172
_directory_kind: "/",
174
'tree-reference': '+',
139
178
def kind_marker(kind):
142
elif kind == _directory_kind:
144
elif kind == 'symlink':
147
raise BzrError('invalid file kind %r' % kind)
180
return _kind_marker_map[kind]
182
raise errors.BzrError('invalid file kind %r' % kind)
149
185
lexists = getattr(os.path, 'lexists', None)
150
186
if lexists is None:
153
if getattr(os, 'lstat') is not None:
189
stat = getattr(os, 'lstat', os.stat)
159
193
if e.errno == errno.ENOENT:
162
raise BzrError("lstat/stat of (%r): %r" % (f, e))
196
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
165
199
def fancy_rename(old, new, rename_func, unlink_func):
166
200
"""A fancy rename, when you don't have atomic rename.
168
202
:param old: The old path, to rename from
169
203
:param new: The new path, to rename to
170
204
:param rename_func: The potentially non-atomic rename function
171
:param unlink_func: A way to delete the target file if the full rename succeeds
205
:param unlink_func: A way to delete the target file if the full rename
174
208
# sftp rename doesn't allow overwriting, so play tricks:
176
209
base = os.path.basename(new)
177
210
dirname = os.path.dirname(new)
178
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
211
# callers use different encodings for the paths so the following MUST
212
# respect that. We rely on python upcasting to unicode if new is unicode
213
# and keeping a str if not.
214
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
215
os.getpid(), rand_chars(10))
179
216
tmp_name = pathjoin(dirname, tmp_name)
181
218
# Rename the file out of the way, but keep track if it didn't exist
482
548
for dirname in dir_list:
483
549
if is_inside(dirname, fname) or is_inside(fname, dirname):
554
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
555
report_activity=None, direction='read'):
556
"""Copy contents of one file to another.
558
The read_length can either be -1 to read to end-of-file (EOF) or
559
it can specify the maximum number of bytes to read.
561
The buff_size represents the maximum size for each read operation
562
performed on from_file.
564
:param report_activity: Call this as bytes are read, see
565
Transport._report_activity
566
:param direction: Will be passed to report_activity
568
:return: The number of bytes copied.
572
# read specified number of bytes
574
while read_length > 0:
575
num_bytes_to_read = min(read_length, buff_size)
577
block = from_file.read(num_bytes_to_read)
581
if report_activity is not None:
582
report_activity(len(block), direction)
585
actual_bytes_read = len(block)
586
read_length -= actual_bytes_read
587
length += actual_bytes_read
489
def pumpfile(fromfile, tofile):
490
"""Copy contents of one file to another."""
493
b = fromfile.read(BUFSIZE)
591
block = from_file.read(buff_size)
595
if report_activity is not None:
596
report_activity(len(block), direction)
602
def pump_string_file(bytes, file_handle, segment_size=None):
603
"""Write bytes to file_handle in many smaller writes.
605
:param bytes: The string to write.
606
:param file_handle: The file to write to.
608
# Write data in chunks rather than all at once, because very large
609
# writes fail on some platforms (e.g. Windows with SMB mounted
612
segment_size = 5242880 # 5MB
613
segments = range(len(bytes) / segment_size + 1)
614
write = file_handle.write
615
for segment_index in segments:
616
segment = buffer(bytes, segment_index * segment_size, segment_size)
499
620
def file_iterator(input_file, readsize=32768):
555
704
def local_time_offset(t=None):
556
705
"""Return offset of local zone from GMT, either at present or at time t."""
557
# python2.3 localtime() can't take None
561
if time.localtime(t).tm_isdst and time.daylight:
564
return -time.timezone
567
def format_date(t, offset=0, timezone='original', date_fmt=None,
708
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
709
return offset.days * 86400 + offset.seconds
711
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
712
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
715
def format_date(t, offset=0, timezone='original', date_fmt=None,
568
716
show_offset=True):
569
## TODO: Perhaps a global option to use either universal or local time?
570
## Or perhaps just let people set $TZ?
571
assert isinstance(t, float)
717
"""Return a formatted date string.
719
:param t: Seconds since the epoch.
720
:param offset: Timezone offset in seconds east of utc.
721
:param timezone: How to display the time: 'utc', 'original' for the
722
timezone specified by offset, or 'local' for the process's current
724
:param date_fmt: strftime format.
725
:param show_offset: Whether to append the timezone.
727
(date_fmt, tt, offset_str) = \
728
_format_date(t, offset, timezone, date_fmt, show_offset)
729
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
730
date_str = time.strftime(date_fmt, tt)
731
return date_str + offset_str
734
# Cache of formatted offset strings
738
def format_date_with_offset_in_original_timezone(t, offset=0,
739
_cache=_offset_cache):
740
"""Return a formatted date string in the original timezone.
742
This routine may be faster then format_date.
744
:param t: Seconds since the epoch.
745
:param offset: Timezone offset in seconds east of utc.
749
tt = time.gmtime(t + offset)
750
date_fmt = _default_format_by_weekday_num[tt[6]]
751
date_str = time.strftime(date_fmt, tt)
752
offset_str = _cache.get(offset, None)
753
if offset_str is None:
754
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
755
_cache[offset] = offset_str
756
return date_str + offset_str
759
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
761
"""Return an unicode date string formatted according to the current locale.
763
:param t: Seconds since the epoch.
764
:param offset: Timezone offset in seconds east of utc.
765
:param timezone: How to display the time: 'utc', 'original' for the
766
timezone specified by offset, or 'local' for the process's current
768
:param date_fmt: strftime format.
769
:param show_offset: Whether to append the timezone.
771
(date_fmt, tt, offset_str) = \
772
_format_date(t, offset, timezone, date_fmt, show_offset)
773
date_str = time.strftime(date_fmt, tt)
774
if not isinstance(date_str, unicode):
775
date_str = date_str.decode(get_user_encoding(), 'replace')
776
return date_str + offset_str
779
def _format_date(t, offset, timezone, date_fmt, show_offset):
573
780
if timezone == 'utc':
574
781
tt = time.gmtime(t)
581
788
tt = time.localtime(t)
582
789
offset = local_time_offset(t)
584
raise BzrError("unsupported timezone format %r" % timezone,
585
['options are "utc", "original", "local"'])
791
raise errors.UnsupportedTimezoneFormat(timezone)
586
792
if date_fmt is None:
587
793
date_fmt = "%a %Y-%m-%d %H:%M:%S"
589
795
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
592
return (time.strftime(date_fmt, tt) + offset_str)
798
return (date_fmt, tt, offset_str)
595
801
def compact_date(when):
596
802
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
805
def format_delta(delta):
806
"""Get a nice looking string for a time delta.
808
:param delta: The time difference in seconds, can be positive or negative.
809
positive indicates time in the past, negative indicates time in the
810
future. (usually time.time() - stored_time)
811
:return: String formatted to show approximate resolution
817
direction = 'in the future'
821
if seconds < 90: # print seconds up to 90 seconds
823
return '%d second %s' % (seconds, direction,)
825
return '%d seconds %s' % (seconds, direction)
827
minutes = int(seconds / 60)
828
seconds -= 60 * minutes
833
if minutes < 90: # print minutes, seconds up to 90 minutes
835
return '%d minute, %d second%s %s' % (
836
minutes, seconds, plural_seconds, direction)
838
return '%d minutes, %d second%s %s' % (
839
minutes, seconds, plural_seconds, direction)
841
hours = int(minutes / 60)
842
minutes -= 60 * hours
849
return '%d hour, %d minute%s %s' % (hours, minutes,
850
plural_minutes, direction)
851
return '%d hours, %d minute%s %s' % (hours, minutes,
852
plural_minutes, direction)
601
855
"""Return size of given open file."""
669
raise BzrError("sorry, %r not allowed in path" % f)
907
raise errors.BzrError("sorry, %r not allowed in path" % f)
670
908
elif (f == '.') or (f == ''):
677
assert isinstance(p, list)
679
917
if (f == '..') or (f is None) or (f == ''):
680
raise BzrError("sorry, %r not allowed in path" % f)
918
raise errors.BzrError("sorry, %r not allowed in path" % f)
681
919
return pathjoin(*p)
684
@deprecated_function(zero_nine)
685
def appendpath(p1, p2):
689
return pathjoin(p1, p2)
922
def parent_directories(filename):
923
"""Return the list of parent directories, deepest first.
925
For example, parent_directories("a/b/c") -> ["a/b", "a"].
928
parts = splitpath(dirname(filename))
930
parents.append(joinpath(parts))
935
_extension_load_failures = []
938
def failed_to_load_extension(exception):
939
"""Handle failing to load a binary extension.
941
This should be called from the ImportError block guarding the attempt to
942
import the native extension. If this function returns, the pure-Python
943
implementation should be loaded instead::
946
>>> import bzrlib._fictional_extension_pyx
947
>>> except ImportError, e:
948
>>> bzrlib.osutils.failed_to_load_extension(e)
949
>>> import bzrlib._fictional_extension_py
951
# NB: This docstring is just an example, not a doctest, because doctest
952
# currently can't cope with the use of lazy imports in this namespace --
955
# This currently doesn't report the failure at the time it occurs, because
956
# they tend to happen very early in startup when we can't check config
957
# files etc, and also we want to report all failures but not spam the user
959
from bzrlib import trace
960
exception_str = str(exception)
961
if exception_str not in _extension_load_failures:
962
trace.mutter("failed to load compiled extension: %s" % exception_str)
963
_extension_load_failures.append(exception_str)
966
def report_extension_load_failures():
967
if not _extension_load_failures:
969
from bzrlib.config import GlobalConfig
970
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
972
# the warnings framework should by default show this only once
973
from bzrlib.trace import warning
975
"bzr: warning: some compiled extensions could not be loaded; "
976
"see <https://answers.launchpad.net/bzr/+faq/703>")
977
# we no longer show the specific missing extensions here, because it makes
978
# the message too long and scary - see
979
# https://bugs.launchpad.net/bzr/+bug/430529
983
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
984
except ImportError, e:
985
failed_to_load_extension(e)
986
from bzrlib._chunks_to_lines_py import chunks_to_lines
692
989
def split_lines(s):
693
990
"""Split s into lines, but without removing the newline characters."""
991
# Trivially convert a fulltext into a 'chunked' representation, and let
992
# chunks_to_lines do the heavy lifting.
993
if isinstance(s, str):
994
# chunks_to_lines only supports 8-bit strings
995
return chunks_to_lines([s])
997
return _split_lines(s)
1000
def _split_lines(s):
1001
"""Split s into lines, but without removing the newline characters.
1003
This supports Unicode or plain string objects.
694
1005
lines = s.split('\n')
695
1006
result = [line + '\n' for line in lines[:-1]]
761
1130
avoids that problem.
764
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
765
' exceed the platform minimum length (which is %d)' %
1133
if len(base) < MIN_ABS_PATHLENGTH:
1134
# must have space for e.g. a drive letter
1135
raise ValueError('%r is too short to calculate a relative path'
768
1138
rp = abspath(path)
772
while len(head) >= len(base):
1143
if len(head) <= len(base) and head != base:
1144
raise errors.PathNotChild(rp, base)
773
1145
if head == base:
775
head, tail = os.path.split(head)
1147
head, tail = split(head)
779
raise PathNotChild(rp, base)
1152
return pathjoin(*reversed(s))
1157
def _cicp_canonical_relpath(base, path):
1158
"""Return the canonical path relative to base.
1160
Like relpath, but on case-insensitive-case-preserving file-systems, this
1161
will return the relpath as stored on the file-system rather than in the
1162
case specified in the input string, for all existing portions of the path.
1164
This will cause O(N) behaviour if called for every path in a tree; if you
1165
have a number of paths to convert, you should use canonical_relpaths().
1167
# TODO: it should be possible to optimize this for Windows by using the
1168
# win32 API FindFiles function to look for the specified name - but using
1169
# os.listdir() still gives us the correct, platform agnostic semantics in
1172
rel = relpath(base, path)
1173
# '.' will have been turned into ''
1177
abs_base = abspath(base)
1179
_listdir = os.listdir
1181
# use an explicit iterator so we can easily consume the rest on early exit.
1182
bit_iter = iter(rel.split('/'))
1183
for bit in bit_iter:
1186
next_entries = _listdir(current)
1187
except OSError: # enoent, eperm, etc
1188
# We can't find this in the filesystem, so just append the
1190
current = pathjoin(current, bit, *list(bit_iter))
1192
for look in next_entries:
1193
if lbit == look.lower():
1194
current = pathjoin(current, look)
1197
# got to the end, nothing matched, so we just return the
1198
# non-existing bits as they were specified (the filename may be
1199
# the target of a move, for example).
1200
current = pathjoin(current, bit, *list(bit_iter))
1202
return current[len(abs_base):].lstrip('/')
1204
# XXX - TODO - we need better detection/integration of case-insensitive
1205
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1206
# filesystems), for example, so could probably benefit from the same basic
1207
# support there. For now though, only Windows and OSX get that support, and
1208
# they get it for *all* file-systems!
1209
if sys.platform in ('win32', 'darwin'):
1210
canonical_relpath = _cicp_canonical_relpath
1212
canonical_relpath = relpath
1214
def canonical_relpaths(base, paths):
1215
"""Create an iterable to canonicalize a sequence of relative paths.
1217
The intent is for this implementation to use a cache, vastly speeding
1218
up multiple transformations in the same directory.
1220
# but for now, we haven't optimized...
1221
return [canonical_relpath(base, p) for p in paths]
787
1223
def safe_unicode(unicode_or_utf8_string):
788
1224
"""Coerce unicode_or_utf8_string into unicode.
790
1226
If it is unicode, it is returned.
791
Otherwise it is decoded from utf-8. If a decoding error
792
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
793
as a BzrBadParameter exception.
1227
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1228
wrapped in a BzrBadParameterNotUnicode exception.
795
1230
if isinstance(unicode_or_utf8_string, unicode):
796
1231
return unicode_or_utf8_string
798
1233
return unicode_or_utf8_string.decode('utf8')
799
1234
except UnicodeDecodeError:
800
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
1235
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1238
def safe_utf8(unicode_or_utf8_string):
1239
"""Coerce unicode_or_utf8_string to a utf8 string.
1241
If it is a str, it is returned.
1242
If it is Unicode, it is encoded into a utf-8 string.
1244
if isinstance(unicode_or_utf8_string, str):
1245
# TODO: jam 20070209 This is overkill, and probably has an impact on
1246
# performance if we are dealing with lots of apis that want a
1249
# Make sure it is a valid utf-8 string
1250
unicode_or_utf8_string.decode('utf-8')
1251
except UnicodeDecodeError:
1252
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1253
return unicode_or_utf8_string
1254
return unicode_or_utf8_string.encode('utf-8')
1257
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1258
' Revision id generators should be creating utf8'
1262
def safe_revision_id(unicode_or_utf8_string, warn=True):
1263
"""Revision ids should now be utf8, but at one point they were unicode.
1265
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1267
:param warn: Functions that are sanitizing user data can set warn=False
1268
:return: None or a utf8 revision id.
1270
if (unicode_or_utf8_string is None
1271
or unicode_or_utf8_string.__class__ == str):
1272
return unicode_or_utf8_string
1274
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1276
return cache_utf8.encode(unicode_or_utf8_string)
1279
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1280
' generators should be creating utf8 file ids.')
1283
def safe_file_id(unicode_or_utf8_string, warn=True):
1284
"""File ids should now be utf8, but at one point they were unicode.
1286
This is the same as safe_utf8, except it uses the cached encode functions
1287
to save a little bit of performance.
1289
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1291
:param warn: Functions that are sanitizing user data can set warn=False
1292
:return: None or a utf8 file id.
1294
if (unicode_or_utf8_string is None
1295
or unicode_or_utf8_string.__class__ == str):
1296
return unicode_or_utf8_string
1298
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1300
return cache_utf8.encode(unicode_or_utf8_string)
803
1303
_platform_normalizes_filenames = False
844
1344
normalized_filename = _inaccessible_normalized_filename
1347
default_terminal_width = 80
1348
"""The default terminal width for ttys.
1350
This is defined so that higher levels can share a common fallback value when
1351
terminal_width() returns None.
847
1355
def terminal_width():
848
"""Return estimated terminal width."""
849
if sys.platform == 'win32':
850
import bzrlib.win32console
851
return bzrlib.win32console.get_console_size()[0]
1356
"""Return terminal width.
1358
None is returned if the width can't established precisely.
1361
- if BZR_COLUMNS is set, returns its value
1362
- if there is no controlling terminal, returns None
1363
- if COLUMNS is set, returns its value,
1365
From there, we need to query the OS to get the size of the controlling
1369
- get termios.TIOCGWINSZ
1370
- if an error occurs or a negative value is obtained, returns None
1374
- win32utils.get_console_size() decides,
1375
- returns None on error (provided default value)
1378
# If BZR_COLUMNS is set, take it, user is always right
1380
return int(os.environ['BZR_COLUMNS'])
1381
except (KeyError, ValueError):
1384
isatty = getattr(sys.stdout, 'isatty', None)
1385
if isatty is None or not isatty():
1386
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1389
# If COLUMNS is set, take it, the terminal knows better (even inside a
1390
# given terminal, the application can decide to set COLUMNS to a lower
1391
# value (splitted screen) or a bigger value (scroll bars))
1393
return int(os.environ['COLUMNS'])
1394
except (KeyError, ValueError):
1397
width, height = _terminal_size(None, None)
1399
# Consider invalid values as meaning no width
1405
def _win32_terminal_size(width, height):
1406
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1407
return width, height
1410
def _ioctl_terminal_size(width, height):
854
1412
import struct, fcntl, termios
855
1413
s = struct.pack('HHHH', 0, 0, 0, 0)
856
1414
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
857
width = struct.unpack('HHHH', x)[1]
1415
height, width = struct.unpack('HHHH', x)[0:2]
1416
except (IOError, AttributeError):
862
width = int(os.environ['COLUMNS'])
1418
return width, height
1420
_terminal_size = None
1421
"""Returns the terminal size as (width, height).
1423
:param width: Default value for width.
1424
:param height: Default value for height.
1426
This is defined specifically for each OS and query the size of the controlling
1427
terminal. If any error occurs, the provided default values should be returned.
1429
if sys.platform == 'win32':
1430
_terminal_size = _win32_terminal_size
1432
_terminal_size = _ioctl_terminal_size
1435
def _terminal_size_changed(signum, frame):
1436
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1437
width, height = _terminal_size(None, None)
1438
if width is not None:
1439
os.environ['COLUMNS'] = str(width)
1441
if sys.platform == 'win32':
1442
# Martin (gz) mentioned WINDOW_BUFFER_SIZE_RECORD from ReadConsoleInput but
1443
# I've no idea how to plug that in the current design -- vila 20091216
1446
signal.signal(signal.SIGWINCH, _terminal_size_changed)
871
1449
def supports_executable():
872
1450
return sys.platform != "win32"
1453
def supports_posix_readonly():
1454
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1456
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1457
directory controls creation/deletion, etc.
1459
And under win32, readonly means that the directory itself cannot be
1460
deleted. The contents of a readonly directory can be changed, unlike POSIX
1461
where files in readonly directories cannot be added, deleted or renamed.
1463
return sys.platform != "win32"
875
1466
def set_or_unset_env(env_variable, value):
876
1467
"""Modify the environment, setting or removing the env_variable.
897
1488
def check_legal_path(path):
898
"""Check whether the supplied path is legal.
1489
"""Check whether the supplied path is legal.
899
1490
This is only required on Windows, so we don't test on other platforms
902
1493
if sys.platform != "win32":
904
1495
if _validWin32PathRE.match(path) is None:
905
raise IllegalPath(path)
1496
raise errors.IllegalPath(path)
1499
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1501
def _is_error_enotdir(e):
1502
"""Check if this exception represents ENOTDIR.
1504
Unfortunately, python is very inconsistent about the exception
1505
here. The cases are:
1506
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1507
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1508
which is the windows error code.
1509
3) Windows, Python2.5 uses errno == EINVAL and
1510
winerror == ERROR_DIRECTORY
1512
:param e: An Exception object (expected to be OSError with an errno
1513
attribute, but we should be able to cope with anything)
1514
:return: True if this represents an ENOTDIR error. False otherwise.
1516
en = getattr(e, 'errno', None)
1517
if (en == errno.ENOTDIR
1518
or (sys.platform == 'win32'
1519
and (en == _WIN32_ERROR_DIRECTORY
1520
or (en == errno.EINVAL
1521
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
908
1527
def walkdirs(top, prefix=""):
909
1528
"""Yield data about all the directories in a tree.
911
1530
This yields all the data about the contents of a directory at a time.
912
1531
After each directory has been yielded, if the caller has mutated the list
913
1532
to exclude some directories, they are then not descended into.
915
1534
The data yielded is of the form:
916
1535
((directory-relpath, directory-path-from-top),
917
[(relpath, basename, kind, lstat), ...]),
1536
[(relpath, basename, kind, lstat, path-from-top), ...]),
918
1537
- directory-relpath is the relative path of the directory being returned
919
1538
with respect to top. prefix is prepended to this.
920
- directory-path-from-root is the path including top for this directory.
1539
- directory-path-from-root is the path including top for this directory.
921
1540
It is suitable for use with os functions.
922
1541
- relpath is the relative path within the subtree being walked.
923
1542
- basename is the basename of the path
925
1544
present within the tree - but it may be recorded as versioned. See
927
1546
- lstat is the stat data *if* the file was statted.
928
- planned, not implemented:
1547
- planned, not implemented:
929
1548
path_from_tree_root is the path from the root of the tree.
931
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1550
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
932
1551
allows one to walk a subtree but get paths that are relative to a tree
933
1552
rooted higher up.
934
1553
:return: an iterator over the dirs.
936
1555
#TODO there is a bit of a smell where the results of the directory-
937
# summary in this, and the path from the root, may not agree
1556
# summary in this, and the path from the root, may not agree
938
1557
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
939
1558
# potentially confusing output. We should make this more robust - but
940
1559
# not at a speed cost. RBC 20060731
943
1561
_directory = _directory_kind
945
pending = [(prefix, "", _directory, None, top)]
1562
_listdir = os.listdir
1563
_kind_from_mode = file_kind_from_stat_mode
1564
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
948
currentdir = pending.pop()
949
1566
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
952
relroot = currentdir[0] + '/'
1567
relroot, _, _, _, top = pending.pop()
1569
relprefix = relroot + u'/'
1572
top_slash = top + u'/'
1575
append = dirblock.append
1577
names = sorted(_listdir(top))
1579
if not _is_error_enotdir(e):
1583
abspath = top_slash + name
1584
statvalue = _lstat(abspath)
1585
kind = _kind_from_mode(statvalue.st_mode)
1586
append((relprefix + name, name, kind, statvalue, abspath))
1587
yield (relroot, top), dirblock
1589
# push the user specified dirs from dirblock
1590
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1593
class DirReader(object):
1594
"""An interface for reading directories."""
1596
def top_prefix_to_starting_dir(self, top, prefix=""):
1597
"""Converts top and prefix to a starting dir entry
1599
:param top: A utf8 path
1600
:param prefix: An optional utf8 path to prefix output relative paths
1602
:return: A tuple starting with prefix, and ending with the native
1605
raise NotImplementedError(self.top_prefix_to_starting_dir)
1607
def read_dir(self, prefix, top):
1608
"""Read a specific dir.
1610
:param prefix: A utf8 prefix to be preprended to the path basenames.
1611
:param top: A natively encoded path to read.
1612
:return: A list of the directories contents. Each item contains:
1613
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1615
raise NotImplementedError(self.read_dir)
1618
_selected_dir_reader = None
1621
def _walkdirs_utf8(top, prefix=""):
1622
"""Yield data about all the directories in a tree.
1624
This yields the same information as walkdirs() only each entry is yielded
1625
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1626
are returned as exact byte-strings.
1628
:return: yields a tuple of (dir_info, [file_info])
1629
dir_info is (utf8_relpath, path-from-top)
1630
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1631
if top is an absolute path, path-from-top is also an absolute path.
1632
path-from-top might be unicode or utf8, but it is the correct path to
1633
pass to os functions to affect the file in question. (such as os.lstat)
1635
global _selected_dir_reader
1636
if _selected_dir_reader is None:
1637
fs_encoding = _fs_enc.upper()
1638
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1639
# Win98 doesn't have unicode apis like FindFirstFileW
1640
# TODO: We possibly could support Win98 by falling back to the
1641
# original FindFirstFile, and using TCHAR instead of WCHAR,
1642
# but that gets a bit tricky, and requires custom compiling
1645
from bzrlib._walkdirs_win32 import Win32ReadDir
1646
_selected_dir_reader = Win32ReadDir()
1649
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1650
# ANSI_X3.4-1968 is a form of ASCII
1652
from bzrlib._readdir_pyx import UTF8DirReader
1653
_selected_dir_reader = UTF8DirReader()
1654
except ImportError, e:
1655
failed_to_load_extension(e)
1658
if _selected_dir_reader is None:
1659
# Fallback to the python version
1660
_selected_dir_reader = UnicodeDirReader()
1662
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1663
# But we don't actually uses 1-3 in pending, so set them to None
1664
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1665
read_dir = _selected_dir_reader.read_dir
1666
_directory = _directory_kind
1668
relroot, _, _, _, top = pending[-1].pop()
1671
dirblock = sorted(read_dir(relroot, top))
1672
yield (relroot, top), dirblock
1673
# push the user specified dirs from dirblock
1674
next = [d for d in reversed(dirblock) if d[2] == _directory]
1676
pending.append(next)
1679
class UnicodeDirReader(DirReader):
1680
"""A dir reader for non-utf8 file systems, which transcodes."""
1682
__slots__ = ['_utf8_encode']
1685
self._utf8_encode = codecs.getencoder('utf8')
1687
def top_prefix_to_starting_dir(self, top, prefix=""):
1688
"""See DirReader.top_prefix_to_starting_dir."""
1689
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1691
def read_dir(self, prefix, top):
1692
"""Read a single directory from a non-utf8 file system.
1694
top, and the abspath element in the output are unicode, all other paths
1695
are utf8. Local disk IO is done via unicode calls to listdir etc.
1697
This is currently the fallback code path when the filesystem encoding is
1698
not UTF-8. It may be better to implement an alternative so that we can
1699
safely handle paths that are not properly decodable in the current
1702
See DirReader.read_dir for details.
1704
_utf8_encode = self._utf8_encode
1706
_listdir = os.listdir
1707
_kind_from_mode = file_kind_from_stat_mode
1710
relprefix = prefix + '/'
1713
top_slash = top + u'/'
1716
append = dirblock.append
955
1717
for name in sorted(_listdir(top)):
956
abspath = top + '/' + name
957
statvalue = lstat(abspath)
958
dirblock.append((relroot + name, name,
959
file_kind_from_stat_mode(statvalue.st_mode),
961
yield (currentdir[0], top), dirblock
962
# push the user specified dirs from dirblock
963
for dir in reversed(dirblock):
964
if dir[2] == _directory:
1719
name_utf8 = _utf8_encode(name)[0]
1720
except UnicodeDecodeError:
1721
raise errors.BadFilenameEncoding(
1722
_utf8_encode(relprefix)[0] + name, _fs_enc)
1723
abspath = top_slash + name
1724
statvalue = _lstat(abspath)
1725
kind = _kind_from_mode(statvalue.st_mode)
1726
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
968
1730
def copy_tree(from_path, to_path, handlers={}):
969
1731
"""Copy all of the entries in from_path into to_path.
971
:param from_path: The base directory to copy.
1733
:param from_path: The base directory to copy.
972
1734
:param to_path: The target directory. If it does not exist, it will
974
1736
:param handlers: A dictionary of functions, which takes a source and
1057
1833
' doesn\'t support the locale set by $LANG (%s)\n'
1058
1834
" Continuing with ascii encoding.\n"
1059
1835
% (e, os.environ.get('LANG')))
1061
if _cached_user_encoding is None:
1062
_cached_user_encoding = 'ascii'
1063
return _cached_user_encoding
1836
user_encoding = 'ascii'
1838
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1839
# treat that as ASCII, and not support printing unicode characters to the
1842
# For python scripts run under vim, we get '', so also treat that as ASCII
1843
if user_encoding in (None, 'cp0', ''):
1844
user_encoding = 'ascii'
1848
codecs.lookup(user_encoding)
1850
sys.stderr.write('bzr: warning:'
1851
' unknown encoding %s.'
1852
' Continuing with ascii encoding.\n'
1855
user_encoding = 'ascii'
1858
_cached_user_encoding = user_encoding
1860
return user_encoding
1863
def get_host_name():
1864
"""Return the current unicode host name.
1866
This is meant to be used in place of socket.gethostname() because that
1867
behaves inconsistently on different platforms.
1869
if sys.platform == "win32":
1871
return win32utils.get_host_name()
1874
return socket.gethostname().decode(get_user_encoding())
1877
def recv_all(socket, bytes):
1878
"""Receive an exact number of bytes.
1880
Regular Socket.recv() may return less than the requested number of bytes,
1881
dependning on what's in the OS buffer. MSG_WAITALL is not available
1882
on all platforms, but this should work everywhere. This will return
1883
less than the requested amount if the remote end closes.
1885
This isn't optimized and is intended mostly for use in testing.
1888
while len(b) < bytes:
1889
new = until_no_eintr(socket.recv, bytes - len(b))
1896
def send_all(socket, bytes, report_activity=None):
1897
"""Send all bytes on a socket.
1899
Regular socket.sendall() can give socket error 10053 on Windows. This
1900
implementation sends no more than 64k at a time, which avoids this problem.
1902
:param report_activity: Call this as bytes are read, see
1903
Transport._report_activity
1906
for pos in xrange(0, len(bytes), chunk_size):
1907
block = bytes[pos:pos+chunk_size]
1908
if report_activity is not None:
1909
report_activity(len(block), 'write')
1910
until_no_eintr(socket.sendall, block)
1913
def dereference_path(path):
1914
"""Determine the real path to a file.
1916
All parent elements are dereferenced. But the file itself is not
1918
:param path: The original path. May be absolute or relative.
1919
:return: the real path *to* the file
1921
parent, base = os.path.split(path)
1922
# The pathjoin for '.' is a workaround for Python bug #1213894.
1923
# (initial path components aren't dereferenced)
1924
return pathjoin(realpath(pathjoin('.', parent)), base)
1927
def supports_mapi():
1928
"""Return True if we can use MAPI to launch a mail client."""
1929
return sys.platform == "win32"
1932
def resource_string(package, resource_name):
1933
"""Load a resource from a package and return it as a string.
1935
Note: Only packages that start with bzrlib are currently supported.
1937
This is designed to be a lightweight implementation of resource
1938
loading in a way which is API compatible with the same API from
1940
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1941
If and when pkg_resources becomes a standard library, this routine
1944
# Check package name is within bzrlib
1945
if package == "bzrlib":
1946
resource_relpath = resource_name
1947
elif package.startswith("bzrlib."):
1948
package = package[len("bzrlib."):].replace('.', os.sep)
1949
resource_relpath = pathjoin(package, resource_name)
1951
raise errors.BzrError('resource package %s not in bzrlib' % package)
1953
# Map the resource to a file and read its contents
1954
base = dirname(bzrlib.__file__)
1955
if getattr(sys, 'frozen', None): # bzr.exe
1956
base = abspath(pathjoin(base, '..', '..'))
1957
filename = pathjoin(base, resource_relpath)
1958
return open(filename, 'rU').read()
1961
def file_kind_from_stat_mode_thunk(mode):
1962
global file_kind_from_stat_mode
1963
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1965
from bzrlib._readdir_pyx import UTF8DirReader
1966
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1967
except ImportError, e:
1968
# This is one time where we won't warn that an extension failed to
1969
# load. The extension is never available on Windows anyway.
1970
from bzrlib._readdir_py import (
1971
_kind_from_mode as file_kind_from_stat_mode
1973
return file_kind_from_stat_mode(mode)
1974
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1977
def file_kind(f, _lstat=os.lstat):
1979
return file_kind_from_stat_mode(_lstat(f).st_mode)
1981
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1982
raise errors.NoSuchFile(f)
1986
def until_no_eintr(f, *a, **kw):
1987
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1988
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1992
except (IOError, OSError), e:
1993
if e.errno == errno.EINTR:
1997
def re_compile_checked(re_string, flags=0, where=""):
1998
"""Return a compiled re, or raise a sensible error.
2000
This should only be used when compiling user-supplied REs.
2002
:param re_string: Text form of regular expression.
2003
:param flags: eg re.IGNORECASE
2004
:param where: Message explaining to the user the context where
2005
it occurred, eg 'log search filter'.
2007
# from https://bugs.launchpad.net/bzr/+bug/251352
2009
re_obj = re.compile(re_string, flags)
2014
where = ' in ' + where
2015
# despite the name 'error' is a type
2016
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2017
% (where, re_string, e))
2020
if sys.platform == "win32":
2023
return msvcrt.getch()
2028
fd = sys.stdin.fileno()
2029
settings = termios.tcgetattr(fd)
2032
ch = sys.stdin.read(1)
2034
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2038
if sys.platform == 'linux2':
2039
def _local_concurrency():
2041
prefix = 'processor'
2042
for line in file('/proc/cpuinfo', 'rb'):
2043
if line.startswith(prefix):
2044
concurrency = int(line[line.find(':')+1:]) + 1
2046
elif sys.platform == 'darwin':
2047
def _local_concurrency():
2048
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2049
stdout=subprocess.PIPE).communicate()[0]
2050
elif sys.platform[0:7] == 'freebsd':
2051
def _local_concurrency():
2052
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2053
stdout=subprocess.PIPE).communicate()[0]
2054
elif sys.platform == 'sunos5':
2055
def _local_concurrency():
2056
return subprocess.Popen(['psrinfo', '-p',],
2057
stdout=subprocess.PIPE).communicate()[0]
2058
elif sys.platform == "win32":
2059
def _local_concurrency():
2060
# This appears to return the number of cores.
2061
return os.environ.get('NUMBER_OF_PROCESSORS')
2063
def _local_concurrency():
2068
_cached_local_concurrency = None
2070
def local_concurrency(use_cache=True):
2071
"""Return how many processes can be run concurrently.
2073
Rely on platform specific implementations and default to 1 (one) if
2074
anything goes wrong.
2076
global _cached_local_concurrency
2078
if _cached_local_concurrency is not None and use_cache:
2079
return _cached_local_concurrency
2081
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2082
if concurrency is None:
2084
concurrency = _local_concurrency()
2085
except (OSError, IOError):
2088
concurrency = int(concurrency)
2089
except (TypeError, ValueError):
2092
_cached_concurrency = concurrency
2096
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2097
"""A stream writer that doesn't decode str arguments."""
2099
def __init__(self, encode, stream, errors='strict'):
2100
codecs.StreamWriter.__init__(self, stream, errors)
2101
self.encode = encode
2103
def write(self, object):
2104
if type(object) is str:
2105
self.stream.write(object)
2107
data, _ = self.encode(object, self.errors)
2108
self.stream.write(data)
added
removed
bzrlib/osutils.py
