utf8_string, str):
1261
# TODO: jam 20070209 This is overkill, and probably has an impact on
1262
# performance if we are dealing with lots of apis that want a
1265
# Make sure it is a valid utf-8 string
1266
unicode_or_utf8_string.decode('utf-8')
1267
except UnicodeDecodeError:
1268
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1269
return unicode_or_utf8_string
1270
return unicode_or_utf8_string.encode('utf-8')
1273
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1274
' Revision id generators should be creating utf8'
1278
def safe_revision_id(unicode_or_utf8_string, warn=True):
1279
"""Revision ids should now be utf8, but at one point they were unicode.
1281
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1283
:param warn: Functions that are sanitizing user data can set warn=False
1284
:return: None or a utf8 revision id.
1286
if (unicode_or_utf8_string is None
1287
or unicode_or_utf8_string.__class__ == str):
1288
return unicode_or_utf8_string
1290
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1292
return cache_utf8.encode(unicode_or_utf8_string)
1295
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1296
' generators should be creating utf8 file ids.')
1299
def safe_file_id(unicode_or_utf8_string, warn=True):
1300
"""File ids should now be utf8, but at one point they were unicode.
1302
This is the same as safe_utf8, except it uses the cached encode functions
1303
to save a little bit of performance.
1305
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1307
:param warn: Functions that are sanitizing user data can set warn=False
1308
:return: None or a utf8 file id.
1310
if (unicode_or_utf8_string is None
1311
or unicode_or_utf8_string.__class__ == str):
1312
return unicode_or_utf8_string
1314
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1316
return cache_utf8.encode(unicode_or_utf8_string)
1319
_platform_normalizes_filenames = False
1320
if sys.platform == 'darwin':
1321
_platform_normalizes_filenames = True
1324
def normalizes_filenames():
1325
"""Return True if this platform normalizes unicode filenames.
1329
return _platform_normalizes_filenames
1332
def _accessible_normalized_filename(path):
1333
"""Get the unicode normalized path, and if you can access the file.
1335
On platforms where the system normalizes filenames (Mac OSX),
1336
you can access a file by any path which will normalize correctly.
1337
On platforms where the system does not normalize filenames
1338
(everything else), you have to access a file by its exact path.
1340
Internally, bzr only supports NFC normalization, since that is
1341
the standard for XML documents.
1343
So return the normalized path, and a flag indicating if the file
1344
can be accessed by that path.
1347
return unicodedata.normalize('NFC', unicode(path)), True
1350
def _inaccessible_normalized_filename(path):
1351
__doc__ = _accessible_normalized_filename.__doc__
1353
normalized = unicodedata.normalize('NFC', unicode(path))
1354
return normalized, normalized == path
1357
if _platform_normalizes_filenames:
1358
normalized_filename = _accessible_normalized_filename
1360
normalized_filename = _inaccessible_normalized_filename
1363
def set_signal_handler(signum, handler, restart_syscall=True):
1364
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1365
on platforms that support that.
1367
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1368
automatically restart (by calling `signal.siginterrupt(signum,
1369
False)`). May be ignored if the feature is not available on this
1370
platform or Python version.
1374
siginterrupt = signal.siginterrupt
1376
# This python implementation doesn't provide signal support, hence no
1379
except AttributeError:
1380
# siginterrupt doesn't exist on this platform, or for this version
1382
siginterrupt = lambda signum, flag: None
1384
def sig_handler(*args):
1385
# Python resets the siginterrupt flag when a signal is
1386
# received. <http://bugs.python.org/issue8354>
1387
# As a workaround for some cases, set it back the way we want it.
1388
siginterrupt(signum, False)
1389
# Now run the handler function passed to set_signal_handler.
1392
sig_handler = handler
1393
old_handler = signal.signal(signum, sig_handler)
1395
siginterrupt(signum, False)
1399
default_terminal_width = 80
1400
"""The default terminal width for ttys.
1402
This is defined so that higher levels can share a common fallback value when
1403
terminal_width() returns None.
1406
# Keep some state so that terminal_width can detect if _terminal_size has
1407
# returned a different size since the process started. See docstring and
1408
# comments of terminal_width for details.
1409
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1410
_terminal_size_state = 'no_data'
1411
_first_terminal_size = None
1413
def terminal_width():
1414
"""Return terminal width.
1416
None is returned if the width can't established precisely.
1419
- if BZR_COLUMNS is set, returns its value
1420
- if there is no controlling terminal, returns None
1421
- query the OS, if the queried size has changed since the last query,
1423
- if COLUMNS is set, returns its value,
1424
- if the OS has a value (even though it's never changed), return its value.
1426
From there, we need to query the OS to get the size of the controlling
1429
On Unices we query the OS by:
1430
- get termios.TIOCGWINSZ
1431
- if an error occurs or a negative value is obtained, returns None
1433
On Windows we query the OS by:
1434
- win32utils.get_console_size() decides,
1435
- returns None on error (provided default value)
1437
# Note to implementors: if changing the rules for determining the width,
1438
# make sure you've considered the behaviour in these cases:
1439
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1440
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1442
# - (add more interesting cases here, if you find any)
1443
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1444
# but we don't want to register a signal handler because it is impossible
1445
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1446
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1447
# time so we can notice if the reported size has changed, which should have
1450
# If BZR_COLUMNS is set, take it, user is always right
1452
return int(os.environ['BZR_COLUMNS'])
1453
except (KeyError, ValueError):
1456
isatty = getattr(sys.stdout, 'isatty', None)
1457
if isatty is None or not isatty():
1458
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1462
width, height = os_size = _terminal_size(None, None)
1463
global _first_terminal_size, _terminal_size_state
1464
if _terminal_size_state == 'no_data':
1465
_first_terminal_size = os_size
1466
_terminal_size_state = 'unchanged'
1467
elif (_terminal_size_state == 'unchanged' and
1468
_first_terminal_size != os_size):
1469
_terminal_size_state = 'changed'
1471
# If the OS claims to know how wide the terminal is, and this value has
1472
# ever changed, use that.
1473
if _terminal_size_state == 'changed':
1474
if width is not None and width > 0:
1477
# If COLUMNS is set, use it.
1479
return int(os.environ['COLUMNS'])
1480
except (KeyError, ValueError):
1483
# Finally, use an unchanged size from the OS, if we have one.
1484
if _terminal_size_state == 'unchanged':
1485
if width is not None and width > 0:
1488
# The width could not be determined.
1492
def _win32_terminal_size(width, height):
1493
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1494
return width, height
1497
def _ioctl_terminal_size(width, height):
1499
import struct, fcntl, termios
1500
s = struct.pack('HHHH', 0, 0, 0, 0)
1501
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1502
height, width = struct.unpack('HHHH', x)[0:2]
1503
except (IOError, AttributeError):
1505
return width, height
1507
_terminal_size = None
1508
"""Returns the terminal size as (width, height).
1510
:param width: Default value for width.
1511
:param height: Default value for height.
1513
This is defined specifically for each OS and query the size of the controlling
1514
terminal. If any error occurs, the provided default values should be returned.
1516
if sys.platform == 'win32':
1517
_terminal_size = _win32_terminal_size
1519
_terminal_size = _ioctl_terminal_size
1522
def supports_executable():
1523
return sys.platform != "win32"
1526
def supports_posix_readonly():
1527
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1529
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1530
directory controls creation/deletion, etc.
1532
And under win32, readonly means that the directory itself cannot be
1533
deleted. The contents of a readonly directory can be changed, unlike POSIX
1534
where files in readonly directories cannot be added, deleted or renamed.
1536
return sys.platform != "win32"
1539
def set_or_unset_env(env_variable, value):
1540
"""Modify the environment, setting or removing the env_variable.
1542
:param env_variable: The environment variable in question
1543
:param value: The value to set the environment to. If None, then
1544
the variable will be removed.
1545
:return: The original value of the environment variable.
1547
orig_val = os.environ.get(env_variable)
1549
if orig_val is not None:
1550
del os.environ[env_variable]
1552
if isinstance(value, unicode):
1553
value = value.encode(get_user_encoding())
1554
os.environ[env_variable] = value
1558
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1561
def check_legal_path(path):
1562
"""Check whether the supplied path is legal.
1563
This is only required on Windows, so we don't test on other platforms
1566
if sys.platform != "win32":
1568
if _validWin32PathRE.match(path) is None:
1569
raise errors.IllegalPath(path)
1572
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1574
def _is_error_enotdir(e):
1575
"""Check if this exception represents ENOTDIR.
1577
Unfortunately, python is very inconsistent about the exception
1578
here. The cases are:
1579
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1580
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1581
which is the windows error code.
1582
3) Windows, Python2.5 uses errno == EINVAL and
1583
winerror == ERROR_DIRECTORY
1585
:param e: An Exception object (expected to be OSError with an errno
1586
attribute, but we should be able to cope with anything)
1587
:return: True if this represents an ENOTDIR error. False otherwise.
1589
en = getattr(e, 'errno', None)
1590
if (en == errno.ENOTDIR
1591
or (sys.platform == 'win32'
1592
and (en == _WIN32_ERROR_DIRECTORY
1593
or (en == errno.EINVAL
1594
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1600
def walkdirs(top, prefix=""):
1601
"""Yield data about all the directories in a tree.
1603
This yields all the data about the contents of a directory at a time.
1604
After each directory has been yielded, if the caller has mutated the list
1605
to exclude some directories, they are then not descended into.
1607
The data yielded is of the form:
1608
((directory-relpath, directory-path-from-top),
1609
[(relpath, basename, kind, lstat, path-from-top), ...]),
1610
- directory-relpath is the relative path of the directory being returned
1611
with respect to top. prefix is prepended to this.
1612
- directory-path-from-root is the path including top for this directory.
1613
It is suitable for use with os functions.
1614
- relpath is the relative path within the subtree being walked.
1615
- basename is the basename of the path
1616
- kind is the kind of the file now. If unknown then the file is not
1617
present within the tree - but it may be recorded as versioned. See
1619
- lstat is the stat data *if* the file was statted.
1620
- planned, not implemented:
1621
path_from_tree_root is the path from the root of the tree.
1623
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1624
allows one to walk a subtree but get paths that are relative to a tree
1626
:return: an iterator over the dirs.
1628
#TODO there is a bit of a smell where the results of the directory-
1629
# summary in this, and the path from the root, may not agree
1630
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1631
# potentially confusing output. We should make this more robust - but
1632
# not at a speed cost. RBC 20060731
1634
_directory = _directory_kind
1635
_listdir = os.listdir
1636
_kind_from_mode = file_kind_from_stat_mode
1637
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1639
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1640
relroot, _, _, _, top = pending.pop()
1642
relprefix = relroot + u'/'
1645
top_slash = top + u'/'
1648
append = dirblock.append
1650
names = sorted(map(decode_filename, _listdir(top)))
1652
if not _is_error_enotdir(e):
1656
abspath = top_slash + name
1657
statvalue = _lstat(abspath)
1658
kind = _kind_from_mode(statvalue.st_mode)
1659
append((relprefix + name, name, kind, statvalue, abspath))
1660
yield (relroot, top), dirblock
1662
# push the user specified dirs from dirblock
1663
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1666
class DirReader(object):
1667
"""An interface for reading directories."""
1669
def top_prefix_to_starting_dir(self, top, prefix=""):
1670
"""Converts top and prefix to a starting dir entry
1672
:param top: A utf8 path
1673
:param prefix: An optional utf8 path to prefix output relative paths
1675
:return: A tuple starting with prefix, and ending with the native
1678
raise NotImplementedError(self.top_prefix_to_starting_dir)
1680
def read_dir(self, prefix, top):
1681
"""Read a specific dir.
1683
:param prefix: A utf8 prefix to be preprended to the path basenames.
1684
:param top: A natively encoded path to read.
1685
:return: A list of the directories contents. Each item contains:
1686
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1688
raise NotImplementedError(self.read_dir)
1691
_selected_dir_reader = None
1694
def _walkdirs_utf8(top, prefix=""):
1695
"""Yield data about all the directories in a tree.
1697
This yields the same information as walkdirs() only each entry is yielded
1698
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1699
are returned as exact byte-strings.
1701
:return: yields a tuple of (dir_info, [file_info])
1702
dir_info is (utf8_relpath, path-from-top)
1703
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1704
if top is an absolute path, path-from-top is also an absolute path.
1705
path-from-top might be unicode or utf8, but it is the correct path to
1706
pass to os functions to affect the file in question. (such as os.lstat)
1708
global _selected_dir_reader
1709
if _selected_dir_reader is None:
1710
fs_encoding = _fs_enc.upper()
1711
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1712
# Win98 doesn't have unicode apis like FindFirstFileW
1713
# TODO: We possibly could support Win98 by falling back to the
1714
# original FindFirstFile, and using TCHAR instead of WCHAR,
1715
# but that gets a bit tricky, and requires custom compiling
1718
from bzrlib._walkdirs_win32 import Win32ReadDir
1719
_selected_dir_reader = Win32ReadDir()
1722
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1723
# ANSI_X3.4-1968 is a form of ASCII
1725
from bzrlib._readdir_pyx import UTF8DirReader
1726
_selected_dir_reader = UTF8DirReader()
1727
except ImportError, e:
1728
failed_to_load_extension(e)
1731
if _selected_dir_reader is None:
1732
# Fallback to the python version
1733
_selected_dir_reader = UnicodeDirReader()
1735
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1736
# But we don't actually uses 1-3 in pending, so set them to None
1737
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1738
read_dir = _selected_dir_reader.read_dir
1739
_directory = _directory_kind
1741
relroot, _, _, _, top = pending[-1].pop()
1744
dirblock = sorted(read_dir(relroot, top))
1745
yield (relroot, top), dirblock
1746
# push the user specified dirs from dirblock
1747
next = [d for d in reversed(dirblock) if d[2] == _directory]
1749
pending.append(next)
1752
class UnicodeDirReader(DirReader):
1753
"""A dir reader for non-utf8 file systems, which transcodes."""
1755
__slots__ = ['_utf8_encode']
1758
self._utf8_encode = codecs.getencoder('utf8')
1760
def top_prefix_to_starting_dir(self, top, prefix=""):
1761
"""See DirReader.top_prefix_to_starting_dir."""
1762
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1764
def read_dir(self, prefix, top):
1765
"""Read a single directory from a non-utf8 file system.
1767
top, and the abspath element in the output are unicode, all other paths
1768
are utf8. Local disk IO is done via unicode calls to listdir etc.
1770
This is currently the fallback code path when the filesystem encoding is
1771
not UTF-8. It may be better to implement an alternative so that we can
1772
safely handle paths that are not properly decodable in the current
1775
See DirReader.read_dir for details.
1777
_utf8_encode = self._utf8_encode
1779
_listdir = os.listdir
1780
_kind_from_mode = file_kind_from_stat_mode
1783
relprefix = prefix + '/'
1786
top_slash = top + u'/'
1789
append = dirblock.append
1790
for name in sorted(_listdir(top)):
1792
name_utf8 = _utf8_encode(name)[0]
1793
except UnicodeDecodeError:
1794
raise errors.BadFilenameEncoding(
1795
_utf8_encode(relprefix)[0] + name, _fs_enc)
1796
abspath = top_slash + name
1797
statvalue = _lstat(abspath)
1798
kind = _kind_from_mode(statvalue.st_mode)
1799
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1803
def copy_tree(from_path, to_path, handlers={}):
1804
"""Copy all of the entries in from_path into to_path.
1806
:param from_path: The base directory to copy.
1807
:param to_path: The target directory. If it does not exist, it will
1809
:param handlers: A dictionary of functions, which takes a source and
1810
destinations for files, directories, etc.
1811
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1812
'file', 'directory', and 'symlink' should always exist.
1813
If they are missing, they will be replaced with 'os.mkdir()',
1814
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1816
# Now, just copy the existing cached tree to the new location
1817
# We use a cheap trick here.
1818
# Absolute paths are prefixed with the first parameter
1819
# relative paths are prefixed with the second.
1820
# So we can get both the source and target returned
1821
# without any extra work.
1823
def copy_dir(source, dest):
1826
def copy_link(source, dest):
1827
"""Copy the contents of a symlink"""
1828
link_to = os.readlink(source)
1829
os.symlink(link_to, dest)
1831
real_handlers = {'file':shutil.copy2,
1832
'symlink':copy_link,
1833
'directory':copy_dir,
1835
real_handlers.update(handlers)
1837
if not os.path.exists(to_path):
1838
real_handlers['directory'](from_path, to_path)
1840
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1841
for relpath, name, kind, st, abspath in entries:
1842
real_handlers[kind](abspath, relpath)
1845
def copy_ownership_from_path(dst, src=None):
1846
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1848
If src is None, the containing directory is used as source. If chown
1849
fails, the error is ignored and a warning is printed.
1851
chown = getattr(os, 'chown', None)
1856
src = os.path.dirname(dst)
1862
chown(dst, s.st_uid, s.st_gid)
1864
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1867
def path_prefix_key(path):
1868
"""Generate a prefix-order path key for path.
1870
This can be used to sort paths in the same way that walkdirs does.
1872
return (dirname(path) , path)
1875
def compare_paths_prefix_order(path_a, path_b):
1876
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1877
key_a = path_prefix_key(path_a)
1878
key_b = path_prefix_key(path_b)
1879
return cmp(key_a, key_b)
1882
_cached_user_encoding = None
1885
def get_user_encoding(use_cache=True):
1886
"""Find out what the preferred user encoding is.
1888
This is generally the encoding that is used for command line parameters
1889
and file contents. This may be different from the terminal encoding
1890
or the filesystem encoding.
1892
:param use_cache: Enable cache for detected encoding.
1893
(This parameter is turned on by default,
1894
and required only for selftesting)
1896
:return: A string defining the preferred user encoding
1898
global _cached_user_encoding
1899
if _cached_user_encoding is not None and use_cache:
1900
return _cached_user_encoding
1902
if sys.platform == 'darwin':
1903
# python locale.getpreferredencoding() always return
1904
# 'mac-roman' on darwin. That's a lie.
1905
sys.platform = 'posix'
1907
if os.environ.get('LANG', None) is None:
1908
# If LANG is not set, we end up with 'ascii', which is bad
1909
# ('mac-roman' is more than ascii), so we set a default which
1910
# will give us UTF-8 (which appears to work in all cases on
1911
# OSX). Users are still free to override LANG of course, as
1912
# long as it give us something meaningful. This work-around
1913
# *may* not be needed with python 3k and/or OSX 10.5, but will
1914
# work with them too -- vila 20080908
1915
os.environ['LANG'] = 'en_US.UTF-8'
1918
sys.platform = 'darwin'
1923
user_encoding = locale.getpreferredencoding()
1924
except locale.Error, e:
1925
sys.stderr.write('bzr: warning: %s\n'
1926
' Could not determine what text encoding to use.\n'
1927
' This error usually means your Python interpreter\n'
1928
' doesn\'t support the locale set by $LANG (%s)\n'
1929
" Continuing with ascii encoding.\n"
1930
% (e, os.environ.get('LANG')))
1931
user_encoding = 'ascii'
1933
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1934
# treat that as ASCII, and not support printing unicode characters to the
1937
# For python scripts run under vim, we get '', so also treat that as ASCII
1938
if user_encoding in (None, 'cp0', ''):
1939
user_encoding = 'ascii'
1943
codecs.lookup(user_encoding)
1945
sys.stderr.write('bzr: warning:'
1946
' unknown encoding %s.'
1947
' Continuing with ascii encoding.\n'
1950
user_encoding = 'ascii'
1953
_cached_user_encoding = user_encoding
1955
return user_encoding
1958
def get_diff_header_encoding():
1959
return get_terminal_encoding()
1962
def get_host_name():
1963
"""Return the current unicode host name.
1965
This is meant to be used in place of socket.gethostname() because that
1966
behaves inconsistently on different platforms.
1968
if sys.platform == "win32":
1970
return win32utils.get_host_name()
1973
return socket.gethostname().decode(get_user_encoding())
1976
# We must not read/write any more than 64k at a time from/to a socket so we
1977
# don't risk "no buffer space available" errors on some platforms. Windows in
1978
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1980
MAX_SOCKET_CHUNK = 64 * 1024
1982
def read_bytes_from_socket(sock, report_activity=None,
1983
max_read_size=MAX_SOCKET_CHUNK):
1984
"""Read up to max_read_size of bytes from sock and notify of progress.
1986
Translates "Connection reset by peer" into file-like EOF (return an
1987
empty string rather than raise an error), and repeats the recv if
1988
interrupted by a signal.
1992
bytes = sock.recv(max_read_size)
1993
except socket.error, e:
1995
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1996
# The connection was closed by the other side. Callers expect
1997
# an empty string to signal end-of-stream.
1999
elif eno == errno.EINTR:
2000
# Retry the interrupted recv.
2004
if report_activity is not None:
2005
report_activity(len(bytes), 'read')
2009
def recv_all(socket, count):
2010
"""Receive an exact number of bytes.
2012
Regular Socket.recv() may return less than the requested number of bytes,
2013
depending on what's in the OS buffer. MSG_WAITALL is not available
2014
on all platforms, but this should work everywhere. This will return
2015
less than the requested amount if the remote end closes.
2017
This isn't optimized and is intended mostly for use in testing.
2020
while len(b) < count:
2021
new = read_bytes_from_socket(socket, None, count - len(b))
2028
def send_all(sock, bytes, report_activity=None):
2029
"""Send all bytes on a socket.
2031
Breaks large blocks in smaller chunks to avoid buffering limitations on
2032
some platforms, and catches EINTR which may be thrown if the send is
2033
interrupted by a signal.
2035
This is preferred to socket.sendall(), because it avoids portability bugs
2036
and provides activity reporting.
2038
:param report_activity: Call this as bytes are read, see
2039
Transport._report_activity
2042
byte_count = len(bytes)
2043
while sent_total < byte_count:
2045
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2046
except socket.error, e:
2047
if e.args[0] != errno.EINTR:
2051
report_activity(sent, 'write')
2054
def dereference_path(path):
2055
"""Determine the real path to a file.
2057
All parent elements are dereferenced. But the file itself is not
2059
:param path: The original path. May be absolute or relative.
2060
:return: the real path *to* the file
2062
parent, base = os.path.split(path)
2063
# The pathjoin for '.' is a workaround for Python bug #1213894.
2064
# (initial path components aren't dereferenced)
2065
return pathjoin(realpath(pathjoin('.', parent)), base)
2068
def supports_mapi():
2069
"""Return True if we can use MAPI to launch a mail client."""
2070
return sys.platform == "win32"
2073
def resource_string(package, resource_name):
2074
"""Load a resource from a package and return it as a string.
2076
Note: Only packages that start with bzrlib are currently supported.
2078
This is designed to be a lightweight implementation of resource
2079
loading in a way which is API compatible with the same API from
2081
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2082
If and when pkg_resources becomes a standard library, this routine
2085
# Check package name is within bzrlib
2086
if package == "bzrlib":
2087
resource_relpath = resource_name
2088
elif package.startswith("bzrlib."):
2089
package = package[len("bzrlib."):].replace('.', os.sep)
2090
resource_relpath = pathjoin(package, resource_name)
2092
raise errors.BzrError('resource package %s not in bzrlib' % package)
2094
# Map the resource to a file and read its contents
2095
base = dirname(bzrlib.__file__)
2096
if getattr(sys, 'frozen', None): # bzr.exe
2097
base = abspath(pathjoin(base, '..', '..'))
2098
f = file(pathjoin(base, resource_relpath), "rU")
2104
def file_kind_from_stat_mode_thunk(mode):
2105
global file_kind_from_stat_mode
2106
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2108
from bzrlib._readdir_pyx import UTF8DirReader
2109
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2110
except ImportError, e:
2111
# This is one time where we won't warn that an extension failed to
2112
# load. The extension is never available on Windows anyway.
2113
from bzrlib._readdir_py import (
2114
_kind_from_mode as file_kind_from_stat_mode
2116
return file_kind_from_stat_mode(mode)
2117
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2120
def file_kind(f, _lstat=os.lstat):
2122
return file_kind_from_stat_mode(_lstat(f).st_mode)
2124
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2125
raise errors.NoSuchFile(f)
2129
def until_no_eintr(f, *a, **kw):
2130
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2132
WARNING: you must be certain that it is safe to retry the call repeatedly
2133
if EINTR does occur. This is typically only true for low-level operations
2134
like os.read. If in any doubt, don't use this.
2136
Keep in mind that this is not a complete solution to EINTR. There is
2137
probably code in the Python standard library and other dependencies that
2138
may encounter EINTR if a signal arrives (and there is signal handler for
2139
that signal). So this function can reduce the impact for IO that bzrlib
2140
directly controls, but it is not a complete solution.
2142
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2146
except (IOError, OSError), e:
2147
if e.errno == errno.EINTR:
2152
def re_compile_checked(re_string, flags=0, where=""):
2153
"""Return a compiled re, or raise a sensible error.
2155
This should only be used when compiling user-supplied REs.
2157
:param re_string: Text form of regular expression.
2158
:param flags: eg re.IGNORECASE
2159
:param where: Message explaining to the user the context where
2160
it occurred, eg 'log search filter'.
2162
# from https://bugs.launchpad.net/bzr/+bug/251352
2164
re_obj = re.compile(re_string, flags)
2169
where = ' in ' + where
2170
# despite the name 'error' is a type
2171
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2172
% (where, re_string, e))
2175
if sys.platform == "win32":
2178
return msvcrt.getch()
2183
fd = sys.stdin.fileno()
2184
settings = termios.tcgetattr(fd)
2187
ch = sys.stdin.read(1)
2189
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2193
if sys.platform == 'linux2':
2194
def _local_concurrency():
2196
prefix = 'processor'
2197
for line in file('/proc/cpuinfo', 'rb'):
2198
if line.startswith(prefix):
2199
concurrency = int(line[line.find(':')+1:]) + 1
2201
elif sys.platform == 'darwin':
2202
def _local_concurrency():
2203
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2204
stdout=subprocess.PIPE).communicate()[0]
2205
elif sys.platform[0:7] == 'freebsd':
2206
def _local_concurrency():
2207
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2208
stdout=subprocess.PIPE).communicate()[0]
2209
elif sys.platform == 'sunos5':
2210
def _local_concurrency():
2211
return subprocess.Popen(['psrinfo', '-p',],
2212
stdout=subprocess.PIPE).communicate()[0]
2213
elif sys.platform == "win32":
2214
def _local_concurrency():
2215
# This appears to return the number of cores.
2216
return os.environ.get('NUMBER_OF_PROCESSORS')
2218
def _local_concurrency():
2223
_cached_local_concurrency = None
2225
def local_concurrency(use_cache=True):
2226
"""Return how many processes can be run concurrently.
2228
Rely on platform specific implementations and default to 1 (one) if
2229
anything goes wrong.
2231
global _cached_local_concurrency
2233
if _cached_local_concurrency is not None and use_cache:
2234
return _cached_local_concurrency
2236
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2237
if concurrency is None:
2239
concurrency = _local_concurrency()
2240
except (OSError, IOError):
2243
concurrency = int(concurrency)
2244
except (TypeError, ValueError):
2247
_cached_concurrency = concurrency
2251
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2252
"""A stream writer that doesn't decode str arguments."""
2254
def __init__(self, encode, stream, errors='strict'):
2255
codecs.StreamWriter.__init__(self, stream, errors)
2256
self.encode = encode
2258
def write(self, object):
2259
if type(object) is str:
2260
self.stream.write(object)
2262
data, _ = self.encode(object, self.errors)
2263
self.stream.write(data)
2265
if sys.platform == 'win32':
2266
def open_file(filename, mode='r', bufsize=-1):
2267
"""This function is used to override the ``open`` builtin.
2269
But it uses O_NOINHERIT flag so the file handle is not inherited by
2270
child processes. Deleting or renaming a closed file opened with this
2271
function is not blocking child processes.
2273
writing = 'w' in mode
2274
appending = 'a' in mode
2275
updating = '+' in mode
2276
binary = 'b' in mode
2279
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2280
# for flags for each modes.
2290
flags |= os.O_WRONLY
2291
flags |= os.O_CREAT | os.O_TRUNC
2296
flags |= os.O_WRONLY
2297
flags |= os.O_CREAT | os.O_APPEND
2302
flags |= os.O_RDONLY
2304
return os.fdopen(os.open(filename, flags), mode, bufsize)
2309
def getuser_unicode():
2310
"""Return the username as unicode.
2313
user_encoding = get_user_encoding()
2314
username = getpass.getuser().decode(user_encoding)
2315
except UnicodeDecodeError:
2316
raise errors.BzrError("Can't decode username as %s." % \
added
removed
bzrlib/osutils.py
