1
# Copyright (C) 2005-2010 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
from stat import (S_ISREG, S_ISDIR, S_ISLNK, ST_MODE, ST_SIZE,
21
S_ISCHR, S_ISBLK, S_ISFIFO, S_ISSOCK)
27
from bzrlib.lazy_import import lazy_import
28
lazy_import(globals(), """
29
from datetime import datetime
31
from ntpath import (abspath as _nt_abspath,
33
normpath as _nt_normpath,
34
realpath as _nt_realpath,
35
splitdrive as _nt_splitdrive,
45
from tempfile import (
58
from bzrlib.symbol_versioning import (
63
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
65
if sys.version_info < (2, 5):
66
import md5 as _mod_md5
68
import sha as _mod_sha
78
from bzrlib import symbol_versioning
81
# Cross platform wall-clock time functionality with decent resolution.
82
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
83
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
84
# synchronized with ``time.time()``, this is only meant to be used to find
85
# delta times by subtracting from another call to this function.
86
timer_func = time.time
87
if sys.platform == 'win32':
88
timer_func = time.clock
90
# On win32, O_BINARY is used to indicate the file should
91
# be opened in binary mode, rather than text mode.
92
# On other platforms, O_BINARY doesn't exist, because
93
# they always open in binary mode, so it is okay to
94
# OR with 0 on those platforms.
95
# O_NOINHERIT and O_TEXT exists only on win32 too.
96
O_BINARY = getattr(os, 'O_BINARY', 0)
97
O_TEXT = getattr(os, 'O_TEXT', 0)
98
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
101
def get_unicode_argv():
103
user_encoding = get_user_encoding()
104
return [a.decode(user_encoding) for a in sys.argv[1:]]
105
except UnicodeDecodeError:
106
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
110
def make_readonly(filename):
111
"""Make a filename read-only."""
112
mod = os.lstat(filename).st_mode
113
if not stat.S_ISLNK(mod):
115
os.chmod(filename, mod)
118
def make_writable(filename):
119
mod = os.lstat(filename).st_mode
120
if not stat.S_ISLNK(mod):
122
os.chmod(filename, mod)
125
def minimum_path_selection(paths):
126
"""Return the smallset subset of paths which are outside paths.
128
:param paths: A container (and hence not None) of paths.
129
:return: A set of paths sufficient to include everything in paths via
130
is_inside, drawn from the paths parameter.
136
return path.split('/')
137
sorted_paths = sorted(list(paths), key=sort_key)
139
search_paths = [sorted_paths[0]]
140
for path in sorted_paths[1:]:
141
if not is_inside(search_paths[-1], path):
142
# This path is unique, add it
143
search_paths.append(path)
145
return set(search_paths)
152
"""Return a quoted filename filename
154
This previously used backslash quoting, but that works poorly on
156
# TODO: I'm not really sure this is the best format either.x
158
if _QUOTE_RE is None:
159
_QUOTE_RE = re.compile(r'([^a-zA-Z0-9.,:/\\_~-])')
161
if _QUOTE_RE.search(f):
167
_directory_kind = 'directory'
170
"""Return the current umask"""
171
# Assume that people aren't messing with the umask while running
172
# XXX: This is not thread safe, but there is no way to get the
173
# umask without setting it
181
_directory_kind: "/",
183
'tree-reference': '+',
187
def kind_marker(kind):
189
return _kind_marker_map[kind]
191
# Slightly faster than using .get(, '') when the common case is that
196
lexists = getattr(os.path, 'lexists', None)
200
stat = getattr(os, 'lstat', os.stat)
204
if e.errno == errno.ENOENT:
207
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
210
def fancy_rename(old, new, rename_func, unlink_func):
211
"""A fancy rename, when you don't have atomic rename.
213
:param old: The old path, to rename from
214
:param new: The new path, to rename to
215
:param rename_func: The potentially non-atomic rename function
216
:param unlink_func: A way to delete the target file if the full rename
219
# sftp rename doesn't allow overwriting, so play tricks:
220
base = os.path.basename(new)
221
dirname = os.path.dirname(new)
222
# callers use different encodings for the paths so the following MUST
223
# respect that. We rely on python upcasting to unicode if new is unicode
224
# and keeping a str if not.
225
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
226
os.getpid(), rand_chars(10))
227
tmp_name = pathjoin(dirname, tmp_name)
229
# Rename the file out of the way, but keep track if it didn't exist
230
# We don't want to grab just any exception
231
# something like EACCES should prevent us from continuing
232
# The downside is that the rename_func has to throw an exception
233
# with an errno = ENOENT, or NoSuchFile
236
rename_func(new, tmp_name)
237
except (errors.NoSuchFile,), e:
240
# RBC 20060103 abstraction leakage: the paramiko SFTP clients rename
241
# function raises an IOError with errno is None when a rename fails.
242
# This then gets caught here.
243
if e.errno not in (None, errno.ENOENT, errno.ENOTDIR):
246
if (getattr(e, 'errno', None) is None
247
or e.errno not in (errno.ENOENT, errno.ENOTDIR)):
256
# This may throw an exception, in which case success will
258
rename_func(old, new)
260
except (IOError, OSError), e:
261
# source and target may be aliases of each other (e.g. on a
262
# case-insensitive filesystem), so we may have accidentally renamed
263
# source by when we tried to rename target
264
failure_exc = sys.exc_info()
265
if (file_existed and e.errno in (None, errno.ENOENT)
266
and old.lower() == new.lower()):
267
# source and target are the same file on a case-insensitive
268
# filesystem, so we don't generate an exception
272
# If the file used to exist, rename it back into place
273
# otherwise just delete it from the tmp location
275
unlink_func(tmp_name)
277
rename_func(tmp_name, new)
278
if failure_exc is not None:
279
raise failure_exc[0], failure_exc[1], failure_exc[2]
282
# In Python 2.4.2 and older, os.path.abspath and os.path.realpath
283
# choke on a Unicode string containing a relative path if
284
# os.getcwd() returns a non-sys.getdefaultencoding()-encoded
286
_fs_enc = sys.getfilesystemencoding() or 'utf-8'
287
def _posix_abspath(path):
288
# jam 20060426 rather than encoding to fsencoding
289
# copy posixpath.abspath, but use os.getcwdu instead
290
if not posixpath.isabs(path):
291
path = posixpath.join(getcwd(), path)
292
return posixpath.normpath(path)
295
def _posix_realpath(path):
296
return posixpath.realpath(path.encode(_fs_enc)).decode(_fs_enc)
299
def _win32_fixdrive(path):
300
"""Force drive letters to be consistent.
302
win32 is inconsistent whether it returns lower or upper case
303
and even if it was consistent the user might type the other
304
so we force it to uppercase
305
running python.exe under cmd.exe return capital C:\\
306
running win32 python inside a cygwin shell returns lowercase c:\\
308
drive, path = _nt_splitdrive(path)
309
return drive.upper() + path
312
def _win32_abspath(path):
313
# Real _nt_abspath doesn't have a problem with a unicode cwd
314
return _win32_fixdrive(_nt_abspath(unicode(path)).replace('\\', '/'))
317
def _win98_abspath(path):
318
"""Return the absolute version of a path.
319
Windows 98 safe implementation (python reimplementation
320
of Win32 API function GetFullPathNameW)
325
# \\HOST\path => //HOST/path
326
# //HOST/path => //HOST/path
327
# path => C:/cwd/path
330
# check for absolute path
331
drive = _nt_splitdrive(path)[0]
332
if drive == '' and path[:2] not in('//','\\\\'):
334
# we cannot simply os.path.join cwd and path
335
# because os.path.join('C:','/path') produce '/path'
336
# and this is incorrect
337
if path[:1] in ('/','\\'):
338
cwd = _nt_splitdrive(cwd)[0]
340
path = cwd + '\\' + path
341
return _win32_fixdrive(_nt_normpath(path).replace('\\', '/'))
344
def _win32_realpath(path):
345
# Real _nt_realpath doesn't have a problem with a unicode cwd
346
return _win32_fixdrive(_nt_realpath(unicode(path)).replace('\\', '/'))
349
def _win32_pathjoin(*args):
350
return _nt_join(*args).replace('\\', '/')
353
def _win32_normpath(path):
354
return _win32_fixdrive(_nt_normpath(unicode(path)).replace('\\', '/'))
358
return _win32_fixdrive(os.getcwdu().replace('\\', '/'))
361
def _win32_mkdtemp(*args, **kwargs):
362
return _win32_fixdrive(tempfile.mkdtemp(*args, **kwargs).replace('\\', '/'))
365
def _win32_rename(old, new):
366
"""We expect to be able to atomically replace 'new' with old.
368
On win32, if new exists, it must be moved out of the way first,
372
fancy_rename(old, new, rename_func=os.rename, unlink_func=os.unlink)
374
if e.errno in (errno.EPERM, errno.EACCES, errno.EBUSY, errno.EINVAL):
375
# If we try to rename a non-existant file onto cwd, we get
376
# EPERM or EACCES instead of ENOENT, this will raise ENOENT
377
# if the old path doesn't exist, sometimes we get EACCES
378
# On Linux, we seem to get EBUSY, on Mac we get EINVAL
384
return unicodedata.normalize('NFC', os.getcwdu())
387
# Default is to just use the python builtins, but these can be rebound on
388
# particular platforms.
389
abspath = _posix_abspath
390
realpath = _posix_realpath
391
pathjoin = os.path.join
392
normpath = os.path.normpath
395
dirname = os.path.dirname
396
basename = os.path.basename
397
split = os.path.split
398
splitext = os.path.splitext
399
# These were already imported into local scope
400
# mkdtemp = tempfile.mkdtemp
401
# rmtree = shutil.rmtree
403
MIN_ABS_PATHLENGTH = 1
406
if sys.platform == 'win32':
407
if win32utils.winver == 'Windows 98':
408
abspath = _win98_abspath
410
abspath = _win32_abspath
411
realpath = _win32_realpath
412
pathjoin = _win32_pathjoin
413
normpath = _win32_normpath
414
getcwd = _win32_getcwd
415
mkdtemp = _win32_mkdtemp
416
rename = _win32_rename
418
MIN_ABS_PATHLENGTH = 3
420
def _win32_delete_readonly(function, path, excinfo):
421
"""Error handler for shutil.rmtree function [for win32]
422
Helps to remove files and dirs marked as read-only.
424
exception = excinfo[1]
425
if function in (os.remove, os.rmdir) \
426
and isinstance(exception, OSError) \
427
and exception.errno == errno.EACCES:
433
def rmtree(path, ignore_errors=False, onerror=_win32_delete_readonly):
434
"""Replacer for shutil.rmtree: could remove readonly dirs/files"""
435
return shutil.rmtree(path, ignore_errors, onerror)
437
f = win32utils.get_unicode_argv # special function or None
441
elif sys.platform == 'darwin':
445
def get_terminal_encoding():
446
"""Find the best encoding for printing to the screen.
448
This attempts to check both sys.stdout and sys.stdin to see
449
what encoding they are in, and if that fails it falls back to
450
osutils.get_user_encoding().
451
The problem is that on Windows, locale.getpreferredencoding()
452
is not the same encoding as that used by the console:
453
http://mail.python.org/pipermail/python-list/2003-May/162357.html
455
On my standard US Windows XP, the preferred encoding is
456
cp1252, but the console is cp437
458
from bzrlib.trace import mutter
459
output_encoding = getattr(sys.stdout, 'encoding', None)
460
if not output_encoding:
461
input_encoding = getattr(sys.stdin, 'encoding', None)
462
if not input_encoding:
463
output_encoding = get_user_encoding()
464
mutter('encoding stdout as osutils.get_user_encoding() %r',
467
output_encoding = input_encoding
468
mutter('encoding stdout as sys.stdin encoding %r', output_encoding)
470
mutter('encoding stdout as sys.stdout encoding %r', output_encoding)
471
if output_encoding == 'cp0':
472
# invalid encoding (cp0 means 'no codepage' on Windows)
473
output_encoding = get_user_encoding()
474
mutter('cp0 is invalid encoding.'
475
' encoding stdout as osutils.get_user_encoding() %r',
479
codecs.lookup(output_encoding)
481
sys.stderr.write('bzr: warning:'
482
' unknown terminal encoding %s.\n'
483
' Using encoding %s instead.\n'
484
% (output_encoding, get_user_encoding())
486
output_encoding = get_user_encoding()
488
return output_encoding
491
def normalizepath(f):
492
if getattr(os.path, 'realpath', None) is not None:
496
[p,e] = os.path.split(f)
497
if e == "" or e == "." or e == "..":
500
return pathjoin(F(p), e)
504
"""True if f is an accessible directory."""
506
return S_ISDIR(os.lstat(f)[ST_MODE])
512
"""True if f is a regular file."""
514
return S_ISREG(os.lstat(f)[ST_MODE])
519
"""True if f is a symlink."""
521
return S_ISLNK(os.lstat(f)[ST_MODE])
525
def is_inside(dir, fname):
526
"""True if fname is inside dir.
528
The parameters should typically be passed to osutils.normpath first, so
529
that . and .. and repeated slashes are eliminated, and the separators
530
are canonical for the platform.
532
The empty string as a dir name is taken as top-of-tree and matches
535
# XXX: Most callers of this can actually do something smarter by
536
# looking at the inventory
546
return fname.startswith(dir)
549
def is_inside_any(dir_list, fname):
550
"""True if fname is inside any of given dirs."""
551
for dirname in dir_list:
552
if is_inside(dirname, fname):
557
def is_inside_or_parent_of_any(dir_list, fname):
558
"""True if fname is a child or a parent of any of the given files."""
559
for dirname in dir_list:
560
if is_inside(dirname, fname) or is_inside(fname, dirname):
565
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
566
report_activity=None, direction='read'):
567
"""Copy contents of one file to another.
569
The read_length can either be -1 to read to end-of-file (EOF) or
570
it can specify the maximum number of bytes to read.
572
The buff_size represents the maximum size for each read operation
573
performed on from_file.
575
:param report_activity: Call this as bytes are read, see
576
Transport._report_activity
577
:param direction: Will be passed to report_activity
579
:return: The number of bytes copied.
583
# read specified number of bytes
585
while read_length > 0:
586
num_bytes_to_read = min(read_length, buff_size)
588
block = from_file.read(num_bytes_to_read)
592
if report_activity is not None:
593
report_activity(len(block), direction)
596
actual_bytes_read = len(block)
597
read_length -= actual_bytes_read
598
length += actual_bytes_read
602
block = from_file.read(buff_size)
606
if report_activity is not None:
607
report_activity(len(block), direction)
613
def pump_string_file(bytes, file_handle, segment_size=None):
614
"""Write bytes to file_handle in many smaller writes.
616
:param bytes: The string to write.
617
:param file_handle: The file to write to.
619
# Write data in chunks rather than all at once, because very large
620
# writes fail on some platforms (e.g. Windows with SMB mounted
623
segment_size = 5242880 # 5MB
624
segments = range(len(bytes) / segment_size + 1)
625
write = file_handle.write
626
for segment_index in segments:
627
segment = buffer(bytes, segment_index * segment_size, segment_size)
631
def file_iterator(input_file, readsize=32768):
633
b = input_file.read(readsize)
640
"""Calculate the hexdigest of an open file.
642
The file cursor should be already at the start.
654
def size_sha_file(f):
655
"""Calculate the size and hexdigest of an open file.
657
The file cursor should be already at the start and
658
the caller is responsible for closing the file afterwards.
669
return size, s.hexdigest()
672
def sha_file_by_name(fname):
673
"""Calculate the SHA1 of a file by reading the full text"""
675
f = os.open(fname, os.O_RDONLY | O_BINARY | O_NOINHERIT)
678
b = os.read(f, 1<<16)
686
def sha_strings(strings, _factory=sha):
687
"""Return the sha-1 of concatenation of strings"""
689
map(s.update, strings)
693
def sha_string(f, _factory=sha):
694
return _factory(f).hexdigest()
697
def fingerprint_file(f):
699
return {'size': len(b),
700
'sha1': sha(b).hexdigest()}
703
def compare_files(a, b):
704
"""Returns true if equal in contents"""
715
def local_time_offset(t=None):
716
"""Return offset of local zone from GMT, either at present or at time t."""
719
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
720
return offset.days * 86400 + offset.seconds
722
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
723
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
726
def format_date(t, offset=0, timezone='original', date_fmt=None,
728
"""Return a formatted date string.
730
:param t: Seconds since the epoch.
731
:param offset: Timezone offset in seconds east of utc.
732
:param timezone: How to display the time: 'utc', 'original' for the
733
timezone specified by offset, or 'local' for the process's current
735
:param date_fmt: strftime format.
736
:param show_offset: Whether to append the timezone.
738
(date_fmt, tt, offset_str) = \
739
_format_date(t, offset, timezone, date_fmt, show_offset)
740
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
741
date_str = time.strftime(date_fmt, tt)
742
return date_str + offset_str
745
# Cache of formatted offset strings
749
def format_date_with_offset_in_original_timezone(t, offset=0,
750
_cache=_offset_cache):
751
"""Return a formatted date string in the original timezone.
753
This routine may be faster then format_date.
755
:param t: Seconds since the epoch.
756
:param offset: Timezone offset in seconds east of utc.
760
tt = time.gmtime(t + offset)
761
date_fmt = _default_format_by_weekday_num[tt[6]]
762
date_str = time.strftime(date_fmt, tt)
763
offset_str = _cache.get(offset, None)
764
if offset_str is None:
765
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
766
_cache[offset] = offset_str
767
return date_str + offset_str
770
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
772
"""Return an unicode date string formatted according to the current locale.
774
:param t: Seconds since the epoch.
775
:param offset: Timezone offset in seconds east of utc.
776
:param timezone: How to display the time: 'utc', 'original' for the
777
timezone specified by offset, or 'local' for the process's current
779
:param date_fmt: strftime format.
780
:param show_offset: Whether to append the timezone.
782
(date_fmt, tt, offset_str) = \
783
_format_date(t, offset, timezone, date_fmt, show_offset)
784
date_str = time.strftime(date_fmt, tt)
785
if not isinstance(date_str, unicode):
786
date_str = date_str.decode(get_user_encoding(), 'replace')
787
return date_str + offset_str
790
def _format_date(t, offset, timezone, date_fmt, show_offset):
791
if timezone == 'utc':
794
elif timezone == 'original':
797
tt = time.gmtime(t + offset)
798
elif timezone == 'local':
799
tt = time.localtime(t)
800
offset = local_time_offset(t)
802
raise errors.UnsupportedTimezoneFormat(timezone)
804
date_fmt = "%a %Y-%m-%d %H:%M:%S"
806
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
809
return (date_fmt, tt, offset_str)
812
def compact_date(when):
813
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
816
def format_delta(delta):
817
"""Get a nice looking string for a time delta.
819
:param delta: The time difference in seconds, can be positive or negative.
820
positive indicates time in the past, negative indicates time in the
821
future. (usually time.time() - stored_time)
822
:return: String formatted to show approximate resolution
828
direction = 'in the future'
832
if seconds < 90: # print seconds up to 90 seconds
834
return '%d second %s' % (seconds, direction,)
836
return '%d seconds %s' % (seconds, direction)
838
minutes = int(seconds / 60)
839
seconds -= 60 * minutes
844
if minutes < 90: # print minutes, seconds up to 90 minutes
846
return '%d minute, %d second%s %s' % (
847
minutes, seconds, plural_seconds, direction)
849
return '%d minutes, %d second%s %s' % (
850
minutes, seconds, plural_seconds, direction)
852
hours = int(minutes / 60)
853
minutes -= 60 * hours
860
return '%d hour, %d minute%s %s' % (hours, minutes,
861
plural_minutes, direction)
862
return '%d hours, %d minute%s %s' % (hours, minutes,
863
plural_minutes, direction)
866
"""Return size of given open file."""
867
return os.fstat(f.fileno())[ST_SIZE]
870
# Define rand_bytes based on platform.
872
# Python 2.4 and later have os.urandom,
873
# but it doesn't work on some arches
875
rand_bytes = os.urandom
876
except (NotImplementedError, AttributeError):
877
# If python doesn't have os.urandom, or it doesn't work,
878
# then try to first pull random data from /dev/urandom
880
rand_bytes = file('/dev/urandom', 'rb').read
881
# Otherwise, use this hack as a last resort
882
except (IOError, OSError):
883
# not well seeded, but better than nothing
888
s += chr(random.randint(0, 255))
893
ALNUM = '0123456789abcdefghijklmnopqrstuvwxyz'
895
"""Return a random string of num alphanumeric characters
897
The result only contains lowercase chars because it may be used on
898
case-insensitive filesystems.
901
for raw_byte in rand_bytes(num):
902
s += ALNUM[ord(raw_byte) % 36]
906
## TODO: We could later have path objects that remember their list
907
## decomposition (might be too tricksy though.)
910
"""Turn string into list of parts."""
911
# split on either delimiter because people might use either on
913
ps = re.split(r'[\\/]', p)
918
raise errors.BzrError("sorry, %r not allowed in path" % f)
919
elif (f == '.') or (f == ''):
928
if (f == '..') or (f is None) or (f == ''):
929
raise errors.BzrError("sorry, %r not allowed in path" % f)
933
def parent_directories(filename):
934
"""Return the list of parent directories, deepest first.
936
For example, parent_directories("a/b/c") -> ["a/b", "a"].
939
parts = splitpath(dirname(filename))
941
parents.append(joinpath(parts))
946
_extension_load_failures = []
949
def failed_to_load_extension(exception):
950
"""Handle failing to load a binary extension.
952
This should be called from the ImportError block guarding the attempt to
953
import the native extension. If this function returns, the pure-Python
954
implementation should be loaded instead::
957
>>> import bzrlib._fictional_extension_pyx
958
>>> except ImportError, e:
959
>>> bzrlib.osutils.failed_to_load_extension(e)
960
>>> import bzrlib._fictional_extension_py
962
# NB: This docstring is just an example, not a doctest, because doctest
963
# currently can't cope with the use of lazy imports in this namespace --
966
# This currently doesn't report the failure at the time it occurs, because
967
# they tend to happen very early in startup when we can't check config
968
# files etc, and also we want to report all failures but not spam the user
970
from bzrlib import trace
971
exception_str = str(exception)
972
if exception_str not in _extension_load_failures:
973
trace.mutter("failed to load compiled extension: %s" % exception_str)
974
_extension_load_failures.append(exception_str)
977
def report_extension_load_failures():
978
if not _extension_load_failures:
980
from bzrlib.config import GlobalConfig
981
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
983
# the warnings framework should by default show this only once
984
from bzrlib.trace import warning
986
"bzr: warning: some compiled extensions could not be loaded; "
987
"see <https://answers.launchpad.net/bzr/+faq/703>")
988
# we no longer show the specific missing extensions here, because it makes
989
# the message too long and scary - see
990
# https://bugs.launchpad.net/bzr/+bug/430529
994
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
995
except ImportError, e:
996
failed_to_load_extension(e)
997
from bzrlib._chunks_to_lines_py import chunks_to_lines
1001
"""Split s into lines, but without removing the newline characters."""
1002
# Trivially convert a fulltext into a 'chunked' representation, and let
1003
# chunks_to_lines do the heavy lifting.
1004
if isinstance(s, str):
1005
# chunks_to_lines only supports 8-bit strings
1006
return chunks_to_lines([s])
1008
return _split_lines(s)
1011
def _split_lines(s):
1012
"""Split s into lines, but without removing the newline characters.
1014
This supports Unicode or plain string objects.
1016
lines = s.split('\n')
1017
result = [line + '\n' for line in lines[:-1]]
1019
result.append(lines[-1])
1023
def hardlinks_good():
1024
return sys.platform not in ('win32', 'cygwin', 'darwin')
1027
def link_or_copy(src, dest):
1028
"""Hardlink a file, or copy it if it can't be hardlinked."""
1029
if not hardlinks_good():
1030
shutil.copyfile(src, dest)
1034
except (OSError, IOError), e:
1035
if e.errno != errno.EXDEV:
1037
shutil.copyfile(src, dest)
1040
def delete_any(path):
1041
"""Delete a file, symlink or directory.
1043
Will delete even if readonly.
1046
_delete_file_or_dir(path)
1047
except (OSError, IOError), e:
1048
if e.errno in (errno.EPERM, errno.EACCES):
1049
# make writable and try again
1052
except (OSError, IOError):
1054
_delete_file_or_dir(path)
1059
def _delete_file_or_dir(path):
1060
# Look Before You Leap (LBYL) is appropriate here instead of Easier to Ask for
1061
# Forgiveness than Permission (EAFP) because:
1062
# - root can damage a solaris file system by using unlink,
1063
# - unlink raises different exceptions on different OSes (linux: EISDIR, win32:
1064
# EACCES, OSX: EPERM) when invoked on a directory.
1065
if isdir(path): # Takes care of symlinks
1072
if getattr(os, 'symlink', None) is not None:
1078
def has_hardlinks():
1079
if getattr(os, 'link', None) is not None:
1085
def host_os_dereferences_symlinks():
1086
return (has_symlinks()
1087
and sys.platform not in ('cygwin', 'win32'))
1090
def readlink(abspath):
1091
"""Return a string representing the path to which the symbolic link points.
1093
:param abspath: The link absolute unicode path.
1095
This his guaranteed to return the symbolic link in unicode in all python
1098
link = abspath.encode(_fs_enc)
1099
target = os.readlink(link)
1100
target = target.decode(_fs_enc)
1104
def contains_whitespace(s):
1105
"""True if there are any whitespace characters in s."""
1106
# string.whitespace can include '\xa0' in certain locales, because it is
1107
# considered "non-breaking-space" as part of ISO-8859-1. But it
1108
# 1) Isn't a breaking whitespace
1109
# 2) Isn't one of ' \t\r\n' which are characters we sometimes use as
1111
# 3) '\xa0' isn't unicode safe since it is >128.
1113
# This should *not* be a unicode set of characters in case the source
1114
# string is not a Unicode string. We can auto-up-cast the characters since
1115
# they are ascii, but we don't want to auto-up-cast the string in case it
1117
for ch in ' \t\n\r\v\f':
1124
def contains_linebreaks(s):
1125
"""True if there is any vertical whitespace in s."""
1133
def relpath(base, path):
1134
"""Return path relative to base, or raise exception.
1136
The path may be either an absolute path or a path relative to the
1137
current working directory.
1139
os.path.commonprefix (python2.4) has a bad bug that it works just
1140
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1141
avoids that problem.
1144
if len(base) < MIN_ABS_PATHLENGTH:
1145
# must have space for e.g. a drive letter
1146
raise ValueError('%r is too short to calculate a relative path'
1154
if len(head) <= len(base) and head != base:
1155
raise errors.PathNotChild(rp, base)
1158
head, tail = split(head)
1163
return pathjoin(*reversed(s))
1168
def _cicp_canonical_relpath(base, path):
1169
"""Return the canonical path relative to base.
1171
Like relpath, but on case-insensitive-case-preserving file-systems, this
1172
will return the relpath as stored on the file-system rather than in the
1173
case specified in the input string, for all existing portions of the path.
1175
This will cause O(N) behaviour if called for every path in a tree; if you
1176
have a number of paths to convert, you should use canonical_relpaths().
1178
# TODO: it should be possible to optimize this for Windows by using the
1179
# win32 API FindFiles function to look for the specified name - but using
1180
# os.listdir() still gives us the correct, platform agnostic semantics in
1183
rel = relpath(base, path)
1184
# '.' will have been turned into ''
1188
abs_base = abspath(base)
1190
_listdir = os.listdir
1192
# use an explicit iterator so we can easily consume the rest on early exit.
1193
bit_iter = iter(rel.split('/'))
1194
for bit in bit_iter:
1197
next_entries = _listdir(current)
1198
except OSError: # enoent, eperm, etc
1199
# We can't find this in the filesystem, so just append the
1201
current = pathjoin(current, bit, *list(bit_iter))
1203
for look in next_entries:
1204
if lbit == look.lower():
1205
current = pathjoin(current, look)
1208
# got to the end, nothing matched, so we just return the
1209
# non-existing bits as they were specified (the filename may be
1210
# the target of a move, for example).
1211
current = pathjoin(current, bit, *list(bit_iter))
1213
return current[len(abs_base):].lstrip('/')
1215
# XXX - TODO - we need better detection/integration of case-insensitive
1216
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1217
# filesystems), for example, so could probably benefit from the same basic
1218
# support there. For now though, only Windows and OSX get that support, and
1219
# they get it for *all* file-systems!
1220
if sys.platform in ('win32', 'darwin'):
1221
canonical_relpath = _cicp_canonical_relpath
1223
canonical_relpath = relpath
1225
def canonical_relpaths(base, paths):
1226
"""Create an iterable to canonicalize a sequence of relative paths.
1228
The intent is for this implementation to use a cache, vastly speeding
1229
up multiple transformations in the same directory.
1231
# but for now, we haven't optimized...
1232
return [canonical_relpath(base, p) for p in paths]
1234
def safe_unicode(unicode_or_utf8_string):
1235
"""Coerce unicode_or_utf8_string into unicode.
1237
If it is unicode, it is returned.
1238
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1239
wrapped in a BzrBadParameterNotUnicode exception.
1241
if isinstance(unicode_or_utf8_string, unicode):
1242
return unicode_or_utf8_string
1244
return unicode_or_utf8_string.decode('utf8')
1245
except UnicodeDecodeError:
1246
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1249
def safe_utf8(unicode_or_utf8_string):
1250
"""Coerce unicode_or_utf8_string to a utf8 string.
1252
If it is a str, it is returned.
1253
If it is Unicode, it is encoded into a utf-8 string.
1255
if isinstance(unicode_or_utf8_string, str):
1256
# TODO: jam 20070209 This is overkill, and probably has an impact on
1257
# performance if we are dealing with lots of apis that want a
1260
# Make sure it is a valid utf-8 string
1261
unicode_or_utf8_string.decode('utf-8')
1262
except UnicodeDecodeError:
1263
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1264
return unicode_or_utf8_string
1265
return unicode_or_utf8_string.encode('utf-8')
1268
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1269
' Revision id generators should be creating utf8'
1273
def safe_revision_id(unicode_or_utf8_string, warn=True):
1274
"""Revision ids should now be utf8, but at one point they were unicode.
1276
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1278
:param warn: Functions that are sanitizing user data can set warn=False
1279
:return: None or a utf8 revision id.
1281
if (unicode_or_utf8_string is None
1282
or unicode_or_utf8_string.__class__ == str):
1283
return unicode_or_utf8_string
1285
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1287
return cache_utf8.encode(unicode_or_utf8_string)
1290
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1291
' generators should be creating utf8 file ids.')
1294
def safe_file_id(unicode_or_utf8_string, warn=True):
1295
"""File ids should now be utf8, but at one point they were unicode.
1297
This is the same as safe_utf8, except it uses the cached encode functions
1298
to save a little bit of performance.
1300
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1302
:param warn: Functions that are sanitizing user data can set warn=False
1303
:return: None or a utf8 file id.
1305
if (unicode_or_utf8_string is None
1306
or unicode_or_utf8_string.__class__ == str):
1307
return unicode_or_utf8_string
1309
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1311
return cache_utf8.encode(unicode_or_utf8_string)
1314
_platform_normalizes_filenames = False
1315
if sys.platform == 'darwin':
1316
_platform_normalizes_filenames = True
1319
def normalizes_filenames():
1320
"""Return True if this platform normalizes unicode filenames.
1322
Mac OSX does, Windows/Linux do not.
1324
return _platform_normalizes_filenames
1327
def _accessible_normalized_filename(path):
1328
"""Get the unicode normalized path, and if you can access the file.
1330
On platforms where the system normalizes filenames (Mac OSX),
1331
you can access a file by any path which will normalize correctly.
1332
On platforms where the system does not normalize filenames
1333
(Windows, Linux), you have to access a file by its exact path.
1335
Internally, bzr only supports NFC normalization, since that is
1336
the standard for XML documents.
1338
So return the normalized path, and a flag indicating if the file
1339
can be accessed by that path.
1342
return unicodedata.normalize('NFC', unicode(path)), True
1345
def _inaccessible_normalized_filename(path):
1346
__doc__ = _accessible_normalized_filename.__doc__
1348
normalized = unicodedata.normalize('NFC', unicode(path))
1349
return normalized, normalized == path
1352
if _platform_normalizes_filenames:
1353
normalized_filename = _accessible_normalized_filename
1355
normalized_filename = _inaccessible_normalized_filename
1358
def set_signal_handler(signum, handler, restart_syscall=True):
1359
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1360
on platforms that support that.
1362
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1363
automatically restart (by calling `signal.siginterrupt(signum,
1364
False)`). May be ignored if the feature is not available on this
1365
platform or Python version.
1369
siginterrupt = signal.siginterrupt
1371
# This python implementation doesn't provide signal support, hence no
1374
except AttributeError:
1375
# siginterrupt doesn't exist on this platform, or for this version
1377
siginterrupt = lambda signum, flag: None
1379
def sig_handler(*args):
1380
# Python resets the siginterrupt flag when a signal is
1381
# received. <http://bugs.python.org/issue8354>
1382
# As a workaround for some cases, set it back the way we want it.
1383
siginterrupt(signum, False)
1384
# Now run the handler function passed to set_signal_handler.
1387
sig_handler = handler
1388
old_handler = signal.signal(signum, sig_handler)
1390
siginterrupt(signum, False)
1394
default_terminal_width = 80
1395
"""The default terminal width for ttys.
1397
This is defined so that higher levels can share a common fallback value when
1398
terminal_width() returns None.
1402
def terminal_width():
1403
"""Return terminal width.
1405
None is returned if the width can't established precisely.
1408
- if BZR_COLUMNS is set, returns its value
1409
- if there is no controlling terminal, returns None
1410
- if COLUMNS is set, returns its value,
1412
From there, we need to query the OS to get the size of the controlling
1416
- get termios.TIOCGWINSZ
1417
- if an error occurs or a negative value is obtained, returns None
1421
- win32utils.get_console_size() decides,
1422
- returns None on error (provided default value)
1425
# If BZR_COLUMNS is set, take it, user is always right
1427
return int(os.environ['BZR_COLUMNS'])
1428
except (KeyError, ValueError):
1431
isatty = getattr(sys.stdout, 'isatty', None)
1432
if isatty is None or not isatty():
1433
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1436
# If COLUMNS is set, take it, the terminal knows better (even inside a
1437
# given terminal, the application can decide to set COLUMNS to a lower
1438
# value (splitted screen) or a bigger value (scroll bars))
1440
return int(os.environ['COLUMNS'])
1441
except (KeyError, ValueError):
1444
width, height = _terminal_size(None, None)
1446
# Consider invalid values as meaning no width
1452
def _win32_terminal_size(width, height):
1453
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1454
return width, height
1457
def _ioctl_terminal_size(width, height):
1459
import struct, fcntl, termios
1460
s = struct.pack('HHHH', 0, 0, 0, 0)
1461
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1462
height, width = struct.unpack('HHHH', x)[0:2]
1463
except (IOError, AttributeError):
1465
return width, height
1467
_terminal_size = None
1468
"""Returns the terminal size as (width, height).
1470
:param width: Default value for width.
1471
:param height: Default value for height.
1473
This is defined specifically for each OS and query the size of the controlling
1474
terminal. If any error occurs, the provided default values should be returned.
1476
if sys.platform == 'win32':
1477
_terminal_size = _win32_terminal_size
1479
_terminal_size = _ioctl_terminal_size
1482
def _terminal_size_changed(signum, frame):
1483
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1484
width, height = _terminal_size(None, None)
1485
if width is not None:
1486
os.environ['COLUMNS'] = str(width)
1489
_registered_sigwinch = False
1490
def watch_sigwinch():
1491
"""Register for SIGWINCH, once and only once.
1493
Do nothing if the signal module is not available.
1495
global _registered_sigwinch
1496
if not _registered_sigwinch:
1499
if getattr(signal, "SIGWINCH", None) is not None:
1500
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1502
# python doesn't provide signal support, nothing we can do about it
1504
_registered_sigwinch = True
1507
def supports_executable():
1508
return sys.platform != "win32"
1511
def supports_posix_readonly():
1512
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1514
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1515
directory controls creation/deletion, etc.
1517
And under win32, readonly means that the directory itself cannot be
1518
deleted. The contents of a readonly directory can be changed, unlike POSIX
1519
where files in readonly directories cannot be added, deleted or renamed.
1521
return sys.platform != "win32"
1524
def set_or_unset_env(env_variable, value):
1525
"""Modify the environment, setting or removing the env_variable.
1527
:param env_variable: The environment variable in question
1528
:param value: The value to set the environment to. If None, then
1529
the variable will be removed.
1530
:return: The original value of the environment variable.
1532
orig_val = os.environ.get(env_variable)
1534
if orig_val is not None:
1535
del os.environ[env_variable]
1537
if isinstance(value, unicode):
1538
value = value.encode(get_user_encoding())
1539
os.environ[env_variable] = value
1543
_validWin32PathRE = re.compile(r'^([A-Za-z]:[/\\])?[^:<>*"?\|]*$')
1546
def check_legal_path(path):
1547
"""Check whether the supplied path is legal.
1548
This is only required on Windows, so we don't test on other platforms
1551
if sys.platform != "win32":
1553
if _validWin32PathRE.match(path) is None:
1554
raise errors.IllegalPath(path)
1557
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1559
def _is_error_enotdir(e):
1560
"""Check if this exception represents ENOTDIR.
1562
Unfortunately, python is very inconsistent about the exception
1563
here. The cases are:
1564
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1565
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1566
which is the windows error code.
1567
3) Windows, Python2.5 uses errno == EINVAL and
1568
winerror == ERROR_DIRECTORY
1570
:param e: An Exception object (expected to be OSError with an errno
1571
attribute, but we should be able to cope with anything)
1572
:return: True if this represents an ENOTDIR error. False otherwise.
1574
en = getattr(e, 'errno', None)
1575
if (en == errno.ENOTDIR
1576
or (sys.platform == 'win32'
1577
and (en == _WIN32_ERROR_DIRECTORY
1578
or (en == errno.EINVAL
1579
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
1585
def walkdirs(top, prefix=""):
1586
"""Yield data about all the directories in a tree.
1588
This yields all the data about the contents of a directory at a time.
1589
After each directory has been yielded, if the caller has mutated the list
1590
to exclude some directories, they are then not descended into.
1592
The data yielded is of the form:
1593
((directory-relpath, directory-path-from-top),
1594
[(relpath, basename, kind, lstat, path-from-top), ...]),
1595
- directory-relpath is the relative path of the directory being returned
1596
with respect to top. prefix is prepended to this.
1597
- directory-path-from-root is the path including top for this directory.
1598
It is suitable for use with os functions.
1599
- relpath is the relative path within the subtree being walked.
1600
- basename is the basename of the path
1601
- kind is the kind of the file now. If unknown then the file is not
1602
present within the tree - but it may be recorded as versioned. See
1604
- lstat is the stat data *if* the file was statted.
1605
- planned, not implemented:
1606
path_from_tree_root is the path from the root of the tree.
1608
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1609
allows one to walk a subtree but get paths that are relative to a tree
1611
:return: an iterator over the dirs.
1613
#TODO there is a bit of a smell where the results of the directory-
1614
# summary in this, and the path from the root, may not agree
1615
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1616
# potentially confusing output. We should make this more robust - but
1617
# not at a speed cost. RBC 20060731
1619
_directory = _directory_kind
1620
_listdir = os.listdir
1621
_kind_from_mode = file_kind_from_stat_mode
1622
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
1624
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1625
relroot, _, _, _, top = pending.pop()
1627
relprefix = relroot + u'/'
1630
top_slash = top + u'/'
1633
append = dirblock.append
1635
names = sorted(_listdir(top))
1637
if not _is_error_enotdir(e):
1641
abspath = top_slash + name
1642
statvalue = _lstat(abspath)
1643
kind = _kind_from_mode(statvalue.st_mode)
1644
append((relprefix + name, name, kind, statvalue, abspath))
1645
yield (relroot, top), dirblock
1647
# push the user specified dirs from dirblock
1648
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1651
class DirReader(object):
1652
"""An interface for reading directories."""
1654
def top_prefix_to_starting_dir(self, top, prefix=""):
1655
"""Converts top and prefix to a starting dir entry
1657
:param top: A utf8 path
1658
:param prefix: An optional utf8 path to prefix output relative paths
1660
:return: A tuple starting with prefix, and ending with the native
1663
raise NotImplementedError(self.top_prefix_to_starting_dir)
1665
def read_dir(self, prefix, top):
1666
"""Read a specific dir.
1668
:param prefix: A utf8 prefix to be preprended to the path basenames.
1669
:param top: A natively encoded path to read.
1670
:return: A list of the directories contents. Each item contains:
1671
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1673
raise NotImplementedError(self.read_dir)
1676
_selected_dir_reader = None
1679
def _walkdirs_utf8(top, prefix=""):
1680
"""Yield data about all the directories in a tree.
1682
This yields the same information as walkdirs() only each entry is yielded
1683
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1684
are returned as exact byte-strings.
1686
:return: yields a tuple of (dir_info, [file_info])
1687
dir_info is (utf8_relpath, path-from-top)
1688
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1689
if top is an absolute path, path-from-top is also an absolute path.
1690
path-from-top might be unicode or utf8, but it is the correct path to
1691
pass to os functions to affect the file in question. (such as os.lstat)
1693
global _selected_dir_reader
1694
if _selected_dir_reader is None:
1695
fs_encoding = _fs_enc.upper()
1696
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1697
# Win98 doesn't have unicode apis like FindFirstFileW
1698
# TODO: We possibly could support Win98 by falling back to the
1699
# original FindFirstFile, and using TCHAR instead of WCHAR,
1700
# but that gets a bit tricky, and requires custom compiling
1703
from bzrlib._walkdirs_win32 import Win32ReadDir
1704
_selected_dir_reader = Win32ReadDir()
1707
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1708
# ANSI_X3.4-1968 is a form of ASCII
1710
from bzrlib._readdir_pyx import UTF8DirReader
1711
_selected_dir_reader = UTF8DirReader()
1712
except ImportError, e:
1713
failed_to_load_extension(e)
1716
if _selected_dir_reader is None:
1717
# Fallback to the python version
1718
_selected_dir_reader = UnicodeDirReader()
1720
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1721
# But we don't actually uses 1-3 in pending, so set them to None
1722
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1723
read_dir = _selected_dir_reader.read_dir
1724
_directory = _directory_kind
1726
relroot, _, _, _, top = pending[-1].pop()
1729
dirblock = sorted(read_dir(relroot, top))
1730
yield (relroot, top), dirblock
1731
# push the user specified dirs from dirblock
1732
next = [d for d in reversed(dirblock) if d[2] == _directory]
1734
pending.append(next)
1737
class UnicodeDirReader(DirReader):
1738
"""A dir reader for non-utf8 file systems, which transcodes."""
1740
__slots__ = ['_utf8_encode']
1743
self._utf8_encode = codecs.getencoder('utf8')
1745
def top_prefix_to_starting_dir(self, top, prefix=""):
1746
"""See DirReader.top_prefix_to_starting_dir."""
1747
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1749
def read_dir(self, prefix, top):
1750
"""Read a single directory from a non-utf8 file system.
1752
top, and the abspath element in the output are unicode, all other paths
1753
are utf8. Local disk IO is done via unicode calls to listdir etc.
1755
This is currently the fallback code path when the filesystem encoding is
1756
not UTF-8. It may be better to implement an alternative so that we can
1757
safely handle paths that are not properly decodable in the current
1760
See DirReader.read_dir for details.
1762
_utf8_encode = self._utf8_encode
1764
_listdir = os.listdir
1765
_kind_from_mode = file_kind_from_stat_mode
1768
relprefix = prefix + '/'
1771
top_slash = top + u'/'
1774
append = dirblock.append
1775
for name in sorted(_listdir(top)):
1777
name_utf8 = _utf8_encode(name)[0]
1778
except UnicodeDecodeError:
1779
raise errors.BadFilenameEncoding(
1780
_utf8_encode(relprefix)[0] + name, _fs_enc)
1781
abspath = top_slash + name
1782
statvalue = _lstat(abspath)
1783
kind = _kind_from_mode(statvalue.st_mode)
1784
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
1788
def copy_tree(from_path, to_path, handlers={}):
1789
"""Copy all of the entries in from_path into to_path.
1791
:param from_path: The base directory to copy.
1792
:param to_path: The target directory. If it does not exist, it will
1794
:param handlers: A dictionary of functions, which takes a source and
1795
destinations for files, directories, etc.
1796
It is keyed on the file kind, such as 'directory', 'symlink', or 'file'
1797
'file', 'directory', and 'symlink' should always exist.
1798
If they are missing, they will be replaced with 'os.mkdir()',
1799
'os.readlink() + os.symlink()', and 'shutil.copy2()', respectively.
1801
# Now, just copy the existing cached tree to the new location
1802
# We use a cheap trick here.
1803
# Absolute paths are prefixed with the first parameter
1804
# relative paths are prefixed with the second.
1805
# So we can get both the source and target returned
1806
# without any extra work.
1808
def copy_dir(source, dest):
1811
def copy_link(source, dest):
1812
"""Copy the contents of a symlink"""
1813
link_to = os.readlink(source)
1814
os.symlink(link_to, dest)
1816
real_handlers = {'file':shutil.copy2,
1817
'symlink':copy_link,
1818
'directory':copy_dir,
1820
real_handlers.update(handlers)
1822
if not os.path.exists(to_path):
1823
real_handlers['directory'](from_path, to_path)
1825
for dir_info, entries in walkdirs(from_path, prefix=to_path):
1826
for relpath, name, kind, st, abspath in entries:
1827
real_handlers[kind](abspath, relpath)
1830
def copy_ownership_from_path(dst, src=None):
1831
"""Copy usr/grp ownership from src file/dir to dst file/dir.
1833
If src is None, the containing directory is used as source. If chown
1834
fails, the error is ignored and a warning is printed.
1836
chown = getattr(os, 'chown', None)
1841
src = os.path.dirname(dst)
1847
chown(dst, s.st_uid, s.st_gid)
1849
trace.warning("Unable to copy ownership from '%s' to '%s': IOError: %s." % (src, dst, e))
1852
def path_prefix_key(path):
1853
"""Generate a prefix-order path key for path.
1855
This can be used to sort paths in the same way that walkdirs does.
1857
return (dirname(path) , path)
1860
def compare_paths_prefix_order(path_a, path_b):
1861
"""Compare path_a and path_b to generate the same order walkdirs uses."""
1862
key_a = path_prefix_key(path_a)
1863
key_b = path_prefix_key(path_b)
1864
return cmp(key_a, key_b)
1867
_cached_user_encoding = None
1870
def get_user_encoding(use_cache=True):
1871
"""Find out what the preferred user encoding is.
1873
This is generally the encoding that is used for command line parameters
1874
and file contents. This may be different from the terminal encoding
1875
or the filesystem encoding.
1877
:param use_cache: Enable cache for detected encoding.
1878
(This parameter is turned on by default,
1879
and required only for selftesting)
1881
:return: A string defining the preferred user encoding
1883
global _cached_user_encoding
1884
if _cached_user_encoding is not None and use_cache:
1885
return _cached_user_encoding
1887
if sys.platform == 'darwin':
1888
# python locale.getpreferredencoding() always return
1889
# 'mac-roman' on darwin. That's a lie.
1890
sys.platform = 'posix'
1892
if os.environ.get('LANG', None) is None:
1893
# If LANG is not set, we end up with 'ascii', which is bad
1894
# ('mac-roman' is more than ascii), so we set a default which
1895
# will give us UTF-8 (which appears to work in all cases on
1896
# OSX). Users are still free to override LANG of course, as
1897
# long as it give us something meaningful. This work-around
1898
# *may* not be needed with python 3k and/or OSX 10.5, but will
1899
# work with them too -- vila 20080908
1900
os.environ['LANG'] = 'en_US.UTF-8'
1903
sys.platform = 'darwin'
1908
user_encoding = locale.getpreferredencoding()
1909
except locale.Error, e:
1910
sys.stderr.write('bzr: warning: %s\n'
1911
' Could not determine what text encoding to use.\n'
1912
' This error usually means your Python interpreter\n'
1913
' doesn\'t support the locale set by $LANG (%s)\n'
1914
" Continuing with ascii encoding.\n"
1915
% (e, os.environ.get('LANG')))
1916
user_encoding = 'ascii'
1918
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1919
# treat that as ASCII, and not support printing unicode characters to the
1922
# For python scripts run under vim, we get '', so also treat that as ASCII
1923
if user_encoding in (None, 'cp0', ''):
1924
user_encoding = 'ascii'
1928
codecs.lookup(user_encoding)
1930
sys.stderr.write('bzr: warning:'
1931
' unknown encoding %s.'
1932
' Continuing with ascii encoding.\n'
1935
user_encoding = 'ascii'
1938
_cached_user_encoding = user_encoding
1940
return user_encoding
1943
def get_host_name():
1944
"""Return the current unicode host name.
1946
This is meant to be used in place of socket.gethostname() because that
1947
behaves inconsistently on different platforms.
1949
if sys.platform == "win32":
1951
return win32utils.get_host_name()
1954
return socket.gethostname().decode(get_user_encoding())
1957
# We must not read/write any more than 64k at a time from/to a socket so we
1958
# don't risk "no buffer space available" errors on some platforms. Windows in
1959
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1961
MAX_SOCKET_CHUNK = 64 * 1024
1963
def read_bytes_from_socket(sock, report_activity=None,
1964
max_read_size=MAX_SOCKET_CHUNK):
1965
"""Read up to max_read_size of bytes from sock and notify of progress.
1967
Translates "Connection reset by peer" into file-like EOF (return an
1968
empty string rather than raise an error), and repeats the recv if
1969
interrupted by a signal.
1973
bytes = sock.recv(max_read_size)
1974
except socket.error, e:
1976
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1977
# The connection was closed by the other side. Callers expect
1978
# an empty string to signal end-of-stream.
1980
elif eno == errno.EINTR:
1981
# Retry the interrupted recv.
1985
if report_activity is not None:
1986
report_activity(len(bytes), 'read')
1990
def recv_all(socket, count):
1991
"""Receive an exact number of bytes.
1993
Regular Socket.recv() may return less than the requested number of bytes,
1994
depending on what's in the OS buffer. MSG_WAITALL is not available
1995
on all platforms, but this should work everywhere. This will return
1996
less than the requested amount if the remote end closes.
1998
This isn't optimized and is intended mostly for use in testing.
2001
while len(b) < count:
2002
new = read_bytes_from_socket(socket, None, count - len(b))
2009
def send_all(sock, bytes, report_activity=None):
2010
"""Send all bytes on a socket.
2012
Breaks large blocks in smaller chunks to avoid buffering limitations on
2013
some platforms, and catches EINTR which may be thrown if the send is
2014
interrupted by a signal.
2016
This is preferred to socket.sendall(), because it avoids portability bugs
2017
and provides activity reporting.
2019
:param report_activity: Call this as bytes are read, see
2020
Transport._report_activity
2023
byte_count = len(bytes)
2024
while sent_total < byte_count:
2026
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2027
except socket.error, e:
2028
if e.args[0] != errno.EINTR:
2032
report_activity(sent, 'write')
2035
def dereference_path(path):
2036
"""Determine the real path to a file.
2038
All parent elements are dereferenced. But the file itself is not
2040
:param path: The original path. May be absolute or relative.
2041
:return: the real path *to* the file
2043
parent, base = os.path.split(path)
2044
# The pathjoin for '.' is a workaround for Python bug #1213894.
2045
# (initial path components aren't dereferenced)
2046
return pathjoin(realpath(pathjoin('.', parent)), base)
2049
def supports_mapi():
2050
"""Return True if we can use MAPI to launch a mail client."""
2051
return sys.platform == "win32"
2054
def resource_string(package, resource_name):
2055
"""Load a resource from a package and return it as a string.
2057
Note: Only packages that start with bzrlib are currently supported.
2059
This is designed to be a lightweight implementation of resource
2060
loading in a way which is API compatible with the same API from
2062
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2063
If and when pkg_resources becomes a standard library, this routine
2066
# Check package name is within bzrlib
2067
if package == "bzrlib":
2068
resource_relpath = resource_name
2069
elif package.startswith("bzrlib."):
2070
package = package[len("bzrlib."):].replace('.', os.sep)
2071
resource_relpath = pathjoin(package, resource_name)
2073
raise errors.BzrError('resource package %s not in bzrlib' % package)
2075
# Map the resource to a file and read its contents
2076
base = dirname(bzrlib.__file__)
2077
if getattr(sys, 'frozen', None): # bzr.exe
2078
base = abspath(pathjoin(base, '..', '..'))
2079
filename = pathjoin(base, resource_relpath)
2080
return open(filename, 'rU').read()
2083
def file_kind_from_stat_mode_thunk(mode):
2084
global file_kind_from_stat_mode
2085
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2087
from bzrlib._readdir_pyx import UTF8DirReader
2088
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2089
except ImportError, e:
2090
# This is one time where we won't warn that an extension failed to
2091
# load. The extension is never available on Windows anyway.
2092
from bzrlib._readdir_py import (
2093
_kind_from_mode as file_kind_from_stat_mode
2095
return file_kind_from_stat_mode(mode)
2096
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2099
def file_kind(f, _lstat=os.lstat):
2101
return file_kind_from_stat_mode(_lstat(f).st_mode)
2103
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2104
raise errors.NoSuchFile(f)
2108
def until_no_eintr(f, *a, **kw):
2109
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2111
WARNING: you must be certain that it is safe to retry the call repeatedly
2112
if EINTR does occur. This is typically only true for low-level operations
2113
like os.read. If in any doubt, don't use this.
2115
Keep in mind that this is not a complete solution to EINTR. There is
2116
probably code in the Python standard library and other dependencies that
2117
may encounter EINTR if a signal arrives (and there is signal handler for
2118
that signal). So this function can reduce the impact for IO that bzrlib
2119
directly controls, but it is not a complete solution.
2121
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2125
except (IOError, OSError), e:
2126
if e.errno == errno.EINTR:
2131
def re_compile_checked(re_string, flags=0, where=""):
2132
"""Return a compiled re, or raise a sensible error.
2134
This should only be used when compiling user-supplied REs.
2136
:param re_string: Text form of regular expression.
2137
:param flags: eg re.IGNORECASE
2138
:param where: Message explaining to the user the context where
2139
it occurred, eg 'log search filter'.
2141
# from https://bugs.launchpad.net/bzr/+bug/251352
2143
re_obj = re.compile(re_string, flags)
2148
where = ' in ' + where
2149
# despite the name 'error' is a type
2150
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2151
% (where, re_string, e))
2154
if sys.platform == "win32":
2157
return msvcrt.getch()
2162
fd = sys.stdin.fileno()
2163
settings = termios.tcgetattr(fd)
2166
ch = sys.stdin.read(1)
2168
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2172
if sys.platform == 'linux2':
2173
def _local_concurrency():
2175
prefix = 'processor'
2176
for line in file('/proc/cpuinfo', 'rb'):
2177
if line.startswith(prefix):
2178
concurrency = int(line[line.find(':')+1:]) + 1
2180
elif sys.platform == 'darwin':
2181
def _local_concurrency():
2182
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2183
stdout=subprocess.PIPE).communicate()[0]
2184
elif sys.platform[0:7] == 'freebsd':
2185
def _local_concurrency():
2186
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2187
stdout=subprocess.PIPE).communicate()[0]
2188
elif sys.platform == 'sunos5':
2189
def _local_concurrency():
2190
return subprocess.Popen(['psrinfo', '-p',],
2191
stdout=subprocess.PIPE).communicate()[0]
2192
elif sys.platform == "win32":
2193
def _local_concurrency():
2194
# This appears to return the number of cores.
2195
return os.environ.get('NUMBER_OF_PROCESSORS')
2197
def _local_concurrency():
2202
_cached_local_concurrency = None
2204
def local_concurrency(use_cache=True):
2205
"""Return how many processes can be run concurrently.
2207
Rely on platform specific implementations and default to 1 (one) if
2208
anything goes wrong.
2210
global _cached_local_concurrency
2212
if _cached_local_concurrency is not None and use_cache:
2213
return _cached_local_concurrency
2215
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2216
if concurrency is None:
2218
concurrency = _local_concurrency()
2219
except (OSError, IOError):
2222
concurrency = int(concurrency)
2223
except (TypeError, ValueError):
2226
_cached_concurrency = concurrency
2230
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2231
"""A stream writer that doesn't decode str arguments."""
2233
def __init__(self, encode, stream, errors='strict'):
2234
codecs.StreamWriter.__init__(self, stream, errors)
2235
self.encode = encode
2237
def write(self, object):
2238
if type(object) is str:
2239
self.stream.write(object)
2241
data, _ = self.encode(object, self.errors)
2242
self.stream.write(data)
2244
if sys.platform == 'win32':
2245
def open_file(filename, mode='r', bufsize=-1):
2246
"""This function is used to override the ``open`` builtin.
2248
But it uses O_NOINHERIT flag so the file handle is not inherited by
2249
child processes. Deleting or renaming a closed file opened with this
2250
function is not blocking child processes.
2252
writing = 'w' in mode
2253
appending = 'a' in mode
2254
updating = '+' in mode
2255
binary = 'b' in mode
2258
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2259
# for flags for each modes.
2269
flags |= os.O_WRONLY
2270
flags |= os.O_CREAT | os.O_TRUNC
2275
flags |= os.O_WRONLY
2276
flags |= os.O_CREAT | os.O_APPEND
2281
flags |= os.O_RDONLY
2283
return os.fdopen(os.open(filename, flags), mode, bufsize)
added
removed
bzrlib/osutils.py
