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,
45
from tempfile import (
59
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
61
if sys.version_info < (2, 5):
62
import md5 as _mod_md5
64
import sha as _mod_sha
74
from bzrlib import symbol_versioning
77
# Cross platform wall-clock time functionality with decent resolution.
78
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
79
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
80
# synchronized with ``time.time()``, this is only meant to be used to find
81
# delta times by subtracting from another call to this function.
82
timer_func = time.time
83
if sys.platform == 'win32':
84
timer_func = time.clock
86
# On win32, O_BINARY is used to indicate the file should
87
# be opened in binary mode, rather than text mode.
88
# On other platforms, O_BINARY doesn't exist, because
89
# they always open in binary mode, so it is okay to
90
# OR with 0 on those platforms.
91
# O_NOINHERIT and O_TEXT exists only on win32 too.
92
O_BINARY = getattr(os, 'O_BINARY', 0)
93
O_TEXT = getattr(os, 'O_TEXT', 0)
94
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
97
def get_unicode_argv():
99
user_encoding = get_user_encoding()
100
return [a.decode(user_encoding) for a in sys.argv[1:]]
101
except UnicodeDecodeError:
102
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
106
def make_readonly(filename):
107
"""Make a filename read-only."""
108
mod = os.lstat(filename).st_mode
109
if not stat.S_ISLNK(mod):
111
os.chmod(filename, mod)
114
def make_writable(filename):
115
mod = os.lstat(filename).st_mode
116
if not stat.S_ISLNK(mod):
118
os.chmod(filename, mod)
121
def minimum_path_selection(paths):
122
"""Return the smallset subset of paths which are outside paths.
124
:param paths: A container (and hence not None) of paths.
125
:return: A set of paths sufficient to include everything in paths via
126
is_inside, drawn from the paths parameter.
132
return path.split('/')
133
sorted_paths = sorted(list(paths), key=sort_key)
135
search_paths = [sorted_paths[0]]
136
for path in sorted_paths[1:]:
137
if not is_inside(search_paths[-1], path):
138
# This path is unique, add it
139
search_paths.append(path)
141
return set(search_paths)
148
"""Return a quoted filename filename
150
This previously used backslash quoting, but that works poorly on
152
# TODO: I'm not really sure this is the best format either.x
154
if _QUOTE_RE is None:
155
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
157
if _QUOTE_RE.search(f):
163
_directory_kind = 'directory'
166
"""Return the current umask"""
167
# Assume that people aren't messing with the umask while running
168
# XXX: This is not thread safe, but there is no way to get the
169
# umask without setting it
177
_directory_kind: "/",
179
'tree-reference': '+',
183
def kind_marker(kind):
185
return _kind_marker_map[kind]
187
# Slightly faster than using .get(, '') when the common case is that
192
lexists = getattr(os.path, 'lexists', None)
196
stat = getattr(os, 'lstat', os.stat)
200
if e.errno == errno.ENOENT:
203
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
206
def fancy_rename(old, new, rename_func, unlink_func):
207
"""A fancy rename, when you don't have atomic rename.
209
:param old: The old path, to rename from
210
:param new: The new path, to rename to
211
:param rename_func: The potentially non-atomic rename function
212
:param unlink_func: A way to delete the target file if the full rename
215
# sftp rename doesn't allow overwriting, so play tricks:
216
base = os.path.basename(new)
217
dirname = os.path.dirname(new)
218
# callers use different encodings for the paths so the following MUST
219
# respect that. We rely on python upcasting to unicode if new is unicode
220
# and keeping a str if not.
221
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
222
os.getpid(), rand_chars(10))
223
tmp_name = pathjoin(dirname, tmp_name)
225
# Rename the file out of the way, but keep track if it didn't exist
226
# We don't want to grab just any exception
227
# something like EACCES should prevent us from continuing
228
# The downside is that the rename_func has to throw an exception
229
# with an errno = ENOENT, or NoSuchFile
232
rename_func(new, tmp_name)
233
except (errors.NoSuchFile,), e:
236
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
237
# function raises an IOError with errno is None when a rename fails.
238
# This then gets caught here.
239
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
242
if (getattr(e, 'errno', None) is None
243
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
252
# This may throw an exception, in which case success will
254
rename_func(old, new)
256
except (IOError, OSError), e:
257
# source and target may be aliases of each other (e.g. on a
258
# case-insensitive filesystem), so we may have accidentally renamed
259
# source by when we tried to rename target
260
failure_exc = sys.exc_info()
261
if (file_existed and e.errno in (None, errno.ENOENT)
262
and old.lower() == new.lower()):
263
# source and target are the same file on a case-insensitive
264
# filesystem, so we don't generate an exception
268
# If the file used to exist, rename it back into place
269
# otherwise just delete it from the tmp location
271
unlink_func(tmp_name)
273
rename_func(tmp_name, new)
274
if failure_exc is not None:
275
raise failure_exc[0], failure_exc[1], failure_exc[2]
278
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
279
# choke on a Unicode string containing a relative path if
280
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
282
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
283
def _posix_abspath(path):
284
# jam 20060426 rather than encoding to fsencoding
285
# copy posixpath.abspath, but use os.getcwdu instead
286
if not posixpath.isabs(path):
287
path = posixpath.join(getcwd(), path)
288
return posixpath.normpath(path)
291
def _posix_realpath(path):
292
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
295
def _win32_fixdrive(path):
296
"""Force drive letters to be consistent.
298
win32 is inconsistent whether it returns lower or upper case
299
and even if it was consistent the user might type the other
300
so we force it to uppercase
301
running python.exe under cmd.exe return capital C:\\
302
running win32 python inside a cygwin shell returns lowercase c:\\
304
drive, path = _nt_splitdrive(path)
305
return drive.upper() + path
308
def _win32_abspath(path):
309
# Real _nt_abspath doesn't have a problem with a unicode cwd
310
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
313
def _win98_abspath(path):
314
"""Return the absolute version of a path.
315
Windows 98 safe implementation (python reimplementation
316
of Win32 API function GetFullPathNameW)
321
# \\HOST\path => //HOST/path
322
# //HOST/path => //HOST/path
323
# path => C:/cwd/path
326
# check for absolute path
327
drive = _nt_splitdrive(path)[0]
328
if drive == '' and path[:2] not in('//','\\\\'):
330
# we cannot simply os.path.join cwd and path
331
# because os.path.join('C:','/path') produce '/path'
332
# and this is incorrect
333
if path[:1] in ('/','\\'):
334
cwd = _nt_splitdrive(cwd)[0]
336
path = cwd + '\\' + path
337
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
340
def _win32_realpath(path):
341
# Real _nt_realpath doesn't have a problem with a unicode cwd
342
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
345
def _win32_pathjoin(*args):
346
return _nt_join(*args).replace('\\', '/')
349
def _win32_normpath(path):
350
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
354
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
357
def _win32_mkdtemp(*args, **kwargs):
358
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
361
def _win32_rename(old, new):
362
"""We expect to be able to atomically replace 'new' with old.
364
On win32, if new exists, it must be moved out of the way first,
368
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
370
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
371
# If we try to rename a non-existant file onto cwd, we get
372
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
373
# if the old path doesn't exist, sometimes we get EACCES
374
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
380
return unicodedata.normalize('NFC', os.getcwdu())
383
# Default is to just use the python builtins, but these can be rebound on
384
# particular platforms.
385
abspath = _posix_abspath
386
realpath = _posix_realpath
387
pathjoin = os.path.join
388
normpath = os.path.normpath
391
dirname = os.path.dirname
392
basename = os.path.basename
393
split = os.path.split
394
splitext = os.path.splitext
395
# These were already imported into local scope
396
# mkdtemp = tempfile.mkdtemp
397
# rmtree = shutil.rmtree
399
MIN_ABS_PATHLENGTH = 1
402
if sys.platform == 'win32':
403
if win32utils.winver == 'Windows 98':
404
abspath = _win98_abspath
406
abspath = _win32_abspath
407
realpath = _win32_realpath
408
pathjoin = _win32_pathjoin
409
normpath = _win32_normpath
410
getcwd = _win32_getcwd
411
mkdtemp = _win32_mkdtemp
412
rename = _win32_rename
414
MIN_ABS_PATHLENGTH = 3
416
def _win32_delete_readonly(function, path, excinfo):
417
"""Error handler for shutil.rmtree function [for win32]
418
Helps to remove files and dirs marked as read-only.
420
exception = excinfo[1]
421
if function in (os.remove, os.rmdir) \
422
and isinstance(exception, OSError) \
423
and exception.errno == errno.EACCES:
429
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
430
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
431
return shutil.rmtree(path, ignore_errors, onerror)
433
f = win32utils.get_unicode_argv # special function or None
437
elif sys.platform == 'darwin':
441
def get_terminal_encoding():
442
"""Find the best encoding for printing to the screen.
444
This attempts to check both sys.stdout and sys.stdin to see
445
what encoding they are in, and if that fails it falls back to
446
osutils.get_user_encoding().
447
The problem is that on Windows, locale.getpreferredencoding()
448
is not the same encoding as that used by the console:
449
http://mail.python.org/pipermail/python-list/2003-May/162357.html
451
On my standard US Windows XP, the preferred encoding is
452
cp1252, but the console is cp437
454
from bzrlib.trace import mutter
455
output_encoding = getattr(sys.stdout, 'encoding', None)
456
if not output_encoding:
457
input_encoding = getattr(sys.stdin, 'encoding', None)
458
if not input_encoding:
459
output_encoding = get_user_encoding()
460
mutter('encoding stdout as osutils.get_user_encoding() %r',
463
output_encoding = input_encoding
464
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
466
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
467
if output_encoding == 'cp0':
468
# invalid encoding (cp0 means 'no codepage' on Windows)
469
output_encoding = get_user_encoding()
470
mutter('cp0 is invalid encoding.'
471
' encoding stdout as osutils.get_user_encoding() %r',
475
codecs.lookup(output_encoding)
477
sys.stderr.write('bzr: warning:'
478
' unknown terminal encoding %s.\n'
479
' Using encoding %s instead.\n'
480
% (output_encoding, get_user_encoding())
482
output_encoding = get_user_encoding()
484
return output_encoding
487
def normalizepath(f):
488
if getattr(os.path, 'realpath', None) is not None:
492
[p,e] = os.path.split(f)
493
if e == "" or e == "." or e == "..":
496
return pathjoin(F(p), e)
500
"""True if f is an accessible directory."""
502
return S_ISDIR(os.lstat(f)[ST_MODE])
508
"""True if f is a regular file."""
510
return S_ISREG(os.lstat(f)[ST_MODE])
515
"""True if f is a symlink."""
517
return S_ISLNK(os.lstat(f)[ST_MODE])
521
def is_inside(dir, fname):
522
"""True if fname is inside dir.
524
The parameters should typically be passed to osutils.normpath first, so
525
that . and .. and repeated slashes are eliminated, and the separators
526
are canonical for the platform.
528
The empty string as a dir name is taken as top-of-tree and matches
531
# XXX: Most callers of this can actually do something smarter by
532
# looking at the inventory
542
return fname.startswith(dir)
545
def is_inside_any(dir_list, fname):
546
"""True if fname is inside any of given dirs."""
547
for dirname in dir_list:
548
if is_inside(dirname, fname):
553
def is_inside_or_parent_of_any(dir_list, fname):
554
"""True if fname is a child or a parent of any of the given files."""
555
for dirname in dir_list:
556
if is_inside(dirname, fname) or is_inside(fname, dirname):
561
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
562
report_activity=None, direction='read'):
563
"""Copy contents of one file to another.
565
The read_length can either be -1 to read to end-of-file (EOF) or
566
it can specify the maximum number of bytes to read.
568
The buff_size represents the maximum size for each read operation
569
performed on from_file.
571
:param report_activity: Call this as bytes are read, see
572
Transport._report_activity
573
:param direction: Will be passed to report_activity
575
:return: The number of bytes copied.
579
# read specified number of bytes
581
while read_length > 0:
582
num_bytes_to_read = min(read_length, buff_size)
584
block = from_file.read(num_bytes_to_read)
588
if report_activity is not None:
589
report_activity(len(block), direction)
592
actual_bytes_read = len(block)
593
read_length -= actual_bytes_read
594
length += actual_bytes_read
598
block = from_file.read(buff_size)
602
if report_activity is not None:
603
report_activity(len(block), direction)
609
def pump_string_file(bytes, file_handle, segment_size=None):
610
"""Write bytes to file_handle in many smaller writes.
612
:param bytes: The string to write.
613
:param file_handle: The file to write to.
615
# Write data in chunks rather than all at once, because very large
616
# writes fail on some platforms (e.g. Windows with SMB mounted
619
segment_size = 5242880 # 5MB
620
segments = range(len(bytes) / segment_size + 1)
621
write = file_handle.write
622
for segment_index in segments:
623
segment = buffer(bytes, segment_index * segment_size, segment_size)
627
def file_iterator(input_file, readsize=32768):
629
b = input_file.read(readsize)
636
"""Calculate the hexdigest of an open file.
638
The file cursor should be already at the start.
650
def size_sha_file(f):
651
"""Calculate the size and hexdigest of an open file.
653
The file cursor should be already at the start and
654
the caller is responsible for closing the file afterwards.
665
return size, s.hexdigest()
668
def sha_file_by_name(fname):
669
"""Calculate the SHA1 of a file by reading the full text"""
671
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
674
b = os.read(f, 1<<16)
682
def sha_strings(strings, _factory=sha):
683
"""Return the sha-1 of concatenation of strings"""
685
map(s.update, strings)
689
def sha_string(f, _factory=sha):
690
return _factory(f).hexdigest()
693
def fingerprint_file(f):
695
return {'size': len(b),
696
'sha1': sha(b).hexdigest()}
699
def compare_files(a, b):
700
"""Returns true if equal in contents"""
711
def local_time_offset(t=None):
712
"""Return offset of local zone from GMT, either at present or at time t."""
715
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
716
return offset.days * 86400 + offset.seconds
718
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
719
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
722
def format_date(t, offset=0, timezone='original', date_fmt=None,
724
"""Return a formatted date string.
726
:param t: Seconds since the epoch.
727
:param offset: Timezone offset in seconds east of utc.
728
:param timezone: How to display the time: 'utc', 'original' for the
729
timezone specified by offset, or 'local' for the process's current
731
:param date_fmt: strftime format.
732
:param show_offset: Whether to append the timezone.
734
(date_fmt, tt, offset_str) = \
735
_format_date(t, offset, timezone, date_fmt, show_offset)
736
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
737
date_str = time.strftime(date_fmt, tt)
738
return date_str + offset_str
741
# Cache of formatted offset strings
745
def format_date_with_offset_in_original_timezone(t, offset=0,
746
_cache=_offset_cache):
747
"""Return a formatted date string in the original timezone.
749
This routine may be faster then format_date.
751
:param t: Seconds since the epoch.
752
:param offset: Timezone offset in seconds east of utc.
756
tt = time.gmtime(t + offset)
757
date_fmt = _default_format_by_weekday_num[tt[6]]
758
date_str = time.strftime(date_fmt, tt)
759
offset_str = _cache.get(offset, None)
760
if offset_str is None:
761
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
762
_cache[offset] = offset_str
763
return date_str + offset_str
766
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
768
"""Return an unicode date string formatted according to the current locale.
770
:param t: Seconds since the epoch.
771
:param offset: Timezone offset in seconds east of utc.
772
:param timezone: How to display the time: 'utc', 'original' for the
773
timezone specified by offset, or 'local' for the process's current
775
:param date_fmt: strftime format.
776
:param show_offset: Whether to append the timezone.
778
(date_fmt, tt, offset_str) = \
779
_format_date(t, offset, timezone, date_fmt, show_offset)
780
date_str = time.strftime(date_fmt, tt)
781
if not isinstance(date_str, unicode):
782
date_str = date_str.decode(get_user_encoding(), 'replace')
783
return date_str + offset_str
786
def _format_date(t, offset, timezone, date_fmt, show_offset):
787
if timezone == 'utc':
790
elif timezone == 'original':
793
tt = time.gmtime(t + offset)
794
elif timezone == 'local':
795
tt = time.localtime(t)
796
offset = local_time_offset(t)
798
raise errors.UnsupportedTimezoneFormat(timezone)
800
date_fmt = "%a %Y-%m-%d %H:%M:%S"
802
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
805
return (date_fmt, tt, offset_str)
808
def compact_date(when):
809
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
812
def format_delta(delta):
813
"""Get a nice looking string for a time delta.
815
:param delta: The time difference in seconds, can be positive or negative.
816
positive indicates time in the past, negative indicates time in the
817
future. (usually time.time() - stored_time)
818
:return: String formatted to show approximate resolution
824
direction = 'in the future'
828
if seconds < 90: # print seconds up to 90 seconds
830
return '%d second %s' % (seconds, direction,)
832
return '%d seconds %s' % (seconds, direction)
834
minutes = int(seconds / 60)
835
seconds -= 60 * minutes
840
if minutes < 90: # print minutes, seconds up to 90 minutes
842
return '%d minute, %d second%s %s' % (
843
minutes, seconds, plural_seconds, direction)
845
return '%d minutes, %d second%s %s' % (
846
minutes, seconds, plural_seconds, direction)
848
hours = int(minutes / 60)
849
minutes -= 60 * hours
856
return '%d hour, %d minute%s %s' % (hours, minutes,
857
plural_minutes, direction)
858
return '%d hours, %d minute%s %s' % (hours, minutes,
859
plural_minutes, direction)
862
"""Return size of given open file."""
863
return os.fstat(f.fileno())[ST_SIZE]
866
# Define rand_bytes based on platform.
868
# Python 2.4 and later have os.urandom,
869
# but it doesn't work on some arches
871
rand_bytes = os.urandom
872
except (NotImplementedError, AttributeError):
873
# If python doesn't have os.urandom, or it doesn't work,
874
# then try to first pull random data from /dev/urandom
876
rand_bytes = file('/dev/urandom', 'rb').read
877
# Otherwise, use this hack as a last resort
878
except (IOError, OSError):
879
# not well seeded, but better than nothing
884
s += chr(random.randint(0, 255))
889
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
891
"""Return a random string of num alphanumeric characters
893
The result only contains lowercase chars because it may be used on
894
case-insensitive filesystems.
897
for raw_byte in rand_bytes(num):
898
s += ALNUM[ord(raw_byte) % 36]
902
## TODO: We could later have path objects that remember their list
903
## decomposition (might be too tricksy though.)
906
"""Turn string into list of parts."""
907
# split on either delimiter because people might use either on
909
ps = re.split(r'[\\/]', p)
914
raise errors.BzrError("sorry, %r not allowed in path" % f)
915
elif (f == '.') or (f == ''):
924
if (f == '..') or (f is None) or (f == ''):
925
raise errors.BzrError("sorry, %r not allowed in path" % f)
929
def parent_directories(filename):
930
"""Return the list of parent directories, deepest first.
932
For example, parent_directories("a/b/c") -> ["a/b", "a"].
935
parts = splitpath(dirname(filename))
937
parents.append(joinpath(parts))
942
_extension_load_failures = []
945
def failed_to_load_extension(exception):
946
"""Handle failing to load a binary extension.
948
This should be called from the ImportError block guarding the attempt to
949
import the native extension. If this function returns, the pure-Python
950
implementation should be loaded instead::
953
>>> import bzrlib._fictional_extension_pyx
954
>>> except ImportError, e:
955
>>> bzrlib.osutils.failed_to_load_extension(e)
956
>>> import bzrlib._fictional_extension_py
958
# NB: This docstring is just an example, not a doctest, because doctest
959
# currently can't cope with the use of lazy imports in this namespace --
962
# This currently doesn't report the failure at the time it occurs, because
963
# they tend to happen very early in startup when we can't check config
964
# files etc, and also we want to report all failures but not spam the user
966
from bzrlib import trace
967
exception_str = str(exception)
968
if exception_str not in _extension_load_failures:
969
trace.mutter("failed to load compiled extension: %s" % exception_str)
970
_extension_load_failures.append(exception_str)
973
def report_extension_load_failures():
974
if not _extension_load_failures:
976
from bzrlib.config import GlobalConfig
977
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
979
# the warnings framework should by default show this only once
980
from bzrlib.trace import warning
982
"bzr: warning: some compiled extensions could not be loaded; "
983
"see <https://answers.launchpad.net/bzr/+faq/703>")
984
# we no longer show the specific missing extensions here, because it makes
985
# the message too long and scary - see
986
# https://bugs.launchpad.net/bzr/+bug/430529
990
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
991
except ImportError, e:
992
failed_to_load_extension(e)
993
from bzrlib._chunks_to_lines_py import chunks_to_lines
997
"""Split s into lines, but without removing the newline characters."""
998
# Trivially convert a fulltext into a 'chunked' representation, and let
999
# chunks_to_lines do the heavy lifting.
1000
if isinstance(s, str):
1001
# chunks_to_lines only supports 8-bit strings
1002
return chunks_to_lines([s])
1004
return _split_lines(s)
1007
def _split_lines(s):
1008
"""Split s into lines, but without removing the newline characters.
1010
This supports Unicode or plain string objects.
1012
lines = s.split('\n')
1013
result = [line + '\n' for line in lines[:-1]]
1015
result.append(lines[-1])
1019
def hardlinks_good():
1020
return sys.platform not in ('win32', 'cygwin', 'darwin')
1023
def link_or_copy(src, dest):
1024
"""Hardlink a file, or copy it if it can't be hardlinked."""
1025
if not hardlinks_good():
1026
shutil.copyfile(src, dest)
1030
except (OSError, IOError), e:
1031
if e.errno != errno.EXDEV:
1033
shutil.copyfile(src, dest)
1036
def delete_any(path):
1037
"""Delete a file, symlink or directory.
1039
Will delete even if readonly.
1042
_delete_file_or_dir(path)
1043
except (OSError, IOError), e:
1044
if e.errno in (errno.EPERM, errno.EACCES):
1045
# make writable and try again
1048
except (OSError, IOError):
1050
_delete_file_or_dir(path)
1055
def _delete_file_or_dir(path):
1056
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1057
# Forgiveness than Permission (EAFP) because:
1058
# - root can damage a solaris file system by using unlink,
1059
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1060
# EACCES, OSX: EPERM) when invoked on a directory.
1061
if isdir(path): # Takes care of symlinks
1068
if getattr(os, 'symlink', None) is not None:
1074
def has_hardlinks():
1075
if getattr(os, 'link', None) is not None:
1081
def host_os_dereferences_symlinks():
1082
return (has_symlinks()
1083
and sys.platform not in ('cygwin', 'win32'))
1086
def readlink(abspath):
1087
"""Return a string representing the path to which the symbolic link points.
1089
:param abspath: The link absolute unicode path.
1091
This his guaranteed to return the symbolic link in unicode in all python
1094
link = abspath.encode(_fs_enc)
1095
target = os.readlink(link)
1096
target = target.decode(_fs_enc)
1100
def contains_whitespace(s):
1101
"""True if there are any whitespace characters in s."""
1102
# string.whitespace can include '\xa0' in certain locales, because it is
1103
# considered "non-breaking-space" as part of ISO-8859-1. But it
1104
# 1) Isn't a breaking whitespace
1105
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1107
# 3) '\xa0' isn't unicode safe since it is >128.
1109
# This should *not* be a unicode set of characters in case the source
1110
# string is not a Unicode string. We can auto-up-cast the characters since
1111
# they are ascii, but we don't want to auto-up-cast the string in case it
1113
for ch in ' \t\n\r\v\f':
1120
def contains_linebreaks(s):
1121
"""True if there is any vertical whitespace in s."""
1129
def relpath(base, path):
1130
"""Return path relative to base, or raise exception.
1132
The path may be either an absolute path or a path relative to the
1133
current working directory.
1135
os.path.commonprefix (python2.4) has a bad bug that it works just
1136
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1137
avoids that problem.
1140
if len(base) < MIN_ABS_PATHLENGTH:
1141
# must have space for e.g. a drive letter
1142
raise ValueError('%r is too short to calculate a relative path'
1150
if len(head) <= len(base) and head != base:
1151
raise errors.PathNotChild(rp, base)
1154
head, tail = split(head)
1159
return pathjoin(*reversed(s))
1164
def _cicp_canonical_relpath(base, path):
1165
"""Return the canonical path relative to base.
1167
Like relpath, but on case-insensitive-case-preserving file-systems, this
1168
will return the relpath as stored on the file-system rather than in the
1169
case specified in the input string, for all existing portions of the path.
1171
This will cause O(N) behaviour if called for every path in a tree; if you
1172
have a number of paths to convert, you should use canonical_relpaths().
1174
# TODO: it should be possible to optimize this for Windows by using the
1175
# win32 API FindFiles function to look for the specified name - but using
1176
# os.listdir() still gives us the correct, platform agnostic semantics in
1179
rel = relpath(base, path)
1180
# '.' will have been turned into ''
1184
abs_base = abspath(base)
1186
_listdir = os.listdir
1188
# use an explicit iterator so we can easily consume the rest on early exit.
1189
bit_iter = iter(rel.split('/'))
1190
for bit in bit_iter:
1193
next_entries = _listdir(current)
1194
except OSError: # enoent, eperm, etc
1195
# We can't find this in the filesystem, so just append the
1197
current = pathjoin(current, bit, *list(bit_iter))
1199
for look in next_entries:
1200
if lbit == look.lower():
1201
current = pathjoin(current, look)
1204
# got to the end, nothing matched, so we just return the
1205
# non-existing bits as they were specified (the filename may be
1206
# the target of a move, for example).
1207
current = pathjoin(current, bit, *list(bit_iter))
1209
return current[len(abs_base):].lstrip('/')
1211
# XXX - TODO - we need better detection/integration of case-insensitive
1212
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1213
# filesystems), for example, so could probably benefit from the same basic
1214
# support there. For now though, only Windows and OSX get that support, and
1215
# they get it for *all* file-systems!
1216
if sys.platform in ('win32', 'darwin'):
1217
canonical_relpath = _cicp_canonical_relpath
1219
canonical_relpath = relpath
1221
def canonical_relpaths(base, paths):
1222
"""Create an iterable to canonicalize a sequence of relative paths.
1224
The intent is for this implementation to use a cache, vastly speeding
1225
up multiple transformations in the same directory.
1227
# but for now, we haven't optimized...
1228
return [canonical_relpath(base, p) for p in paths]
1230
def safe_unicode(unicode_or_utf8_string):
1231
"""Coerce unicode_or_utf8_string into unicode.
1233
If it is unicode, it is returned.
1234
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1235
wrapped in a BzrBadParameterNotUnicode exception.
1237
if isinstance(unicode_or_utf8_string, unicode):
1238
return unicode_or_utf8_string
1240
return unicode_or_utf8_string.decode('utf8')
1241
except UnicodeDecodeError:
1242
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1245
def safe_utf8(unicode_or_utf8_string):
1246
"""Coerce unicode_or_utf8_string to a utf8 string.
1248
If it is a str, it is returned.
1249
If it is Unicode, it is encoded into a utf-8 string.
1251
if isinstance(unicode_or_utf8_string, str):
1252
# TODO: jam 20070209 This is overkill, and probably has an impact on
1253
# performance if we are dealing with lots of apis that want a
1256
# Make sure it is a valid utf-8 string
1257
unicode_or_utf8_string.decode('utf-8')
1258
except UnicodeDecodeError:
1259
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1260
return unicode_or_utf8_string
1261
return unicode_or_utf8_string.encode('utf-8')
1264
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1265
' Revision id generators should be creating utf8'
1269
def safe_revision_id(unicode_or_utf8_string, warn=True):
1270
"""Revision ids should now be utf8, but at one point they were unicode.
1272
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1274
:param warn: Functions that are sanitizing user data can set warn=False
1275
:return: None or a utf8 revision id.
1277
if (unicode_or_utf8_string is None
1278
or unicode_or_utf8_string.__class__ == str):
1279
return unicode_or_utf8_string
1281
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1283
return cache_utf8.encode(unicode_or_utf8_string)
1286
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1287
' generators should be creating utf8 file ids.')
1290
def safe_file_id(unicode_or_utf8_string, warn=True):
1291
"""File ids should now be utf8, but at one point they were unicode.
1293
This is the same as safe_utf8, except it uses the cached encode functions
1294
to save a little bit of performance.
1296
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1298
:param warn: Functions that are sanitizing user data can set warn=False
1299
:return: None or a utf8 file id.
1301
if (unicode_or_utf8_string is None
1302
or unicode_or_utf8_string.__class__ == str):
1303
return unicode_or_utf8_string
1305
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1307
return cache_utf8.encode(unicode_or_utf8_string)
1310
_platform_normalizes_filenames = False
1311
if sys.platform == 'darwin':
1312
_platform_normalizes_filenames = True
1315
def normalizes_filenames():
1316
"""Return True if this platform normalizes unicode filenames.
1318
Mac OSX does, Windows/Linux do not.
1320
return _platform_normalizes_filenames
1323
def _accessible_normalized_filename(path):
1324
"""Get the unicode normalized path, and if you can access the file.
1326
On platforms where the system normalizes filenames (Mac OSX),
1327
you can access a file by any path which will normalize correctly.
1328
On platforms where the system does not normalize filenames
1329
(Windows, Linux), you have to access a file by its exact path.
1331
Internally, bzr only supports NFC normalization, since that is
1332
the standard for XML documents.
1334
So return the normalized path, and a flag indicating if the file
1335
can be accessed by that path.
1338
return unicodedata.normalize('NFC', unicode(path)), True
1341
def _inaccessible_normalized_filename(path):
1342
__doc__ = _accessible_normalized_filename.__doc__
1344
normalized = unicodedata.normalize('NFC', unicode(path))
1345
return normalized, normalized == path
1348
if _platform_normalizes_filenames:
1349
normalized_filename = _accessible_normalized_filename
1351
normalized_filename = _inaccessible_normalized_filename
1354
def set_signal_handler(signum, handler, restart_syscall=True):
1355
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1356
on platforms that support that.
1358
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1359
automatically restart (by calling `signal.siginterrupt(signum,
1360
False)`). May be ignored if the feature is not available on this
1361
platform or Python version.
1363
old_handler = signal.signal(signum, handler)
1366
siginterrupt = signal.siginterrupt
1367
except AttributeError: # siginterrupt doesn't exist on this platform, or for this version of
1371
siginterrupt(signum, False)
1375
default_terminal_width = 80
1376
"""The default terminal width for ttys.
1378
This is defined so that higher levels can share a common fallback value when
1379
terminal_width() returns None.
1383
def terminal_width():
1384
"""Return terminal width.
1386
None is returned if the width can't established precisely.
1389
- if BZR_COLUMNS is set, returns its value
1390
- if there is no controlling terminal, returns None
1391
- if COLUMNS is set, returns its value,
1393
From there, we need to query the OS to get the size of the controlling
1397
- get termios.TIOCGWINSZ
1398
- if an error occurs or a negative value is obtained, returns None
1402
- win32utils.get_console_size() decides,
1403
- returns None on error (provided default value)
1406
# If BZR_COLUMNS is set, take it, user is always right
1408
return int(os.environ['BZR_COLUMNS'])
1409
except (KeyError, ValueError):
1412
isatty = getattr(sys.stdout, 'isatty', None)
1413
if isatty is None or not isatty():
1414
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1417
# If COLUMNS is set, take it, the terminal knows better (even inside a
1418
# given terminal, the application can decide to set COLUMNS to a lower
1419
# value (splitted screen) or a bigger value (scroll bars))
1421
return int(os.environ['COLUMNS'])
1422
except (KeyError, ValueError):
1425
width, height = _terminal_size(None, None)
1427
# Consider invalid values as meaning no width
1433
def _win32_terminal_size(width, height):
1434
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1435
return width, height
1438
def _ioctl_terminal_size(width, height):
1440
import struct, fcntl, termios
1441
s = struct.pack('HHHH', 0, 0, 0, 0)
1442
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1443
height, width = struct.unpack('HHHH', x)[0:2]
1444
except (IOError, AttributeError):
1446
return width, height
1448
_terminal_size = None
1449
"""Returns the terminal size as (width, height).
1451
:param width: Default value for width.
1452
:param height: Default value for height.
1454
This is defined specifically for each OS and query the size of the controlling
1455
terminal. If any error occurs, the provided default values should be returned.
1457
if sys.platform == 'win32':
1458
_terminal_size = _win32_terminal_size
1460
_terminal_size = _ioctl_terminal_size
1463
def _terminal_size_changed(signum, frame):
1464
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1465
width, height = _terminal_size(None, None)
1466
if width is not None:
1467
os.environ['COLUMNS'] = str(width)
1470
_registered_sigwinch = False
1472
def watch_sigwinch():
1473
"""Register for SIGWINCH, once and only once."""
1474
global _registered_sigwinch
1475
if not _registered_sigwinch:
1476
if sys.platform == 'win32':
1477
# Martin (gz) mentioned WINDOW_BUFFER_SIZE_RECORD from
1478
# ReadConsoleInput but I've no idea how to plug that in
1479
# the current design -- vila 20091216
1482
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1483
_registered_sigwinch = True
1486
def supports_executable():
1487
return sys.platform != "win32"
1490
def supports_posix_readonly():
1491
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1493
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1494
directory controls creation/deletion, etc.
1496
And under win32, readonly means that the directory itself cannot be
1497
deleted. The contents of a readonly directory can be changed, unlike POSIX
1498
where files in readonly directories cannot be added, deleted or renamed.
1500
return sys.platform != "win32"
1503
def set_or_unset_env(env_variable, value):
1504
"""Modify the environment, setting or removing the env_variable.
1506
:param env_variable: The environment variable in question
1507
:param value: The value to set the environment to. If None, then
1508
the variable will be removed.
1509
:return: The original value of the environment variable.
1511
orig_val = os.environ.get(env_variable)
1513
if orig_val is not None:
1514
del os.environ[env_variable]
1516
if isinstance(value, unicode):
1517
value = value.encode(get_user_encoding())
1518
os.environ[env_variable] = value
1522
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1525
def check_legal_path(path):
1526
"""Check whether the supplied path is legal.
1527
This is only required on Windows, so we don't test on other platforms
1530
if sys.platform != "win32":
1532
if _validWin32PathRE.match(path) is None:
1533
raise errors.IllegalPath(path)
1536
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1538
def _is_error_enotdir(e):
1539
"""Check if this exception represents ENOTDIR.
1541
Unfortunately, python is very inconsistent about the exception
1542
here. The cases are:
1543
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1544
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1545
which is the windows error code.
1546
3) Windows, Python2.5 uses errno == EINVAL and
1547
winerror == ERROR_DIRECTORY
1549
:param e: An Exception object (expected to be OSError with an errno
1550
attribute, but we should be able to cope with anything)
1551
:return: True if this represents an ENOTDIR error. False otherwise.
1553
en = getattr(e, 'errno', None)
1554
if (en == errno.ENOTDIR
1555
or (sys.platform == 'win32'
1556
and (en == _WIN32_ERROR_DIRECTORY
1557
or (en == errno.EINVAL
1558
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1564
def walkdirs(top, prefix=""):
1565
"""Yield data about all the directories in a tree.
1567
This yields all the data about the contents of a directory at a time.
1568
After each directory has been yielded, if the caller has mutated the list
1569
to exclude some directories, they are then not descended into.
1571
The data yielded is of the form:
1572
((directory-relpath, directory-path-from-top),
1573
[(relpath, basename, kind, lstat, path-from-top), ...]),
1574
- directory-relpath is the relative path of the directory being returned
1575
with respect to top. prefix is prepended to this.
1576
- directory-path-from-root is the path including top for this directory.
1577
It is suitable for use with os functions.
1578
- relpath is the relative path within the subtree being walked.
1579
- basename is the basename of the path
1580
- kind is the kind of the file now. If unknown then the file is not
1581
present within the tree - but it may be recorded as versioned. See
1583
- lstat is the stat data *if* the file was statted.
1584
- planned, not implemented:
1585
path_from_tree_root is the path from the root of the tree.
1587
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1588
allows one to walk a subtree but get paths that are relative to a tree
1590
:return: an iterator over the dirs.
1592
#TODO there is a bit of a smell where the results of the directory-
1593
# summary in this, and the path from the root, may not agree
1594
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1595
# potentially confusing output. We should make this more robust - but
1596
# not at a speed cost. RBC 20060731
1598
_directory = _directory_kind
1599
_listdir = os.listdir
1600
_kind_from_mode = file_kind_from_stat_mode
1601
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1603
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1604
relroot, _, _, _, top = pending.pop()
1606
relprefix = relroot + u'/'
1609
top_slash = top + u'/'
1612
append = dirblock.append
1614
names = sorted(_listdir(top))
1616
if not _is_error_enotdir(e):
1620
abspath = top_slash + name
1621
statvalue = _lstat(abspath)
1622
kind = _kind_from_mode(statvalue.st_mode)
1623
append((relprefix + name, name, kind, statvalue, abspath))
1624
yield (relroot, top), dirblock
1626
# push the user specified dirs from dirblock
1627
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1630
class DirReader(object):
1631
"""An interface for reading directories."""
1633
def top_prefix_to_starting_dir(self, top, prefix=""):
1634
"""Converts top and prefix to a starting dir entry
1636
:param top: A utf8 path
1637
:param prefix: An optional utf8 path to prefix output relative paths
1639
:return: A tuple starting with prefix, and ending with the native
1642
raise NotImplementedError(self.top_prefix_to_starting_dir)
1644
def read_dir(self, prefix, top):
1645
"""Read a specific dir.
1647
:param prefix: A utf8 prefix to be preprended to the path basenames.
1648
:param top: A natively encoded path to read.
1649
:return: A list of the directories contents. Each item contains:
1650
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1652
raise NotImplementedError(self.read_dir)
1655
_selected_dir_reader = None
1658
def _walkdirs_utf8(top, prefix=""):
1659
"""Yield data about all the directories in a tree.
1661
This yields the same information as walkdirs() only each entry is yielded
1662
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1663
are returned as exact byte-strings.
1665
:return: yields a tuple of (dir_info, [file_info])
1666
dir_info is (utf8_relpath, path-from-top)
1667
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1668
if top is an absolute path, path-from-top is also an absolute path.
1669
path-from-top might be unicode or utf8, but it is the correct path to
1670
pass to os functions to affect the file in question. (such as os.lstat)
1672
global _selected_dir_reader
1673
if _selected_dir_reader is None:
1674
fs_encoding = _fs_enc.upper()
1675
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1676
# Win98 doesn't have unicode apis like FindFirstFileW
1677
# TODO: We possibly could support Win98 by falling back to the
1678
# original FindFirstFile, and using TCHAR instead of WCHAR,
1679
# but that gets a bit tricky, and requires custom compiling
1682
from bzrlib._walkdirs_win32 import Win32ReadDir
1683
_selected_dir_reader = Win32ReadDir()
1686
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1687
# ANSI_X3.4-1968 is a form of ASCII
1689
from bzrlib._readdir_pyx import UTF8DirReader
1690
_selected_dir_reader = UTF8DirReader()
1691
except ImportError, e:
1692
failed_to_load_extension(e)
1695
if _selected_dir_reader is None:
1696
# Fallback to the python version
1697
_selected_dir_reader = UnicodeDirReader()
1699
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1700
# But we don't actually uses 1-3 in pending, so set them to None
1701
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1702
read_dir = _selected_dir_reader.read_dir
1703
_directory = _directory_kind
1705
relroot, _, _, _, top = pending[-1].pop()
1708
dirblock = sorted(read_dir(relroot, top))
1709
yield (relroot, top), dirblock
1710
# push the user specified dirs from dirblock
1711
next = [d for d in reversed(dirblock) if d[2] == _directory]
1713
pending.append(next)
1716
class UnicodeDirReader(DirReader):
1717
"""A dir reader for non-utf8 file systems, which transcodes."""
1719
__slots__ = ['_utf8_encode']
1722
self._utf8_encode = codecs.getencoder('utf8')
1724
def top_prefix_to_starting_dir(self, top, prefix=""):
1725
"""See DirReader.top_prefix_to_starting_dir."""
1726
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1728
def read_dir(self, prefix, top):
1729
"""Read a single directory from a non-utf8 file system.
1731
top, and the abspath element in the output are unicode, all other paths
1732
are utf8. Local disk IO is done via unicode calls to listdir etc.
1734
This is currently the fallback code path when the filesystem encoding is
1735
not UTF-8. It may be better to implement an alternative so that we can
1736
safely handle paths that are not properly decodable in the current
1739
See DirReader.read_dir for details.
1741
_utf8_encode = self._utf8_encode
1743
_listdir = os.listdir
1744
_kind_from_mode = file_kind_from_stat_mode
1747
relprefix = prefix + '/'
1750
top_slash = top + u'/'
1753
append = dirblock.append
1754
for name in sorted(_listdir(top)):
1756
name_utf8 = _utf8_encode(name)[0]
1757
except UnicodeDecodeError:
1758
raise errors.BadFilenameEncoding(
1759
_utf8_encode(relprefix)[0] + name, _fs_enc)
1760
abspath = top_slash + name
1761
statvalue = _lstat(abspath)
1762
kind = _kind_from_mode(statvalue.st_mode)
1763
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1767
def copy_tree(from_path, to_path, handlers={}):
1768
"""Copy all of the entries in from_path into to_path.
1770
:param from_path: The base directory to copy.
1771
:param to_path: The target directory. If it does not exist, it will
1773
:param handlers: A dictionary of functions, which takes a source and
1774
destinations for files, directories, etc.
1775
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1776
'file', 'directory', and 'symlink' should always exist.
1777
If they are missing, they will be replaced with 'os.mkdir()',
1778
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1780
# Now, just copy the existing cached tree to the new location
1781
# We use a cheap trick here.
1782
# Absolute paths are prefixed with the first parameter
1783
# relative paths are prefixed with the second.
1784
# So we can get both the source and target returned
1785
# without any extra work.
1787
def copy_dir(source, dest):
1790
def copy_link(source, dest):
1791
"""Copy the contents of a symlink"""
1792
link_to = os.readlink(source)
1793
os.symlink(link_to, dest)
1795
real_handlers = {'file':shutil.copy2,
1796
'symlink':copy_link,
1797
'directory':copy_dir,
1799
real_handlers.update(handlers)
1801
if not os.path.exists(to_path):
1802
real_handlers['directory'](from_path, to_path)
1804
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1805
for relpath, name, kind, st, abspath in entries:
1806
real_handlers[kind](abspath, relpath)
1809
def copy_ownership(dst, src=None):
1810
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1812
If src is None, the containing directory is used as source. If chown
1813
fails, the error is ignored and a warning is printed.
1815
has_chown = getattr(os, 'chown')
1816
if has_chown is None: return
1819
src = os.path.dirname(dst)
1825
os.chown(dst, s.st_uid, s.st_gid)
1827
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1830
def mkdir_with_ownership(path, ownership_src=None):
1831
"""Create the directory 'path' with specified ownership.
1833
If ownership_src is given, copies (chown) usr/grp ownership
1834
from 'ownership_src' to 'path'. If ownership_src is None, use the
1835
containing dir ownership.
1838
copy_ownership(path, ownership_src)
1841
def open_with_ownership(filename, mode='r', bufsize=-1, ownership_src=None):
1842
"""Open the file 'filename' with the specified ownership.
1844
If ownership_src is specified, copy usr/grp ownership from ownership_src
1845
to filename. If ownership_src is None, copy ownership from containing
1847
Returns the opened file object.
1849
f = open(filename, mode, bufsize)
1850
copy_ownership(filename, ownership_src)
1854
def path_prefix_key(path):
1855
"""Generate a prefix-order path key for path.
1857
This can be used to sort paths in the same way that walkdirs does.
1859
return (dirname(path) , path)
1862
def compare_paths_prefix_order(path_a, path_b):
1863
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1864
key_a = path_prefix_key(path_a)
1865
key_b = path_prefix_key(path_b)
1866
return cmp(key_a, key_b)
1869
_cached_user_encoding = None
1872
def get_user_encoding(use_cache=True):
1873
"""Find out what the preferred user encoding is.
1875
This is generally the encoding that is used for command line parameters
1876
and file contents. This may be different from the terminal encoding
1877
or the filesystem encoding.
1879
:param use_cache: Enable cache for detected encoding.
1880
(This parameter is turned on by default,
1881
and required only for selftesting)
1883
:return: A string defining the preferred user encoding
1885
global _cached_user_encoding
1886
if _cached_user_encoding is not None and use_cache:
1887
return _cached_user_encoding
1889
if sys.platform == 'darwin':
1890
# python locale.getpreferredencoding() always return
1891
# 'mac-roman' on darwin. That's a lie.
1892
sys.platform = 'posix'
1894
if os.environ.get('LANG', None) is None:
1895
# If LANG is not set, we end up with 'ascii', which is bad
1896
# ('mac-roman' is more than ascii), so we set a default which
1897
# will give us UTF-8 (which appears to work in all cases on
1898
# OSX). Users are still free to override LANG of course, as
1899
# long as it give us something meaningful. This work-around
1900
# *may* not be needed with python 3k and/or OSX 10.5, but will
1901
# work with them too -- vila 20080908
1902
os.environ['LANG'] = 'en_US.UTF-8'
1905
sys.platform = 'darwin'
1910
user_encoding = locale.getpreferredencoding()
1911
except locale.Error, e:
1912
sys.stderr.write('bzr: warning: %s\n'
1913
' Could not determine what text encoding to use.\n'
1914
' This error usually means your Python interpreter\n'
1915
' doesn\'t support the locale set by $LANG (%s)\n'
1916
" Continuing with ascii encoding.\n"
1917
% (e, os.environ.get('LANG')))
1918
user_encoding = 'ascii'
1920
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1921
# treat that as ASCII, and not support printing unicode characters to the
1924
# For python scripts run under vim, we get '', so also treat that as ASCII
1925
if user_encoding in (None, 'cp0', ''):
1926
user_encoding = 'ascii'
1930
codecs.lookup(user_encoding)
1932
sys.stderr.write('bzr: warning:'
1933
' unknown encoding %s.'
1934
' Continuing with ascii encoding.\n'
1937
user_encoding = 'ascii'
1940
_cached_user_encoding = user_encoding
1942
return user_encoding
1945
def get_host_name():
1946
"""Return the current unicode host name.
1948
This is meant to be used in place of socket.gethostname() because that
1949
behaves inconsistently on different platforms.
1951
if sys.platform == "win32":
1953
return win32utils.get_host_name()
1956
return socket.gethostname().decode(get_user_encoding())
1959
def recv_all(socket, bytes):
1960
"""Receive an exact number of bytes.
1962
Regular Socket.recv() may return less than the requested number of bytes,
1963
dependning on what's in the OS buffer. MSG_WAITALL is not available
1964
on all platforms, but this should work everywhere. This will return
1965
less than the requested amount if the remote end closes.
1967
This isn't optimized and is intended mostly for use in testing.
1970
while len(b) < bytes:
1971
new = until_no_eintr(socket.recv, bytes - len(b))
1978
def send_all(socket, bytes, report_activity=None):
1979
"""Send all bytes on a socket.
1981
Regular socket.sendall() can give socket error 10053 on Windows. This
1982
implementation sends no more than 64k at a time, which avoids this problem.
1984
:param report_activity: Call this as bytes are read, see
1985
Transport._report_activity
1988
for pos in xrange(0, len(bytes), chunk_size):
1989
block = bytes[pos:pos+chunk_size]
1990
if report_activity is not None:
1991
report_activity(len(block), 'write')
1992
until_no_eintr(socket.sendall, block)
1995
def dereference_path(path):
1996
"""Determine the real path to a file.
1998
All parent elements are dereferenced. But the file itself is not
2000
:param path: The original path. May be absolute or relative.
2001
:return: the real path *to* the file
2003
parent, base = os.path.split(path)
2004
# The pathjoin for '.' is a workaround for Python bug #1213894.
2005
# (initial path components aren't dereferenced)
2006
return pathjoin(realpath(pathjoin('.', parent)), base)
2009
def supports_mapi():
2010
"""Return True if we can use MAPI to launch a mail client."""
2011
return sys.platform == "win32"
2014
def resource_string(package, resource_name):
2015
"""Load a resource from a package and return it as a string.
2017
Note: Only packages that start with bzrlib are currently supported.
2019
This is designed to be a lightweight implementation of resource
2020
loading in a way which is API compatible with the same API from
2022
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2023
If and when pkg_resources becomes a standard library, this routine
2026
# Check package name is within bzrlib
2027
if package == "bzrlib":
2028
resource_relpath = resource_name
2029
elif package.startswith("bzrlib."):
2030
package = package[len("bzrlib."):].replace('.', os.sep)
2031
resource_relpath = pathjoin(package, resource_name)
2033
raise errors.BzrError('resource package %s not in bzrlib' % package)
2035
# Map the resource to a file and read its contents
2036
base = dirname(bzrlib.__file__)
2037
if getattr(sys, 'frozen', None): # bzr.exe
2038
base = abspath(pathjoin(base, '..', '..'))
2039
filename = pathjoin(base, resource_relpath)
2040
return open(filename, 'rU').read()
2043
def file_kind_from_stat_mode_thunk(mode):
2044
global file_kind_from_stat_mode
2045
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2047
from bzrlib._readdir_pyx import UTF8DirReader
2048
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2049
except ImportError, e:
2050
# This is one time where we won't warn that an extension failed to
2051
# load. The extension is never available on Windows anyway.
2052
from bzrlib._readdir_py import (
2053
_kind_from_mode as file_kind_from_stat_mode
2055
return file_kind_from_stat_mode(mode)
2056
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2059
def file_kind(f, _lstat=os.lstat):
2061
return file_kind_from_stat_mode(_lstat(f).st_mode)
2063
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2064
raise errors.NoSuchFile(f)
2068
def until_no_eintr(f, *a, **kw):
2069
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
2070
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2074
except (IOError, OSError), e:
2075
if e.errno == errno.EINTR:
2079
def re_compile_checked(re_string, flags=0, where=""):
2080
"""Return a compiled re, or raise a sensible error.
2082
This should only be used when compiling user-supplied REs.
2084
:param re_string: Text form of regular expression.
2085
:param flags: eg re.IGNORECASE
2086
:param where: Message explaining to the user the context where
2087
it occurred, eg 'log search filter'.
2089
# from https://bugs.launchpad.net/bzr/+bug/251352
2091
re_obj = re.compile(re_string, flags)
2096
where = ' in ' + where
2097
# despite the name 'error' is a type
2098
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2099
% (where, re_string, e))
2102
if sys.platform == "win32":
2105
return msvcrt.getch()
2110
fd = sys.stdin.fileno()
2111
settings = termios.tcgetattr(fd)
2114
ch = sys.stdin.read(1)
2116
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2120
if sys.platform == 'linux2':
2121
def _local_concurrency():
2123
prefix = 'processor'
2124
for line in file('/proc/cpuinfo', 'rb'):
2125
if line.startswith(prefix):
2126
concurrency = int(line[line.find(':')+1:]) + 1
2128
elif sys.platform == 'darwin':
2129
def _local_concurrency():
2130
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2131
stdout=subprocess.PIPE).communicate()[0]
2132
elif sys.platform[0:7] == 'freebsd':
2133
def _local_concurrency():
2134
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2135
stdout=subprocess.PIPE).communicate()[0]
2136
elif sys.platform == 'sunos5':
2137
def _local_concurrency():
2138
return subprocess.Popen(['psrinfo', '-p',],
2139
stdout=subprocess.PIPE).communicate()[0]
2140
elif sys.platform == "win32":
2141
def _local_concurrency():
2142
# This appears to return the number of cores.
2143
return os.environ.get('NUMBER_OF_PROCESSORS')
2145
def _local_concurrency():
2150
_cached_local_concurrency = None
2152
def local_concurrency(use_cache=True):
2153
"""Return how many processes can be run concurrently.
2155
Rely on platform specific implementations and default to 1 (one) if
2156
anything goes wrong.
2158
global _cached_local_concurrency
2160
if _cached_local_concurrency is not None and use_cache:
2161
return _cached_local_concurrency
2163
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2164
if concurrency is None:
2166
concurrency = _local_concurrency()
2167
except (OSError, IOError):
2170
concurrency = int(concurrency)
2171
except (TypeError, ValueError):
2174
_cached_concurrency = concurrency
2178
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2179
"""A stream writer that doesn't decode str arguments."""
2181
def __init__(self, encode, stream, errors='strict'):
2182
codecs.StreamWriter.__init__(self, stream, errors)
2183
self.encode = encode
2185
def write(self, object):
2186
if type(object) is str:
2187
self.stream.write(object)
2189
data, _ = self.encode(object, self.errors)
2190
self.stream.write(data)
2192
if sys.platform == 'win32':
2193
def open_file(filename, mode='r', bufsize=-1):
2194
"""This function is used to override the ``open`` builtin.
2196
But it uses O_NOINHERIT flag so the file handle is not inherited by
2197
child processes. Deleting or renaming a closed file opened with this
2198
function is not blocking child processes.
2200
writing = 'w' in mode
2201
appending = 'a' in mode
2202
updating = '+' in mode
2203
binary = 'b' in mode
2206
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2207
# for flags for each modes.
2217
flags |= os.O_WRONLY
2218
flags |= os.O_CREAT | os.O_TRUNC
2223
flags |= os.O_WRONLY
2224
flags |= os.O_CREAT | os.O_APPEND
2229
flags |= os.O_RDONLY
2231
return os.fdopen(os.open(filename, flags), mode, bufsize)
added
removed
bzrlib/osutils.py
