172
_directory_kind: "/",
174
'tree-reference': '+',
139
178
def kind_marker(kind):
180
return _kind_marker_map[kind]
182
# Slightly faster than using .get(, '') when the common case is that
142
elif kind == _directory_kind:
144
elif kind == 'symlink':
147
raise BzrError('invalid file kind %r' % kind)
149
187
lexists = getattr(os.path, 'lexists', None)
150
188
if lexists is None:
153
if getattr(os, 'lstat') is not None:
191
stat = getattr(os, 'lstat', os.stat)
159
195
if e.errno == errno.ENOENT:
162
raise BzrError("lstat/stat of (%r): %r" % (f, e))
198
raise errors.BzrError("lstat/stat of (%r): %r" % (f, e))
165
201
def fancy_rename(old, new, rename_func, unlink_func):
166
202
"""A fancy rename, when you don't have atomic rename.
168
204
:param old: The old path, to rename from
169
205
:param new: The new path, to rename to
170
206
: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
207
:param unlink_func: A way to delete the target file if the full rename
174
210
# sftp rename doesn't allow overwriting, so play tricks:
176
211
base = os.path.basename(new)
177
212
dirname = os.path.dirname(new)
178
tmp_name = u'tmp.%s.%.9f.%d.%s' % (base, time.time(), os.getpid(), rand_chars(10))
213
# callers use different encodings for the paths so the following MUST
214
# respect that. We rely on python upcasting to unicode if new is unicode
215
# and keeping a str if not.
216
tmp_name = 'tmp.%s.%.9f.%d.%s' % (base, time.time(),
217
os.getpid(), rand_chars(10))
179
218
tmp_name = pathjoin(dirname, tmp_name)
181
220
# Rename the file out of the way, but keep track if it didn't exist
482
550
for dirname in dir_list:
483
551
if is_inside(dirname, fname) or is_inside(fname, dirname):
556
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
557
report_activity=None, direction='read'):
558
"""Copy contents of one file to another.
560
The read_length can either be -1 to read to end-of-file (EOF) or
561
it can specify the maximum number of bytes to read.
563
The buff_size represents the maximum size for each read operation
564
performed on from_file.
566
:param report_activity: Call this as bytes are read, see
567
Transport._report_activity
568
:param direction: Will be passed to report_activity
570
:return: The number of bytes copied.
574
# read specified number of bytes
576
while read_length > 0:
577
num_bytes_to_read = min(read_length, buff_size)
579
block = from_file.read(num_bytes_to_read)
583
if report_activity is not None:
584
report_activity(len(block), direction)
587
actual_bytes_read = len(block)
588
read_length -= actual_bytes_read
589
length += actual_bytes_read
489
def pumpfile(fromfile, tofile):
490
"""Copy contents of one file to another."""
493
b = fromfile.read(BUFSIZE)
593
block = from_file.read(buff_size)
597
if report_activity is not None:
598
report_activity(len(block), direction)
604
def pump_string_file(bytes, file_handle, segment_size=None):
605
"""Write bytes to file_handle in many smaller writes.
607
:param bytes: The string to write.
608
:param file_handle: The file to write to.
610
# Write data in chunks rather than all at once, because very large
611
# writes fail on some platforms (e.g. Windows with SMB mounted
614
segment_size = 5242880 # 5MB
615
segments = range(len(bytes) / segment_size + 1)
616
write = file_handle.write
617
for segment_index in segments:
618
segment = buffer(bytes, segment_index * segment_size, segment_size)
499
622
def file_iterator(input_file, readsize=32768):
555
706
def local_time_offset(t=None):
556
707
"""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,
710
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
711
return offset.days * 86400 + offset.seconds
713
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
714
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
717
def format_date(t, offset=0, timezone='original', date_fmt=None,
568
718
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)
719
"""Return a formatted date string.
721
:param t: Seconds since the epoch.
722
:param offset: Timezone offset in seconds east of utc.
723
:param timezone: How to display the time: 'utc', 'original' for the
724
timezone specified by offset, or 'local' for the process's current
726
:param date_fmt: strftime format.
727
:param show_offset: Whether to append the timezone.
729
(date_fmt, tt, offset_str) = \
730
_format_date(t, offset, timezone, date_fmt, show_offset)
731
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
732
date_str = time.strftime(date_fmt, tt)
733
return date_str + offset_str
736
# Cache of formatted offset strings
740
def format_date_with_offset_in_original_timezone(t, offset=0,
741
_cache=_offset_cache):
742
"""Return a formatted date string in the original timezone.
744
This routine may be faster then format_date.
746
:param t: Seconds since the epoch.
747
:param offset: Timezone offset in seconds east of utc.
751
tt = time.gmtime(t + offset)
752
date_fmt = _default_format_by_weekday_num[tt[6]]
753
date_str = time.strftime(date_fmt, tt)
754
offset_str = _cache.get(offset, None)
755
if offset_str is None:
756
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
757
_cache[offset] = offset_str
758
return date_str + offset_str
761
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
763
"""Return an unicode date string formatted according to the current locale.
765
:param t: Seconds since the epoch.
766
:param offset: Timezone offset in seconds east of utc.
767
:param timezone: How to display the time: 'utc', 'original' for the
768
timezone specified by offset, or 'local' for the process's current
770
:param date_fmt: strftime format.
771
:param show_offset: Whether to append the timezone.
773
(date_fmt, tt, offset_str) = \
774
_format_date(t, offset, timezone, date_fmt, show_offset)
775
date_str = time.strftime(date_fmt, tt)
776
if not isinstance(date_str, unicode):
777
date_str = date_str.decode(get_user_encoding(), 'replace')
778
return date_str + offset_str
781
def _format_date(t, offset, timezone, date_fmt, show_offset):
573
782
if timezone == 'utc':
574
783
tt = time.gmtime(t)
581
790
tt = time.localtime(t)
582
791
offset = local_time_offset(t)
584
raise BzrError("unsupported timezone format %r" % timezone,
585
['options are "utc", "original", "local"'])
793
raise errors.UnsupportedTimezoneFormat(timezone)
586
794
if date_fmt is None:
587
795
date_fmt = "%a %Y-%m-%d %H:%M:%S"
589
797
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
592
return (time.strftime(date_fmt, tt) + offset_str)
800
return (date_fmt, tt, offset_str)
595
803
def compact_date(when):
596
804
return time.strftime('%Y%m%d%H%M%S', time.gmtime(when))
807
def format_delta(delta):
808
"""Get a nice looking string for a time delta.
810
:param delta: The time difference in seconds, can be positive or negative.
811
positive indicates time in the past, negative indicates time in the
812
future. (usually time.time() - stored_time)
813
:return: String formatted to show approximate resolution
819
direction = 'in the future'
823
if seconds < 90: # print seconds up to 90 seconds
825
return '%d second %s' % (seconds, direction,)
827
return '%d seconds %s' % (seconds, direction)
829
minutes = int(seconds / 60)
830
seconds -= 60 * minutes
835
if minutes < 90: # print minutes, seconds up to 90 minutes
837
return '%d minute, %d second%s %s' % (
838
minutes, seconds, plural_seconds, direction)
840
return '%d minutes, %d second%s %s' % (
841
minutes, seconds, plural_seconds, direction)
843
hours = int(minutes / 60)
844
minutes -= 60 * hours
851
return '%d hour, %d minute%s %s' % (hours, minutes,
852
plural_minutes, direction)
853
return '%d hours, %d minute%s %s' % (hours, minutes,
854
plural_minutes, direction)
601
857
"""Return size of given open file."""
669
raise BzrError("sorry, %r not allowed in path" % f)
909
raise errors.BzrError("sorry, %r not allowed in path" % f)
670
910
elif (f == '.') or (f == ''):
677
assert isinstance(p, list)
679
919
if (f == '..') or (f is None) or (f == ''):
680
raise BzrError("sorry, %r not allowed in path" % f)
920
raise errors.BzrError("sorry, %r not allowed in path" % f)
681
921
return pathjoin(*p)
684
@deprecated_function(zero_nine)
685
def appendpath(p1, p2):
689
return pathjoin(p1, p2)
924
def parent_directories(filename):
925
"""Return the list of parent directories, deepest first.
927
For example, parent_directories("a/b/c") -> ["a/b", "a"].
930
parts = splitpath(dirname(filename))
932
parents.append(joinpath(parts))
937
_extension_load_failures = []
940
def failed_to_load_extension(exception):
941
"""Handle failing to load a binary extension.
943
This should be called from the ImportError block guarding the attempt to
944
import the native extension. If this function returns, the pure-Python
945
implementation should be loaded instead::
948
>>> import bzrlib._fictional_extension_pyx
949
>>> except ImportError, e:
950
>>> bzrlib.osutils.failed_to_load_extension(e)
951
>>> import bzrlib._fictional_extension_py
953
# NB: This docstring is just an example, not a doctest, because doctest
954
# currently can't cope with the use of lazy imports in this namespace --
957
# This currently doesn't report the failure at the time it occurs, because
958
# they tend to happen very early in startup when we can't check config
959
# files etc, and also we want to report all failures but not spam the user
961
from bzrlib import trace
962
exception_str = str(exception)
963
if exception_str not in _extension_load_failures:
964
trace.mutter("failed to load compiled extension: %s" % exception_str)
965
_extension_load_failures.append(exception_str)
968
def report_extension_load_failures():
969
if not _extension_load_failures:
971
from bzrlib.config import GlobalConfig
972
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
974
# the warnings framework should by default show this only once
975
from bzrlib.trace import warning
977
"bzr: warning: some compiled extensions could not be loaded; "
978
"see <https://answers.launchpad.net/bzr/+faq/703>")
979
# we no longer show the specific missing extensions here, because it makes
980
# the message too long and scary - see
981
# https://bugs.launchpad.net/bzr/+bug/430529
985
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
986
except ImportError, e:
987
failed_to_load_extension(e)
988
from bzrlib._chunks_to_lines_py import chunks_to_lines
692
991
def split_lines(s):
693
992
"""Split s into lines, but without removing the newline characters."""
993
# Trivially convert a fulltext into a 'chunked' representation, and let
994
# chunks_to_lines do the heavy lifting.
995
if isinstance(s, str):
996
# chunks_to_lines only supports 8-bit strings
997
return chunks_to_lines([s])
999
return _split_lines(s)
1002
def _split_lines(s):
1003
"""Split s into lines, but without removing the newline characters.
1005
This supports Unicode or plain string objects.
694
1007
lines = s.split('\n')
695
1008
result = [line + '\n' for line in lines[:-1]]
761
1132
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)' %
1135
if len(base) < MIN_ABS_PATHLENGTH:
1136
# must have space for e.g. a drive letter
1137
raise ValueError('%r is too short to calculate a relative path'
768
1140
rp = abspath(path)
772
while len(head) >= len(base):
1145
if len(head) <= len(base) and head != base:
1146
raise errors.PathNotChild(rp, base)
773
1147
if head == base:
775
head, tail = os.path.split(head)
1149
head, tail = split(head)
779
raise PathNotChild(rp, base)
1154
return pathjoin(*reversed(s))
1159
def _cicp_canonical_relpath(base, path):
1160
"""Return the canonical path relative to base.
1162
Like relpath, but on case-insensitive-case-preserving file-systems, this
1163
will return the relpath as stored on the file-system rather than in the
1164
case specified in the input string, for all existing portions of the path.
1166
This will cause O(N) behaviour if called for every path in a tree; if you
1167
have a number of paths to convert, you should use canonical_relpaths().
1169
# TODO: it should be possible to optimize this for Windows by using the
1170
# win32 API FindFiles function to look for the specified name - but using
1171
# os.listdir() still gives us the correct, platform agnostic semantics in
1174
rel = relpath(base, path)
1175
# '.' will have been turned into ''
1179
abs_base = abspath(base)
1181
_listdir = os.listdir
1183
# use an explicit iterator so we can easily consume the rest on early exit.
1184
bit_iter = iter(rel.split('/'))
1185
for bit in bit_iter:
1188
next_entries = _listdir(current)
1189
except OSError: # enoent, eperm, etc
1190
# We can't find this in the filesystem, so just append the
1192
current = pathjoin(current, bit, *list(bit_iter))
1194
for look in next_entries:
1195
if lbit == look.lower():
1196
current = pathjoin(current, look)
1199
# got to the end, nothing matched, so we just return the
1200
# non-existing bits as they were specified (the filename may be
1201
# the target of a move, for example).
1202
current = pathjoin(current, bit, *list(bit_iter))
1204
return current[len(abs_base):].lstrip('/')
1206
# XXX - TODO - we need better detection/integration of case-insensitive
1207
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1208
# filesystems), for example, so could probably benefit from the same basic
1209
# support there. For now though, only Windows and OSX get that support, and
1210
# they get it for *all* file-systems!
1211
if sys.platform in ('win32', 'darwin'):
1212
canonical_relpath = _cicp_canonical_relpath
1214
canonical_relpath = relpath
1216
def canonical_relpaths(base, paths):
1217
"""Create an iterable to canonicalize a sequence of relative paths.
1219
The intent is for this implementation to use a cache, vastly speeding
1220
up multiple transformations in the same directory.
1222
# but for now, we haven't optimized...
1223
return [canonical_relpath(base, p) for p in paths]
787
1225
def safe_unicode(unicode_or_utf8_string):
788
1226
"""Coerce unicode_or_utf8_string into unicode.
790
1228
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.
1229
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1230
wrapped in a BzrBadParameterNotUnicode exception.
795
1232
if isinstance(unicode_or_utf8_string, unicode):
796
1233
return unicode_or_utf8_string
798
1235
return unicode_or_utf8_string.decode('utf8')
799
1236
except UnicodeDecodeError:
800
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
1237
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1240
def safe_utf8(unicode_or_utf8_string):
1241
"""Coerce unicode_or_utf8_string to a utf8 string.
1243
If it is a str, it is returned.
1244
If it is Unicode, it is encoded into a utf-8 string.
1246
if isinstance(unicode_or_utf8_string, str):
1247
# TODO: jam 20070209 This is overkill, and probably has an impact on
1248
# performance if we are dealing with lots of apis that want a
1251
# Make sure it is a valid utf-8 string
1252
unicode_or_utf8_string.decode('utf-8')
1253
except UnicodeDecodeError:
1254
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1255
return unicode_or_utf8_string
1256
return unicode_or_utf8_string.encode('utf-8')
1259
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1260
' Revision id generators should be creating utf8'
1264
def safe_revision_id(unicode_or_utf8_string, warn=True):
1265
"""Revision ids should now be utf8, but at one point they were unicode.
1267
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1269
:param warn: Functions that are sanitizing user data can set warn=False
1270
:return: None or a utf8 revision id.
1272
if (unicode_or_utf8_string is None
1273
or unicode_or_utf8_string.__class__ == str):
1274
return unicode_or_utf8_string
1276
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1278
return cache_utf8.encode(unicode_or_utf8_string)
1281
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1282
' generators should be creating utf8 file ids.')
1285
def safe_file_id(unicode_or_utf8_string, warn=True):
1286
"""File ids should now be utf8, but at one point they were unicode.
1288
This is the same as safe_utf8, except it uses the cached encode functions
1289
to save a little bit of performance.
1291
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1293
:param warn: Functions that are sanitizing user data can set warn=False
1294
:return: None or a utf8 file id.
1296
if (unicode_or_utf8_string is None
1297
or unicode_or_utf8_string.__class__ == str):
1298
return unicode_or_utf8_string
1300
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1302
return cache_utf8.encode(unicode_or_utf8_string)
803
1305
_platform_normalizes_filenames = False
844
1346
normalized_filename = _inaccessible_normalized_filename
1349
default_terminal_width = 80
1350
"""The default terminal width for ttys.
1352
This is defined so that higher levels can share a common fallback value when
1353
terminal_width() returns None.
847
1357
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]
1358
"""Return terminal width.
1360
None is returned if the width can't established precisely.
1363
- if BZR_COLUMNS is set, returns its value
1364
- if there is no controlling terminal, returns None
1365
- if COLUMNS is set, returns its value,
1367
From there, we need to query the OS to get the size of the controlling
1371
- get termios.TIOCGWINSZ
1372
- if an error occurs or a negative value is obtained, returns None
1376
- win32utils.get_console_size() decides,
1377
- returns None on error (provided default value)
1380
# If BZR_COLUMNS is set, take it, user is always right
1382
return int(os.environ['BZR_COLUMNS'])
1383
except (KeyError, ValueError):
1386
isatty = getattr(sys.stdout, 'isatty', None)
1387
if isatty is None or not isatty():
1388
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1391
# If COLUMNS is set, take it, the terminal knows better (even inside a
1392
# given terminal, the application can decide to set COLUMNS to a lower
1393
# value (splitted screen) or a bigger value (scroll bars))
1395
return int(os.environ['COLUMNS'])
1396
except (KeyError, ValueError):
1399
width, height = _terminal_size(None, None)
1401
# Consider invalid values as meaning no width
1407
def _win32_terminal_size(width, height):
1408
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1409
return width, height
1412
def _ioctl_terminal_size(width, height):
854
1414
import struct, fcntl, termios
855
1415
s = struct.pack('HHHH', 0, 0, 0, 0)
856
1416
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
857
width = struct.unpack('HHHH', x)[1]
1417
height, width = struct.unpack('HHHH', x)[0:2]
1418
except (IOError, AttributeError):
862
width = int(os.environ['COLUMNS'])
1420
return width, height
1422
_terminal_size = None
1423
"""Returns the terminal size as (width, height).
1425
:param width: Default value for width.
1426
:param height: Default value for height.
1428
This is defined specifically for each OS and query the size of the controlling
1429
terminal. If any error occurs, the provided default values should be returned.
1431
if sys.platform == 'win32':
1432
_terminal_size = _win32_terminal_size
1434
_terminal_size = _ioctl_terminal_size
1437
def _terminal_size_changed(signum, frame):
1438
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1439
width, height = _terminal_size(None, None)
1440
if width is not None:
1441
os.environ['COLUMNS'] = str(width)
1443
if sys.platform == 'win32':
1444
# Martin (gz) mentioned WINDOW_BUFFER_SIZE_RECORD from ReadConsoleInput but
1445
# I've no idea how to plug that in the current design -- vila 20091216
1448
signal.signal(signal.SIGWINCH, _terminal_size_changed)
871
1451
def supports_executable():
872
1452
return sys.platform != "win32"
1455
def supports_posix_readonly():
1456
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1458
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1459
directory controls creation/deletion, etc.
1461
And under win32, readonly means that the directory itself cannot be
1462
deleted. The contents of a readonly directory can be changed, unlike POSIX
1463
where files in readonly directories cannot be added, deleted or renamed.
1465
return sys.platform != "win32"
875
1468
def set_or_unset_env(env_variable, value):
876
1469
"""Modify the environment, setting or removing the env_variable.
897
1490
def check_legal_path(path):
898
"""Check whether the supplied path is legal.
1491
"""Check whether the supplied path is legal.
899
1492
This is only required on Windows, so we don't test on other platforms
902
1495
if sys.platform != "win32":
904
1497
if _validWin32PathRE.match(path) is None:
905
raise IllegalPath(path)
1498
raise errors.IllegalPath(path)
1501
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1503
def _is_error_enotdir(e):
1504
"""Check if this exception represents ENOTDIR.
1506
Unfortunately, python is very inconsistent about the exception
1507
here. The cases are:
1508
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1509
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1510
which is the windows error code.
1511
3) Windows, Python2.5 uses errno == EINVAL and
1512
winerror == ERROR_DIRECTORY
1514
:param e: An Exception object (expected to be OSError with an errno
1515
attribute, but we should be able to cope with anything)
1516
:return: True if this represents an ENOTDIR error. False otherwise.
1518
en = getattr(e, 'errno', None)
1519
if (en == errno.ENOTDIR
1520
or (sys.platform == 'win32'
1521
and (en == _WIN32_ERROR_DIRECTORY
1522
or (en == errno.EINVAL
1523
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
908
1529
def walkdirs(top, prefix=""):
909
1530
"""Yield data about all the directories in a tree.
911
1532
This yields all the data about the contents of a directory at a time.
912
1533
After each directory has been yielded, if the caller has mutated the list
913
1534
to exclude some directories, they are then not descended into.
915
1536
The data yielded is of the form:
916
1537
((directory-relpath, directory-path-from-top),
917
[(relpath, basename, kind, lstat), ...]),
1538
[(relpath, basename, kind, lstat, path-from-top), ...]),
918
1539
- directory-relpath is the relative path of the directory being returned
919
1540
with respect to top. prefix is prepended to this.
920
- directory-path-from-root is the path including top for this directory.
1541
- directory-path-from-root is the path including top for this directory.
921
1542
It is suitable for use with os functions.
922
1543
- relpath is the relative path within the subtree being walked.
923
1544
- basename is the basename of the path
925
1546
present within the tree - but it may be recorded as versioned. See
927
1548
- lstat is the stat data *if* the file was statted.
928
- planned, not implemented:
1549
- planned, not implemented:
929
1550
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
1552
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
932
1553
allows one to walk a subtree but get paths that are relative to a tree
933
1554
rooted higher up.
934
1555
:return: an iterator over the dirs.
936
1557
#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
1558
# summary in this, and the path from the root, may not agree
938
1559
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
939
1560
# potentially confusing output. We should make this more robust - but
940
1561
# not at a speed cost. RBC 20060731
943
1563
_directory = _directory_kind
945
pending = [(prefix, "", _directory, None, top)]
1564
_listdir = os.listdir
1565
_kind_from_mode = file_kind_from_stat_mode
1566
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
948
currentdir = pending.pop()
949
1568
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
952
relroot = currentdir[0] + '/'
1569
relroot, _, _, _, top = pending.pop()
1571
relprefix = relroot + u'/'
1574
top_slash = top + u'/'
1577
append = dirblock.append
1579
names = sorted(_listdir(top))
1581
if not _is_error_enotdir(e):
1585
abspath = top_slash + name
1586
statvalue = _lstat(abspath)
1587
kind = _kind_from_mode(statvalue.st_mode)
1588
append((relprefix + name, name, kind, statvalue, abspath))
1589
yield (relroot, top), dirblock
1591
# push the user specified dirs from dirblock
1592
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1595
class DirReader(object):
1596
"""An interface for reading directories."""
1598
def top_prefix_to_starting_dir(self, top, prefix=""):
1599
"""Converts top and prefix to a starting dir entry
1601
:param top: A utf8 path
1602
:param prefix: An optional utf8 path to prefix output relative paths
1604
:return: A tuple starting with prefix, and ending with the native
1607
raise NotImplementedError(self.top_prefix_to_starting_dir)
1609
def read_dir(self, prefix, top):
1610
"""Read a specific dir.
1612
:param prefix: A utf8 prefix to be preprended to the path basenames.
1613
:param top: A natively encoded path to read.
1614
:return: A list of the directories contents. Each item contains:
1615
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1617
raise NotImplementedError(self.read_dir)
1620
_selected_dir_reader = None
1623
def _walkdirs_utf8(top, prefix=""):
1624
"""Yield data about all the directories in a tree.
1626
This yields the same information as walkdirs() only each entry is yielded
1627
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1628
are returned as exact byte-strings.
1630
:return: yields a tuple of (dir_info, [file_info])
1631
dir_info is (utf8_relpath, path-from-top)
1632
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1633
if top is an absolute path, path-from-top is also an absolute path.
1634
path-from-top might be unicode or utf8, but it is the correct path to
1635
pass to os functions to affect the file in question. (such as os.lstat)
1637
global _selected_dir_reader
1638
if _selected_dir_reader is None:
1639
fs_encoding = _fs_enc.upper()
1640
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1641
# Win98 doesn't have unicode apis like FindFirstFileW
1642
# TODO: We possibly could support Win98 by falling back to the
1643
# original FindFirstFile, and using TCHAR instead of WCHAR,
1644
# but that gets a bit tricky, and requires custom compiling
1647
from bzrlib._walkdirs_win32 import Win32ReadDir
1648
_selected_dir_reader = Win32ReadDir()
1651
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1652
# ANSI_X3.4-1968 is a form of ASCII
1654
from bzrlib._readdir_pyx import UTF8DirReader
1655
_selected_dir_reader = UTF8DirReader()
1656
except ImportError, e:
1657
failed_to_load_extension(e)
1660
if _selected_dir_reader is None:
1661
# Fallback to the python version
1662
_selected_dir_reader = UnicodeDirReader()
1664
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1665
# But we don't actually uses 1-3 in pending, so set them to None
1666
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1667
read_dir = _selected_dir_reader.read_dir
1668
_directory = _directory_kind
1670
relroot, _, _, _, top = pending[-1].pop()
1673
dirblock = sorted(read_dir(relroot, top))
1674
yield (relroot, top), dirblock
1675
# push the user specified dirs from dirblock
1676
next = [d for d in reversed(dirblock) if d[2] == _directory]
1678
pending.append(next)
1681
class UnicodeDirReader(DirReader):
1682
"""A dir reader for non-utf8 file systems, which transcodes."""
1684
__slots__ = ['_utf8_encode']
1687
self._utf8_encode = codecs.getencoder('utf8')
1689
def top_prefix_to_starting_dir(self, top, prefix=""):
1690
"""See DirReader.top_prefix_to_starting_dir."""
1691
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1693
def read_dir(self, prefix, top):
1694
"""Read a single directory from a non-utf8 file system.
1696
top, and the abspath element in the output are unicode, all other paths
1697
are utf8. Local disk IO is done via unicode calls to listdir etc.
1699
This is currently the fallback code path when the filesystem encoding is
1700
not UTF-8. It may be better to implement an alternative so that we can
1701
safely handle paths that are not properly decodable in the current
1704
See DirReader.read_dir for details.
1706
_utf8_encode = self._utf8_encode
1708
_listdir = os.listdir
1709
_kind_from_mode = file_kind_from_stat_mode
1712
relprefix = prefix + '/'
1715
top_slash = top + u'/'
1718
append = dirblock.append
955
1719
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:
1721
name_utf8 = _utf8_encode(name)[0]
1722
except UnicodeDecodeError:
1723
raise errors.BadFilenameEncoding(
1724
_utf8_encode(relprefix)[0] + name, _fs_enc)
1725
abspath = top_slash + name
1726
statvalue = _lstat(abspath)
1727
kind = _kind_from_mode(statvalue.st_mode)
1728
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
968
1732
def copy_tree(from_path, to_path, handlers={}):
969
1733
"""Copy all of the entries in from_path into to_path.
971
:param from_path: The base directory to copy.
1735
:param from_path: The base directory to copy.
972
1736
:param to_path: The target directory. If it does not exist, it will
974
1738
:param handlers: A dictionary of functions, which takes a source and
1057
1835
' doesn\'t support the locale set by $LANG (%s)\n'
1058
1836
" Continuing with ascii encoding.\n"
1059
1837
% (e, os.environ.get('LANG')))
1061
if _cached_user_encoding is None:
1062
_cached_user_encoding = 'ascii'
1063
return _cached_user_encoding
1838
user_encoding = 'ascii'
1840
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1841
# treat that as ASCII, and not support printing unicode characters to the
1844
# For python scripts run under vim, we get '', so also treat that as ASCII
1845
if user_encoding in (None, 'cp0', ''):
1846
user_encoding = 'ascii'
1850
codecs.lookup(user_encoding)
1852
sys.stderr.write('bzr: warning:'
1853
' unknown encoding %s.'
1854
' Continuing with ascii encoding.\n'
1857
user_encoding = 'ascii'
1860
_cached_user_encoding = user_encoding
1862
return user_encoding
1865
def get_host_name():
1866
"""Return the current unicode host name.
1868
This is meant to be used in place of socket.gethostname() because that
1869
behaves inconsistently on different platforms.
1871
if sys.platform == "win32":
1873
return win32utils.get_host_name()
1876
return socket.gethostname().decode(get_user_encoding())
1879
def recv_all(socket, bytes):
1880
"""Receive an exact number of bytes.
1882
Regular Socket.recv() may return less than the requested number of bytes,
1883
dependning on what's in the OS buffer. MSG_WAITALL is not available
1884
on all platforms, but this should work everywhere. This will return
1885
less than the requested amount if the remote end closes.
1887
This isn't optimized and is intended mostly for use in testing.
1890
while len(b) < bytes:
1891
new = until_no_eintr(socket.recv, bytes - len(b))
1898
def send_all(socket, bytes, report_activity=None):
1899
"""Send all bytes on a socket.
1901
Regular socket.sendall() can give socket error 10053 on Windows. This
1902
implementation sends no more than 64k at a time, which avoids this problem.
1904
:param report_activity: Call this as bytes are read, see
1905
Transport._report_activity
1908
for pos in xrange(0, len(bytes), chunk_size):
1909
block = bytes[pos:pos+chunk_size]
1910
if report_activity is not None:
1911
report_activity(len(block), 'write')
1912
until_no_eintr(socket.sendall, block)
1915
def dereference_path(path):
1916
"""Determine the real path to a file.
1918
All parent elements are dereferenced. But the file itself is not
1920
:param path: The original path. May be absolute or relative.
1921
:return: the real path *to* the file
1923
parent, base = os.path.split(path)
1924
# The pathjoin for '.' is a workaround for Python bug #1213894.
1925
# (initial path components aren't dereferenced)
1926
return pathjoin(realpath(pathjoin('.', parent)), base)
1929
def supports_mapi():
1930
"""Return True if we can use MAPI to launch a mail client."""
1931
return sys.platform == "win32"
1934
def resource_string(package, resource_name):
1935
"""Load a resource from a package and return it as a string.
1937
Note: Only packages that start with bzrlib are currently supported.
1939
This is designed to be a lightweight implementation of resource
1940
loading in a way which is API compatible with the same API from
1942
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
1943
If and when pkg_resources becomes a standard library, this routine
1946
# Check package name is within bzrlib
1947
if package == "bzrlib":
1948
resource_relpath = resource_name
1949
elif package.startswith("bzrlib."):
1950
package = package[len("bzrlib."):].replace('.', os.sep)
1951
resource_relpath = pathjoin(package, resource_name)
1953
raise errors.BzrError('resource package %s not in bzrlib' % package)
1955
# Map the resource to a file and read its contents
1956
base = dirname(bzrlib.__file__)
1957
if getattr(sys, 'frozen', None): # bzr.exe
1958
base = abspath(pathjoin(base, '..', '..'))
1959
filename = pathjoin(base, resource_relpath)
1960
return open(filename, 'rU').read()
1963
def file_kind_from_stat_mode_thunk(mode):
1964
global file_kind_from_stat_mode
1965
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
1967
from bzrlib._readdir_pyx import UTF8DirReader
1968
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
1969
except ImportError, e:
1970
# This is one time where we won't warn that an extension failed to
1971
# load. The extension is never available on Windows anyway.
1972
from bzrlib._readdir_py import (
1973
_kind_from_mode as file_kind_from_stat_mode
1975
return file_kind_from_stat_mode(mode)
1976
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
1979
def file_kind(f, _lstat=os.lstat):
1981
return file_kind_from_stat_mode(_lstat(f).st_mode)
1983
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
1984
raise errors.NoSuchFile(f)
1988
def until_no_eintr(f, *a, **kw):
1989
"""Run f(*a, **kw), retrying if an EINTR error occurs."""
1990
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
1994
except (IOError, OSError), e:
1995
if e.errno == errno.EINTR:
1999
def re_compile_checked(re_string, flags=0, where=""):
2000
"""Return a compiled re, or raise a sensible error.
2002
This should only be used when compiling user-supplied REs.
2004
:param re_string: Text form of regular expression.
2005
:param flags: eg re.IGNORECASE
2006
:param where: Message explaining to the user the context where
2007
it occurred, eg 'log search filter'.
2009
# from https://bugs.launchpad.net/bzr/+bug/251352
2011
re_obj = re.compile(re_string, flags)
2016
where = ' in ' + where
2017
# despite the name 'error' is a type
2018
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2019
% (where, re_string, e))
2022
if sys.platform == "win32":
2025
return msvcrt.getch()
2030
fd = sys.stdin.fileno()
2031
settings = termios.tcgetattr(fd)
2034
ch = sys.stdin.read(1)
2036
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2040
if sys.platform == 'linux2':
2041
def _local_concurrency():
2043
prefix = 'processor'
2044
for line in file('/proc/cpuinfo', 'rb'):
2045
if line.startswith(prefix):
2046
concurrency = int(line[line.find(':')+1:]) + 1
2048
elif sys.platform == 'darwin':
2049
def _local_concurrency():
2050
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2051
stdout=subprocess.PIPE).communicate()[0]
2052
elif sys.platform[0:7] == 'freebsd':
2053
def _local_concurrency():
2054
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2055
stdout=subprocess.PIPE).communicate()[0]
2056
elif sys.platform == 'sunos5':
2057
def _local_concurrency():
2058
return subprocess.Popen(['psrinfo', '-p',],
2059
stdout=subprocess.PIPE).communicate()[0]
2060
elif sys.platform == "win32":
2061
def _local_concurrency():
2062
# This appears to return the number of cores.
2063
return os.environ.get('NUMBER_OF_PROCESSORS')
2065
def _local_concurrency():
2070
_cached_local_concurrency = None
2072
def local_concurrency(use_cache=True):
2073
"""Return how many processes can be run concurrently.
2075
Rely on platform specific implementations and default to 1 (one) if
2076
anything goes wrong.
2078
global _cached_local_concurrency
2080
if _cached_local_concurrency is not None and use_cache:
2081
return _cached_local_concurrency
2083
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2084
if concurrency is None:
2086
concurrency = _local_concurrency()
2087
except (OSError, IOError):
2090
concurrency = int(concurrency)
2091
except (TypeError, ValueError):
2094
_cached_concurrency = concurrency
2098
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2099
"""A stream writer that doesn't decode str arguments."""
2101
def __init__(self, encode, stream, errors='strict'):
2102
codecs.StreamWriter.__init__(self, stream, errors)
2103
self.encode = encode
2105
def write(self, object):
2106
if type(object) is str:
2107
self.stream.write(object)
2109
data, _ = self.encode(object, self.errors)
2110
self.stream.write(data)
added
removed
bzrlib/osutils.py
