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
21
from stat import S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE
26
from bzrlib.lazy_import import lazy_import
27
lazy_import(globals(), """
28
from datetime import datetime
30
from ntpath import (abspath as _nt_abspath,
32
normpath as _nt_normpath,
33
realpath as _nt_realpath,
34
splitdrive as _nt_splitdrive,
44
from tempfile import (
57
from bzrlib.symbol_versioning import (
62
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
64
if sys.version_info < (2, 5):
65
import md5 as _mod_md5
67
import sha as _mod_sha
77
from bzrlib import symbol_versioning
80
# Cross platform wall-clock time functionality with decent resolution.
81
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
82
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
83
# synchronized with ``time.time()``, this is only meant to be used to find
84
# delta times by subtracting from another call to this function.
85
timer_func = time.time
86
if sys.platform == 'win32':
87
timer_func = time.clock
89
# On win32, O_BINARY is used to indicate the file should
90
# be opened in binary mode, rather than text mode.
91
# On other platforms, O_BINARY doesn't exist, because
92
# they always open in binary mode, so it is okay to
93
# OR with 0 on those platforms.
94
# O_NOINHERIT and O_TEXT exists only on win32 too.
95
O_BINARY = getattr(os, 'O_BINARY', 0)
96
O_TEXT = getattr(os, 'O_TEXT', 0)
97
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
100
def get_unicode_argv():
102
user_encoding = get_user_encoding()
103
return [a.decode(user_encoding) for a in sys.argv[1:]]
104
except UnicodeDecodeError:
105
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
109
def make_readonly(filename):
110
"""Make a filename read-only."""
111
mod = os.lstat(filename).st_mode
112
if not stat.S_ISLNK(mod):
114
os.chmod(filename, mod)
117
def make_writable(filename):
118
mod = os.lstat(filename).st_mode
119
if not stat.S_ISLNK(mod):
121
os.chmod(filename, mod)
124
def minimum_path_selection(paths):
125
"""Return the smallset subset of paths which are outside paths.
127
:param paths: A container (and hence not None) of paths.
128
:return: A set of paths sufficient to include everything in paths via
129
is_inside, drawn from the paths parameter.
135
return path.split('/')
136
sorted_paths = sorted(list(paths), key=sort_key)
138
search_paths = [sorted_paths[0]]
139
for path in sorted_paths[1:]:
140
if not is_inside(search_paths[-1], path):
141
# This path is unique, add it
142
search_paths.append(path)
144
return set(search_paths)
151
"""Return a quoted filename filename
153
This previously used backslash quoting, but that works poorly on
155
# TODO: I'm not really sure this is the best format either.x
157
if _QUOTE_RE is None:
158
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
160
if _QUOTE_RE.search(f):
166
_directory_kind = 'directory'
169
"""Return the current umask"""
170
# Assume that people aren't messing with the umask while running
171
# XXX: This is not thread safe, but there is no way to get the
172
# umask without setting it
180
_directory_kind: "/",
182
'tree-reference': '+',
186
def kind_marker(kind):
188
return _kind_marker_map[kind]
190
# Slightly faster than using .get(, '') when the common case is that
195
lexists = getattr(os.path, 'lexists', None)
199
stat = getattr(os, 'lstat', os.stat)
203
if e.errno == errno.ENOENT:
206
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
209
def fancy_rename(old, new, rename_func, unlink_func):
210
"""A fancy rename, when you don't have atomic rename.
212
:param old: The old path, to rename from
213
:param new: The new path, to rename to
214
:param rename_func: The potentially non-atomic rename function
215
:param unlink_func: A way to delete the target file if the full rename
218
# sftp rename doesn't allow overwriting, so play tricks:
219
base = os.path.basename(new)
220
dirname = os.path.dirname(new)
221
# callers use different encodings for the paths so the following MUST
222
# respect that. We rely on python upcasting to unicode if new is unicode
223
# and keeping a str if not.
224
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
225
os.getpid(), rand_chars(10))
226
tmp_name = pathjoin(dirname, tmp_name)
228
# Rename the file out of the way, but keep track if it didn't exist
229
# We don't want to grab just any exception
230
# something like EACCES should prevent us from continuing
231
# The downside is that the rename_func has to throw an exception
232
# with an errno = ENOENT, or NoSuchFile
235
rename_func(new, tmp_name)
236
except (errors.NoSuchFile,), e:
239
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
240
# function raises an IOError with errno is None when a rename fails.
241
# This then gets caught here.
242
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
245
if (getattr(e, 'errno', None) is None
246
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
255
# This may throw an exception, in which case success will
257
rename_func(old, new)
259
except (IOError, OSError), e:
260
# source and target may be aliases of each other (e.g. on a
261
# case-insensitive filesystem), so we may have accidentally renamed
262
# source by when we tried to rename target
263
failure_exc = sys.exc_info()
264
if (file_existed and e.errno in (None, errno.ENOENT)
265
and old.lower() == new.lower()):
266
# source and target are the same file on a case-insensitive
267
# filesystem, so we don't generate an exception
271
# If the file used to exist, rename it back into place
272
# otherwise just delete it from the tmp location
274
unlink_func(tmp_name)
276
rename_func(tmp_name, new)
277
if failure_exc is not None:
278
raise failure_exc[0], failure_exc[1], failure_exc[2]
281
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
282
# choke on a Unicode string containing a relative path if
283
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
285
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
286
def _posix_abspath(path):
287
# jam 20060426 rather than encoding to fsencoding
288
# copy posixpath.abspath, but use os.getcwdu instead
289
if not posixpath.isabs(path):
290
path = posixpath.join(getcwd(), path)
291
return posixpath.normpath(path)
294
def _posix_realpath(path):
295
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
298
def _win32_fixdrive(path):
299
"""Force drive letters to be consistent.
301
win32 is inconsistent whether it returns lower or upper case
302
and even if it was consistent the user might type the other
303
so we force it to uppercase
304
running python.exe under cmd.exe return capital C:\\
305
running win32 python inside a cygwin shell returns lowercase c:\\
307
drive, path = _nt_splitdrive(path)
308
return drive.upper() + path
311
def _win32_abspath(path):
312
# Real _nt_abspath doesn't have a problem with a unicode cwd
313
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
316
def _win98_abspath(path):
317
"""Return the absolute version of a path.
318
Windows 98 safe implementation (python reimplementation
319
of Win32 API function GetFullPathNameW)
324
# \\HOST\path => //HOST/path
325
# //HOST/path => //HOST/path
326
# path => C:/cwd/path
329
# check for absolute path
330
drive = _nt_splitdrive(path)[0]
331
if drive == '' and path[:2] not in('//','\\\\'):
333
# we cannot simply os.path.join cwd and path
334
# because os.path.join('C:','/path') produce '/path'
335
# and this is incorrect
336
if path[:1] in ('/','\\'):
337
cwd = _nt_splitdrive(cwd)[0]
339
path = cwd + '\\' + path
340
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
343
def _win32_realpath(path):
344
# Real _nt_realpath doesn't have a problem with a unicode cwd
345
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
348
def _win32_pathjoin(*args):
349
return _nt_join(*args).replace('\\', '/')
352
def _win32_normpath(path):
353
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
357
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
360
def _win32_mkdtemp(*args, **kwargs):
361
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
364
def _win32_rename(old, new):
365
"""We expect to be able to atomically replace 'new' with old.
367
On win32, if new exists, it must be moved out of the way first,
371
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
373
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
374
# If we try to rename a non-existant file onto cwd, we get
375
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
376
# if the old path doesn't exist, sometimes we get EACCES
377
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
383
return unicodedata.normalize('NFC', os.getcwdu())
386
# Default is to just use the python builtins, but these can be rebound on
387
# particular platforms.
388
abspath = _posix_abspath
389
realpath = _posix_realpath
390
pathjoin = os.path.join
391
normpath = os.path.normpath
394
dirname = os.path.dirname
395
basename = os.path.basename
396
split = os.path.split
397
splitext = os.path.splitext
398
# These were already imported into local scope
399
# mkdtemp = tempfile.mkdtemp
400
# rmtree = shutil.rmtree
402
MIN_ABS_PATHLENGTH = 1
405
if sys.platform == 'win32':
406
if win32utils.winver == 'Windows 98':
407
abspath = _win98_abspath
409
abspath = _win32_abspath
410
realpath = _win32_realpath
411
pathjoin = _win32_pathjoin
412
normpath = _win32_normpath
413
getcwd = _win32_getcwd
414
mkdtemp = _win32_mkdtemp
415
rename = _win32_rename
417
MIN_ABS_PATHLENGTH = 3
419
def _win32_delete_readonly(function, path, excinfo):
420
"""Error handler for shutil.rmtree function [for win32]
421
Helps to remove files and dirs marked as read-only.
423
exception = excinfo[1]
424
if function in (os.remove, os.rmdir) \
425
and isinstance(exception, OSError) \
426
and exception.errno == errno.EACCES:
432
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
433
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
434
return shutil.rmtree(path, ignore_errors, onerror)
436
f = win32utils.get_unicode_argv # special function or None
440
elif sys.platform == 'darwin':
444
def get_terminal_encoding(trace=False):
445
"""Find the best encoding for printing to the screen.
447
This attempts to check both sys.stdout and sys.stdin to see
448
what encoding they are in, and if that fails it falls back to
449
osutils.get_user_encoding().
450
The problem is that on Windows, locale.getpreferredencoding()
451
is not the same encoding as that used by the console:
452
http://mail.python.org/pipermail/python-list/2003-May/162357.html
454
On my standard US Windows XP, the preferred encoding is
455
cp1252, but the console is cp437
457
:param trace: If True trace the selected encoding via mutter().
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()
466
mutter('encoding stdout as osutils.get_user_encoding() %r',
469
output_encoding = input_encoding
471
mutter('encoding stdout as sys.stdin encoding %r',
475
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
476
if output_encoding == 'cp0':
477
# invalid encoding (cp0 means 'no codepage' on Windows)
478
output_encoding = get_user_encoding()
480
mutter('cp0 is invalid encoding.'
481
' encoding stdout as osutils.get_user_encoding() %r',
485
codecs.lookup(output_encoding)
487
sys.stderr.write('bzr: warning:'
488
' unknown terminal encoding %s.\n'
489
' Using encoding %s instead.\n'
490
% (output_encoding, get_user_encoding())
492
output_encoding = get_user_encoding()
494
return output_encoding
497
def normalizepath(f):
498
if getattr(os.path, 'realpath', None) is not None:
502
[p,e] = os.path.split(f)
503
if e == "" or e == "." or e == "..":
506
return pathjoin(F(p), e)
510
"""True if f is an accessible directory."""
512
return S_ISDIR(os.lstat(f)[ST_MODE])
518
"""True if f is a regular file."""
520
return S_ISREG(os.lstat(f)[ST_MODE])
525
"""True if f is a symlink."""
527
return S_ISLNK(os.lstat(f)[ST_MODE])
531
def is_inside(dir, fname):
532
"""True if fname is inside dir.
534
The parameters should typically be passed to osutils.normpath first, so
535
that . and .. and repeated slashes are eliminated, and the separators
536
are canonical for the platform.
538
The empty string as a dir name is taken as top-of-tree and matches
541
# XXX: Most callers of this can actually do something smarter by
542
# looking at the inventory
552
return fname.startswith(dir)
555
def is_inside_any(dir_list, fname):
556
"""True if fname is inside any of given dirs."""
557
for dirname in dir_list:
558
if is_inside(dirname, fname):
563
def is_inside_or_parent_of_any(dir_list, fname):
564
"""True if fname is a child or a parent of any of the given files."""
565
for dirname in dir_list:
566
if is_inside(dirname, fname) or is_inside(fname, dirname):
571
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
572
report_activity=None, direction='read'):
573
"""Copy contents of one file to another.
575
The read_length can either be -1 to read to end-of-file (EOF) or
576
it can specify the maximum number of bytes to read.
578
The buff_size represents the maximum size for each read operation
579
performed on from_file.
581
:param report_activity: Call this as bytes are read, see
582
Transport._report_activity
583
:param direction: Will be passed to report_activity
585
:return: The number of bytes copied.
589
# read specified number of bytes
591
while read_length > 0:
592
num_bytes_to_read = min(read_length, buff_size)
594
block = from_file.read(num_bytes_to_read)
598
if report_activity is not None:
599
report_activity(len(block), direction)
602
actual_bytes_read = len(block)
603
read_length -= actual_bytes_read
604
length += actual_bytes_read
608
block = from_file.read(buff_size)
612
if report_activity is not None:
613
report_activity(len(block), direction)
619
def pump_string_file(bytes, file_handle, segment_size=None):
620
"""Write bytes to file_handle in many smaller writes.
622
:param bytes: The string to write.
623
:param file_handle: The file to write to.
625
# Write data in chunks rather than all at once, because very large
626
# writes fail on some platforms (e.g. Windows with SMB mounted
629
segment_size = 5242880 # 5MB
630
segments = range(len(bytes) / segment_size + 1)
631
write = file_handle.write
632
for segment_index in segments:
633
segment = buffer(bytes, segment_index * segment_size, segment_size)
637
def file_iterator(input_file, readsize=32768):
639
b = input_file.read(readsize)
646
"""Calculate the hexdigest of an open file.
648
The file cursor should be already at the start.
660
def size_sha_file(f):
661
"""Calculate the size and hexdigest of an open file.
663
The file cursor should be already at the start and
664
the caller is responsible for closing the file afterwards.
675
return size, s.hexdigest()
678
def sha_file_by_name(fname):
679
"""Calculate the SHA1 of a file by reading the full text"""
681
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
684
b = os.read(f, 1<<16)
692
def sha_strings(strings, _factory=sha):
693
"""Return the sha-1 of concatenation of strings"""
695
map(s.update, strings)
699
def sha_string(f, _factory=sha):
700
return _factory(f).hexdigest()
703
def fingerprint_file(f):
705
return {'size': len(b),
706
'sha1': sha(b).hexdigest()}
709
def compare_files(a, b):
710
"""Returns true if equal in contents"""
721
def local_time_offset(t=None):
722
"""Return offset of local zone from GMT, either at present or at time t."""
725
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
726
return offset.days * 86400 + offset.seconds
728
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
729
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
732
def format_date(t, offset=0, timezone='original', date_fmt=None,
734
"""Return a formatted date string.
736
:param t: Seconds since the epoch.
737
:param offset: Timezone offset in seconds east of utc.
738
:param timezone: How to display the time: 'utc', 'original' for the
739
timezone specified by offset, or 'local' for the process's current
741
:param date_fmt: strftime format.
742
:param show_offset: Whether to append the timezone.
744
(date_fmt, tt, offset_str) = \
745
_format_date(t, offset, timezone, date_fmt, show_offset)
746
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
747
date_str = time.strftime(date_fmt, tt)
748
return date_str + offset_str
751
# Cache of formatted offset strings
755
def format_date_with_offset_in_original_timezone(t, offset=0,
756
_cache=_offset_cache):
757
"""Return a formatted date string in the original timezone.
759
This routine may be faster then format_date.
761
:param t: Seconds since the epoch.
762
:param offset: Timezone offset in seconds east of utc.
766
tt = time.gmtime(t + offset)
767
date_fmt = _default_format_by_weekday_num[tt[6]]
768
date_str = time.strftime(date_fmt, tt)
769
offset_str = _cache.get(offset, None)
770
if offset_str is None:
771
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
772
_cache[offset] = offset_str
773
return date_str + offset_str
776
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
778
"""Return an unicode date string formatted according to the current locale.
780
:param t: Seconds since the epoch.
781
:param offset: Timezone offset in seconds east of utc.
782
:param timezone: How to display the time: 'utc', 'original' for the
783
timezone specified by offset, or 'local' for the process's current
785
:param date_fmt: strftime format.
786
:param show_offset: Whether to append the timezone.
788
(date_fmt, tt, offset_str) = \
789
_format_date(t, offset, timezone, date_fmt, show_offset)
790
date_str = time.strftime(date_fmt, tt)
791
if not isinstance(date_str, unicode):
792
date_str = date_str.decode(get_user_encoding(), 'replace')
793
return date_str + offset_str
796
def _format_date(t, offset, timezone, date_fmt, show_offset):
797
if timezone == 'utc':
800
elif timezone == 'original':
803
tt = time.gmtime(t + offset)
804
elif timezone == 'local':
805
tt = time.localtime(t)
806
offset = local_time_offset(t)
808
raise errors.UnsupportedTimezoneFormat(timezone)
810
date_fmt = "%a %Y-%m-%d %H:%M:%S"
812
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
815
return (date_fmt, tt, offset_str)
818
def compact_date(when):
819
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
822
def format_delta(delta):
823
"""Get a nice looking string for a time delta.
825
:param delta: The time difference in seconds, can be positive or negative.
826
positive indicates time in the past, negative indicates time in the
827
future. (usually time.time() - stored_time)
828
:return: String formatted to show approximate resolution
834
direction = 'in the future'
838
if seconds < 90: # print seconds up to 90 seconds
840
return '%d second %s' % (seconds, direction,)
842
return '%d seconds %s' % (seconds, direction)
844
minutes = int(seconds / 60)
845
seconds -= 60 * minutes
850
if minutes < 90: # print minutes, seconds up to 90 minutes
852
return '%d minute, %d second%s %s' % (
853
minutes, seconds, plural_seconds, direction)
855
return '%d minutes, %d second%s %s' % (
856
minutes, seconds, plural_seconds, direction)
858
hours = int(minutes / 60)
859
minutes -= 60 * hours
866
return '%d hour, %d minute%s %s' % (hours, minutes,
867
plural_minutes, direction)
868
return '%d hours, %d minute%s %s' % (hours, minutes,
869
plural_minutes, direction)
872
"""Return size of given open file."""
873
return os.fstat(f.fileno())[ST_SIZE]
876
# Define rand_bytes based on platform.
878
# Python 2.4 and later have os.urandom,
879
# but it doesn't work on some arches
881
rand_bytes = os.urandom
882
except (NotImplementedError, AttributeError):
883
# If python doesn't have os.urandom, or it doesn't work,
884
# then try to first pull random data from /dev/urandom
886
rand_bytes = file('/dev/urandom', 'rb').read
887
# Otherwise, use this hack as a last resort
888
except (IOError, OSError):
889
# not well seeded, but better than nothing
894
s += chr(random.randint(0, 255))
899
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
901
"""Return a random string of num alphanumeric characters
903
The result only contains lowercase chars because it may be used on
904
case-insensitive filesystems.
907
for raw_byte in rand_bytes(num):
908
s += ALNUM[ord(raw_byte) % 36]
912
## TODO: We could later have path objects that remember their list
913
## decomposition (might be too tricksy though.)
916
"""Turn string into list of parts."""
917
# split on either delimiter because people might use either on
919
ps = re.split(r'[\\/]', p)
924
raise errors.BzrError("sorry, %r not allowed in path" % f)
925
elif (f == '.') or (f == ''):
934
if (f == '..') or (f is None) or (f == ''):
935
raise errors.BzrError("sorry, %r not allowed in path" % f)
939
def parent_directories(filename):
940
"""Return the list of parent directories, deepest first.
942
For example, parent_directories("a/b/c") -> ["a/b", "a"].
945
parts = splitpath(dirname(filename))
947
parents.append(joinpath(parts))
952
_extension_load_failures = []
955
def failed_to_load_extension(exception):
956
"""Handle failing to load a binary extension.
958
This should be called from the ImportError block guarding the attempt to
959
import the native extension. If this function returns, the pure-Python
960
implementation should be loaded instead::
963
>>> import bzrlib._fictional_extension_pyx
964
>>> except ImportError, e:
965
>>> bzrlib.osutils.failed_to_load_extension(e)
966
>>> import bzrlib._fictional_extension_py
968
# NB: This docstring is just an example, not a doctest, because doctest
969
# currently can't cope with the use of lazy imports in this namespace --
972
# This currently doesn't report the failure at the time it occurs, because
973
# they tend to happen very early in startup when we can't check config
974
# files etc, and also we want to report all failures but not spam the user
976
from bzrlib import trace
977
exception_str = str(exception)
978
if exception_str not in _extension_load_failures:
979
trace.mutter("failed to load compiled extension: %s" % exception_str)
980
_extension_load_failures.append(exception_str)
983
def report_extension_load_failures():
984
if not _extension_load_failures:
986
from bzrlib.config import GlobalConfig
987
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
989
# the warnings framework should by default show this only once
990
from bzrlib.trace import warning
992
"bzr: warning: some compiled extensions could not be loaded; "
993
"see <https://answers.launchpad.net/bzr/+faq/703>")
994
# we no longer show the specific missing extensions here, because it makes
995
# the message too long and scary - see
996
# https://bugs.launchpad.net/bzr/+bug/430529
1000
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1001
except ImportError, e:
1002
failed_to_load_extension(e)
1003
from bzrlib._chunks_to_lines_py import chunks_to_lines
1007
"""Split s into lines, but without removing the newline characters."""
1008
# Trivially convert a fulltext into a 'chunked' representation, and let
1009
# chunks_to_lines do the heavy lifting.
1010
if isinstance(s, str):
1011
# chunks_to_lines only supports 8-bit strings
1012
return chunks_to_lines([s])
1014
return _split_lines(s)
1017
def _split_lines(s):
1018
"""Split s into lines, but without removing the newline characters.
1020
This supports Unicode or plain string objects.
1022
lines = s.split('\n')
1023
result = [line + '\n' for line in lines[:-1]]
1025
result.append(lines[-1])
1029
def hardlinks_good():
1030
return sys.platform not in ('win32', 'cygwin', 'darwin')
1033
def link_or_copy(src, dest):
1034
"""Hardlink a file, or copy it if it can't be hardlinked."""
1035
if not hardlinks_good():
1036
shutil.copyfile(src, dest)
1040
except (OSError, IOError), e:
1041
if e.errno != errno.EXDEV:
1043
shutil.copyfile(src, dest)
1046
def delete_any(path):
1047
"""Delete a file, symlink or directory.
1049
Will delete even if readonly.
1052
_delete_file_or_dir(path)
1053
except (OSError, IOError), e:
1054
if e.errno in (errno.EPERM, errno.EACCES):
1055
# make writable and try again
1058
except (OSError, IOError):
1060
_delete_file_or_dir(path)
1065
def _delete_file_or_dir(path):
1066
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1067
# Forgiveness than Permission (EAFP) because:
1068
# - root can damage a solaris file system by using unlink,
1069
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1070
# EACCES, OSX: EPERM) when invoked on a directory.
1071
if isdir(path): # Takes care of symlinks
1078
if getattr(os, 'symlink', None) is not None:
1084
def has_hardlinks():
1085
if getattr(os, 'link', None) is not None:
1091
def host_os_dereferences_symlinks():
1092
return (has_symlinks()
1093
and sys.platform not in ('cygwin', 'win32'))
1096
def readlink(abspath):
1097
"""Return a string representing the path to which the symbolic link points.
1099
:param abspath: The link absolute unicode path.
1101
This his guaranteed to return the symbolic link in unicode in all python
1104
link = abspath.encode(_fs_enc)
1105
target = os.readlink(link)
1106
target = target.decode(_fs_enc)
1110
def contains_whitespace(s):
1111
"""True if there are any whitespace characters in s."""
1112
# string.whitespace can include '\xa0' in certain locales, because it is
1113
# considered "non-breaking-space" as part of ISO-8859-1. But it
1114
# 1) Isn't a breaking whitespace
1115
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1117
# 3) '\xa0' isn't unicode safe since it is >128.
1119
# This should *not* be a unicode set of characters in case the source
1120
# string is not a Unicode string. We can auto-up-cast the characters since
1121
# they are ascii, but we don't want to auto-up-cast the string in case it
1123
for ch in ' \t\n\r\v\f':
1130
def contains_linebreaks(s):
1131
"""True if there is any vertical whitespace in s."""
1139
def relpath(base, path):
1140
"""Return path relative to base, or raise PathNotChild exception.
1142
The path may be either an absolute path or a path relative to the
1143
current working directory.
1145
os.path.commonprefix (python2.4) has a bad bug that it works just
1146
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1147
avoids that problem.
1149
NOTE: `base` should not have a trailing slash otherwise you'll get
1150
PathNotChild exceptions regardless of `path`.
1153
if len(base) < MIN_ABS_PATHLENGTH:
1154
# must have space for e.g. a drive letter
1155
raise ValueError('%r is too short to calculate a relative path'
1163
if len(head) <= len(base) and head != base:
1164
raise errors.PathNotChild(rp, base)
1167
head, tail = split(head)
1172
return pathjoin(*reversed(s))
1177
def _cicp_canonical_relpath(base, path):
1178
"""Return the canonical path relative to base.
1180
Like relpath, but on case-insensitive-case-preserving file-systems, this
1181
will return the relpath as stored on the file-system rather than in the
1182
case specified in the input string, for all existing portions of the path.
1184
This will cause O(N) behaviour if called for every path in a tree; if you
1185
have a number of paths to convert, you should use canonical_relpaths().
1187
# TODO: it should be possible to optimize this for Windows by using the
1188
# win32 API FindFiles function to look for the specified name - but using
1189
# os.listdir() still gives us the correct, platform agnostic semantics in
1192
rel = relpath(base, path)
1193
# '.' will have been turned into ''
1197
abs_base = abspath(base)
1199
_listdir = os.listdir
1201
# use an explicit iterator so we can easily consume the rest on early exit.
1202
bit_iter = iter(rel.split('/'))
1203
for bit in bit_iter:
1206
next_entries = _listdir(current)
1207
except OSError: # enoent, eperm, etc
1208
# We can't find this in the filesystem, so just append the
1210
current = pathjoin(current, bit, *list(bit_iter))
1212
for look in next_entries:
1213
if lbit == look.lower():
1214
current = pathjoin(current, look)
1217
# got to the end, nothing matched, so we just return the
1218
# non-existing bits as they were specified (the filename may be
1219
# the target of a move, for example).
1220
current = pathjoin(current, bit, *list(bit_iter))
1222
return current[len(abs_base):].lstrip('/')
1224
# XXX - TODO - we need better detection/integration of case-insensitive
1225
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1226
# filesystems), for example, so could probably benefit from the same basic
1227
# support there. For now though, only Windows and OSX get that support, and
1228
# they get it for *all* file-systems!
1229
if sys.platform in ('win32', 'darwin'):
1230
canonical_relpath = _cicp_canonical_relpath
1232
canonical_relpath = relpath
1234
def canonical_relpaths(base, paths):
1235
"""Create an iterable to canonicalize a sequence of relative paths.
1237
The intent is for this implementation to use a cache, vastly speeding
1238
up multiple transformations in the same directory.
1240
# but for now, we haven't optimized...
1241
return [canonical_relpath(base, p) for p in paths]
1244
def decode_filename(filename):
1245
"""Decode the filename using the filesystem encoding
1247
If it is unicode, it is returned.
1248
Otherwise it is decoded from the the filesystem's encoding. If decoding
1249
fails, a errors.BadFilenameEncoding exception is raised.
1251
if type(filename) is unicode:
1254
return filename.decode(_fs_enc)
1255
except UnicodeDecodeError:
1256
raise errors.BadFilenameEncoding(filename, _fs_enc)
1259
def safe_unicode(unicode_or_utf8_string):
1260
"""Coerce unicode_or_utf8_string into unicode.
1262
If it is unicode, it is returned.
1263
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1264
wrapped in a BzrBadParameterNotUnicode exception.
1266
if isinstance(unicode_or_utf8_string, unicode):
1267
return unicode_or_utf8_string
1269
return unicode_or_utf8_string.decode('utf8')
1270
except UnicodeDecodeError:
1271
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1274
def safe_utf8(unicode_or_utf8_string):
1275
"""Coerce unicode_or_utf8_string to a utf8 string.
1277
If it is a str, it is returned.
1278
If it is Unicode, it is encoded into a utf-8 string.
1280
if isinstance(unicode_or_utf8_string, str):
1281
# TODO: jam 20070209 This is overkill, and probably has an impact on
1282
# performance if we are dealing with lots of apis that want a
1285
# Make sure it is a valid utf-8 string
1286
unicode_or_utf8_string.decode('utf-8')
1287
except UnicodeDecodeError:
1288
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1289
return unicode_or_utf8_string
1290
return unicode_or_utf8_string.encode('utf-8')
1293
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1294
' Revision id generators should be creating utf8'
1298
def safe_revision_id(unicode_or_utf8_string, warn=True):
1299
"""Revision ids should now be utf8, but at one point they were unicode.
1301
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1303
:param warn: Functions that are sanitizing user data can set warn=False
1304
:return: None or a utf8 revision 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(_revision_id_warning, DeprecationWarning,
1312
return cache_utf8.encode(unicode_or_utf8_string)
1315
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1316
' generators should be creating utf8 file ids.')
1319
def safe_file_id(unicode_or_utf8_string, warn=True):
1320
"""File ids should now be utf8, but at one point they were unicode.
1322
This is the same as safe_utf8, except it uses the cached encode functions
1323
to save a little bit of performance.
1325
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1327
:param warn: Functions that are sanitizing user data can set warn=False
1328
:return: None or a utf8 file id.
1330
if (unicode_or_utf8_string is None
1331
or unicode_or_utf8_string.__class__ == str):
1332
return unicode_or_utf8_string
1334
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1336
return cache_utf8.encode(unicode_or_utf8_string)
1339
_platform_normalizes_filenames = False
1340
if sys.platform == 'darwin':
1341
_platform_normalizes_filenames = True
1344
def normalizes_filenames():
1345
"""Return True if this platform normalizes unicode filenames.
1349
return _platform_normalizes_filenames
1352
def _accessible_normalized_filename(path):
1353
"""Get the unicode normalized path, and if you can access the file.
1355
On platforms where the system normalizes filenames (Mac OSX),
1356
you can access a file by any path which will normalize correctly.
1357
On platforms where the system does not normalize filenames
1358
(everything else), you have to access a file by its exact path.
1360
Internally, bzr only supports NFC normalization, since that is
1361
the standard for XML documents.
1363
So return the normalized path, and a flag indicating if the file
1364
can be accessed by that path.
1367
return unicodedata.normalize('NFC', unicode(path)), True
1370
def _inaccessible_normalized_filename(path):
1371
__doc__ = _accessible_normalized_filename.__doc__
1373
normalized = unicodedata.normalize('NFC', unicode(path))
1374
return normalized, normalized == path
1377
if _platform_normalizes_filenames:
1378
normalized_filename = _accessible_normalized_filename
1380
normalized_filename = _inaccessible_normalized_filename
1383
def set_signal_handler(signum, handler, restart_syscall=True):
1384
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1385
on platforms that support that.
1387
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1388
automatically restart (by calling `signal.siginterrupt(signum,
1389
False)`). May be ignored if the feature is not available on this
1390
platform or Python version.
1394
siginterrupt = signal.siginterrupt
1396
# This python implementation doesn't provide signal support, hence no
1399
except AttributeError:
1400
# siginterrupt doesn't exist on this platform, or for this version
1402
siginterrupt = lambda signum, flag: None
1404
def sig_handler(*args):
1405
# Python resets the siginterrupt flag when a signal is
1406
# received. <http://bugs.python.org/issue8354>
1407
# As a workaround for some cases, set it back the way we want it.
1408
siginterrupt(signum, False)
1409
# Now run the handler function passed to set_signal_handler.
1412
sig_handler = handler
1413
old_handler = signal.signal(signum, sig_handler)
1415
siginterrupt(signum, False)
1419
default_terminal_width = 80
1420
"""The default terminal width for ttys.
1422
This is defined so that higher levels can share a common fallback value when
1423
terminal_width() returns None.
1426
# Keep some state so that terminal_width can detect if _terminal_size has
1427
# returned a different size since the process started. See docstring and
1428
# comments of terminal_width for details.
1429
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1430
_terminal_size_state = 'no_data'
1431
_first_terminal_size = None
1433
def terminal_width():
1434
"""Return terminal width.
1436
None is returned if the width can't established precisely.
1439
- if BZR_COLUMNS is set, returns its value
1440
- if there is no controlling terminal, returns None
1441
- query the OS, if the queried size has changed since the last query,
1443
- if COLUMNS is set, returns its value,
1444
- if the OS has a value (even though it's never changed), return its value.
1446
From there, we need to query the OS to get the size of the controlling
1449
On Unices we query the OS by:
1450
- get termios.TIOCGWINSZ
1451
- if an error occurs or a negative value is obtained, returns None
1453
On Windows we query the OS by:
1454
- win32utils.get_console_size() decides,
1455
- returns None on error (provided default value)
1457
# Note to implementors: if changing the rules for determining the width,
1458
# make sure you've considered the behaviour in these cases:
1459
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1460
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1462
# - (add more interesting cases here, if you find any)
1463
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1464
# but we don't want to register a signal handler because it is impossible
1465
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1466
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1467
# time so we can notice if the reported size has changed, which should have
1470
# If BZR_COLUMNS is set, take it, user is always right
1472
return int(os.environ['BZR_COLUMNS'])
1473
except (KeyError, ValueError):
1476
isatty = getattr(sys.stdout, 'isatty', None)
1477
if isatty is None or not isatty():
1478
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1482
width, height = os_size = _terminal_size(None, None)
1483
global _first_terminal_size, _terminal_size_state
1484
if _terminal_size_state == 'no_data':
1485
_first_terminal_size = os_size
1486
_terminal_size_state = 'unchanged'
1487
elif (_terminal_size_state == 'unchanged' and
1488
_first_terminal_size != os_size):
1489
_terminal_size_state = 'changed'
1491
# If the OS claims to know how wide the terminal is, and this value has
1492
# ever changed, use that.
1493
if _terminal_size_state == 'changed':
1494
if width is not None and width > 0:
1497
# If COLUMNS is set, use it.
1499
return int(os.environ['COLUMNS'])
1500
except (KeyError, ValueError):
1503
# Finally, use an unchanged size from the OS, if we have one.
1504
if _terminal_size_state == 'unchanged':
1505
if width is not None and width > 0:
1508
# The width could not be determined.
1512
def _win32_terminal_size(width, height):
1513
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1514
return width, height
1517
def _ioctl_terminal_size(width, height):
1519
import struct, fcntl, termios
1520
s = struct.pack('HHHH', 0, 0, 0, 0)
1521
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1522
height, width = struct.unpack('HHHH', x)[0:2]
1523
except (IOError, AttributeError):
1525
return width, height
1527
_terminal_size = None
1528
"""Returns the terminal size as (width, height).
1530
:param width: Default value for width.
1531
:param height: Default value for height.
1533
This is defined specifically for each OS and query the size of the controlling
1534
terminal. If any error occurs, the provided default values should be returned.
1536
if sys.platform == 'win32':
1537
_terminal_size = _win32_terminal_size
1539
_terminal_size = _ioctl_terminal_size
1542
def supports_executable():
1543
return sys.platform != "win32"
1546
def supports_posix_readonly():
1547
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1549
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1550
directory controls creation/deletion, etc.
1552
And under win32, readonly means that the directory itself cannot be
1553
deleted. The contents of a readonly directory can be changed, unlike POSIX
1554
where files in readonly directories cannot be added, deleted or renamed.
1556
return sys.platform != "win32"
1559
def set_or_unset_env(env_variable, value):
1560
"""Modify the environment, setting or removing the env_variable.
1562
:param env_variable: The environment variable in question
1563
:param value: The value to set the environment to. If None, then
1564
the variable will be removed.
1565
:return: The original value of the environment variable.
1567
orig_val = os.environ.get(env_variable)
1569
if orig_val is not None:
1570
del os.environ[env_variable]
1572
if isinstance(value, unicode):
1573
value = value.encode(get_user_encoding())
1574
os.environ[env_variable] = value
1578
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1581
def check_legal_path(path):
1582
"""Check whether the supplied path is legal.
1583
This is only required on Windows, so we don't test on other platforms
1586
if sys.platform != "win32":
1588
if _validWin32PathRE.match(path) is None:
1589
raise errors.IllegalPath(path)
1592
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1594
def _is_error_enotdir(e):
1595
"""Check if this exception represents ENOTDIR.
1597
Unfortunately, python is very inconsistent about the exception
1598
here. The cases are:
1599
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1600
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1601
which is the windows error code.
1602
3) Windows, Python2.5 uses errno == EINVAL and
1603
winerror == ERROR_DIRECTORY
1605
:param e: An Exception object (expected to be OSError with an errno
1606
attribute, but we should be able to cope with anything)
1607
:return: True if this represents an ENOTDIR error. False otherwise.
1609
en = getattr(e, 'errno', None)
1610
if (en == errno.ENOTDIR
1611
or (sys.platform == 'win32'
1612
and (en == _WIN32_ERROR_DIRECTORY
1613
or (en == errno.EINVAL
1614
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1620
def walkdirs(top, prefix=""):
1621
"""Yield data about all the directories in a tree.
1623
This yields all the data about the contents of a directory at a time.
1624
After each directory has been yielded, if the caller has mutated the list
1625
to exclude some directories, they are then not descended into.
1627
The data yielded is of the form:
1628
((directory-relpath, directory-path-from-top),
1629
[(relpath, basename, kind, lstat, path-from-top), ...]),
1630
- directory-relpath is the relative path of the directory being returned
1631
with respect to top. prefix is prepended to this.
1632
- directory-path-from-root is the path including top for this directory.
1633
It is suitable for use with os functions.
1634
- relpath is the relative path within the subtree being walked.
1635
- basename is the basename of the path
1636
- kind is the kind of the file now. If unknown then the file is not
1637
present within the tree - but it may be recorded as versioned. See
1639
- lstat is the stat data *if* the file was statted.
1640
- planned, not implemented:
1641
path_from_tree_root is the path from the root of the tree.
1643
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1644
allows one to walk a subtree but get paths that are relative to a tree
1646
:return: an iterator over the dirs.
1648
#TODO there is a bit of a smell where the results of the directory-
1649
# summary in this, and the path from the root, may not agree
1650
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1651
# potentially confusing output. We should make this more robust - but
1652
# not at a speed cost. RBC 20060731
1654
_directory = _directory_kind
1655
_listdir = os.listdir
1656
_kind_from_mode = file_kind_from_stat_mode
1657
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1659
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1660
relroot, _, _, _, top = pending.pop()
1662
relprefix = relroot + u'/'
1665
top_slash = top + u'/'
1668
append = dirblock.append
1670
names = sorted(map(decode_filename, _listdir(top)))
1672
if not _is_error_enotdir(e):
1676
abspath = top_slash + name
1677
statvalue = _lstat(abspath)
1678
kind = _kind_from_mode(statvalue.st_mode)
1679
append((relprefix + name, name, kind, statvalue, abspath))
1680
yield (relroot, top), dirblock
1682
# push the user specified dirs from dirblock
1683
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1686
class DirReader(object):
1687
"""An interface for reading directories."""
1689
def top_prefix_to_starting_dir(self, top, prefix=""):
1690
"""Converts top and prefix to a starting dir entry
1692
:param top: A utf8 path
1693
:param prefix: An optional utf8 path to prefix output relative paths
1695
:return: A tuple starting with prefix, and ending with the native
1698
raise NotImplementedError(self.top_prefix_to_starting_dir)
1700
def read_dir(self, prefix, top):
1701
"""Read a specific dir.
1703
:param prefix: A utf8 prefix to be preprended to the path basenames.
1704
:param top: A natively encoded path to read.
1705
:return: A list of the directories contents. Each item contains:
1706
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1708
raise NotImplementedError(self.read_dir)
1711
_selected_dir_reader = None
1714
def _walkdirs_utf8(top, prefix=""):
1715
"""Yield data about all the directories in a tree.
1717
This yields the same information as walkdirs() only each entry is yielded
1718
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1719
are returned as exact byte-strings.
1721
:return: yields a tuple of (dir_info, [file_info])
1722
dir_info is (utf8_relpath, path-from-top)
1723
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1724
if top is an absolute path, path-from-top is also an absolute path.
1725
path-from-top might be unicode or utf8, but it is the correct path to
1726
pass to os functions to affect the file in question. (such as os.lstat)
1728
global _selected_dir_reader
1729
if _selected_dir_reader is None:
1730
fs_encoding = _fs_enc.upper()
1731
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1732
# Win98 doesn't have unicode apis like FindFirstFileW
1733
# TODO: We possibly could support Win98 by falling back to the
1734
# original FindFirstFile, and using TCHAR instead of WCHAR,
1735
# but that gets a bit tricky, and requires custom compiling
1738
from bzrlib._walkdirs_win32 import Win32ReadDir
1739
_selected_dir_reader = Win32ReadDir()
1742
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1743
# ANSI_X3.4-1968 is a form of ASCII
1745
from bzrlib._readdir_pyx import UTF8DirReader
1746
_selected_dir_reader = UTF8DirReader()
1747
except ImportError, e:
1748
failed_to_load_extension(e)
1751
if _selected_dir_reader is None:
1752
# Fallback to the python version
1753
_selected_dir_reader = UnicodeDirReader()
1755
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1756
# But we don't actually uses 1-3 in pending, so set them to None
1757
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1758
read_dir = _selected_dir_reader.read_dir
1759
_directory = _directory_kind
1761
relroot, _, _, _, top = pending[-1].pop()
1764
dirblock = sorted(read_dir(relroot, top))
1765
yield (relroot, top), dirblock
1766
# push the user specified dirs from dirblock
1767
next = [d for d in reversed(dirblock) if d[2] == _directory]
1769
pending.append(next)
1772
class UnicodeDirReader(DirReader):
1773
"""A dir reader for non-utf8 file systems, which transcodes."""
1775
__slots__ = ['_utf8_encode']
1778
self._utf8_encode = codecs.getencoder('utf8')
1780
def top_prefix_to_starting_dir(self, top, prefix=""):
1781
"""See DirReader.top_prefix_to_starting_dir."""
1782
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1784
def read_dir(self, prefix, top):
1785
"""Read a single directory from a non-utf8 file system.
1787
top, and the abspath element in the output are unicode, all other paths
1788
are utf8. Local disk IO is done via unicode calls to listdir etc.
1790
This is currently the fallback code path when the filesystem encoding is
1791
not UTF-8. It may be better to implement an alternative so that we can
1792
safely handle paths that are not properly decodable in the current
1795
See DirReader.read_dir for details.
1797
_utf8_encode = self._utf8_encode
1799
_listdir = os.listdir
1800
_kind_from_mode = file_kind_from_stat_mode
1803
relprefix = prefix + '/'
1806
top_slash = top + u'/'
1809
append = dirblock.append
1810
for name in sorted(_listdir(top)):
1812
name_utf8 = _utf8_encode(name)[0]
1813
except UnicodeDecodeError:
1814
raise errors.BadFilenameEncoding(
1815
_utf8_encode(relprefix)[0] + name, _fs_enc)
1816
abspath = top_slash + name
1817
statvalue = _lstat(abspath)
1818
kind = _kind_from_mode(statvalue.st_mode)
1819
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1823
def copy_tree(from_path, to_path, handlers={}):
1824
"""Copy all of the entries in from_path into to_path.
1826
:param from_path: The base directory to copy.
1827
:param to_path: The target directory. If it does not exist, it will
1829
:param handlers: A dictionary of functions, which takes a source and
1830
destinations for files, directories, etc.
1831
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1832
'file', 'directory', and 'symlink' should always exist.
1833
If they are missing, they will be replaced with 'os.mkdir()',
1834
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1836
# Now, just copy the existing cached tree to the new location
1837
# We use a cheap trick here.
1838
# Absolute paths are prefixed with the first parameter
1839
# relative paths are prefixed with the second.
1840
# So we can get both the source and target returned
1841
# without any extra work.
1843
def copy_dir(source, dest):
1846
def copy_link(source, dest):
1847
"""Copy the contents of a symlink"""
1848
link_to = os.readlink(source)
1849
os.symlink(link_to, dest)
1851
real_handlers = {'file':shutil.copy2,
1852
'symlink':copy_link,
1853
'directory':copy_dir,
1855
real_handlers.update(handlers)
1857
if not os.path.exists(to_path):
1858
real_handlers['directory'](from_path, to_path)
1860
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1861
for relpath, name, kind, st, abspath in entries:
1862
real_handlers[kind](abspath, relpath)
1865
def copy_ownership_from_path(dst, src=None):
1866
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1868
If src is None, the containing directory is used as source. If chown
1869
fails, the error is ignored and a warning is printed.
1871
chown = getattr(os, 'chown', None)
1876
src = os.path.dirname(dst)
1882
chown(dst, s.st_uid, s.st_gid)
1884
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1887
def path_prefix_key(path):
1888
"""Generate a prefix-order path key for path.
1890
This can be used to sort paths in the same way that walkdirs does.
1892
return (dirname(path) , path)
1895
def compare_paths_prefix_order(path_a, path_b):
1896
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1897
key_a = path_prefix_key(path_a)
1898
key_b = path_prefix_key(path_b)
1899
return cmp(key_a, key_b)
1902
_cached_user_encoding = None
1905
def get_user_encoding(use_cache=True):
1906
"""Find out what the preferred user encoding is.
1908
This is generally the encoding that is used for command line parameters
1909
and file contents. This may be different from the terminal encoding
1910
or the filesystem encoding.
1912
:param use_cache: Enable cache for detected encoding.
1913
(This parameter is turned on by default,
1914
and required only for selftesting)
1916
:return: A string defining the preferred user encoding
1918
global _cached_user_encoding
1919
if _cached_user_encoding is not None and use_cache:
1920
return _cached_user_encoding
1922
if sys.platform == 'darwin':
1923
# python locale.getpreferredencoding() always return
1924
# 'mac-roman' on darwin. That's a lie.
1925
sys.platform = 'posix'
1927
if os.environ.get('LANG', None) is None:
1928
# If LANG is not set, we end up with 'ascii', which is bad
1929
# ('mac-roman' is more than ascii), so we set a default which
1930
# will give us UTF-8 (which appears to work in all cases on
1931
# OSX). Users are still free to override LANG of course, as
1932
# long as it give us something meaningful. This work-around
1933
# *may* not be needed with python 3k and/or OSX 10.5, but will
1934
# work with them too -- vila 20080908
1935
os.environ['LANG'] = 'en_US.UTF-8'
1938
sys.platform = 'darwin'
1943
user_encoding = locale.getpreferredencoding()
1944
except locale.Error, e:
1945
sys.stderr.write('bzr: warning: %s\n'
1946
' Could not determine what text encoding to use.\n'
1947
' This error usually means your Python interpreter\n'
1948
' doesn\'t support the locale set by $LANG (%s)\n'
1949
" Continuing with ascii encoding.\n"
1950
% (e, os.environ.get('LANG')))
1951
user_encoding = 'ascii'
1953
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1954
# treat that as ASCII, and not support printing unicode characters to the
1957
# For python scripts run under vim, we get '', so also treat that as ASCII
1958
if user_encoding in (None, 'cp0', ''):
1959
user_encoding = 'ascii'
1963
codecs.lookup(user_encoding)
1965
sys.stderr.write('bzr: warning:'
1966
' unknown encoding %s.'
1967
' Continuing with ascii encoding.\n'
1970
user_encoding = 'ascii'
1973
_cached_user_encoding = user_encoding
1975
return user_encoding
1978
def get_diff_header_encoding():
1979
return get_terminal_encoding()
1982
def get_host_name():
1983
"""Return the current unicode host name.
1985
This is meant to be used in place of socket.gethostname() because that
1986
behaves inconsistently on different platforms.
1988
if sys.platform == "win32":
1990
return win32utils.get_host_name()
1993
return socket.gethostname().decode(get_user_encoding())
1996
# We must not read/write any more than 64k at a time from/to a socket so we
1997
# don't risk "no buffer space available" errors on some platforms. Windows in
1998
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
2000
MAX_SOCKET_CHUNK = 64 * 1024
2002
def read_bytes_from_socket(sock, report_activity=None,
2003
max_read_size=MAX_SOCKET_CHUNK):
2004
"""Read up to max_read_size of bytes from sock and notify of progress.
2006
Translates "Connection reset by peer" into file-like EOF (return an
2007
empty string rather than raise an error), and repeats the recv if
2008
interrupted by a signal.
2012
bytes = sock.recv(max_read_size)
2013
except socket.error, e:
2015
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
2016
# The connection was closed by the other side. Callers expect
2017
# an empty string to signal end-of-stream.
2019
elif eno == errno.EINTR:
2020
# Retry the interrupted recv.
2024
if report_activity is not None:
2025
report_activity(len(bytes), 'read')
2029
def recv_all(socket, count):
2030
"""Receive an exact number of bytes.
2032
Regular Socket.recv() may return less than the requested number of bytes,
2033
depending on what's in the OS buffer. MSG_WAITALL is not available
2034
on all platforms, but this should work everywhere. This will return
2035
less than the requested amount if the remote end closes.
2037
This isn't optimized and is intended mostly for use in testing.
2040
while len(b) < count:
2041
new = read_bytes_from_socket(socket, None, count - len(b))
2048
def send_all(sock, bytes, report_activity=None):
2049
"""Send all bytes on a socket.
2051
Breaks large blocks in smaller chunks to avoid buffering limitations on
2052
some platforms, and catches EINTR which may be thrown if the send is
2053
interrupted by a signal.
2055
This is preferred to socket.sendall(), because it avoids portability bugs
2056
and provides activity reporting.
2058
:param report_activity: Call this as bytes are read, see
2059
Transport._report_activity
2062
byte_count = len(bytes)
2063
while sent_total < byte_count:
2065
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2066
except socket.error, e:
2067
if e.args[0] != errno.EINTR:
2071
report_activity(sent, 'write')
2074
def dereference_path(path):
2075
"""Determine the real path to a file.
2077
All parent elements are dereferenced. But the file itself is not
2079
:param path: The original path. May be absolute or relative.
2080
:return: the real path *to* the file
2082
parent, base = os.path.split(path)
2083
# The pathjoin for '.' is a workaround for Python bug #1213894.
2084
# (initial path components aren't dereferenced)
2085
return pathjoin(realpath(pathjoin('.', parent)), base)
2088
def supports_mapi():
2089
"""Return True if we can use MAPI to launch a mail client."""
2090
return sys.platform == "win32"
2093
def resource_string(package, resource_name):
2094
"""Load a resource from a package and return it as a string.
2096
Note: Only packages that start with bzrlib are currently supported.
2098
This is designed to be a lightweight implementation of resource
2099
loading in a way which is API compatible with the same API from
2101
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2102
If and when pkg_resources becomes a standard library, this routine
2105
# Check package name is within bzrlib
2106
if package == "bzrlib":
2107
resource_relpath = resource_name
2108
elif package.startswith("bzrlib."):
2109
package = package[len("bzrlib."):].replace('.', os.sep)
2110
resource_relpath = pathjoin(package, resource_name)
2112
raise errors.BzrError('resource package %s not in bzrlib' % package)
2114
# Map the resource to a file and read its contents
2115
base = dirname(bzrlib.__file__)
2116
if getattr(sys, 'frozen', None): # bzr.exe
2117
base = abspath(pathjoin(base, '..', '..'))
2118
f = file(pathjoin(base, resource_relpath), "rU")
2124
def file_kind_from_stat_mode_thunk(mode):
2125
global file_kind_from_stat_mode
2126
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2128
from bzrlib._readdir_pyx import UTF8DirReader
2129
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2130
except ImportError, e:
2131
# This is one time where we won't warn that an extension failed to
2132
# load. The extension is never available on Windows anyway.
2133
from bzrlib._readdir_py import (
2134
_kind_from_mode as file_kind_from_stat_mode
2136
return file_kind_from_stat_mode(mode)
2137
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2140
def file_kind(f, _lstat=os.lstat):
2142
return file_kind_from_stat_mode(_lstat(f).st_mode)
2144
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2145
raise errors.NoSuchFile(f)
2149
def until_no_eintr(f, *a, **kw):
2150
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2152
WARNING: you must be certain that it is safe to retry the call repeatedly
2153
if EINTR does occur. This is typically only true for low-level operations
2154
like os.read. If in any doubt, don't use this.
2156
Keep in mind that this is not a complete solution to EINTR. There is
2157
probably code in the Python standard library and other dependencies that
2158
may encounter EINTR if a signal arrives (and there is signal handler for
2159
that signal). So this function can reduce the impact for IO that bzrlib
2160
directly controls, but it is not a complete solution.
2162
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2166
except (IOError, OSError), e:
2167
if e.errno == errno.EINTR:
2172
def re_compile_checked(re_string, flags=0, where=""):
2173
"""Return a compiled re, or raise a sensible error.
2175
This should only be used when compiling user-supplied REs.
2177
:param re_string: Text form of regular expression.
2178
:param flags: eg re.IGNORECASE
2179
:param where: Message explaining to the user the context where
2180
it occurred, eg 'log search filter'.
2182
# from https://bugs.launchpad.net/bzr/+bug/251352
2184
re_obj = re.compile(re_string, flags)
2189
where = ' in ' + where
2190
# despite the name 'error' is a type
2191
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2192
% (where, re_string, e))
2195
if sys.platform == "win32":
2198
return msvcrt.getch()
2203
fd = sys.stdin.fileno()
2204
settings = termios.tcgetattr(fd)
2207
ch = sys.stdin.read(1)
2209
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2213
if sys.platform == 'linux2':
2214
def _local_concurrency():
2216
prefix = 'processor'
2217
for line in file('/proc/cpuinfo', 'rb'):
2218
if line.startswith(prefix):
2219
concurrency = int(line[line.find(':')+1:]) + 1
2221
elif sys.platform == 'darwin':
2222
def _local_concurrency():
2223
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2224
stdout=subprocess.PIPE).communicate()[0]
2225
elif sys.platform[0:7] == 'freebsd':
2226
def _local_concurrency():
2227
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2228
stdout=subprocess.PIPE).communicate()[0]
2229
elif sys.platform == 'sunos5':
2230
def _local_concurrency():
2231
return subprocess.Popen(['psrinfo', '-p',],
2232
stdout=subprocess.PIPE).communicate()[0]
2233
elif sys.platform == "win32":
2234
def _local_concurrency():
2235
# This appears to return the number of cores.
2236
return os.environ.get('NUMBER_OF_PROCESSORS')
2238
def _local_concurrency():
2243
_cached_local_concurrency = None
2245
def local_concurrency(use_cache=True):
2246
"""Return how many processes can be run concurrently.
2248
Rely on platform specific implementations and default to 1 (one) if
2249
anything goes wrong.
2251
global _cached_local_concurrency
2253
if _cached_local_concurrency is not None and use_cache:
2254
return _cached_local_concurrency
2256
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2257
if concurrency is None:
2259
concurrency = _local_concurrency()
2260
except (OSError, IOError):
2263
concurrency = int(concurrency)
2264
except (TypeError, ValueError):
2267
_cached_concurrency = concurrency
2271
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2272
"""A stream writer that doesn't decode str arguments."""
2274
def __init__(self, encode, stream, errors='strict'):
2275
codecs.StreamWriter.__init__(self, stream, errors)
2276
self.encode = encode
2278
def write(self, object):
2279
if type(object) is str:
2280
self.stream.write(object)
2282
data, _ = self.encode(object, self.errors)
2283
self.stream.write(data)
2285
if sys.platform == 'win32':
2286
def open_file(filename, mode='r', bufsize=-1):
2287
"""This function is used to override the ``open`` builtin.
2289
But it uses O_NOINHERIT flag so the file handle is not inherited by
2290
child processes. Deleting or renaming a closed file opened with this
2291
function is not blocking child processes.
2293
writing = 'w' in mode
2294
appending = 'a' in mode
2295
updating = '+' in mode
2296
binary = 'b' in mode
2299
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2300
# for flags for each modes.
2310
flags |= os.O_WRONLY
2311
flags |= os.O_CREAT | os.O_TRUNC
2316
flags |= os.O_WRONLY
2317
flags |= os.O_CREAT | os.O_APPEND
2322
flags |= os.O_RDONLY
2324
return os.fdopen(os.open(filename, flags), mode, bufsize)
2329
def getuser_unicode():
2330
"""Return the username as unicode.
2333
user_encoding = get_user_encoding()
2334
username = getpass.getuser().decode(user_encoding)
2335
except UnicodeDecodeError:
2336
raise errors.BzrError("Can't decode username as %s." % \
added
removed
bzrlib/osutils.py
