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():
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
from bzrlib.trace import mutter
458
output_encoding = getattr(sys.stdout, 'encoding', None)
459
if not output_encoding:
460
input_encoding = getattr(sys.stdin, 'encoding', None)
461
if not input_encoding:
462
output_encoding = get_user_encoding()
463
mutter('encoding stdout as osutils.get_user_encoding() %r',
466
output_encoding = input_encoding
467
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
469
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
470
if output_encoding == 'cp0':
471
# invalid encoding (cp0 means 'no codepage' on Windows)
472
output_encoding = get_user_encoding()
473
mutter('cp0 is invalid encoding.'
474
' encoding stdout as osutils.get_user_encoding() %r',
478
codecs.lookup(output_encoding)
480
sys.stderr.write('bzr: warning:'
481
' unknown terminal encoding %s.\n'
482
' Using encoding %s instead.\n'
483
% (output_encoding, get_user_encoding())
485
output_encoding = get_user_encoding()
487
return output_encoding
490
def normalizepath(f):
491
if getattr(os.path, 'realpath', None) is not None:
495
[p,e] = os.path.split(f)
496
if e == "" or e == "." or e == "..":
499
return pathjoin(F(p), e)
503
"""True if f is an accessible directory."""
505
return S_ISDIR(os.lstat(f)[ST_MODE])
511
"""True if f is a regular file."""
513
return S_ISREG(os.lstat(f)[ST_MODE])
518
"""True if f is a symlink."""
520
return S_ISLNK(os.lstat(f)[ST_MODE])
524
def is_inside(dir, fname):
525
"""True if fname is inside dir.
527
The parameters should typically be passed to osutils.normpath first, so
528
that . and .. and repeated slashes are eliminated, and the separators
529
are canonical for the platform.
531
The empty string as a dir name is taken as top-of-tree and matches
534
# XXX: Most callers of this can actually do something smarter by
535
# looking at the inventory
545
return fname.startswith(dir)
548
def is_inside_any(dir_list, fname):
549
"""True if fname is inside any of given dirs."""
550
for dirname in dir_list:
551
if is_inside(dirname, fname):
556
def is_inside_or_parent_of_any(dir_list, fname):
557
"""True if fname is a child or a parent of any of the given files."""
558
for dirname in dir_list:
559
if is_inside(dirname, fname) or is_inside(fname, dirname):
564
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
565
report_activity=None, direction='read'):
566
"""Copy contents of one file to another.
568
The read_length can either be -1 to read to end-of-file (EOF) or
569
it can specify the maximum number of bytes to read.
571
The buff_size represents the maximum size for each read operation
572
performed on from_file.
574
:param report_activity: Call this as bytes are read, see
575
Transport._report_activity
576
:param direction: Will be passed to report_activity
578
:return: The number of bytes copied.
582
# read specified number of bytes
584
while read_length > 0:
585
num_bytes_to_read = min(read_length, buff_size)
587
block = from_file.read(num_bytes_to_read)
591
if report_activity is not None:
592
report_activity(len(block), direction)
595
actual_bytes_read = len(block)
596
read_length -= actual_bytes_read
597
length += actual_bytes_read
601
block = from_file.read(buff_size)
605
if report_activity is not None:
606
report_activity(len(block), direction)
612
def pump_string_file(bytes, file_handle, segment_size=None):
613
"""Write bytes to file_handle in many smaller writes.
615
:param bytes: The string to write.
616
:param file_handle: The file to write to.
618
# Write data in chunks rather than all at once, because very large
619
# writes fail on some platforms (e.g. Windows with SMB mounted
622
segment_size = 5242880 # 5MB
623
segments = range(len(bytes) / segment_size + 1)
624
write = file_handle.write
625
for segment_index in segments:
626
segment = buffer(bytes, segment_index * segment_size, segment_size)
630
def file_iterator(input_file, readsize=32768):
632
b = input_file.read(readsize)
639
"""Calculate the hexdigest of an open file.
641
The file cursor should be already at the start.
653
def size_sha_file(f):
654
"""Calculate the size and hexdigest of an open file.
656
The file cursor should be already at the start and
657
the caller is responsible for closing the file afterwards.
668
return size, s.hexdigest()
671
def sha_file_by_name(fname):
672
"""Calculate the SHA1 of a file by reading the full text"""
674
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
677
b = os.read(f, 1<<16)
685
def sha_strings(strings, _factory=sha):
686
"""Return the sha-1 of concatenation of strings"""
688
map(s.update, strings)
692
def sha_string(f, _factory=sha):
693
return _factory(f).hexdigest()
696
def fingerprint_file(f):
698
return {'size': len(b),
699
'sha1': sha(b).hexdigest()}
702
def compare_files(a, b):
703
"""Returns true if equal in contents"""
714
def local_time_offset(t=None):
715
"""Return offset of local zone from GMT, either at present or at time t."""
718
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
719
return offset.days * 86400 + offset.seconds
721
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
722
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
725
def format_date(t, offset=0, timezone='original', date_fmt=None,
727
"""Return a formatted date string.
729
:param t: Seconds since the epoch.
730
:param offset: Timezone offset in seconds east of utc.
731
:param timezone: How to display the time: 'utc', 'original' for the
732
timezone specified by offset, or 'local' for the process's current
734
:param date_fmt: strftime format.
735
:param show_offset: Whether to append the timezone.
737
(date_fmt, tt, offset_str) = \
738
_format_date(t, offset, timezone, date_fmt, show_offset)
739
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
740
date_str = time.strftime(date_fmt, tt)
741
return date_str + offset_str
744
# Cache of formatted offset strings
748
def format_date_with_offset_in_original_timezone(t, offset=0,
749
_cache=_offset_cache):
750
"""Return a formatted date string in the original timezone.
752
This routine may be faster then format_date.
754
:param t: Seconds since the epoch.
755
:param offset: Timezone offset in seconds east of utc.
759
tt = time.gmtime(t + offset)
760
date_fmt = _default_format_by_weekday_num[tt[6]]
761
date_str = time.strftime(date_fmt, tt)
762
offset_str = _cache.get(offset, None)
763
if offset_str is None:
764
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
765
_cache[offset] = offset_str
766
return date_str + offset_str
769
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
771
"""Return an unicode date string formatted according to the current locale.
773
:param t: Seconds since the epoch.
774
:param offset: Timezone offset in seconds east of utc.
775
:param timezone: How to display the time: 'utc', 'original' for the
776
timezone specified by offset, or 'local' for the process's current
778
:param date_fmt: strftime format.
779
:param show_offset: Whether to append the timezone.
781
(date_fmt, tt, offset_str) = \
782
_format_date(t, offset, timezone, date_fmt, show_offset)
783
date_str = time.strftime(date_fmt, tt)
784
if not isinstance(date_str, unicode):
785
date_str = date_str.decode(get_user_encoding(), 'replace')
786
return date_str + offset_str
789
def _format_date(t, offset, timezone, date_fmt, show_offset):
790
if timezone == 'utc':
793
elif timezone == 'original':
796
tt = time.gmtime(t + offset)
797
elif timezone == 'local':
798
tt = time.localtime(t)
799
offset = local_time_offset(t)
801
raise errors.UnsupportedTimezoneFormat(timezone)
803
date_fmt = "%a %Y-%m-%d %H:%M:%S"
805
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
808
return (date_fmt, tt, offset_str)
811
def compact_date(when):
812
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
815
def format_delta(delta):
816
"""Get a nice looking string for a time delta.
818
:param delta: The time difference in seconds, can be positive or negative.
819
positive indicates time in the past, negative indicates time in the
820
future. (usually time.time() - stored_time)
821
:return: String formatted to show approximate resolution
827
direction = 'in the future'
831
if seconds < 90: # print seconds up to 90 seconds
833
return '%d second %s' % (seconds, direction,)
835
return '%d seconds %s' % (seconds, direction)
837
minutes = int(seconds / 60)
838
seconds -= 60 * minutes
843
if minutes < 90: # print minutes, seconds up to 90 minutes
845
return '%d minute, %d second%s %s' % (
846
minutes, seconds, plural_seconds, direction)
848
return '%d minutes, %d second%s %s' % (
849
minutes, seconds, plural_seconds, direction)
851
hours = int(minutes / 60)
852
minutes -= 60 * hours
859
return '%d hour, %d minute%s %s' % (hours, minutes,
860
plural_minutes, direction)
861
return '%d hours, %d minute%s %s' % (hours, minutes,
862
plural_minutes, direction)
865
"""Return size of given open file."""
866
return os.fstat(f.fileno())[ST_SIZE]
869
# Define rand_bytes based on platform.
871
# Python 2.4 and later have os.urandom,
872
# but it doesn't work on some arches
874
rand_bytes = os.urandom
875
except (NotImplementedError, AttributeError):
876
# If python doesn't have os.urandom, or it doesn't work,
877
# then try to first pull random data from /dev/urandom
879
rand_bytes = file('/dev/urandom', 'rb').read
880
# Otherwise, use this hack as a last resort
881
except (IOError, OSError):
882
# not well seeded, but better than nothing
887
s += chr(random.randint(0, 255))
892
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
894
"""Return a random string of num alphanumeric characters
896
The result only contains lowercase chars because it may be used on
897
case-insensitive filesystems.
900
for raw_byte in rand_bytes(num):
901
s += ALNUM[ord(raw_byte) % 36]
905
## TODO: We could later have path objects that remember their list
906
## decomposition (might be too tricksy though.)
909
"""Turn string into list of parts."""
910
# split on either delimiter because people might use either on
912
ps = re.split(r'[\\/]', p)
917
raise errors.BzrError("sorry, %r not allowed in path" % f)
918
elif (f == '.') or (f == ''):
927
if (f == '..') or (f is None) or (f == ''):
928
raise errors.BzrError("sorry, %r not allowed in path" % f)
932
def parent_directories(filename):
933
"""Return the list of parent directories, deepest first.
935
For example, parent_directories("a/b/c") -> ["a/b", "a"].
938
parts = splitpath(dirname(filename))
940
parents.append(joinpath(parts))
945
_extension_load_failures = []
948
def failed_to_load_extension(exception):
949
"""Handle failing to load a binary extension.
951
This should be called from the ImportError block guarding the attempt to
952
import the native extension. If this function returns, the pure-Python
953
implementation should be loaded instead::
956
>>> import bzrlib._fictional_extension_pyx
957
>>> except ImportError, e:
958
>>> bzrlib.osutils.failed_to_load_extension(e)
959
>>> import bzrlib._fictional_extension_py
961
# NB: This docstring is just an example, not a doctest, because doctest
962
# currently can't cope with the use of lazy imports in this namespace --
965
# This currently doesn't report the failure at the time it occurs, because
966
# they tend to happen very early in startup when we can't check config
967
# files etc, and also we want to report all failures but not spam the user
969
from bzrlib import trace
970
exception_str = str(exception)
971
if exception_str not in _extension_load_failures:
972
trace.mutter("failed to load compiled extension: %s" % exception_str)
973
_extension_load_failures.append(exception_str)
976
def report_extension_load_failures():
977
if not _extension_load_failures:
979
from bzrlib.config import GlobalConfig
980
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
982
# the warnings framework should by default show this only once
983
from bzrlib.trace import warning
985
"bzr: warning: some compiled extensions could not be loaded; "
986
"see <https://answers.launchpad.net/bzr/+faq/703>")
987
# we no longer show the specific missing extensions here, because it makes
988
# the message too long and scary - see
989
# https://bugs.launchpad.net/bzr/+bug/430529
993
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
994
except ImportError, e:
995
failed_to_load_extension(e)
996
from bzrlib._chunks_to_lines_py import chunks_to_lines
1000
"""Split s into lines, but without removing the newline characters."""
1001
# Trivially convert a fulltext into a 'chunked' representation, and let
1002
# chunks_to_lines do the heavy lifting.
1003
if isinstance(s, str):
1004
# chunks_to_lines only supports 8-bit strings
1005
return chunks_to_lines([s])
1007
return _split_lines(s)
1010
def _split_lines(s):
1011
"""Split s into lines, but without removing the newline characters.
1013
This supports Unicode or plain string objects.
1015
lines = s.split('\n')
1016
result = [line + '\n' for line in lines[:-1]]
1018
result.append(lines[-1])
1022
def hardlinks_good():
1023
return sys.platform not in ('win32', 'cygwin', 'darwin')
1026
def link_or_copy(src, dest):
1027
"""Hardlink a file, or copy it if it can't be hardlinked."""
1028
if not hardlinks_good():
1029
shutil.copyfile(src, dest)
1033
except (OSError, IOError), e:
1034
if e.errno != errno.EXDEV:
1036
shutil.copyfile(src, dest)
1039
def delete_any(path):
1040
"""Delete a file, symlink or directory.
1042
Will delete even if readonly.
1045
_delete_file_or_dir(path)
1046
except (OSError, IOError), e:
1047
if e.errno in (errno.EPERM, errno.EACCES):
1048
# make writable and try again
1051
except (OSError, IOError):
1053
_delete_file_or_dir(path)
1058
def _delete_file_or_dir(path):
1059
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1060
# Forgiveness than Permission (EAFP) because:
1061
# - root can damage a solaris file system by using unlink,
1062
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1063
# EACCES, OSX: EPERM) when invoked on a directory.
1064
if isdir(path): # Takes care of symlinks
1071
if getattr(os, 'symlink', None) is not None:
1077
def has_hardlinks():
1078
if getattr(os, 'link', None) is not None:
1084
def host_os_dereferences_symlinks():
1085
return (has_symlinks()
1086
and sys.platform not in ('cygwin', 'win32'))
1089
def readlink(abspath):
1090
"""Return a string representing the path to which the symbolic link points.
1092
:param abspath: The link absolute unicode path.
1094
This his guaranteed to return the symbolic link in unicode in all python
1097
link = abspath.encode(_fs_enc)
1098
target = os.readlink(link)
1099
target = target.decode(_fs_enc)
1103
def contains_whitespace(s):
1104
"""True if there are any whitespace characters in s."""
1105
# string.whitespace can include '\xa0' in certain locales, because it is
1106
# considered "non-breaking-space" as part of ISO-8859-1. But it
1107
# 1) Isn't a breaking whitespace
1108
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1110
# 3) '\xa0' isn't unicode safe since it is >128.
1112
# This should *not* be a unicode set of characters in case the source
1113
# string is not a Unicode string. We can auto-up-cast the characters since
1114
# they are ascii, but we don't want to auto-up-cast the string in case it
1116
for ch in ' \t\n\r\v\f':
1123
def contains_linebreaks(s):
1124
"""True if there is any vertical whitespace in s."""
1132
def relpath(base, path):
1133
"""Return path relative to base, or raise PathNotChild exception.
1135
The path may be either an absolute path or a path relative to the
1136
current working directory.
1138
os.path.commonprefix (python2.4) has a bad bug that it works just
1139
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1140
avoids that problem.
1142
NOTE: `base` should not have a trailing slash otherwise you'll get
1143
PathNotChild exceptions regardless of `path`.
1146
if len(base) < MIN_ABS_PATHLENGTH:
1147
# must have space for e.g. a drive letter
1148
raise ValueError('%r is too short to calculate a relative path'
1156
if len(head) <= len(base) and head != base:
1157
raise errors.PathNotChild(rp, base)
1160
head, tail = split(head)
1165
return pathjoin(*reversed(s))
1170
def _cicp_canonical_relpath(base, path):
1171
"""Return the canonical path relative to base.
1173
Like relpath, but on case-insensitive-case-preserving file-systems, this
1174
will return the relpath as stored on the file-system rather than in the
1175
case specified in the input string, for all existing portions of the path.
1177
This will cause O(N) behaviour if called for every path in a tree; if you
1178
have a number of paths to convert, you should use canonical_relpaths().
1180
# TODO: it should be possible to optimize this for Windows by using the
1181
# win32 API FindFiles function to look for the specified name - but using
1182
# os.listdir() still gives us the correct, platform agnostic semantics in
1185
rel = relpath(base, path)
1186
# '.' will have been turned into ''
1190
abs_base = abspath(base)
1192
_listdir = os.listdir
1194
# use an explicit iterator so we can easily consume the rest on early exit.
1195
bit_iter = iter(rel.split('/'))
1196
for bit in bit_iter:
1199
next_entries = _listdir(current)
1200
except OSError: # enoent, eperm, etc
1201
# We can't find this in the filesystem, so just append the
1203
current = pathjoin(current, bit, *list(bit_iter))
1205
for look in next_entries:
1206
if lbit == look.lower():
1207
current = pathjoin(current, look)
1210
# got to the end, nothing matched, so we just return the
1211
# non-existing bits as they were specified (the filename may be
1212
# the target of a move, for example).
1213
current = pathjoin(current, bit, *list(bit_iter))
1215
return current[len(abs_base):].lstrip('/')
1217
# XXX - TODO - we need better detection/integration of case-insensitive
1218
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1219
# filesystems), for example, so could probably benefit from the same basic
1220
# support there. For now though, only Windows and OSX get that support, and
1221
# they get it for *all* file-systems!
1222
if sys.platform in ('win32', 'darwin'):
1223
canonical_relpath = _cicp_canonical_relpath
1225
canonical_relpath = relpath
1227
def canonical_relpaths(base, paths):
1228
"""Create an iterable to canonicalize a sequence of relative paths.
1230
The intent is for this implementation to use a cache, vastly speeding
1231
up multiple transformations in the same directory.
1233
# but for now, we haven't optimized...
1234
return [canonical_relpath(base, p) for p in paths]
1237
def decode_filename(filename):
1238
"""Decode the filename using the filesystem encoding
1240
If it is unicode, it is returned.
1241
Otherwise it is decoded from the the filesystem's encoding. If decoding
1242
fails, a errors.BadFilenameEncoding exception is raised.
1244
if type(filename) is unicode:
1247
return filename.decode(_fs_enc)
1248
except UnicodeDecodeError:
1249
raise errors.BadFilenameEncoding(filename, _fs_enc)
1252
def safe_unicode(unicode_or_utf8_string):
1253
"""Coerce unicode_or_utf8_string into unicode.
1255
If it is unicode, it is returned.
1256
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1257
wrapped in a BzrBadParameterNotUnicode exception.
1259
if isinstance(unicode_or_utf8_string, unicode):
1260
return unicode_or_utf8_string
1262
return unicode_or_utf8_string.decode('utf8')
1263
except UnicodeDecodeError:
1264
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1267
def safe_utf8(unicode_or_utf8_string):
1268
"""Coerce unicode_or_utf8_string to a utf8 string.
1270
If it is a str, it is returned.
1271
If it is Unicode, it is encoded into a utf-8 string.
1273
if isinstance(unicode_or_utf8_string, str):
1274
# TODO: jam 20070209 This is overkill, and probably has an impact on
1275
# performance if we are dealing with lots of apis that want a
1278
# Make sure it is a valid utf-8 string
1279
unicode_or_utf8_string.decode('utf-8')
1280
except UnicodeDecodeError:
1281
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1282
return unicode_or_utf8_string
1283
return unicode_or_utf8_string.encode('utf-8')
1286
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1287
' Revision id generators should be creating utf8'
1291
def safe_revision_id(unicode_or_utf8_string, warn=True):
1292
"""Revision ids should now be utf8, but at one point they were unicode.
1294
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1296
:param warn: Functions that are sanitizing user data can set warn=False
1297
:return: None or a utf8 revision id.
1299
if (unicode_or_utf8_string is None
1300
or unicode_or_utf8_string.__class__ == str):
1301
return unicode_or_utf8_string
1303
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1305
return cache_utf8.encode(unicode_or_utf8_string)
1308
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1309
' generators should be creating utf8 file ids.')
1312
def safe_file_id(unicode_or_utf8_string, warn=True):
1313
"""File ids should now be utf8, but at one point they were unicode.
1315
This is the same as safe_utf8, except it uses the cached encode functions
1316
to save a little bit of performance.
1318
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1320
:param warn: Functions that are sanitizing user data can set warn=False
1321
:return: None or a utf8 file id.
1323
if (unicode_or_utf8_string is None
1324
or unicode_or_utf8_string.__class__ == str):
1325
return unicode_or_utf8_string
1327
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1329
return cache_utf8.encode(unicode_or_utf8_string)
1332
_platform_normalizes_filenames = False
1333
if sys.platform == 'darwin':
1334
_platform_normalizes_filenames = True
1337
def normalizes_filenames():
1338
"""Return True if this platform normalizes unicode filenames.
1342
return _platform_normalizes_filenames
1345
def _accessible_normalized_filename(path):
1346
"""Get the unicode normalized path, and if you can access the file.
1348
On platforms where the system normalizes filenames (Mac OSX),
1349
you can access a file by any path which will normalize correctly.
1350
On platforms where the system does not normalize filenames
1351
(everything else), you have to access a file by its exact path.
1353
Internally, bzr only supports NFC normalization, since that is
1354
the standard for XML documents.
1356
So return the normalized path, and a flag indicating if the file
1357
can be accessed by that path.
1360
return unicodedata.normalize('NFC', unicode(path)), True
1363
def _inaccessible_normalized_filename(path):
1364
__doc__ = _accessible_normalized_filename.__doc__
1366
normalized = unicodedata.normalize('NFC', unicode(path))
1367
return normalized, normalized == path
1370
if _platform_normalizes_filenames:
1371
normalized_filename = _accessible_normalized_filename
1373
normalized_filename = _inaccessible_normalized_filename
1376
def set_signal_handler(signum, handler, restart_syscall=True):
1377
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1378
on platforms that support that.
1380
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1381
automatically restart (by calling `signal.siginterrupt(signum,
1382
False)`). May be ignored if the feature is not available on this
1383
platform or Python version.
1387
siginterrupt = signal.siginterrupt
1389
# This python implementation doesn't provide signal support, hence no
1392
except AttributeError:
1393
# siginterrupt doesn't exist on this platform, or for this version
1395
siginterrupt = lambda signum, flag: None
1397
def sig_handler(*args):
1398
# Python resets the siginterrupt flag when a signal is
1399
# received. <http://bugs.python.org/issue8354>
1400
# As a workaround for some cases, set it back the way we want it.
1401
siginterrupt(signum, False)
1402
# Now run the handler function passed to set_signal_handler.
1405
sig_handler = handler
1406
old_handler = signal.signal(signum, sig_handler)
1408
siginterrupt(signum, False)
1412
default_terminal_width = 80
1413
"""The default terminal width for ttys.
1415
This is defined so that higher levels can share a common fallback value when
1416
terminal_width() returns None.
1419
# Keep some state so that terminal_width can detect if _terminal_size has
1420
# returned a different size since the process started. See docstring and
1421
# comments of terminal_width for details.
1422
# _terminal_size_state has 3 possible values: no_data, unchanged, and changed.
1423
_terminal_size_state = 'no_data'
1424
_first_terminal_size = None
1426
def terminal_width():
1427
"""Return terminal width.
1429
None is returned if the width can't established precisely.
1432
- if BZR_COLUMNS is set, returns its value
1433
- if there is no controlling terminal, returns None
1434
- query the OS, if the queried size has changed since the last query,
1436
- if COLUMNS is set, returns its value,
1437
- if the OS has a value (even though it's never changed), return its value.
1439
From there, we need to query the OS to get the size of the controlling
1442
On Unices we query the OS by:
1443
- get termios.TIOCGWINSZ
1444
- if an error occurs or a negative value is obtained, returns None
1446
On Windows we query the OS by:
1447
- win32utils.get_console_size() decides,
1448
- returns None on error (provided default value)
1450
# Note to implementors: if changing the rules for determining the width,
1451
# make sure you've considered the behaviour in these cases:
1452
# - M-x shell in emacs, where $COLUMNS is set and TIOCGWINSZ returns 0,0.
1453
# - bzr log | less, in bash, where $COLUMNS not set and TIOCGWINSZ returns
1455
# - (add more interesting cases here, if you find any)
1456
# Some programs implement "Use $COLUMNS (if set) until SIGWINCH occurs",
1457
# but we don't want to register a signal handler because it is impossible
1458
# to do so without risking EINTR errors in Python <= 2.6.5 (see
1459
# <http://bugs.python.org/issue8354>). Instead we check TIOCGWINSZ every
1460
# time so we can notice if the reported size has changed, which should have
1463
# If BZR_COLUMNS is set, take it, user is always right
1465
return int(os.environ['BZR_COLUMNS'])
1466
except (KeyError, ValueError):
1469
isatty = getattr(sys.stdout, 'isatty', None)
1470
if isatty is None or not isatty():
1471
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1475
width, height = os_size = _terminal_size(None, None)
1476
global _first_terminal_size, _terminal_size_state
1477
if _terminal_size_state == 'no_data':
1478
_first_terminal_size = os_size
1479
_terminal_size_state = 'unchanged'
1480
elif (_terminal_size_state == 'unchanged' and
1481
_first_terminal_size != os_size):
1482
_terminal_size_state = 'changed'
1484
# If the OS claims to know how wide the terminal is, and this value has
1485
# ever changed, use that.
1486
if _terminal_size_state == 'changed':
1487
if width is not None and width > 0:
1490
# If COLUMNS is set, use it.
1492
return int(os.environ['COLUMNS'])
1493
except (KeyError, ValueError):
1496
# Finally, use an unchanged size from the OS, if we have one.
1497
if _terminal_size_state == 'unchanged':
1498
if width is not None and width > 0:
1501
# The width could not be determined.
1505
def _win32_terminal_size(width, height):
1506
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1507
return width, height
1510
def _ioctl_terminal_size(width, height):
1512
import struct, fcntl, termios
1513
s = struct.pack('HHHH', 0, 0, 0, 0)
1514
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1515
height, width = struct.unpack('HHHH', x)[0:2]
1516
except (IOError, AttributeError):
1518
return width, height
1520
_terminal_size = None
1521
"""Returns the terminal size as (width, height).
1523
:param width: Default value for width.
1524
:param height: Default value for height.
1526
This is defined specifically for each OS and query the size of the controlling
1527
terminal. If any error occurs, the provided default values should be returned.
1529
if sys.platform == 'win32':
1530
_terminal_size = _win32_terminal_size
1532
_terminal_size = _ioctl_terminal_size
1535
def supports_executable():
1536
return sys.platform != "win32"
1539
def supports_posix_readonly():
1540
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1542
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1543
directory controls creation/deletion, etc.
1545
And under win32, readonly means that the directory itself cannot be
1546
deleted. The contents of a readonly directory can be changed, unlike POSIX
1547
where files in readonly directories cannot be added, deleted or renamed.
1549
return sys.platform != "win32"
1552
def set_or_unset_env(env_variable, value):
1553
"""Modify the environment, setting or removing the env_variable.
1555
:param env_variable: The environment variable in question
1556
:param value: The value to set the environment to. If None, then
1557
the variable will be removed.
1558
:return: The original value of the environment variable.
1560
orig_val = os.environ.get(env_variable)
1562
if orig_val is not None:
1563
del os.environ[env_variable]
1565
if isinstance(value, unicode):
1566
value = value.encode(get_user_encoding())
1567
os.environ[env_variable] = value
1571
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1574
def check_legal_path(path):
1575
"""Check whether the supplied path is legal.
1576
This is only required on Windows, so we don't test on other platforms
1579
if sys.platform != "win32":
1581
if _validWin32PathRE.match(path) is None:
1582
raise errors.IllegalPath(path)
1585
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1587
def _is_error_enotdir(e):
1588
"""Check if this exception represents ENOTDIR.
1590
Unfortunately, python is very inconsistent about the exception
1591
here. The cases are:
1592
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1593
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1594
which is the windows error code.
1595
3) Windows, Python2.5 uses errno == EINVAL and
1596
winerror == ERROR_DIRECTORY
1598
:param e: An Exception object (expected to be OSError with an errno
1599
attribute, but we should be able to cope with anything)
1600
:return: True if this represents an ENOTDIR error. False otherwise.
1602
en = getattr(e, 'errno', None)
1603
if (en == errno.ENOTDIR
1604
or (sys.platform == 'win32'
1605
and (en == _WIN32_ERROR_DIRECTORY
1606
or (en == errno.EINVAL
1607
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1613
def walkdirs(top, prefix=""):
1614
"""Yield data about all the directories in a tree.
1616
This yields all the data about the contents of a directory at a time.
1617
After each directory has been yielded, if the caller has mutated the list
1618
to exclude some directories, they are then not descended into.
1620
The data yielded is of the form:
1621
((directory-relpath, directory-path-from-top),
1622
[(relpath, basename, kind, lstat, path-from-top), ...]),
1623
- directory-relpath is the relative path of the directory being returned
1624
with respect to top. prefix is prepended to this.
1625
- directory-path-from-root is the path including top for this directory.
1626
It is suitable for use with os functions.
1627
- relpath is the relative path within the subtree being walked.
1628
- basename is the basename of the path
1629
- kind is the kind of the file now. If unknown then the file is not
1630
present within the tree - but it may be recorded as versioned. See
1632
- lstat is the stat data *if* the file was statted.
1633
- planned, not implemented:
1634
path_from_tree_root is the path from the root of the tree.
1636
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1637
allows one to walk a subtree but get paths that are relative to a tree
1639
:return: an iterator over the dirs.
1641
#TODO there is a bit of a smell where the results of the directory-
1642
# summary in this, and the path from the root, may not agree
1643
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1644
# potentially confusing output. We should make this more robust - but
1645
# not at a speed cost. RBC 20060731
1647
_directory = _directory_kind
1648
_listdir = os.listdir
1649
_kind_from_mode = file_kind_from_stat_mode
1650
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1652
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1653
relroot, _, _, _, top = pending.pop()
1655
relprefix = relroot + u'/'
1658
top_slash = top + u'/'
1661
append = dirblock.append
1663
names = sorted(map(decode_filename, _listdir(top)))
1665
if not _is_error_enotdir(e):
1669
abspath = top_slash + name
1670
statvalue = _lstat(abspath)
1671
kind = _kind_from_mode(statvalue.st_mode)
1672
append((relprefix + name, name, kind, statvalue, abspath))
1673
yield (relroot, top), dirblock
1675
# push the user specified dirs from dirblock
1676
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1679
class DirReader(object):
1680
"""An interface for reading directories."""
1682
def top_prefix_to_starting_dir(self, top, prefix=""):
1683
"""Converts top and prefix to a starting dir entry
1685
:param top: A utf8 path
1686
:param prefix: An optional utf8 path to prefix output relative paths
1688
:return: A tuple starting with prefix, and ending with the native
1691
raise NotImplementedError(self.top_prefix_to_starting_dir)
1693
def read_dir(self, prefix, top):
1694
"""Read a specific dir.
1696
:param prefix: A utf8 prefix to be preprended to the path basenames.
1697
:param top: A natively encoded path to read.
1698
:return: A list of the directories contents. Each item contains:
1699
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1701
raise NotImplementedError(self.read_dir)
1704
_selected_dir_reader = None
1707
def _walkdirs_utf8(top, prefix=""):
1708
"""Yield data about all the directories in a tree.
1710
This yields the same information as walkdirs() only each entry is yielded
1711
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1712
are returned as exact byte-strings.
1714
:return: yields a tuple of (dir_info, [file_info])
1715
dir_info is (utf8_relpath, path-from-top)
1716
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1717
if top is an absolute path, path-from-top is also an absolute path.
1718
path-from-top might be unicode or utf8, but it is the correct path to
1719
pass to os functions to affect the file in question. (such as os.lstat)
1721
global _selected_dir_reader
1722
if _selected_dir_reader is None:
1723
fs_encoding = _fs_enc.upper()
1724
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1725
# Win98 doesn't have unicode apis like FindFirstFileW
1726
# TODO: We possibly could support Win98 by falling back to the
1727
# original FindFirstFile, and using TCHAR instead of WCHAR,
1728
# but that gets a bit tricky, and requires custom compiling
1731
from bzrlib._walkdirs_win32 import Win32ReadDir
1732
_selected_dir_reader = Win32ReadDir()
1735
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1736
# ANSI_X3.4-1968 is a form of ASCII
1738
from bzrlib._readdir_pyx import UTF8DirReader
1739
_selected_dir_reader = UTF8DirReader()
1740
except ImportError, e:
1741
failed_to_load_extension(e)
1744
if _selected_dir_reader is None:
1745
# Fallback to the python version
1746
_selected_dir_reader = UnicodeDirReader()
1748
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1749
# But we don't actually uses 1-3 in pending, so set them to None
1750
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1751
read_dir = _selected_dir_reader.read_dir
1752
_directory = _directory_kind
1754
relroot, _, _, _, top = pending[-1].pop()
1757
dirblock = sorted(read_dir(relroot, top))
1758
yield (relroot, top), dirblock
1759
# push the user specified dirs from dirblock
1760
next = [d for d in reversed(dirblock) if d[2] == _directory]
1762
pending.append(next)
1765
class UnicodeDirReader(DirReader):
1766
"""A dir reader for non-utf8 file systems, which transcodes."""
1768
__slots__ = ['_utf8_encode']
1771
self._utf8_encode = codecs.getencoder('utf8')
1773
def top_prefix_to_starting_dir(self, top, prefix=""):
1774
"""See DirReader.top_prefix_to_starting_dir."""
1775
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1777
def read_dir(self, prefix, top):
1778
"""Read a single directory from a non-utf8 file system.
1780
top, and the abspath element in the output are unicode, all other paths
1781
are utf8. Local disk IO is done via unicode calls to listdir etc.
1783
This is currently the fallback code path when the filesystem encoding is
1784
not UTF-8. It may be better to implement an alternative so that we can
1785
safely handle paths that are not properly decodable in the current
1788
See DirReader.read_dir for details.
1790
_utf8_encode = self._utf8_encode
1792
_listdir = os.listdir
1793
_kind_from_mode = file_kind_from_stat_mode
1796
relprefix = prefix + '/'
1799
top_slash = top + u'/'
1802
append = dirblock.append
1803
for name in sorted(_listdir(top)):
1805
name_utf8 = _utf8_encode(name)[0]
1806
except UnicodeDecodeError:
1807
raise errors.BadFilenameEncoding(
1808
_utf8_encode(relprefix)[0] + name, _fs_enc)
1809
abspath = top_slash + name
1810
statvalue = _lstat(abspath)
1811
kind = _kind_from_mode(statvalue.st_mode)
1812
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1816
def copy_tree(from_path, to_path, handlers={}):
1817
"""Copy all of the entries in from_path into to_path.
1819
:param from_path: The base directory to copy.
1820
:param to_path: The target directory. If it does not exist, it will
1822
:param handlers: A dictionary of functions, which takes a source and
1823
destinations for files, directories, etc.
1824
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1825
'file', 'directory', and 'symlink' should always exist.
1826
If they are missing, they will be replaced with 'os.mkdir()',
1827
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1829
# Now, just copy the existing cached tree to the new location
1830
# We use a cheap trick here.
1831
# Absolute paths are prefixed with the first parameter
1832
# relative paths are prefixed with the second.
1833
# So we can get both the source and target returned
1834
# without any extra work.
1836
def copy_dir(source, dest):
1839
def copy_link(source, dest):
1840
"""Copy the contents of a symlink"""
1841
link_to = os.readlink(source)
1842
os.symlink(link_to, dest)
1844
real_handlers = {'file':shutil.copy2,
1845
'symlink':copy_link,
1846
'directory':copy_dir,
1848
real_handlers.update(handlers)
1850
if not os.path.exists(to_path):
1851
real_handlers['directory'](from_path, to_path)
1853
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1854
for relpath, name, kind, st, abspath in entries:
1855
real_handlers[kind](abspath, relpath)
1858
def copy_ownership_from_path(dst, src=None):
1859
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1861
If src is None, the containing directory is used as source. If chown
1862
fails, the error is ignored and a warning is printed.
1864
chown = getattr(os, 'chown', None)
1869
src = os.path.dirname(dst)
1875
chown(dst, s.st_uid, s.st_gid)
1877
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1880
def path_prefix_key(path):
1881
"""Generate a prefix-order path key for path.
1883
This can be used to sort paths in the same way that walkdirs does.
1885
return (dirname(path) , path)
1888
def compare_paths_prefix_order(path_a, path_b):
1889
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1890
key_a = path_prefix_key(path_a)
1891
key_b = path_prefix_key(path_b)
1892
return cmp(key_a, key_b)
1895
_cached_user_encoding = None
1898
def get_user_encoding(use_cache=True):
1899
"""Find out what the preferred user encoding is.
1901
This is generally the encoding that is used for command line parameters
1902
and file contents. This may be different from the terminal encoding
1903
or the filesystem encoding.
1905
:param use_cache: Enable cache for detected encoding.
1906
(This parameter is turned on by default,
1907
and required only for selftesting)
1909
:return: A string defining the preferred user encoding
1911
global _cached_user_encoding
1912
if _cached_user_encoding is not None and use_cache:
1913
return _cached_user_encoding
1915
if sys.platform == 'darwin':
1916
# python locale.getpreferredencoding() always return
1917
# 'mac-roman' on darwin. That's a lie.
1918
sys.platform = 'posix'
1920
if os.environ.get('LANG', None) is None:
1921
# If LANG is not set, we end up with 'ascii', which is bad
1922
# ('mac-roman' is more than ascii), so we set a default which
1923
# will give us UTF-8 (which appears to work in all cases on
1924
# OSX). Users are still free to override LANG of course, as
1925
# long as it give us something meaningful. This work-around
1926
# *may* not be needed with python 3k and/or OSX 10.5, but will
1927
# work with them too -- vila 20080908
1928
os.environ['LANG'] = 'en_US.UTF-8'
1931
sys.platform = 'darwin'
1936
user_encoding = locale.getpreferredencoding()
1937
except locale.Error, e:
1938
sys.stderr.write('bzr: warning: %s\n'
1939
' Could not determine what text encoding to use.\n'
1940
' This error usually means your Python interpreter\n'
1941
' doesn\'t support the locale set by $LANG (%s)\n'
1942
" Continuing with ascii encoding.\n"
1943
% (e, os.environ.get('LANG')))
1944
user_encoding = 'ascii'
1946
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1947
# treat that as ASCII, and not support printing unicode characters to the
1950
# For python scripts run under vim, we get '', so also treat that as ASCII
1951
if user_encoding in (None, 'cp0', ''):
1952
user_encoding = 'ascii'
1956
codecs.lookup(user_encoding)
1958
sys.stderr.write('bzr: warning:'
1959
' unknown encoding %s.'
1960
' Continuing with ascii encoding.\n'
1963
user_encoding = 'ascii'
1966
_cached_user_encoding = user_encoding
1968
return user_encoding
1971
def get_diff_header_encoding():
1972
return get_terminal_encoding()
1975
def get_host_name():
1976
"""Return the current unicode host name.
1978
This is meant to be used in place of socket.gethostname() because that
1979
behaves inconsistently on different platforms.
1981
if sys.platform == "win32":
1983
return win32utils.get_host_name()
1986
return socket.gethostname().decode(get_user_encoding())
1989
# We must not read/write any more than 64k at a time from/to a socket so we
1990
# don't risk "no buffer space available" errors on some platforms. Windows in
1991
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1993
MAX_SOCKET_CHUNK = 64 * 1024
1995
def read_bytes_from_socket(sock, report_activity=None,
1996
max_read_size=MAX_SOCKET_CHUNK):
1997
"""Read up to max_read_size of bytes from sock and notify of progress.
1999
Translates "Connection reset by peer" into file-like EOF (return an
2000
empty string rather than raise an error), and repeats the recv if
2001
interrupted by a signal.
2005
bytes = sock.recv(max_read_size)
2006
except socket.error, e:
2008
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
2009
# The connection was closed by the other side. Callers expect
2010
# an empty string to signal end-of-stream.
2012
elif eno == errno.EINTR:
2013
# Retry the interrupted recv.
2017
if report_activity is not None:
2018
report_activity(len(bytes), 'read')
2022
def recv_all(socket, count):
2023
"""Receive an exact number of bytes.
2025
Regular Socket.recv() may return less than the requested number of bytes,
2026
depending on what's in the OS buffer. MSG_WAITALL is not available
2027
on all platforms, but this should work everywhere. This will return
2028
less than the requested amount if the remote end closes.
2030
This isn't optimized and is intended mostly for use in testing.
2033
while len(b) < count:
2034
new = read_bytes_from_socket(socket, None, count - len(b))
2041
def send_all(sock, bytes, report_activity=None):
2042
"""Send all bytes on a socket.
2044
Breaks large blocks in smaller chunks to avoid buffering limitations on
2045
some platforms, and catches EINTR which may be thrown if the send is
2046
interrupted by a signal.
2048
This is preferred to socket.sendall(), because it avoids portability bugs
2049
and provides activity reporting.
2051
:param report_activity: Call this as bytes are read, see
2052
Transport._report_activity
2055
byte_count = len(bytes)
2056
while sent_total < byte_count:
2058
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2059
except socket.error, e:
2060
if e.args[0] != errno.EINTR:
2064
report_activity(sent, 'write')
2067
def dereference_path(path):
2068
"""Determine the real path to a file.
2070
All parent elements are dereferenced. But the file itself is not
2072
:param path: The original path. May be absolute or relative.
2073
:return: the real path *to* the file
2075
parent, base = os.path.split(path)
2076
# The pathjoin for '.' is a workaround for Python bug #1213894.
2077
# (initial path components aren't dereferenced)
2078
return pathjoin(realpath(pathjoin('.', parent)), base)
2081
def supports_mapi():
2082
"""Return True if we can use MAPI to launch a mail client."""
2083
return sys.platform == "win32"
2086
def resource_string(package, resource_name):
2087
"""Load a resource from a package and return it as a string.
2089
Note: Only packages that start with bzrlib are currently supported.
2091
This is designed to be a lightweight implementation of resource
2092
loading in a way which is API compatible with the same API from
2094
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2095
If and when pkg_resources becomes a standard library, this routine
2098
# Check package name is within bzrlib
2099
if package == "bzrlib":
2100
resource_relpath = resource_name
2101
elif package.startswith("bzrlib."):
2102
package = package[len("bzrlib."):].replace('.', os.sep)
2103
resource_relpath = pathjoin(package, resource_name)
2105
raise errors.BzrError('resource package %s not in bzrlib' % package)
2107
# Map the resource to a file and read its contents
2108
base = dirname(bzrlib.__file__)
2109
if getattr(sys, 'frozen', None): # bzr.exe
2110
base = abspath(pathjoin(base, '..', '..'))
2111
f = file(pathjoin(base, resource_relpath), "rU")
2117
def file_kind_from_stat_mode_thunk(mode):
2118
global file_kind_from_stat_mode
2119
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2121
from bzrlib._readdir_pyx import UTF8DirReader
2122
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2123
except ImportError, e:
2124
# This is one time where we won't warn that an extension failed to
2125
# load. The extension is never available on Windows anyway.
2126
from bzrlib._readdir_py import (
2127
_kind_from_mode as file_kind_from_stat_mode
2129
return file_kind_from_stat_mode(mode)
2130
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2133
def file_kind(f, _lstat=os.lstat):
2135
return file_kind_from_stat_mode(_lstat(f).st_mode)
2137
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2138
raise errors.NoSuchFile(f)
2142
def until_no_eintr(f, *a, **kw):
2143
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2145
WARNING: you must be certain that it is safe to retry the call repeatedly
2146
if EINTR does occur. This is typically only true for low-level operations
2147
like os.read. If in any doubt, don't use this.
2149
Keep in mind that this is not a complete solution to EINTR. There is
2150
probably code in the Python standard library and other dependencies that
2151
may encounter EINTR if a signal arrives (and there is signal handler for
2152
that signal). So this function can reduce the impact for IO that bzrlib
2153
directly controls, but it is not a complete solution.
2155
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2159
except (IOError, OSError), e:
2160
if e.errno == errno.EINTR:
2165
def re_compile_checked(re_string, flags=0, where=""):
2166
"""Return a compiled re, or raise a sensible error.
2168
This should only be used when compiling user-supplied REs.
2170
:param re_string: Text form of regular expression.
2171
:param flags: eg re.IGNORECASE
2172
:param where: Message explaining to the user the context where
2173
it occurred, eg 'log search filter'.
2175
# from https://bugs.launchpad.net/bzr/+bug/251352
2177
re_obj = re.compile(re_string, flags)
2182
where = ' in ' + where
2183
# despite the name 'error' is a type
2184
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2185
% (where, re_string, e))
2188
if sys.platform == "win32":
2191
return msvcrt.getch()
2196
fd = sys.stdin.fileno()
2197
settings = termios.tcgetattr(fd)
2200
ch = sys.stdin.read(1)
2202
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2206
if sys.platform == 'linux2':
2207
def _local_concurrency():
2209
prefix = 'processor'
2210
for line in file('/proc/cpuinfo', 'rb'):
2211
if line.startswith(prefix):
2212
concurrency = int(line[line.find(':')+1:]) + 1
2214
elif sys.platform == 'darwin':
2215
def _local_concurrency():
2216
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2217
stdout=subprocess.PIPE).communicate()[0]
2218
elif sys.platform[0:7] == 'freebsd':
2219
def _local_concurrency():
2220
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2221
stdout=subprocess.PIPE).communicate()[0]
2222
elif sys.platform == 'sunos5':
2223
def _local_concurrency():
2224
return subprocess.Popen(['psrinfo', '-p',],
2225
stdout=subprocess.PIPE).communicate()[0]
2226
elif sys.platform == "win32":
2227
def _local_concurrency():
2228
# This appears to return the number of cores.
2229
return os.environ.get('NUMBER_OF_PROCESSORS')
2231
def _local_concurrency():
2236
_cached_local_concurrency = None
2238
def local_concurrency(use_cache=True):
2239
"""Return how many processes can be run concurrently.
2241
Rely on platform specific implementations and default to 1 (one) if
2242
anything goes wrong.
2244
global _cached_local_concurrency
2246
if _cached_local_concurrency is not None and use_cache:
2247
return _cached_local_concurrency
2249
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2250
if concurrency is None:
2252
concurrency = _local_concurrency()
2253
except (OSError, IOError):
2256
concurrency = int(concurrency)
2257
except (TypeError, ValueError):
2260
_cached_concurrency = concurrency
2264
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2265
"""A stream writer that doesn't decode str arguments."""
2267
def __init__(self, encode, stream, errors='strict'):
2268
codecs.StreamWriter.__init__(self, stream, errors)
2269
self.encode = encode
2271
def write(self, object):
2272
if type(object) is str:
2273
self.stream.write(object)
2275
data, _ = self.encode(object, self.errors)
2276
self.stream.write(data)
2278
if sys.platform == 'win32':
2279
def open_file(filename, mode='r', bufsize=-1):
2280
"""This function is used to override the ``open`` builtin.
2282
But it uses O_NOINHERIT flag so the file handle is not inherited by
2283
child processes. Deleting or renaming a closed file opened with this
2284
function is not blocking child processes.
2286
writing = 'w' in mode
2287
appending = 'a' in mode
2288
updating = '+' in mode
2289
binary = 'b' in mode
2292
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2293
# for flags for each modes.
2303
flags |= os.O_WRONLY
2304
flags |= os.O_CREAT | os.O_TRUNC
2309
flags |= os.O_WRONLY
2310
flags |= os.O_CREAT | os.O_APPEND
2315
flags |= os.O_RDONLY
2317
return os.fdopen(os.open(filename, flags), mode, bufsize)
2322
def getuser_unicode():
2323
"""Return the username as unicode.
2326
user_encoding = get_user_encoding()
2327
username = getpass.getuser().decode(user_encoding)
2328
except UnicodeDecodeError:
2329
raise errors.BzrError("Can't decode username as %s." % \
added
removed
bzrlib/osutils.py
