34
35
splitdrive as _nt_splitdrive,
38
40
from shutil import (
44
45
from tempfile import (
49
52
from bzrlib import (
55
from bzrlib.errors import (
57
BzrBadParameterNotUnicode,
57
62
from bzrlib.symbol_versioning import (
58
63
deprecated_function,
62
# sha and md5 modules are deprecated in python2.6 but hashlib is available as
64
if sys.version_info < (2, 5):
65
import md5 as _mod_md5
67
import sha as _mod_sha
77
from bzrlib import symbol_versioning
80
# Cross platform wall-clock time functionality with decent resolution.
81
# On Linux ``time.clock`` returns only CPU time. On Windows, ``time.time()``
82
# only has a resolution of ~15ms. Note that ``time.clock()`` is not
83
# synchronized with ``time.time()``, this is only meant to be used to find
84
# delta times by subtracting from another call to this function.
85
timer_func = time.time
86
if sys.platform == 'win32':
87
timer_func = time.clock
66
from bzrlib.trace import mutter
89
69
# On win32, O_BINARY is used to indicate the file should
90
70
# be opened in binary mode, rather than text mode.
91
71
# On other platforms, O_BINARY doesn't exist, because
92
72
# they always open in binary mode, so it is okay to
93
# OR with 0 on those platforms.
94
# O_NOINHERIT and O_TEXT exists only on win32 too.
73
# OR with 0 on those platforms
95
74
O_BINARY = getattr(os, 'O_BINARY', 0)
96
O_TEXT = getattr(os, 'O_TEXT', 0)
97
O_NOINHERIT = getattr(os, 'O_NOINHERIT', 0)
100
def get_unicode_argv():
102
user_encoding = get_user_encoding()
103
return [a.decode(user_encoding) for a in sys.argv[1:]]
104
except UnicodeDecodeError:
105
raise errors.BzrError(("Parameter '%r' is unsupported by the current "
109
77
def make_readonly(filename):
110
78
"""Make a filename read-only."""
111
mod = os.lstat(filename).st_mode
112
if not stat.S_ISLNK(mod):
114
os.chmod(filename, mod)
79
mod = os.stat(filename).st_mode
81
os.chmod(filename, mod)
117
84
def make_writable(filename):
118
mod = os.lstat(filename).st_mode
119
if not stat.S_ISLNK(mod):
121
os.chmod(filename, mod)
124
def minimum_path_selection(paths):
125
"""Return the smallset subset of paths which are outside paths.
127
:param paths: A container (and hence not None) of paths.
128
:return: A set of paths sufficient to include everything in paths via
129
is_inside, drawn from the paths parameter.
135
return path.split('/')
136
sorted_paths = sorted(list(paths), key=sort_key)
138
search_paths = [sorted_paths[0]]
139
for path in sorted_paths[1:]:
140
if not is_inside(search_paths[-1], path):
141
# This path is unique, add it
142
search_paths.append(path)
144
return set(search_paths)
85
mod = os.stat(filename).st_mode
87
os.chmod(filename, mod)
576
483
for dirname in dir_list:
577
484
if is_inside(dirname, fname) or is_inside(fname, dirname):
582
def pumpfile(from_file, to_file, read_length=-1, buff_size=32768,
583
report_activity=None, direction='read'):
584
"""Copy contents of one file to another.
586
The read_length can either be -1 to read to end-of-file (EOF) or
587
it can specify the maximum number of bytes to read.
589
The buff_size represents the maximum size for each read operation
590
performed on from_file.
592
:param report_activity: Call this as bytes are read, see
593
Transport._report_activity
594
:param direction: Will be passed to report_activity
596
:return: The number of bytes copied.
600
# read specified number of bytes
602
while read_length > 0:
603
num_bytes_to_read = min(read_length, buff_size)
605
block = from_file.read(num_bytes_to_read)
609
if report_activity is not None:
610
report_activity(len(block), direction)
613
actual_bytes_read = len(block)
614
read_length -= actual_bytes_read
615
length += actual_bytes_read
619
block = from_file.read(buff_size)
623
if report_activity is not None:
624
report_activity(len(block), direction)
630
def pump_string_file(bytes, file_handle, segment_size=None):
631
"""Write bytes to file_handle in many smaller writes.
633
:param bytes: The string to write.
634
:param file_handle: The file to write to.
636
# Write data in chunks rather than all at once, because very large
637
# writes fail on some platforms (e.g. Windows with SMB mounted
640
segment_size = 5242880 # 5MB
641
segments = range(len(bytes) / segment_size + 1)
642
write = file_handle.write
643
for segment_index in segments:
644
segment = buffer(bytes, segment_index * segment_size, segment_size)
490
def pumpfile(fromfile, tofile):
491
"""Copy contents of one file to another."""
494
b = fromfile.read(BUFSIZE)
648
500
def file_iterator(input_file, readsize=32768):
732
556
def local_time_offset(t=None):
733
557
"""Return offset of local zone from GMT, either at present or at time t."""
558
# python2.3 localtime() can't take None
736
offset = datetime.fromtimestamp(t) - datetime.utcfromtimestamp(t)
737
return offset.days * 86400 + offset.seconds
739
weekdays = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
740
_default_format_by_weekday_num = [wd + " %Y-%m-%d %H:%M:%S" for wd in weekdays]
743
def format_date(t, offset=0, timezone='original', date_fmt=None,
562
if time.localtime(t).tm_isdst and time.daylight:
565
return -time.timezone
568
def format_date(t, offset=0, timezone='original', date_fmt=None,
744
569
show_offset=True):
745
"""Return a formatted date string.
747
:param t: Seconds since the epoch.
748
:param offset: Timezone offset in seconds east of utc.
749
:param timezone: How to display the time: 'utc', 'original' for the
750
timezone specified by offset, or 'local' for the process's current
752
:param date_fmt: strftime format.
753
:param show_offset: Whether to append the timezone.
755
(date_fmt, tt, offset_str) = \
756
_format_date(t, offset, timezone, date_fmt, show_offset)
757
date_fmt = date_fmt.replace('%a', weekdays[tt[6]])
758
date_str = time.strftime(date_fmt, tt)
759
return date_str + offset_str
762
# Cache of formatted offset strings
766
def format_date_with_offset_in_original_timezone(t, offset=0,
767
_cache=_offset_cache):
768
"""Return a formatted date string in the original timezone.
770
This routine may be faster then format_date.
772
:param t: Seconds since the epoch.
773
:param offset: Timezone offset in seconds east of utc.
777
tt = time.gmtime(t + offset)
778
date_fmt = _default_format_by_weekday_num[tt[6]]
779
date_str = time.strftime(date_fmt, tt)
780
offset_str = _cache.get(offset, None)
781
if offset_str is None:
782
offset_str = ' %+03d%02d' % (offset / 3600, (offset / 60) % 60)
783
_cache[offset] = offset_str
784
return date_str + offset_str
787
def format_local_date(t, offset=0, timezone='original', date_fmt=None,
789
"""Return an unicode date string formatted according to the current locale.
791
:param t: Seconds since the epoch.
792
:param offset: Timezone offset in seconds east of utc.
793
:param timezone: How to display the time: 'utc', 'original' for the
794
timezone specified by offset, or 'local' for the process's current
796
:param date_fmt: strftime format.
797
:param show_offset: Whether to append the timezone.
799
(date_fmt, tt, offset_str) = \
800
_format_date(t, offset, timezone, date_fmt, show_offset)
801
date_str = time.strftime(date_fmt, tt)
802
if not isinstance(date_str, unicode):
803
date_str = date_str.decode(get_user_encoding(), 'replace')
804
return date_str + offset_str
807
def _format_date(t, offset, timezone, date_fmt, show_offset):
570
## TODO: Perhaps a global option to use either universal or local time?
571
## Or perhaps just let people set $TZ?
572
assert isinstance(t, float)
808
574
if timezone == 'utc':
809
575
tt = time.gmtime(t)
935
raise errors.BzrError("sorry, %r not allowed in path" % f)
704
raise BzrError("sorry, %r not allowed in path" % f)
936
705
elif (f == '.') or (f == ''):
712
assert isinstance(p, list)
945
714
if (f == '..') or (f is None) or (f == ''):
946
raise errors.BzrError("sorry, %r not allowed in path" % f)
715
raise BzrError("sorry, %r not allowed in path" % f)
947
716
return pathjoin(*p)
950
def parent_directories(filename):
951
"""Return the list of parent directories, deepest first.
953
For example, parent_directories("a/b/c") -> ["a/b", "a"].
956
parts = splitpath(dirname(filename))
958
parents.append(joinpath(parts))
963
_extension_load_failures = []
966
def failed_to_load_extension(exception):
967
"""Handle failing to load a binary extension.
969
This should be called from the ImportError block guarding the attempt to
970
import the native extension. If this function returns, the pure-Python
971
implementation should be loaded instead::
974
>>> import bzrlib._fictional_extension_pyx
975
>>> except ImportError, e:
976
>>> bzrlib.osutils.failed_to_load_extension(e)
977
>>> import bzrlib._fictional_extension_py
979
# NB: This docstring is just an example, not a doctest, because doctest
980
# currently can't cope with the use of lazy imports in this namespace --
983
# This currently doesn't report the failure at the time it occurs, because
984
# they tend to happen very early in startup when we can't check config
985
# files etc, and also we want to report all failures but not spam the user
987
from bzrlib import trace
988
exception_str = str(exception)
989
if exception_str not in _extension_load_failures:
990
trace.mutter("failed to load compiled extension: %s" % exception_str)
991
_extension_load_failures.append(exception_str)
994
def report_extension_load_failures():
995
if not _extension_load_failures:
997
from bzrlib.config import GlobalConfig
998
if GlobalConfig().get_user_option_as_bool('ignore_missing_extensions'):
1000
# the warnings framework should by default show this only once
1001
from bzrlib.trace import warning
1003
"bzr: warning: some compiled extensions could not be loaded; "
1004
"see <https://answers.launchpad.net/bzr/+faq/703>")
1005
# we no longer show the specific missing extensions here, because it makes
1006
# the message too long and scary - see
1007
# https://bugs.launchpad.net/bzr/+bug/430529
1011
from bzrlib._chunks_to_lines_pyx import chunks_to_lines
1012
except ImportError, e:
1013
failed_to_load_extension(e)
1014
from bzrlib._chunks_to_lines_py import chunks_to_lines
719
@deprecated_function(zero_nine)
720
def appendpath(p1, p2):
724
return pathjoin(p1, p2)
1017
727
def split_lines(s):
1018
728
"""Split s into lines, but without removing the newline characters."""
1019
# Trivially convert a fulltext into a 'chunked' representation, and let
1020
# chunks_to_lines do the heavy lifting.
1021
if isinstance(s, str):
1022
# chunks_to_lines only supports 8-bit strings
1023
return chunks_to_lines([s])
1025
return _split_lines(s)
1028
def _split_lines(s):
1029
"""Split s into lines, but without removing the newline characters.
1031
This supports Unicode or plain string objects.
1033
729
lines = s.split('\n')
1034
730
result = [line + '\n' for line in lines[:-1]]
1156
794
os.path.commonprefix (python2.4) has a bad bug that it works just
1157
795
on string prefixes, assuming that '/u' is a prefix of '/u2'. This
1158
796
avoids that problem.
1160
NOTE: `base` should not have a trailing slash otherwise you'll get
1161
PathNotChild exceptions regardless of `path`.
1164
if len(base) < MIN_ABS_PATHLENGTH:
1165
# must have space for e.g. a drive letter
1166
raise ValueError('%r is too short to calculate a relative path'
799
assert len(base) >= MIN_ABS_PATHLENGTH, ('Length of base must be equal or'
800
' exceed the platform minimum length (which is %d)' %
1169
803
rp = abspath(path)
1174
if len(head) <= len(base) and head != base:
1175
raise errors.PathNotChild(rp, base)
807
while len(head) >= len(base):
1176
808
if head == base:
1178
head, tail = split(head)
810
head, tail = os.path.split(head)
814
raise PathNotChild(rp, base)
1183
return pathjoin(*reversed(s))
1188
def _cicp_canonical_relpath(base, path):
1189
"""Return the canonical path relative to base.
1191
Like relpath, but on case-insensitive-case-preserving file-systems, this
1192
will return the relpath as stored on the file-system rather than in the
1193
case specified in the input string, for all existing portions of the path.
1195
This will cause O(N) behaviour if called for every path in a tree; if you
1196
have a number of paths to convert, you should use canonical_relpaths().
1198
# TODO: it should be possible to optimize this for Windows by using the
1199
# win32 API FindFiles function to look for the specified name - but using
1200
# os.listdir() still gives us the correct, platform agnostic semantics in
1203
rel = relpath(base, path)
1204
# '.' will have been turned into ''
1208
abs_base = abspath(base)
1210
_listdir = os.listdir
1212
# use an explicit iterator so we can easily consume the rest on early exit.
1213
bit_iter = iter(rel.split('/'))
1214
for bit in bit_iter:
1217
next_entries = _listdir(current)
1218
except OSError: # enoent, eperm, etc
1219
# We can't find this in the filesystem, so just append the
1221
current = pathjoin(current, bit, *list(bit_iter))
1223
for look in next_entries:
1224
if lbit == look.lower():
1225
current = pathjoin(current, look)
1228
# got to the end, nothing matched, so we just return the
1229
# non-existing bits as they were specified (the filename may be
1230
# the target of a move, for example).
1231
current = pathjoin(current, bit, *list(bit_iter))
1233
return current[len(abs_base):].lstrip('/')
1235
# XXX - TODO - we need better detection/integration of case-insensitive
1236
# file-systems; Linux often sees FAT32 devices (or NFS-mounted OSX
1237
# filesystems), for example, so could probably benefit from the same basic
1238
# support there. For now though, only Windows and OSX get that support, and
1239
# they get it for *all* file-systems!
1240
if sys.platform in ('win32', 'darwin'):
1241
canonical_relpath = _cicp_canonical_relpath
1243
canonical_relpath = relpath
1245
def canonical_relpaths(base, paths):
1246
"""Create an iterable to canonicalize a sequence of relative paths.
1248
The intent is for this implementation to use a cache, vastly speeding
1249
up multiple transformations in the same directory.
1251
# but for now, we haven't optimized...
1252
return [canonical_relpath(base, p) for p in paths]
1254
822
def safe_unicode(unicode_or_utf8_string):
1255
823
"""Coerce unicode_or_utf8_string into unicode.
1257
825
If it is unicode, it is returned.
1258
Otherwise it is decoded from utf-8. If decoding fails, the exception is
1259
wrapped in a BzrBadParameterNotUnicode exception.
826
Otherwise it is decoded from utf-8. If a decoding error
827
occurs, it is wrapped as a If the decoding fails, the exception is wrapped
828
as a BzrBadParameter exception.
1261
830
if isinstance(unicode_or_utf8_string, unicode):
1262
831
return unicode_or_utf8_string
1264
833
return unicode_or_utf8_string.decode('utf8')
1265
834
except UnicodeDecodeError:
1266
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1269
def safe_utf8(unicode_or_utf8_string):
1270
"""Coerce unicode_or_utf8_string to a utf8 string.
1272
If it is a str, it is returned.
1273
If it is Unicode, it is encoded into a utf-8 string.
1275
if isinstance(unicode_or_utf8_string, str):
1276
# TODO: jam 20070209 This is overkill, and probably has an impact on
1277
# performance if we are dealing with lots of apis that want a
1280
# Make sure it is a valid utf-8 string
1281
unicode_or_utf8_string.decode('utf-8')
1282
except UnicodeDecodeError:
1283
raise errors.BzrBadParameterNotUnicode(unicode_or_utf8_string)
1284
return unicode_or_utf8_string
1285
return unicode_or_utf8_string.encode('utf-8')
1288
_revision_id_warning = ('Unicode revision ids were deprecated in bzr 0.15.'
1289
' Revision id generators should be creating utf8'
1293
def safe_revision_id(unicode_or_utf8_string, warn=True):
1294
"""Revision ids should now be utf8, but at one point they were unicode.
1296
:param unicode_or_utf8_string: A possibly Unicode revision_id. (can also be
1298
:param warn: Functions that are sanitizing user data can set warn=False
1299
:return: None or a utf8 revision id.
1301
if (unicode_or_utf8_string is None
1302
or unicode_or_utf8_string.__class__ == str):
1303
return unicode_or_utf8_string
1305
symbol_versioning.warn(_revision_id_warning, DeprecationWarning,
1307
return cache_utf8.encode(unicode_or_utf8_string)
1310
_file_id_warning = ('Unicode file ids were deprecated in bzr 0.15. File id'
1311
' generators should be creating utf8 file ids.')
1314
def safe_file_id(unicode_or_utf8_string, warn=True):
1315
"""File ids should now be utf8, but at one point they were unicode.
1317
This is the same as safe_utf8, except it uses the cached encode functions
1318
to save a little bit of performance.
1320
:param unicode_or_utf8_string: A possibly Unicode file_id. (can also be
1322
:param warn: Functions that are sanitizing user data can set warn=False
1323
:return: None or a utf8 file id.
1325
if (unicode_or_utf8_string is None
1326
or unicode_or_utf8_string.__class__ == str):
1327
return unicode_or_utf8_string
1329
symbol_versioning.warn(_file_id_warning, DeprecationWarning,
1331
return cache_utf8.encode(unicode_or_utf8_string)
835
raise BzrBadParameterNotUnicode(unicode_or_utf8_string)
1334
838
_platform_normalizes_filenames = False
1375
879
normalized_filename = _inaccessible_normalized_filename
1378
def set_signal_handler(signum, handler, restart_syscall=True):
1379
"""A wrapper for signal.signal that also calls siginterrupt(signum, False)
1380
on platforms that support that.
1382
:param restart_syscall: if set, allow syscalls interrupted by a signal to
1383
automatically restart (by calling `signal.siginterrupt(signum,
1384
False)`). May be ignored if the feature is not available on this
1385
platform or Python version.
1389
siginterrupt = signal.siginterrupt
1391
# This python implementation doesn't provide signal support, hence no
1394
except AttributeError:
1395
# siginterrupt doesn't exist on this platform, or for this version
1397
siginterrupt = lambda signum, flag: None
1399
def sig_handler(*args):
1400
# Python resets the siginterrupt flag when a signal is
1401
# received. <http://bugs.python.org/issue8354>
1402
# As a workaround for some cases, set it back the way we want it.
1403
siginterrupt(signum, False)
1404
# Now run the handler function passed to set_signal_handler.
1407
sig_handler = handler
1408
old_handler = signal.signal(signum, sig_handler)
1410
siginterrupt(signum, False)
1414
default_terminal_width = 80
1415
"""The default terminal width for ttys.
1417
This is defined so that higher levels can share a common fallback value when
1418
terminal_width() returns None.
1422
882
def terminal_width():
1423
"""Return terminal width.
1425
None is returned if the width can't established precisely.
1428
- if BZR_COLUMNS is set, returns its value
1429
- if there is no controlling terminal, returns None
1430
- if COLUMNS is set, returns its value,
1432
From there, we need to query the OS to get the size of the controlling
1436
- get termios.TIOCGWINSZ
1437
- if an error occurs or a negative value is obtained, returns None
1441
- win32utils.get_console_size() decides,
1442
- returns None on error (provided default value)
1445
# If BZR_COLUMNS is set, take it, user is always right
1447
return int(os.environ['BZR_COLUMNS'])
1448
except (KeyError, ValueError):
1451
isatty = getattr(sys.stdout, 'isatty', None)
1452
if isatty is None or not isatty():
1453
# Don't guess, setting BZR_COLUMNS is the recommended way to override.
1456
# If COLUMNS is set, take it, the terminal knows better (even inside a
1457
# given terminal, the application can decide to set COLUMNS to a lower
1458
# value (splitted screen) or a bigger value (scroll bars))
1460
return int(os.environ['COLUMNS'])
1461
except (KeyError, ValueError):
1464
width, height = _terminal_size(None, None)
1466
# Consider invalid values as meaning no width
1472
def _win32_terminal_size(width, height):
1473
width, height = win32utils.get_console_size(defaultx=width, defaulty=height)
1474
return width, height
1477
def _ioctl_terminal_size(width, height):
883
"""Return estimated terminal width."""
884
if sys.platform == 'win32':
885
import bzrlib.win32console
886
return bzrlib.win32console.get_console_size()[0]
1479
889
import struct, fcntl, termios
1480
890
s = struct.pack('HHHH', 0, 0, 0, 0)
1481
891
x = fcntl.ioctl(1, termios.TIOCGWINSZ, s)
1482
height, width = struct.unpack('HHHH', x)[0:2]
1483
except (IOError, AttributeError):
892
width = struct.unpack('HHHH', x)[1]
1485
return width, height
1487
_terminal_size = None
1488
"""Returns the terminal size as (width, height).
1490
:param width: Default value for width.
1491
:param height: Default value for height.
1493
This is defined specifically for each OS and query the size of the controlling
1494
terminal. If any error occurs, the provided default values should be returned.
1496
if sys.platform == 'win32':
1497
_terminal_size = _win32_terminal_size
1499
_terminal_size = _ioctl_terminal_size
1502
def _terminal_size_changed(signum, frame):
1503
"""Set COLUMNS upon receiving a SIGnal for WINdow size CHange."""
1504
width, height = _terminal_size(None, None)
1505
if width is not None:
1506
os.environ['COLUMNS'] = str(width)
1509
_registered_sigwinch = False
1510
def watch_sigwinch():
1511
"""Register for SIGWINCH, once and only once.
1513
Do nothing if the signal module is not available.
1515
global _registered_sigwinch
1516
if not _registered_sigwinch:
1519
if getattr(signal, "SIGWINCH", None) is not None:
1520
set_signal_handler(signal.SIGWINCH, _terminal_size_changed)
1522
# python doesn't provide signal support, nothing we can do about it
897
width = int(os.environ['COLUMNS'])
1524
_registered_sigwinch = True
1527
906
def supports_executable():
1528
907
return sys.platform != "win32"
1531
def supports_posix_readonly():
1532
"""Return True if 'readonly' has POSIX semantics, False otherwise.
1534
Notably, a win32 readonly file cannot be deleted, unlike POSIX where the
1535
directory controls creation/deletion, etc.
1537
And under win32, readonly means that the directory itself cannot be
1538
deleted. The contents of a readonly directory can be changed, unlike POSIX
1539
where files in readonly directories cannot be added, deleted or renamed.
1541
return sys.platform != "win32"
1544
910
def set_or_unset_env(env_variable, value):
1545
911
"""Modify the environment, setting or removing the env_variable.
1566
932
def check_legal_path(path):
1567
"""Check whether the supplied path is legal.
933
"""Check whether the supplied path is legal.
1568
934
This is only required on Windows, so we don't test on other platforms
1571
937
if sys.platform != "win32":
1573
939
if _validWin32PathRE.match(path) is None:
1574
raise errors.IllegalPath(path)
1577
_WIN32_ERROR_DIRECTORY = 267 # Similar to errno.ENOTDIR
1579
def _is_error_enotdir(e):
1580
"""Check if this exception represents ENOTDIR.
1582
Unfortunately, python is very inconsistent about the exception
1583
here. The cases are:
1584
1) Linux, Mac OSX all versions seem to set errno == ENOTDIR
1585
2) Windows, Python2.4, uses errno == ERROR_DIRECTORY (267)
1586
which is the windows error code.
1587
3) Windows, Python2.5 uses errno == EINVAL and
1588
winerror == ERROR_DIRECTORY
1590
:param e: An Exception object (expected to be OSError with an errno
1591
attribute, but we should be able to cope with anything)
1592
:return: True if this represents an ENOTDIR error. False otherwise.
1594
en = getattr(e, 'errno', None)
1595
if (en == errno.ENOTDIR
1596
or (sys.platform == 'win32'
1597
and (en == _WIN32_ERROR_DIRECTORY
1598
or (en == errno.EINVAL
1599
and getattr(e, 'winerror', None) == _WIN32_ERROR_DIRECTORY)
940
raise IllegalPath(path)
1605
943
def walkdirs(top, prefix=""):
1606
944
"""Yield data about all the directories in a tree.
1608
946
This yields all the data about the contents of a directory at a time.
1609
947
After each directory has been yielded, if the caller has mutated the list
1610
948
to exclude some directories, they are then not descended into.
1612
950
The data yielded is of the form:
1613
951
((directory-relpath, directory-path-from-top),
1614
[(relpath, basename, kind, lstat, path-from-top), ...]),
952
[(relpath, basename, kind, lstat), ...]),
1615
953
- directory-relpath is the relative path of the directory being returned
1616
954
with respect to top. prefix is prepended to this.
1617
- directory-path-from-root is the path including top for this directory.
955
- directory-path-from-root is the path including top for this directory.
1618
956
It is suitable for use with os functions.
1619
957
- relpath is the relative path within the subtree being walked.
1620
958
- basename is the basename of the path
1622
960
present within the tree - but it may be recorded as versioned. See
1624
962
- lstat is the stat data *if* the file was statted.
1625
- planned, not implemented:
963
- planned, not implemented:
1626
964
path_from_tree_root is the path from the root of the tree.
1628
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
966
:param prefix: Prefix the relpaths that are yielded with 'prefix'. This
1629
967
allows one to walk a subtree but get paths that are relative to a tree
1630
968
rooted higher up.
1631
969
:return: an iterator over the dirs.
1633
971
#TODO there is a bit of a smell where the results of the directory-
1634
# summary in this, and the path from the root, may not agree
972
# summary in this, and the path from the root, may not agree
1635
973
# depending on top and prefix - i.e. ./foo and foo as a pair leads to
1636
974
# potentially confusing output. We should make this more robust - but
1637
975
# not at a speed cost. RBC 20060731
1639
978
_directory = _directory_kind
1640
979
_listdir = os.listdir
1641
_kind_from_mode = file_kind_from_stat_mode
1642
pending = [(safe_unicode(prefix), "", _directory, None, safe_unicode(top))]
980
pending = [(prefix, "", _directory, None, top)]
983
currentdir = pending.pop()
1644
984
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1645
relroot, _, _, _, top = pending.pop()
1647
relprefix = relroot + u'/'
1650
top_slash = top + u'/'
1653
append = dirblock.append
1655
names = sorted(_listdir(top))
1657
if not _is_error_enotdir(e):
1661
abspath = top_slash + name
1662
statvalue = _lstat(abspath)
1663
kind = _kind_from_mode(statvalue.st_mode)
1664
append((relprefix + name, name, kind, statvalue, abspath))
1665
yield (relroot, top), dirblock
1667
# push the user specified dirs from dirblock
1668
pending.extend(d for d in reversed(dirblock) if d[2] == _directory)
1671
class DirReader(object):
1672
"""An interface for reading directories."""
1674
def top_prefix_to_starting_dir(self, top, prefix=""):
1675
"""Converts top and prefix to a starting dir entry
1677
:param top: A utf8 path
1678
:param prefix: An optional utf8 path to prefix output relative paths
1680
:return: A tuple starting with prefix, and ending with the native
1683
raise NotImplementedError(self.top_prefix_to_starting_dir)
1685
def read_dir(self, prefix, top):
1686
"""Read a specific dir.
1688
:param prefix: A utf8 prefix to be preprended to the path basenames.
1689
:param top: A natively encoded path to read.
1690
:return: A list of the directories contents. Each item contains:
1691
(utf8_relpath, utf8_name, kind, lstatvalue, native_abspath)
1693
raise NotImplementedError(self.read_dir)
1696
_selected_dir_reader = None
1699
def _walkdirs_utf8(top, prefix=""):
1700
"""Yield data about all the directories in a tree.
1702
This yields the same information as walkdirs() only each entry is yielded
1703
in utf-8. On platforms which have a filesystem encoding of utf8 the paths
1704
are returned as exact byte-strings.
1706
:return: yields a tuple of (dir_info, [file_info])
1707
dir_info is (utf8_relpath, path-from-top)
1708
file_info is (utf8_relpath, utf8_name, kind, lstat, path-from-top)
1709
if top is an absolute path, path-from-top is also an absolute path.
1710
path-from-top might be unicode or utf8, but it is the correct path to
1711
pass to os functions to affect the file in question. (such as os.lstat)
1713
global _selected_dir_reader
1714
if _selected_dir_reader is None:
1715
fs_encoding = _fs_enc.upper()
1716
if sys.platform == "win32" and win32utils.winver == 'Windows NT':
1717
# Win98 doesn't have unicode apis like FindFirstFileW
1718
# TODO: We possibly could support Win98 by falling back to the
1719
# original FindFirstFile, and using TCHAR instead of WCHAR,
1720
# but that gets a bit tricky, and requires custom compiling
1723
from bzrlib._walkdirs_win32 import Win32ReadDir
1724
_selected_dir_reader = Win32ReadDir()
1727
elif fs_encoding in ('UTF-8', 'US-ASCII', 'ANSI_X3.4-1968'):
1728
# ANSI_X3.4-1968 is a form of ASCII
1730
from bzrlib._readdir_pyx import UTF8DirReader
1731
_selected_dir_reader = UTF8DirReader()
1732
except ImportError, e:
1733
failed_to_load_extension(e)
1736
if _selected_dir_reader is None:
1737
# Fallback to the python version
1738
_selected_dir_reader = UnicodeDirReader()
1740
# 0 - relpath, 1- basename, 2- kind, 3- stat, 4-toppath
1741
# But we don't actually uses 1-3 in pending, so set them to None
1742
pending = [[_selected_dir_reader.top_prefix_to_starting_dir(top, prefix)]]
1743
read_dir = _selected_dir_reader.read_dir
1744
_directory = _directory_kind
1746
relroot, _, _, _, top = pending[-1].pop()
1749
dirblock = sorted(read_dir(relroot, top))
1750
yield (relroot, top), dirblock
1751
# push the user specified dirs from dirblock
1752
next = [d for d in reversed(dirblock) if d[2] == _directory]
1754
pending.append(next)
1757
class UnicodeDirReader(DirReader):
1758
"""A dir reader for non-utf8 file systems, which transcodes."""
1760
__slots__ = ['_utf8_encode']
1763
self._utf8_encode = codecs.getencoder('utf8')
1765
def top_prefix_to_starting_dir(self, top, prefix=""):
1766
"""See DirReader.top_prefix_to_starting_dir."""
1767
return (safe_utf8(prefix), None, None, None, safe_unicode(top))
1769
def read_dir(self, prefix, top):
1770
"""Read a single directory from a non-utf8 file system.
1772
top, and the abspath element in the output are unicode, all other paths
1773
are utf8. Local disk IO is done via unicode calls to listdir etc.
1775
This is currently the fallback code path when the filesystem encoding is
1776
not UTF-8. It may be better to implement an alternative so that we can
1777
safely handle paths that are not properly decodable in the current
1780
See DirReader.read_dir for details.
1782
_utf8_encode = self._utf8_encode
1784
_listdir = os.listdir
1785
_kind_from_mode = file_kind_from_stat_mode
1788
relprefix = prefix + '/'
1791
top_slash = top + u'/'
1794
append = dirblock.append
987
relroot = currentdir[0] + '/'
1795
990
for name in sorted(_listdir(top)):
1797
name_utf8 = _utf8_encode(name)[0]
1798
except UnicodeDecodeError:
1799
raise errors.BadFilenameEncoding(
1800
_utf8_encode(relprefix)[0] + name, _fs_enc)
1801
abspath = top_slash + name
1802
statvalue = _lstat(abspath)
1803
kind = _kind_from_mode(statvalue.st_mode)
1804
append((relprefix + name_utf8, name_utf8, kind, statvalue, abspath))
991
abspath = top + '/' + name
992
statvalue = lstat(abspath)
993
dirblock.append((relroot + name, name,
994
file_kind_from_stat_mode(statvalue.st_mode),
996
yield (currentdir[0], top), dirblock
997
# push the user specified dirs from dirblock
998
for dir in reversed(dirblock):
999
if dir[2] == _directory:
1808
1003
def copy_tree(from_path, to_path, handlers={}):
1809
1004
"""Copy all of the entries in from_path into to_path.
1811
:param from_path: The base directory to copy.
1006
:param from_path: The base directory to copy.
1812
1007
:param to_path: The target directory. If it does not exist, it will
1814
1009
:param handlers: A dictionary of functions, which takes a source and
1933
1092
' doesn\'t support the locale set by $LANG (%s)\n'
1934
1093
" Continuing with ascii encoding.\n"
1935
1094
% (e, os.environ.get('LANG')))
1936
user_encoding = 'ascii'
1938
# Windows returns 'cp0' to indicate there is no code page. So we'll just
1939
# treat that as ASCII, and not support printing unicode characters to the
1942
# For python scripts run under vim, we get '', so also treat that as ASCII
1943
if user_encoding in (None, 'cp0', ''):
1944
user_encoding = 'ascii'
1948
codecs.lookup(user_encoding)
1950
sys.stderr.write('bzr: warning:'
1951
' unknown encoding %s.'
1952
' Continuing with ascii encoding.\n'
1955
user_encoding = 'ascii'
1958
_cached_user_encoding = user_encoding
1960
return user_encoding
1963
def get_host_name():
1964
"""Return the current unicode host name.
1966
This is meant to be used in place of socket.gethostname() because that
1967
behaves inconsistently on different platforms.
1969
if sys.platform == "win32":
1971
return win32utils.get_host_name()
1974
return socket.gethostname().decode(get_user_encoding())
1977
# We must not read/write any more than 64k at a time from/to a socket so we
1978
# don't risk "no buffer space available" errors on some platforms. Windows in
1979
# particular is likely to throw WSAECONNABORTED or WSAENOBUFS if given too much
1981
MAX_SOCKET_CHUNK = 64 * 1024
1983
def read_bytes_from_socket(sock, report_activity=None,
1984
max_read_size=MAX_SOCKET_CHUNK):
1985
"""Read up to max_read_size of bytes from sock and notify of progress.
1987
Translates "Connection reset by peer" into file-like EOF (return an
1988
empty string rather than raise an error), and repeats the recv if
1989
interrupted by a signal.
1993
bytes = sock.recv(max_read_size)
1994
except socket.error, e:
1996
if eno == getattr(errno, "WSAECONNRESET", errno.ECONNRESET):
1997
# The connection was closed by the other side. Callers expect
1998
# an empty string to signal end-of-stream.
2000
elif eno == errno.EINTR:
2001
# Retry the interrupted recv.
2005
if report_activity is not None:
2006
report_activity(len(bytes), 'read')
2010
def recv_all(socket, count):
2011
"""Receive an exact number of bytes.
2013
Regular Socket.recv() may return less than the requested number of bytes,
2014
depending on what's in the OS buffer. MSG_WAITALL is not available
2015
on all platforms, but this should work everywhere. This will return
2016
less than the requested amount if the remote end closes.
2018
This isn't optimized and is intended mostly for use in testing.
2021
while len(b) < count:
2022
new = read_bytes_from_socket(socket, None, count - len(b))
2029
def send_all(sock, bytes, report_activity=None):
2030
"""Send all bytes on a socket.
2032
Breaks large blocks in smaller chunks to avoid buffering limitations on
2033
some platforms, and catches EINTR which may be thrown if the send is
2034
interrupted by a signal.
2036
This is preferred to socket.sendall(), because it avoids portability bugs
2037
and provides activity reporting.
2039
:param report_activity: Call this as bytes are read, see
2040
Transport._report_activity
2043
byte_count = len(bytes)
2044
while sent_total < byte_count:
2046
sent = sock.send(buffer(bytes, sent_total, MAX_SOCKET_CHUNK))
2047
except socket.error, e:
2048
if e.args[0] != errno.EINTR:
2052
report_activity(sent, 'write')
2055
def dereference_path(path):
2056
"""Determine the real path to a file.
2058
All parent elements are dereferenced. But the file itself is not
2060
:param path: The original path. May be absolute or relative.
2061
:return: the real path *to* the file
2063
parent, base = os.path.split(path)
2064
# The pathjoin for '.' is a workaround for Python bug #1213894.
2065
# (initial path components aren't dereferenced)
2066
return pathjoin(realpath(pathjoin('.', parent)), base)
2069
def supports_mapi():
2070
"""Return True if we can use MAPI to launch a mail client."""
2071
return sys.platform == "win32"
2074
def resource_string(package, resource_name):
2075
"""Load a resource from a package and return it as a string.
2077
Note: Only packages that start with bzrlib are currently supported.
2079
This is designed to be a lightweight implementation of resource
2080
loading in a way which is API compatible with the same API from
2082
http://peak.telecommunity.com/DevCenter/PkgResources#basic-resource-access.
2083
If and when pkg_resources becomes a standard library, this routine
2086
# Check package name is within bzrlib
2087
if package == "bzrlib":
2088
resource_relpath = resource_name
2089
elif package.startswith("bzrlib."):
2090
package = package[len("bzrlib."):].replace('.', os.sep)
2091
resource_relpath = pathjoin(package, resource_name)
2093
raise errors.BzrError('resource package %s not in bzrlib' % package)
2095
# Map the resource to a file and read its contents
2096
base = dirname(bzrlib.__file__)
2097
if getattr(sys, 'frozen', None): # bzr.exe
2098
base = abspath(pathjoin(base, '..', '..'))
2099
filename = pathjoin(base, resource_relpath)
2100
return open(filename, 'rU').read()
2103
def file_kind_from_stat_mode_thunk(mode):
2104
global file_kind_from_stat_mode
2105
if file_kind_from_stat_mode is file_kind_from_stat_mode_thunk:
2107
from bzrlib._readdir_pyx import UTF8DirReader
2108
file_kind_from_stat_mode = UTF8DirReader().kind_from_mode
2109
except ImportError, e:
2110
# This is one time where we won't warn that an extension failed to
2111
# load. The extension is never available on Windows anyway.
2112
from bzrlib._readdir_py import (
2113
_kind_from_mode as file_kind_from_stat_mode
2115
return file_kind_from_stat_mode(mode)
2116
file_kind_from_stat_mode = file_kind_from_stat_mode_thunk
2119
def file_kind(f, _lstat=os.lstat):
2121
return file_kind_from_stat_mode(_lstat(f).st_mode)
2123
if getattr(e, 'errno', None) in (errno.ENOENT, errno.ENOTDIR):
2124
raise errors.NoSuchFile(f)
2128
def until_no_eintr(f, *a, **kw):
2129
"""Run f(*a, **kw), retrying if an EINTR error occurs.
2131
WARNING: you must be certain that it is safe to retry the call repeatedly
2132
if EINTR does occur. This is typically only true for low-level operations
2133
like os.read. If in any doubt, don't use this.
2135
Keep in mind that this is not a complete solution to EINTR. There is
2136
probably code in the Python standard library and other dependencies that
2137
may encounter EINTR if a signal arrives (and there is signal handler for
2138
that signal). So this function can reduce the impact for IO that bzrlib
2139
directly controls, but it is not a complete solution.
2141
# Borrowed from Twisted's twisted.python.util.untilConcludes function.
2145
except (IOError, OSError), e:
2146
if e.errno == errno.EINTR:
2151
def re_compile_checked(re_string, flags=0, where=""):
2152
"""Return a compiled re, or raise a sensible error.
2154
This should only be used when compiling user-supplied REs.
2156
:param re_string: Text form of regular expression.
2157
:param flags: eg re.IGNORECASE
2158
:param where: Message explaining to the user the context where
2159
it occurred, eg 'log search filter'.
2161
# from https://bugs.launchpad.net/bzr/+bug/251352
2163
re_obj = re.compile(re_string, flags)
2168
where = ' in ' + where
2169
# despite the name 'error' is a type
2170
raise errors.BzrCommandError('Invalid regular expression%s: %r: %s'
2171
% (where, re_string, e))
2174
if sys.platform == "win32":
2177
return msvcrt.getch()
2182
fd = sys.stdin.fileno()
2183
settings = termios.tcgetattr(fd)
2186
ch = sys.stdin.read(1)
2188
termios.tcsetattr(fd, termios.TCSADRAIN, settings)
2192
if sys.platform == 'linux2':
2193
def _local_concurrency():
2195
prefix = 'processor'
2196
for line in file('/proc/cpuinfo', 'rb'):
2197
if line.startswith(prefix):
2198
concurrency = int(line[line.find(':')+1:]) + 1
2200
elif sys.platform == 'darwin':
2201
def _local_concurrency():
2202
return subprocess.Popen(['sysctl', '-n', 'hw.availcpu'],
2203
stdout=subprocess.PIPE).communicate()[0]
2204
elif sys.platform[0:7] == 'freebsd':
2205
def _local_concurrency():
2206
return subprocess.Popen(['sysctl', '-n', 'hw.ncpu'],
2207
stdout=subprocess.PIPE).communicate()[0]
2208
elif sys.platform == 'sunos5':
2209
def _local_concurrency():
2210
return subprocess.Popen(['psrinfo', '-p',],
2211
stdout=subprocess.PIPE).communicate()[0]
2212
elif sys.platform == "win32":
2213
def _local_concurrency():
2214
# This appears to return the number of cores.
2215
return os.environ.get('NUMBER_OF_PROCESSORS')
2217
def _local_concurrency():
2222
_cached_local_concurrency = None
2224
def local_concurrency(use_cache=True):
2225
"""Return how many processes can be run concurrently.
2227
Rely on platform specific implementations and default to 1 (one) if
2228
anything goes wrong.
2230
global _cached_local_concurrency
2232
if _cached_local_concurrency is not None and use_cache:
2233
return _cached_local_concurrency
2235
concurrency = os.environ.get('BZR_CONCURRENCY', None)
2236
if concurrency is None:
2238
concurrency = _local_concurrency()
2239
except (OSError, IOError):
2242
concurrency = int(concurrency)
2243
except (TypeError, ValueError):
2246
_cached_concurrency = concurrency
2250
class UnicodeOrBytesToBytesWriter(codecs.StreamWriter):
2251
"""A stream writer that doesn't decode str arguments."""
2253
def __init__(self, encode, stream, errors='strict'):
2254
codecs.StreamWriter.__init__(self, stream, errors)
2255
self.encode = encode
2257
def write(self, object):
2258
if type(object) is str:
2259
self.stream.write(object)
2261
data, _ = self.encode(object, self.errors)
2262
self.stream.write(data)
2264
if sys.platform == 'win32':
2265
def open_file(filename, mode='r', bufsize=-1):
2266
"""This function is used to override the ``open`` builtin.
2268
But it uses O_NOINHERIT flag so the file handle is not inherited by
2269
child processes. Deleting or renaming a closed file opened with this
2270
function is not blocking child processes.
2272
writing = 'w' in mode
2273
appending = 'a' in mode
2274
updating = '+' in mode
2275
binary = 'b' in mode
2278
# see http://msdn.microsoft.com/en-us/library/yeby3zcb%28VS.71%29.aspx
2279
# for flags for each modes.
2289
flags |= os.O_WRONLY
2290
flags |= os.O_CREAT | os.O_TRUNC
2295
flags |= os.O_WRONLY
2296
flags |= os.O_CREAT | os.O_APPEND
2301
flags |= os.O_RDONLY
2303
return os.fdopen(os.open(filename, flags), mode, bufsize)
2308
def getuser_unicode():
2309
"""Return the username as unicode.
2312
user_encoding = get_user_encoding()
2313
username = getpass.getuser().decode(user_encoding)
2314
except UnicodeDecodeError:
2315
raise errors.BzrError("Can't decode username as %s." % \
1096
if _cached_user_encoding is None:
1097
_cached_user_encoding = 'ascii'
1098
return _cached_user_encoding
added
removed
bzrlib/osutils.py
