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
20
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
21
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
27
from bzrlib.lazy_import import lazy_import
28
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
46
from tempfile import (
59
from bzrlib.symbol_versioning import (
64
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
66
if sys.version_info < (2, 5):
67
import md5 as _mod_md5
69
import sha as _mod_sha
79
from bzrlib import symbol_versioning
82
# Cross platform wall-clock time functionality with decent resolution.
83
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
84
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
85
# synchronized with ``time.time()``, this is only meant to be used to find
86
# delta times by subtracting from another call to this function.
87
timer_func = time.time
88
if sys.platform == 'win32':
89
timer_func = time.clock
91
# On win32, O_BINARY is used to indicate the file should
92
# be opened in binary mode, rather than text mode.
93
# On other platforms, O_BINARY doesn't exist, because
94
# they always open in binary mode, so it is okay to
95
# OR with 0 on those platforms.
96
# O_NOINHERIT and O_TEXT exists only on win32 too.
97
O_BINARY = getattr(os, 'O_BINARY', 0)
98
O_TEXT = getattr(os, 'O_TEXT', 0)
99
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
102
def get_unicode_argv():
104
user_encoding = get_user_encoding()
105
return [a.decode(user_encoding) for a in sys.argv[1:]]
106
except UnicodeDecodeError:
107
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
111
def make_readonly(filename):
112
"""Make a filename read-only."""
113
mod = os.lstat(filename).st_mode
114
if not stat.S_ISLNK(mod):
116
os.chmod(filename, mod)
119
def make_writable(filename):
120
mod = os.lstat(filename).st_mode
121
if not stat.S_ISLNK(mod):
123
os.chmod(filename, mod)
126
def minimum_path_selection(paths):
127
"""Return the smallset subset of paths which are outside paths.
129
:param paths: A container (and hence not None) of paths.
130
:return: A set of paths sufficient to include everything in paths via
131
is_inside, drawn from the paths parameter.
137
return path.split('/')
138
sorted_paths = sorted(list(paths), key=sort_key)
140
search_paths = [sorted_paths[0]]
141
for path in sorted_paths[1:]:
142
if not is_inside(search_paths[-1], path):
143
# This path is unique, add it
144
search_paths.append(path)
146
return set(search_paths)
153
"""Return a quoted filename filename
155
This previously used backslash quoting, but that works poorly on
157
# TODO: I'm not really sure this is the best format either.x
159
if _QUOTE_RE is None:
160
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
162
if _QUOTE_RE.search(f):
168
_directory_kind = 'directory'
171
"""Return the current umask"""
172
# Assume that people aren't messing with the umask while running
173
# XXX: This is not thread safe, but there is no way to get the
174
# umask without setting it
182
_directory_kind: "/",
184
'tree-reference': '+',
188
def kind_marker(kind):
190
return _kind_marker_map[kind]
192
# Slightly faster than using .get(, '') when the common case is that
197
lexists = getattr(os.path, 'lexists', None)
201
stat = getattr(os, 'lstat', os.stat)
205
if e.errno == errno.ENOENT:
208
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
211
def fancy_rename(old, new, rename_func, unlink_func):
212
"""A fancy rename, when you don't have atomic rename.
214
:param old: The old path, to rename from
215
:param new: The new path, to rename to
216
:param rename_func: The potentially non-atomic rename function
217
:param unlink_func: A way to delete the target file if the full rename
220
# sftp rename doesn't allow overwriting, so play tricks:
221
base = os.path.basename(new)
222
dirname = os.path.dirname(new)
223
# callers use different encodings for the paths so the following MUST
224
# respect that. We rely on python upcasting to unicode if new is unicode
225
# and keeping a str if not.
226
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
227
os.getpid(), rand_chars(10))
228
tmp_name = pathjoin(dirname, tmp_name)
230
# Rename the file out of the way, but keep track if it didn't exist
231
# We don't want to grab just any exception
232
# something like EACCES should prevent us from continuing
233
# The downside is that the rename_func has to throw an exception
234
# with an errno = ENOENT, or NoSuchFile
237
rename_func(new, tmp_name)
238
except (errors.NoSuchFile,), e:
241
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
242
# function raises an IOError with errno is None when a rename fails.
243
# This then gets caught here.
244
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
247
if (getattr(e, 'errno', None) is None
248
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
257
# This may throw an exception, in which case success will
259
rename_func(old, new)
261
except (IOError, OSError), e:
262
# source and target may be aliases of each other (e.g. on a
263
# case-insensitive filesystem), so we may have accidentally renamed
264
# source by when we tried to rename target
265
failure_exc = sys.exc_info()
266
if (file_existed and e.errno in (None, errno.ENOENT)
267
and old.lower() == new.lower()):
268
# source and target are the same file on a case-insensitive
269
# filesystem, so we don't generate an exception
273
# If the file used to exist, rename it back into place
274
# otherwise just delete it from the tmp location
276
unlink_func(tmp_name)
278
rename_func(tmp_name, new)
279
if failure_exc is not None:
280
raise failure_exc[0], failure_exc[1], failure_exc[2]
283
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
284
# choke on a Unicode string containing a relative path if
285
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
287
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
288
def _posix_abspath(path):
289
# jam 20060426 rather than encoding to fsencoding
290
# copy posixpath.abspath, but use os.getcwdu instead
291
if not posixpath.isabs(path):
292
path = posixpath.join(getcwd(), path)
293
return posixpath.normpath(path)
296
def _posix_realpath(path):
297
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
300
def _win32_fixdrive(path):
301
"""Force drive letters to be consistent.
303
win32 is inconsistent whether it returns lower or upper case
304
and even if it was consistent the user might type the other
305
so we force it to uppercase
306
running python.exe under cmd.exe return capital C:\\
307
running win32 python inside a cygwin shell returns lowercase c:\\
309
drive, path = _nt_splitdrive(path)
310
return drive.upper() + path
313
def _win32_abspath(path):
314
# Real _nt_abspath doesn't have a problem with a unicode cwd
315
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
318
def _win98_abspath(path):
319
"""Return the absolute version of a path.
320
Windows 98 safe implementation (python reimplementation
321
of Win32 API function GetFullPathNameW)
326
# \\HOST\path => //HOST/path
327
# //HOST/path => //HOST/path
328
# path => C:/cwd/path
331
# check for absolute path
332
drive = _nt_splitdrive(path)[0]
333
if drive == '' and path[:2] not in('//','\\\\'):
335
# we cannot simply os.path.join cwd and path
336
# because os.path.join('C:','/path') produce '/path'
337
# and this is incorrect
338
if path[:1] in ('/','\\'):
339
cwd = _nt_splitdrive(cwd)[0]
341
path = cwd + '\\' + path
342
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
345
def _win32_realpath(path):
346
# Real _nt_realpath doesn't have a problem with a unicode cwd
347
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
350
def _win32_pathjoin(*args):
351
return _nt_join(*args).replace('\\', '/')
354
def _win32_normpath(path):
355
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
359
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
362
def _win32_mkdtemp(*args, **kwargs):
363
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
366
def _win32_rename(old, new):
367
"""We expect to be able to atomically replace 'new' with old.
369
On win32, if new exists, it must be moved out of the way first,
373
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
375
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
376
# If we try to rename a non-existant file onto cwd, we get
377
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
378
# if the old path doesn't exist, sometimes we get EACCES
379
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
385
return unicodedata.normalize('NFC', os.getcwdu())
388
# Default is to just use the python builtins, but these can be rebound on
389
# particular platforms.
390
abspath = _posix_abspath
391
realpath = _posix_realpath
392
pathjoin = os.path.join
393
normpath = os.path.normpath
396
dirname = os.path.dirname
397
basename = os.path.basename
398
split = os.path.split
399
splitext = os.path.splitext
400
# These were already imported into local scope
401
# mkdtemp = tempfile.mkdtemp
402
# rmtree = shutil.rmtree
404
MIN_ABS_PATHLENGTH = 1
407
if sys.platform == 'win32':
408
if win32utils.winver == 'Windows 98':
409
abspath = _win98_abspath
411
abspath = _win32_abspath
412
realpath = _win32_realpath
413
pathjoin = _win32_pathjoin
414
normpath = _win32_normpath
415
getcwd = _win32_getcwd
416
mkdtemp = _win32_mkdtemp
417
rename = _win32_rename
419
MIN_ABS_PATHLENGTH = 3
421
def _win32_delete_readonly(function, path, excinfo):
422
"""Error handler for shutil.rmtree function [for win32]
423
Helps to remove files and dirs marked as read-only.
425
exception = excinfo[1]
426
if function in (os.remove, os.rmdir) \
427
and isinstance(exception, OSError) \
428
and exception.errno == errno.EACCES:
434
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
435
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
436
return shutil.rmtree(path, ignore_errors, onerror)
438
f = win32utils.get_unicode_argv # special function or None
442
elif sys.platform == 'darwin':
446
def get_terminal_encoding():
447
"""Find the best encoding for printing to the screen.
449
This attempts to check both sys.stdout and sys.stdin to see
450
what encoding they are in, and if that fails it falls back to
451
osutils.get_user_encoding().
452
The problem is that on Windows, locale.getpreferredencoding()
453
is not the same encoding as that used by the console:
454
http://mail.python.org/pipermail/python-list/2003-May/162357.html
456
On my standard US Windows XP, the preferred encoding is
457
cp1252, but the console is cp437
459
from bzrlib.trace import mutter
460
output_encoding = getattr(sys.stdout, 'encoding', None)
461
if not output_encoding:
462
input_encoding = getattr(sys.stdin, 'encoding', None)
463
if not input_encoding:
464
output_encoding = get_user_encoding()
465
mutter('encoding stdout as osutils.get_user_encoding() %r',
468
output_encoding = input_encoding
469
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
471
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
472
if output_encoding == 'cp0':
473
# invalid encoding (cp0 means 'no codepage' on Windows)
474
output_encoding = get_user_encoding()
475
mutter('cp0 is invalid encoding.'
476
' encoding stdout as osutils.get_user_encoding() %r',
480
codecs.lookup(output_encoding)
482
sys.stderr.write('bzr: warning:'
483
' unknown terminal encoding %s.\n'
484
' Using encoding %s instead.\n'
485
% (output_encoding, get_user_encoding())
487
output_encoding = get_user_encoding()
489
return output_encoding
492
def normalizepath(f):
493
if getattr(os.path, 'realpath', None) is not None:
497
[p,e] = os.path.split(f)
498
if e == "" or e == "." or e == "..":
501
return pathjoin(F(p), e)
505
"""True if f is an accessible directory."""
507
return S_ISDIR(os.lstat(f)[ST_MODE])
513
"""True if f is a regular file."""
515
return S_ISREG(os.lstat(f)[ST_MODE])
520
"""True if f is a symlink."""
522
return S_ISLNK(os.lstat(f)[ST_MODE])
526
def is_inside(dir, fname):
527
"""True if fname is inside dir.
529
The parameters should typically be passed to osutils.normpath first, so
530
that . and .. and repeated slashes are eliminated, and the separators
531
are canonical for the platform.
533
The empty string as a dir name is taken as top-of-tree and matches
536
# XXX: Most callers of this can actually do something smarter by
537
# looking at the inventory
547
return fname.startswith(dir)
550
def is_inside_any(dir_list, fname):
551
"""True if fname is inside any of given dirs."""
552
for dirname in dir_list:
553
if is_inside(dirname, fname):
558
def is_inside_or_parent_of_any(dir_list, fname):
559
"""True if fname is a child or a parent of any of the given files."""
560
for dirname in dir_list:
561
if is_inside(dirname, fname) or is_inside(fname, dirname):
566
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
567
report_activity=None, direction='read'):
568
"""Copy contents of one file to another.
570
The read_length can either be -1 to read to end-of-file (EOF) or
571
it can specify the maximum number of bytes to read.
573
The buff_size represents the maximum size for each read operation
574
performed on from_file.
576
:param report_activity: Call this as bytes are read, see
577
Transport._report_activity
578
:param direction: Will be passed to report_activity
580
:return: The number of bytes copied.
584
# read specified number of bytes
586
while read_length > 0:
587
num_bytes_to_read = min(read_length, buff_size)
589
block = from_file.read(num_bytes_to_read)
593
if report_activity is not None:
594
report_activity(len(block), direction)
597
actual_bytes_read = len(block)
598
read_length -= actual_bytes_read
599
length += actual_bytes_read
603
block = from_file.read(buff_size)
607
if report_activity is not None:
608
report_activity(len(block), direction)
614
def pump_string_file(bytes, file_handle, segment_size=None):
615
"""Write bytes to file_handle in many smaller writes.
617
:param bytes: The string to write.
618
:param file_handle: The file to write to.
620
# Write data in chunks rather than all at once, because very large
621
# writes fail on some platforms (e.g. Windows with SMB mounted
624
segment_size = 5242880 # 5MB
625
segments = range(len(bytes) / segment_size + 1)
626
write = file_handle.write
627
for segment_index in segments:
628
segment = buffer(bytes, segment_index * segment_size, segment_size)
632
def file_iterator(input_file, readsize=32768):
634
b = input_file.read(readsize)
641
"""Calculate the hexdigest of an open file.
643
The file cursor should be already at the start.
655
def size_sha_file(f):
656
"""Calculate the size and hexdigest of an open file.
658
The file cursor should be already at the start and
659
the caller is responsible for closing the file afterwards.
670
return size, s.hexdigest()
673
def sha_file_by_name(fname):
674
"""Calculate the SHA1 of a file by reading the full text"""
676
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
679
b = os.read(f, 1<<16)
687
def sha_strings(strings, _factory=sha):
688
"""Return the sha-1 of concatenation of strings"""
690
map(s.update, strings)
694
def sha_string(f, _factory=sha):
695
return _factory(f).hexdigest()
698
def fingerprint_file(f):
700
return {'size': len(b),
701
'sha1': sha(b).hexdigest()}
704
def compare_files(a, b):
705
"""Returns true if equal in contents"""
716
def local_time_offset(t=None):
717
"""Return offset of local zone from GMT, either at present or at time t."""
720
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
721
return offset.days * 86400 + offset.seconds
723
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
724
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
727
def format_date(t, offset=0, timezone='original', date_fmt=None,
729
"""Return a formatted date string.
731
:param t: Seconds since the epoch.
732
:param offset: Timezone offset in seconds east of utc.
733
:param timezone: How to display the time: 'utc', 'original' for the
734
timezone specified by offset, or 'local' for the process's current
736
:param date_fmt: strftime format.
737
:param show_offset: Whether to append the timezone.
739
(date_fmt, tt, offset_str) = \
740
_format_date(t, offset, timezone, date_fmt, show_offset)
741
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
742
date_str = time.strftime(date_fmt, tt)
743
return date_str + offset_str
746
# Cache of formatted offset strings
750
def format_date_with_offset_in_original_timezone(t, offset=0,
751
_cache=_offset_cache):
752
"""Return a formatted date string in the original timezone.
754
This routine may be faster then format_date.
756
:param t: Seconds since the epoch.
757
:param offset: Timezone offset in seconds east of utc.
761
tt = time.gmtime(t + offset)
762
date_fmt = _default_format_by_weekday_num[tt[6]]
763
date_str = time.strftime(date_fmt, tt)
764
offset_str = _cache.get(offset, None)
765
if offset_str is None:
766
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
767
_cache[offset] = offset_str
768
return date_str + offset_str
771
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
773
"""Return an unicode date string formatted according to the current locale.
775
:param t: Seconds since the epoch.
776
:param offset: Timezone offset in seconds east of utc.
777
:param timezone: How to display the time: 'utc', 'original' for the
778
timezone specified by offset, or 'local' for the process's current
780
:param date_fmt: strftime format.
781
:param show_offset: Whether to append the timezone.
783
(date_fmt, tt, offset_str) = \
784
_format_date(t, offset, timezone, date_fmt, show_offset)
785
date_str = time.strftime(date_fmt, tt)
786
if not isinstance(date_str, unicode):
787
date_str = date_str.decode(get_user_encoding(), 'replace')
788
return date_str + offset_str
791
def _format_date(t, offset, timezone, date_fmt, show_offset):
792
if timezone == 'utc':
795
elif timezone == 'original':
798
tt = time.gmtime(t + offset)
799
elif timezone == 'local':
800
tt = time.localtime(t)
801
offset = local_time_offset(t)
803
raise errors.UnsupportedTimezoneFormat(timezone)
805
date_fmt = "%a %Y-%m-%d %H:%M:%S"
807
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
810
return (date_fmt, tt, offset_str)
813
def compact_date(when):
814
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
817
def format_delta(delta):
818
"""Get a nice looking string for a time delta.
820
:param delta: The time difference in seconds, can be positive or negative.
821
positive indicates time in the past, negative indicates time in the
822
future. (usually time.time() - stored_time)
823
:return: String formatted to show approximate resolution
829
direction = 'in the future'
833
if seconds < 90: # print seconds up to 90 seconds
835
return '%d second %s' % (seconds, direction,)
837
return '%d seconds %s' % (seconds, direction)
839
minutes = int(seconds / 60)
840
seconds -= 60 * minutes
845
if minutes < 90: # print minutes, seconds up to 90 minutes
847
return '%d minute, %d second%s %s' % (
848
minutes, seconds, plural_seconds, direction)
850
return '%d minutes, %d second%s %s' % (
851
minutes, seconds, plural_seconds, direction)
853
hours = int(minutes / 60)
854
minutes -= 60 * hours
861
return '%d hour, %d minute%s %s' % (hours, minutes,
862
plural_minutes, direction)
863
return '%d hours, %d minute%s %s' % (hours, minutes,
864
plural_minutes, direction)
867
"""Return size of given open file."""
868
return os.fstat(f.fileno())[ST_SIZE]
871
# Define rand_bytes based on platform.
873
# Python 2.4 and later have os.urandom,
874
# but it doesn't work on some arches
876
rand_bytes = os.urandom
877
except (NotImplementedError, AttributeError):
878
# If python doesn't have os.urandom, or it doesn't work,
879
# then try to first pull random data from /dev/urandom
881
rand_bytes = file('/dev/urandom', 'rb').read
882
# Otherwise, use this hack as a last resort
883
except (IOError, OSError):
884
# not well seeded, but better than nothing
889
s += chr(random.randint(0, 255))
894
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
896
"""Return a random string of num alphanumeric characters
898
The result only contains lowercase chars because it may be used on
899
case-insensitive filesystems.
902
for raw_byte in rand_bytes(num):
903
s += ALNUM[ord(raw_byte) % 36]
907
## TODO: We could later have path objects that remember their list
908
## decomposition (might be too tricksy though.)
911
"""Turn string into list of parts."""
912
# split on either delimiter because people might use either on
914
ps = re.split(r'[\\/]', p)
919
raise errors.BzrError("sorry, %r not allowed in path" % f)
920
elif (f == '.') or (f == ''):
929
if (f == '..') or (f is None) or (f == ''):
930
raise errors.BzrError("sorry, %r not allowed in path" % f)
934
def parent_directories(filename):
935
"""Return the list of parent directories, deepest first.
937
For example, parent_directories("a/b/c") -> ["a/b", "a"].
940
parts = splitpath(dirname(filename))
942
parents.append(joinpath(parts))
947
_extension_load_failures = []
950
def failed_to_load_extension(exception):
951
"""Handle failing to load a binary extension.
953
This should be called from the ImportError block guarding the attempt to
954
import the native extension. If this function returns, the pure-Python
955
implementation should be loaded instead::
958
>>> import bzrlib._fictional_extension_pyx
959
>>> except ImportError, e:
960
>>> bzrlib.osutils.failed_to_load_extension(e)
961
>>> import bzrlib._fictional_extension_py
963
# NB: This docstring is just an example, not a doctest, because doctest
964
# currently can't cope with the use of lazy imports in this namespace --
967
# This currently doesn't report the failure at the time it occurs, because
968
# they tend to happen very early in startup when we can't check config
969
# files etc, and also we want to report all failures but not spam the user
971
from bzrlib import trace
972
exception_str = str(exception)
973
if exception_str not in _extension_load_failures:
974
trace.mutter("failed to load compiled extension: %s" % exception_str)
975
_extension_load_failures.append(exception_str)
978
def report_extension_load_failures():
979
if not _extension_load_failures:
981
from bzrlib.config import GlobalConfig
982
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
984
# the warnings framework should by default show this only once
985
from bzrlib.trace import warning
987
"bzr: warning: some compiled extensions could not be loaded; "
988
"see <https://answers.launchpad.net/bzr/+faq/703>")
989
# we no longer show the specific missing extensions here, because it makes
990
# the message too long and scary - see
991
# https://bugs.launchpad.net/bzr/+bug/430529
995
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
996
except ImportError, e:
997
failed_to_load_extension(e)
998
from bzrlib._chunks_to_lines_py import chunks_to_lines
1002
"""Split s into lines, but without removing the newline characters."""
1003
# Trivially convert a fulltext into a 'chunked' representation, and let
1004
# chunks_to_lines do the heavy lifting.
1005
if isinstance(s, str):
1006
# chunks_to_lines only supports 8-bit strings
1007
return chunks_to_lines([s])
1009
return _split_lines(s)
1012
def _split_lines(s):
1013
"""Split s into lines, but without removing the newline characters.
1015
This supports Unicode or plain string objects.
1017
lines = s.split('\n')
1018
result = [line + '\n' for line in lines[:-1]]
1020
result.append(lines[-1])
1024
def hardlinks_good():
1025
return sys.platform not in ('win32', 'cygwin', 'darwin')
1028
def link_or_copy(src, dest):
1029
"""Hardlink a file, or copy it if it can't be hardlinked."""
1030
if not hardlinks_good():
1031
shutil.copyfile(src, dest)
1035
except (OSError, IOError), e:
1036
if e.errno != errno.EXDEV:
1038
shutil.copyfile(src, dest)
1041
def delete_any(path):
1042
"""Delete a file, symlink or directory.
1044
Will delete even if readonly.
1047
_delete_file_or_dir(path)
1048
except (OSError, IOError), e:
1049
if e.errno in (errno.EPERM, errno.EACCES):
1050
# make writable and try again
1053
except (OSError, IOError):
1055
_delete_file_or_dir(path)
1060
def _delete_file_or_dir(path):
1061
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1062
# Forgiveness than Permission (EAFP) because:
1063
# - root can damage a solaris file system by using unlink,
1064
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1065
# EACCES, OSX: EPERM) when invoked on a directory.
1066
if isdir(path): # Takes care of symlinks
1073
if getattr(os, 'symlink', None) is not None:
1079
def has_hardlinks():
1080
if getattr(os, 'link', None) is not None:
1086
def host_os_dereferences_symlinks():
1087
return (has_symlinks()
1088
and sys.platform not in ('cygwin', 'win32'))
1091
def readlink(abspath):
1092
"""Return a string representing the path to which the symbolic link points.
1094
:param abspath: The link absolute unicode path.
1096
This his guaranteed to return the symbolic link in unicode in all python
1099
link = abspath.encode(_fs_enc)
1100
target = os.readlink(link)
1101
target = target.decode(_fs_enc)
1105
def contains_whitespace(s):
1106
"""True if there are any whitespace characters in s."""
1107
# string.whitespace can include '\xa0' in certain locales, because it is
1108
# considered "non-breaking-space" as part of ISO-8859-1. But it
1109
# 1) Isn't a breaking whitespace
1110
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1112
# 3) '\xa0' isn't unicode safe since it is >128.
1114
# This should *not* be a unicode set of characters in case the source
1115
# string is not a Unicode string. We can auto-up-cast the characters since
1116
# they are ascii, but we don't want to auto-up-cast the string in case it
1118
for ch in ' \t\n\r\v\f':
1125
def contains_linebreaks(s):
1126
"""True if there is any vertical whitespace in s."""
1134
def relpath(base, path):
1135
"""Return path relative to base, or raise exception.
1137
The path may be either an absolute path or a path relative to the
1138
current working directory.
1140
os.path.commonprefix (python2.4) has a bad bug that it works just
1141
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1142
avoids that problem.
1145
if len(base) < MIN_ABS_PATHLENGTH:
1146
# must have space for e.g. a drive letter
1147
raise ValueError('%r is too short to calculate a relative path'
1155
if len(head) <= len(base) and head != base:
1156
raise errors.PathNotChild(rp, base)
1159
head, tail = split(head)
1164
return pathjoin(*reversed(s))
1169
def _cicp_canonical_relpath(base, path):
1170
"""Return the canonical path relative to base.
1172
Like relpath, but on case-insensitive-case-preserving file-systems, this
1173
will return the relpath as stored on the file-system rather than in the
1174
case specified in the input string, for all existing portions of the path.
1176
This will cause O(N) behaviour if called for every path in a tree; if you
1177
have a number of paths to convert, you should use canonical_relpaths().
1179
# TODO: it should be possible to optimize this for Windows by using the
1180
# win32 API FindFiles function to look for the specified name - but using
1181
# os.listdir() still gives us the correct, platform agnostic semantics in
1184
rel = relpath(base, path)
1185
# '.' will have been turned into ''
1189
abs_base = abspath(base)
1191
_listdir = os.listdir
1193
# use an explicit iterator so we can easily consume the rest on early exit.
1194
bit_iter = iter(rel.split('/'))
1195
for bit in bit_iter:
1198
next_entries = _listdir(current)
1199
except OSError: # enoent, eperm, etc
1200
# We can't find this in the filesystem, so just append the
1202
current = pathjoin(current, bit, *list(bit_iter))
1204
for look in next_entries:
1205
if lbit == look.lower():
1206
current = pathjoin(current, look)
1209
# got to the end, nothing matched, so we just return the
1210
# non-existing bits as they were specified (the filename may be
1211
# the target of a move, for example).
1212
current = pathjoin(current, bit, *list(bit_iter))
1214
return current[len(abs_base):].lstrip('/')
1216
# XXX - TODO - we need better detection/integration of case-insensitive
1217
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1218
# filesystems), for example, so could probably benefit from the same basic
1219
# support there. For now though, only Windows and OSX get that support, and
1220
# they get it for *all* file-systems!
1221
if sys.platform in ('win32', 'darwin'):
1222
canonical_relpath = _cicp_canonical_relpath
1224
canonical_relpath = relpath
1226
def canonical_relpaths(base, paths):
1227
"""Create an iterable to canonicalize a sequence of relative paths.
1229
The intent is for this implementation to use a cache, vastly speeding
1230
up multiple transformations in the same directory.
1232
# but for now, we haven't optimized...
1233
return [canonical_relpath(base, p) for p in paths]
1235
def safe_unicode(unicode_or_utf8_string):
1236
"""Coerce unicode_or_utf8_string into unicode.
1238
If it is unicode, it is returned.
1239
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1240
wrapped in a BzrBadParameterNotUnicode exception.
1242
if isinstance(unicode_or_utf8_string, unicode):
1243
return unicode_or_utf8_string
1245
return unicode_or_utf8_string.decode('utf8')
1246
except UnicodeDecodeError:
1247
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1250
def safe_utf8(unicode_or_utf8_string):
1251
"""Coerce unicode_or_utf8_string to a utf8 string.
1253
If it is a str, it is returned.
1254
If it is Unicode, it is encoded into a utf-8 string.
1256
if isinstance(unicode_or_utf8_string, str):
1257
# TODO: jam 20070209 This is overkill, and probably has an impact on
1258
# performance if we are dealing with lots of apis that want a
1261
# Make sure it is a valid utf-8 string
1262
unicode_or_utf8_string.decode('utf-8')
1263
except UnicodeDecodeError:
1264
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1265
return unicode_or_utf8_string
1266
return unicode_or_utf8_string.encode('utf-8')
1269
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1270
' Revision id generators should be creating utf8'
1274
def safe_revision_id(unicode_or_utf8_string, warn=True):
1275
"""Revision ids should now be utf8, but at one point they were unicode.
1277
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1279
:param warn: Functions that are sanitizing user data can set warn=False
1280
:return: None or a utf8 revision id.
1282
if (unicode_or_utf8_string is None
1283
or unicode_or_utf8_string.__class__ == str):
1284
return unicode_or_utf8_string
1286
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1288
return cache_utf8.encode(unicode_or_utf8_string)
1291
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1292
' generators should be creating utf8 file ids.')
1295
def safe_file_id(unicode_or_utf8_string, warn=True):
1296
"""File ids should now be utf8, but at one point they were unicode.
1298
This is the same as safe_utf8, except it uses the cached encode functions
1299
to save a little bit of performance.
1301
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1303
:param warn: Functions that are sanitizing user data can set warn=False
1304
:return: None or a utf8 file id.
1306
if (unicode_or_utf8_string is None
1307
or unicode_or_utf8_string.__class__ == str):
1308
return unicode_or_utf8_string
1310
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1312
return cache_utf8.encode(unicode_or_utf8_string)
1315
_platform_normalizes_filenames = False
1316
if sys.platform == 'darwin':
1317
_platform_normalizes_filenames = True
1320
def normalizes_filenames():
1321
"""Return True if this platform normalizes unicode filenames.
1323
Mac OSX does, Windows/Linux do not.
1325
return _platform_normalizes_filenames
1328
def _accessible_normalized_filename(path):
1329
"""Get the unicode normalized path, and if you can access the file.
1331
On platforms where the system normalizes filenames (Mac OSX),
1332
you can access a file by any path which will normalize correctly.
1333
On platforms where the system does not normalize filenames
1334
(Windows, Linux), you have to access a file by its exact path.
1336
Internally, bzr only supports NFC normalization, since that is
1337
the standard for XML documents.
1339
So return the normalized path, and a flag indicating if the file
1340
can be accessed by that path.
1343
return unicodedata.normalize('NFC', unicode(path)), True
1346
def _inaccessible_normalized_filename(path):
1347
__doc__ = _accessible_normalized_filename.__doc__
1349
normalized = unicodedata.normalize('NFC', unicode(path))
1350
return normalized, normalized == path
1353
if _platform_normalizes_filenames:
1354
normalized_filename = _accessible_normalized_filename
1356
normalized_filename = _inaccessible_normalized_filename
1359
def set_signal_handler(signum, handler, restart_syscall=True):
1360
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1361
on platforms that support that.
1363
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1364
automatically restart (by calling `signal.siginterrupt(signum,
1365
False)`). May be ignored if the feature is not available on this
1366
platform or Python version.
1369
siginterrupt = signal.siginterrupt
1370
except AttributeError:
1371
# siginterrupt doesn't exist on this platform, or for this version
1373
siginterrupt = lambda signum, flag: None
1375
def sig_handler(*args):
1376
# Python resets the siginterrupt flag when a signal is
1377
# received. <http://bugs.python.org/issue8354>
1378
# As a workaround for some cases, set it back the way we want it.
1379
siginterrupt(signum, False)
1380
# Now run the handler function passed to set_signal_handler.
1383
sig_handler = handler
1384
old_handler = signal.signal(signum, sig_handler)
1386
siginterrupt(signum, False)
1390
default_terminal_width = 80
1391
"""The default terminal width for ttys.
1393
This is defined so that higher levels can share a common fallback value when
1394
terminal_width() returns None.
1398
def terminal_width():
1399
"""Return terminal width.
1401
None is returned if the width can't established precisely.
1404
- if BZR_COLUMNS is set, returns its value
1405
- if there is no controlling terminal, returns None
1406
- if COLUMNS is set, returns its value,
1408
From there, we need to query the OS to get the size of the controlling
1412
- get termios.TIOCGWINSZ
1413
- if an error occurs or a negative value is obtained, returns None
1417
- win32utils.get_console_size() decides,
1418
- returns None on error (provided default value)
1421
# If BZR_COLUMNS is set, take it, user is always right
1423
return int(os.environ['BZR_COLUMNS'])
1424
except (KeyError, ValueError):
1427
isatty = getattr(sys.stdout, 'isatty', None)
1428
if isatty is None or not isatty():
1429
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1432
# If COLUMNS is set, take it, the terminal knows better (even inside a
1433
# given terminal, the application can decide to set COLUMNS to a lower
1434
# value (splitted screen) or a bigger value (scroll bars))
1436
return int(os.environ['COLUMNS'])
1437
except (KeyError, ValueError):
1440
width, height = _terminal_size(None, None)
1442
# Consider invalid values as meaning no width
1448
def _win32_terminal_size(width, height):
1449
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1450
return width, height
1453
def _ioctl_terminal_size(width, height):
1455
import struct, fcntl, termios
1456
s = struct.pack('HHHH', 0, 0, 0, 0)
1457
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1458
height, width = struct.unpack('HHHH', x)[0:2]
1459
except (IOError, AttributeError):
1461
return width, height
1463
_terminal_size = None
1464
"""Returns the terminal size as (width, height).
1466
:param width: Default value for width.
1467
:param height: Default value for height.
1469
This is defined specifically for each OS and query the size of the controlling
1470
terminal. If any error occurs, the provided default values should be returned.
1472
if sys.platform == 'win32':
1473
_terminal_size = _win32_terminal_size
1475
_terminal_size = _ioctl_terminal_size
1478
def _terminal_size_changed(signum, frame):
1479
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1480
width, height = _terminal_size(None, None)
1481
if width is not None:
1482
os.environ['COLUMNS'] = str(width)
1485
_registered_sigwinch = False
1487
def watch_sigwinch():
1488
"""Register for SIGWINCH, once and only once."""
1489
global _registered_sigwinch
1490
if not _registered_sigwinch:
1491
if sys.platform == 'win32':
1492
# Martin (gz) mentioned WINDOW_BUFFER_SIZE_RECORD from
1493
# ReadConsoleInput but I've no idea how to plug that in
1494
# the current design -- vila 20091216
1497
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1498
_registered_sigwinch = True
1501
def supports_executable():
1502
return sys.platform != "win32"
1505
def supports_posix_readonly():
1506
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1508
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1509
directory controls creation/deletion, etc.
1511
And under win32, readonly means that the directory itself cannot be
1512
deleted. The contents of a readonly directory can be changed, unlike POSIX
1513
where files in readonly directories cannot be added, deleted or renamed.
1515
return sys.platform != "win32"
1518
def set_or_unset_env(env_variable, value):
1519
"""Modify the environment, setting or removing the env_variable.
1521
:param env_variable: The environment variable in question
1522
:param value: The value to set the environment to. If None, then
1523
the variable will be removed.
1524
:return: The original value of the environment variable.
1526
orig_val = os.environ.get(env_variable)
1528
if orig_val is not None:
1529
del os.environ[env_variable]
1531
if isinstance(value, unicode):
1532
value = value.encode(get_user_encoding())
1533
os.environ[env_variable] = value
1537
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1540
def check_legal_path(path):
1541
"""Check whether the supplied path is legal.
1542
This is only required on Windows, so we don't test on other platforms
1545
if sys.platform != "win32":
1547
if _validWin32PathRE.match(path) is None:
1548
raise errors.IllegalPath(path)
1551
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1553
def _is_error_enotdir(e):
1554
"""Check if this exception represents ENOTDIR.
1556
Unfortunately, python is very inconsistent about the exception
1557
here. The cases are:
1558
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1559
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1560
which is the windows error code.
1561
3) Windows, Python2.5 uses errno == EINVAL and
1562
winerror == ERROR_DIRECTORY
1564
:param e: An Exception object (expected to be OSError with an errno
1565
attribute, but we should be able to cope with anything)
1566
:return: True if this represents an ENOTDIR error. False otherwise.
1568
en = getattr(e, 'errno', None)
1569
if (en == errno.ENOTDIR
1570
or (sys.platform == 'win32'
1571
and (en == _WIN32_ERROR_DIRECTORY
1572
or (en == errno.EINVAL
1573
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1579
def walkdirs(top, prefix=""):
1580
"""Yield data about all the directories in a tree.
1582
This yields all the data about the contents of a directory at a time.
1583
After each directory has been yielded, if the caller has mutated the list
1584
to exclude some directories, they are then not descended into.
1586
The data yielded is of the form:
1587
((directory-relpath, directory-path-from-top),
1588
[(relpath, basename, kind, lstat, path-from-top), ...]),
1589
- directory-relpath is the relative path of the directory being returned
1590
with respect to top. prefix is prepended to this.
1591
- directory-path-from-root is the path including top for this directory.
1592
It is suitable for use with os functions.
1593
- relpath is the relative path within the subtree being walked.
1594
- basename is the basename of the path
1595
- kind is the kind of the file now. If unknown then the file is not
1596
present within the tree - but it may be recorded as versioned. See
1598
- lstat is the stat data *if* the file was statted.
1599
- planned, not implemented:
1600
path_from_tree_root is the path from the root of the tree.
1602
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1603
allows one to walk a subtree but get paths that are relative to a tree
1605
:return: an iterator over the dirs.
1607
#TODO there is a bit of a smell where the results of the directory-
1608
# summary in this, and the path from the root, may not agree
1609
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1610
# potentially confusing output. We should make this more robust - but
1611
# not at a speed cost. RBC 20060731
1613
_directory = _directory_kind
1614
_listdir = os.listdir
1615
_kind_from_mode = file_kind_from_stat_mode
1616
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1618
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1619
relroot, _, _, _, top = pending.pop()
1621
relprefix = relroot + u'/'
1624
top_slash = top + u'/'
1627
append = dirblock.append
1629
names = sorted(_listdir(top))
1631
if not _is_error_enotdir(e):
1635
abspath = top_slash + name
1636
statvalue = _lstat(abspath)
1637
kind = _kind_from_mode(statvalue.st_mode)
1638
append((relprefix + name, name, kind, statvalue, abspath))
1639
yield (relroot, top), dirblock
1641
# push the user specified dirs from dirblock
1642
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1645
class DirReader(object):
1646
"""An interface for reading directories."""
1648
def top_prefix_to_starting_dir(self, top, prefix=""):
1649
"""Converts top and prefix to a starting dir entry
1651
:param top: A utf8 path
1652
:param prefix: An optional utf8 path to prefix output relative paths
1654
:return: A tuple starting with prefix, and ending with the native
1657
raise NotImplementedError(self.top_prefix_to_starting_dir)
1659
def read_dir(self, prefix, top):
1660
"""Read a specific dir.
1662
:param prefix: A utf8 prefix to be preprended to the path basenames.
1663
:param top: A natively encoded path to read.
1664
:return: A list of the directories contents. Each item contains:
1665
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1667
raise NotImplementedError(self.read_dir)
1670
_selected_dir_reader = None
1673
def _walkdirs_utf8(top, prefix=""):
1674
"""Yield data about all the directories in a tree.
1676
This yields the same information as walkdirs() only each entry is yielded
1677
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1678
are returned as exact byte-strings.
1680
:return: yields a tuple of (dir_info, [file_info])
1681
dir_info is (utf8_relpath, path-from-top)
1682
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1683
if top is an absolute path, path-from-top is also an absolute path.
1684
path-from-top might be unicode or utf8, but it is the correct path to
1685
pass to os functions to affect the file in question. (such as os.lstat)
1687
global _selected_dir_reader
1688
if _selected_dir_reader is None:
1689
fs_encoding = _fs_enc.upper()
1690
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1691
# Win98 doesn't have unicode apis like FindFirstFileW
1692
# TODO: We possibly could support Win98 by falling back to the
1693
# original FindFirstFile, and using TCHAR instead of WCHAR,
1694
# but that gets a bit tricky, and requires custom compiling
1697
from bzrlib._walkdirs_win32 import Win32ReadDir
1698
_selected_dir_reader = Win32ReadDir()
1701
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1702
# ANSI_X3.4-1968 is a form of ASCII
1704
from bzrlib._readdir_pyx import UTF8DirReader
1705
_selected_dir_reader = UTF8DirReader()
1706
except ImportError, e:
1707
failed_to_load_extension(e)
1710
if _selected_dir_reader is None:
1711
# Fallback to the python version
1712
_selected_dir_reader = UnicodeDirReader()
1714
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1715
# But we don't actually uses 1-3 in pending, so set them to None
1716
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1717
read_dir = _selected_dir_reader.read_dir
1718
_directory = _directory_kind
1720
relroot, _, _, _, top = pending[-1].pop()
1723
dirblock = sorted(read_dir(relroot, top))
1724
yield (relroot, top), dirblock
1725
# push the user specified dirs from dirblock
1726
next = [d for d in reversed(dirblock) if d[2] == _directory]
1728
pending.append(next)
1731
class UnicodeDirReader(DirReader):
1732
"""A dir reader for non-utf8 file systems, which transcodes."""
1734
__slots__ = ['_utf8_encode']
1737
self._utf8_encode = codecs.getencoder('utf8')
1739
def top_prefix_to_starting_dir(self, top, prefix=""):
1740
"""See DirReader.top_prefix_to_starting_dir."""
1741
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1743
def read_dir(self, prefix, top):
1744
"""Read a single directory from a non-utf8 file system.
1746
top, and the abspath element in the output are unicode, all other paths
1747
are utf8. Local disk IO is done via unicode calls to listdir etc.
1749
This is currently the fallback code path when the filesystem encoding is
1750
not UTF-8. It may be better to implement an alternative so that we can
1751
safely handle paths that are not properly decodable in the current
1754
See DirReader.read_dir for details.
1756
_utf8_encode = self._utf8_encode
1758
_listdir = os.listdir
1759
_kind_from_mode = file_kind_from_stat_mode
1762
relprefix = prefix + '/'
1765
top_slash = top + u'/'
1768
append = dirblock.append
1769
for name in sorted(_listdir(top)):
1771
name_utf8 = _utf8_encode(name)[0]
1772
except UnicodeDecodeError:
1773
raise errors.BadFilenameEncoding(
1774
_utf8_encode(relprefix)[0] + name, _fs_enc)
1775
abspath = top_slash + name
1776
statvalue = _lstat(abspath)
1777
kind = _kind_from_mode(statvalue.st_mode)
1778
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1782
def copy_tree(from_path, to_path, handlers={}):
1783
"""Copy all of the entries in from_path into to_path.
1785
:param from_path: The base directory to copy.
1786
:param to_path: The target directory. If it does not exist, it will
1788
:param handlers: A dictionary of functions, which takes a source and
1789
destinations for files, directories, etc.
1790
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1791
'file', 'directory', and 'symlink' should always exist.
1792
If they are missing, they will be replaced with 'os.mkdir()',
1793
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1795
# Now, just copy the existing cached tree to the new location
1796
# We use a cheap trick here.
1797
# Absolute paths are prefixed with the first parameter
1798
# relative paths are prefixed with the second.
1799
# So we can get both the source and target returned
1800
# without any extra work.
1802
def copy_dir(source, dest):
1805
def copy_link(source, dest):
1806
"""Copy the contents of a symlink"""
1807
link_to = os.readlink(source)
1808
os.symlink(link_to, dest)
1810
real_handlers = {'file':shutil.copy2,
1811
'symlink':copy_link,
1812
'directory':copy_dir,
1814
real_handlers.update(handlers)
1816
if not os.path.exists(to_path):
1817
real_handlers['directory'](from_path, to_path)
1819
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1820
for relpath, name, kind, st, abspath in entries:
1821
real_handlers[kind](abspath, relpath)
1824
def copy_ownership(dst, src=None):
1825
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1827
If src is None, the containing directory is used as source. If chown
1828
fails, the error is ignored and a warning is printed.
1830
chown = getattr(os, 'chown', None)
1835
src = os.path.dirname(dst)
1841
chown(dst, s.st_uid, s.st_gid)
1843
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1846
def mkdir_with_ownership(path, ownership_src=None):
1847
"""Create the directory 'path' with specified ownership.
1849
If ownership_src is given, copies (chown) usr/grp ownership
1850
from 'ownership_src' to 'path'. If ownership_src is None, use the
1851
containing dir ownership.
1854
copy_ownership(path, ownership_src)
1857
def open_with_ownership(filename, mode='r', bufsize=-1, ownership_src=None):
1858
"""Open the file 'filename' with the specified ownership.
1860
If ownership_src is specified, copy usr/grp ownership from ownership_src
1861
to filename. If ownership_src is None, copy ownership from containing
1863
Returns the opened file object.
1865
f = open(filename, mode, bufsize)
1866
copy_ownership(filename, ownership_src)
1870
def path_prefix_key(path):
1871
"""Generate a prefix-order path key for path.
1873
This can be used to sort paths in the same way that walkdirs does.
1875
return (dirname(path) , path)
1878
def compare_paths_prefix_order(path_a, path_b):
1879
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1880
key_a = path_prefix_key(path_a)
1881
key_b = path_prefix_key(path_b)
1882
return cmp(key_a, key_b)
1885
_cached_user_encoding = None
1888
def get_user_encoding(use_cache=True):
1889
"""Find out what the preferred user encoding is.
1891
This is generally the encoding that is used for command line parameters
1892
and file contents. This may be different from the terminal encoding
1893
or the filesystem encoding.
1895
:param use_cache: Enable cache for detected encoding.
1896
(This parameter is turned on by default,
1897
and required only for selftesting)
1899
:return: A string defining the preferred user encoding
1901
global _cached_user_encoding
1902
if _cached_user_encoding is not None and use_cache:
1903
return _cached_user_encoding
1905
if sys.platform == 'darwin':
1906
# python locale.getpreferredencoding() always return
1907
# 'mac-roman' on darwin. That's a lie.
1908
sys.platform = 'posix'
1910
if os.environ.get('LANG', None) is None:
1911
# If LANG is not set, we end up with 'ascii', which is bad
1912
# ('mac-roman' is more than ascii), so we set a default which
1913
# will give us UTF-8 (which appears to work in all cases on
1914
# OSX). Users are still free to override LANG of course, as
1915
# long as it give us something meaningful. This work-around
1916
# *may* not be needed with python 3k and/or OSX 10.5, but will
1917
# work with them too -- vila 20080908
1918
os.environ['LANG'] = 'en_US.UTF-8'
1921
sys.platform = 'darwin'
1926
user_encoding = locale.getpreferredencoding()
1927
except locale.Error, e:
1928
sys.stderr.write('bzr: warning: %s\n'
1929
' Could not determine what text encoding to use.\n'
1930
' This error usually means your Python interpreter\n'
1931
' doesn\'t support the locale set by $LANG (%s)\n'
1932
" Continuing with ascii encoding.\n"
1933
% (e, os.environ.get('LANG')))
1934
user_encoding = 'ascii'
1936
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1937
# treat that as ASCII, and not support printing unicode characters to the
1940
# For python scripts run under vim, we get '', so also treat that as ASCII
1941
if user_encoding in (None, 'cp0', ''):
1942
user_encoding = 'ascii'
1946
codecs.lookup(user_encoding)
1948
sys.stderr.write('bzr: warning:'
1949
' unknown encoding %s.'
1950
' Continuing with ascii encoding.\n'
1953
user_encoding = 'ascii'
1956
_cached_user_encoding = user_encoding
1958
return user_encoding
1961
def get_host_name():
1962
"""Return the current unicode host name.
1964
This is meant to be used in place of socket.gethostname() because that
1965
behaves inconsistently on different platforms.
1967
if sys.platform == "win32":
1969
return win32utils.get_host_name()
1972
return socket.gethostname().decode(get_user_encoding())
1975
# We must not read/write any more than 64k at a time from/to a socket so we
1976
# don't risk "no buffer space available" errors on some platforms. Windows in
1977
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1979
MAX_SOCKET_CHUNK = 64 * 1024
1981
def read_bytes_from_socket(sock, report_activity=None,
1982
max_read_size=MAX_SOCKET_CHUNK):
1983
"""Read up to max_read_size of bytes from sock and notify of progress.
1985
Translates "Connection reset by peer" into file-like EOF (return an
1986
empty string rather than raise an error), and repeats the recv if
1987
interrupted by a signal.
1991
bytes = sock.recv(max_read_size)
1992
except socket.error, e:
1994
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1995
# The connection was closed by the other side. Callers expect
1996
# an empty string to signal end-of-stream.
1998
elif eno == errno.EINTR:
1999
# Retry the interrupted recv.
2003
if report_activity is not None:
2004
report_activity(len(bytes), 'read')
2008
def recv_all(socket, count):
2009
"""Receive an exact number of bytes.
2011
Regular Socket.recv() may return less than the requested number of bytes,
2012
depending on what's in the OS buffer. MSG_WAITALL is not available
2013
on all platforms, but this should work everywhere. This will return
2014
less than the requested amount if the remote end closes.
2016
This isn't optimized and is intended mostly for use in testing.
2019
while len(b) < count:
2020
new = read_bytes_from_socket(socket, None, count - len(b))
2027
def send_all(sock, bytes, report_activity=None):
2028
"""Send all bytes on a socket.
2030
Breaks large blocks in smaller chunks to avoid buffering limitations on
2031
some platforms, and catches EINTR which may be thrown if the send is
2032
interrupted by a signal.
2034
This is preferred to socket.sendall(), because it avoids portability bugs
2035
and provides activity reporting.
2037
:param report_activity: Call this as bytes are read, see
2038
Transport._report_activity
2041
byte_count = len(bytes)
2042
while sent_total < byte_count:
2044
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2045
except socket.error, e:
2046
if e.args[0] != errno.EINTR:
2050
report_activity(sent, 'write')
2053
def dereference_path(path):
2054
"""Determine the real path to a file.
2056
All parent elements are dereferenced. But the file itself is not
2058
:param path: The original path. May be absolute or relative.
2059
:return: the real path *to* the file
2061
parent, base = os.path.split(path)
2062
# The pathjoin for '.' is a workaround for Python bug #1213894.
2063
# (initial path components aren't dereferenced)
2064
return pathjoin(realpath(pathjoin('.', parent)), base)
2067
def supports_mapi():
2068
"""Return True if we can use MAPI to launch a mail client."""
2069
return sys.platform == "win32"
2072
def resource_string(package, resource_name):
2073
"""Load a resource from a package and return it as a string.
2075
Note: Only packages that start with bzrlib are currently supported.
2077
This is designed to be a lightweight implementation of resource
2078
loading in a way which is API compatible with the same API from
2080
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2081
If and when pkg_resources becomes a standard library, this routine
2084
# Check package name is within bzrlib
2085
if package == "bzrlib":
2086
resource_relpath = resource_name
2087
elif package.startswith("bzrlib."):
2088
package = package[len("bzrlib."):].replace('.', os.sep)
2089
resource_relpath = pathjoin(package, resource_name)
2091
raise errors.BzrError('resource package %s not in bzrlib' % package)
2093
# Map the resource to a file and read its contents
2094
base = dirname(bzrlib.__file__)
2095
if getattr(sys, 'frozen', None): # bzr.exe
2096
base = abspath(pathjoin(base, '..', '..'))
2097
filename = pathjoin(base, resource_relpath)
2098
return open(filename, 'rU').read()
2101
def file_kind_from_stat_mode_thunk(mode):
2102
global file_kind_from_stat_mode
2103
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2105
from bzrlib._readdir_pyx import UTF8DirReader
2106
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2107
except ImportError, e:
2108
# This is one time where we won't warn that an extension failed to
2109
# load. The extension is never available on Windows anyway.
2110
from bzrlib._readdir_py import (
2111
_kind_from_mode as file_kind_from_stat_mode
2113
return file_kind_from_stat_mode(mode)
2114
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2117
def file_kind(f, _lstat=os.lstat):
2119
return file_kind_from_stat_mode(_lstat(f).st_mode)
2121
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2122
raise errors.NoSuchFile(f)
2126
def until_no_eintr(f, *a, **kw):
2127
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2129
WARNING: you must be certain that it is safe to retry the call repeatedly
2130
if EINTR does occur. This is typically only true for low-level operations
2131
like os.read. If in any doubt, don't use this.
2133
Keep in mind that this is not a complete solution to EINTR. There is
2134
probably code in the Python standard library and other dependencies that
2135
may encounter EINTR if a signal arrives (and there is signal handler for
2136
that signal). So this function can reduce the impact for IO that bzrlib
2137
directly controls, but it is not a complete solution.
2139
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2143
except (IOError, OSError), e:
2144
if e.errno == errno.EINTR:
2149
def re_compile_checked(re_string, flags=0, where=""):
2150
"""Return a compiled re, or raise a sensible error.
2152
This should only be used when compiling user-supplied REs.
2154
:param re_string: Text form of regular expression.
2155
:param flags: eg re.IGNORECASE
2156
:param where: Message explaining to the user the context where
2157
it occurred, eg 'log search filter'.
2159
# from https://bugs.launchpad.net/bzr/+bug/251352
2161
re_obj = re.compile(re_string, flags)
2166
where = ' in ' + where
2167
# despite the name 'error' is a type
2168
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2169
% (where, re_string, e))
2172
if sys.platform == "win32":
2175
return msvcrt.getch()
2180
fd = sys.stdin.fileno()
2181
settings = termios.tcgetattr(fd)
2184
ch = sys.stdin.read(1)
2186
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2190
if sys.platform == 'linux2':
2191
def _local_concurrency():
2193
prefix = 'processor'
2194
for line in file('/proc/cpuinfo', 'rb'):
2195
if line.startswith(prefix):
2196
concurrency = int(line[line.find(':')+1:]) + 1
2198
elif sys.platform == 'darwin':
2199
def _local_concurrency():
2200
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2201
stdout=subprocess.PIPE).communicate()[0]
2202
elif sys.platform[0:7] == 'freebsd':
2203
def _local_concurrency():
2204
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2205
stdout=subprocess.PIPE).communicate()[0]
2206
elif sys.platform == 'sunos5':
2207
def _local_concurrency():
2208
return subprocess.Popen(['psrinfo', '-p',],
2209
stdout=subprocess.PIPE).communicate()[0]
2210
elif sys.platform == "win32":
2211
def _local_concurrency():
2212
# This appears to return the number of cores.
2213
return os.environ.get('NUMBER_OF_PROCESSORS')
2215
def _local_concurrency():
2220
_cached_local_concurrency = None
2222
def local_concurrency(use_cache=True):
2223
"""Return how many processes can be run concurrently.
2225
Rely on platform specific implementations and default to 1 (one) if
2226
anything goes wrong.
2228
global _cached_local_concurrency
2230
if _cached_local_concurrency is not None and use_cache:
2231
return _cached_local_concurrency
2233
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2234
if concurrency is None:
2236
concurrency = _local_concurrency()
2237
except (OSError, IOError):
2240
concurrency = int(concurrency)
2241
except (TypeError, ValueError):
2244
_cached_concurrency = concurrency
2248
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2249
"""A stream writer that doesn't decode str arguments."""
2251
def __init__(self, encode, stream, errors='strict'):
2252
codecs.StreamWriter.__init__(self, stream, errors)
2253
self.encode = encode
2255
def write(self, object):
2256
if type(object) is str:
2257
self.stream.write(object)
2259
data, _ = self.encode(object, self.errors)
2260
self.stream.write(data)
2262
if sys.platform == 'win32':
2263
def open_file(filename, mode='r', bufsize=-1):
2264
"""This function is used to override the ``open`` builtin.
2266
But it uses O_NOINHERIT flag so the file handle is not inherited by
2267
child processes. Deleting or renaming a closed file opened with this
2268
function is not blocking child processes.
2270
writing = 'w' in mode
2271
appending = 'a' in mode
2272
updating = '+' in mode
2273
binary = 'b' in mode
2276
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2277
# for flags for each modes.
2287
flags |= os.O_WRONLY
2288
flags |= os.O_CREAT | os.O_TRUNC
2293
flags |= os.O_WRONLY
2294
flags |= os.O_CREAT | os.O_APPEND
2299
flags |= os.O_RDONLY
2301
return os.fdopen(os.open(filename, flags), mode, bufsize)
added
removed
bzrlib/osutils.py
