65
58
return unichr(int(code[1:])).encode('utf8')
61
_unescape_re = lazy_regex.lazy_compile('\&([^;]*);')
71
63
def _unescape_xml(data):
72
64
"""Unescape predefined XML entities in a string of data."""
74
if _unescape_re is None:
75
_unescape_re = re.compile('\&([^;]*);')
76
65
return _unescape_re.sub(_unescaper, data)
79
def _ensure_utf8_re():
80
"""Make sure the _utf8_re and _unicode_re regexes have been compiled."""
81
global _utf8_re, _unicode_re
83
_utf8_re = re.compile('[&<>\'\"]|[\x80-\xff]+')
84
if _unicode_re is None:
85
_unicode_re = re.compile(u'[&<>\'\"\u0080-\uffff]')
88
def _unicode_escape_replace(match, _map=_xml_escape_map):
89
"""Replace a string of non-ascii, non XML safe characters with their escape
91
This will escape both Standard XML escapes, like <>"', etc.
92
As well as escaping non ascii characters, because ElementTree did.
93
This helps us remain compatible to older versions of bzr. We may change
94
our policy in the future, though.
96
# jam 20060816 Benchmarks show that try/KeyError is faster if you
97
# expect the entity to rarely miss. There is about a 10% difference
98
# in overall time. But if you miss frequently, then if None is much
99
# faster. For our use case, we *rarely* have a revision id, file id
100
# or path name that is unicode. So use try/KeyError.
102
return _map[match.group()]
104
return "&#%d;" % ord(match.group())
107
def _utf8_escape_replace(match, _map=_xml_escape_map):
108
"""Escape utf8 characters into XML safe ones.
110
This uses 2 tricks. It is either escaping "standard" characters, like "&<>,
111
or it is handling characters with the high-bit set. For ascii characters,
112
we just lookup the replacement in the dictionary. For everything else, we
113
decode back into Unicode, and then use the XML escape code.
116
return _map[match.group()]
118
return ''.join('&#%d;' % ord(uni_chr)
119
for uni_chr in match.group().decode('utf8'))
124
def _encode_and_escape(unicode_or_utf8_str, _map=_to_escaped_map):
125
"""Encode the string into utf8, and escape invalid XML characters"""
126
# We frequently get entities we have not seen before, so it is better
127
# to check if None, rather than try/KeyError
128
text = _map.get(unicode_or_utf8_str)
130
if unicode_or_utf8_str.__class__ is unicode:
131
# The alternative policy is to do a regular UTF8 encoding
132
# and then escape only XML meta characters.
133
# Performance is equivalent once you use cache_utf8. *However*
134
# this makes the serialized texts incompatible with old versions
135
# of bzr. So no net gain. (Perhaps the read code would handle utf8
136
# better than entity escapes, but cElementTree seems to do just fine
138
text = str(_unicode_re.sub(_unicode_escape_replace,
139
unicode_or_utf8_str)) + '"'
141
# Plain strings are considered to already be in utf-8 so we do a
142
# slightly different method for escaping.
143
text = _utf8_re.sub(_utf8_escape_replace,
144
unicode_or_utf8_str) + '"'
145
_map[unicode_or_utf8_str] = text
149
def _get_utf8_or_ascii(a_str,
150
_encode_utf8=cache_utf8.encode,
151
_get_cached_ascii=cache_utf8.get_cached_ascii):
152
"""Return a cached version of the string.
154
cElementTree will return a plain string if the XML is plain ascii. It only
155
returns Unicode when it needs to. We want to work in utf-8 strings. So if
156
cElementTree returns a plain string, we can just return the cached version.
157
If it is Unicode, then we need to encode it.
159
:param a_str: An 8-bit string or Unicode as returned by
160
cElementTree.Element.get()
161
:return: A utf-8 encoded 8-bit string.
163
# This is fairly optimized because we know what cElementTree does, this is
164
# not meant as a generic function for all cases. Because it is possible for
165
# an 8-bit string to not be ascii or valid utf8.
166
if a_str.__class__ is unicode:
167
return _encode_utf8(a_str)
173
"""Clean out the unicode => escaped map"""
174
_to_escaped_map.clear()
177
68
class Serializer_v8(XMLSerializer):
178
69
"""This serialiser adds rich roots.
261
152
reference_revision, symlink_target.
262
153
:return: The inventory as a list of lines.
265
self._check_revisions(inv)
267
156
append = output.append
268
157
self._append_inventory_root(append, inv)
269
entries = inv.iter_entries()
271
root_path, root_ie = entries.next()
272
for path, ie in entries:
273
if ie.parent_id != self.root_id:
274
parent_str = ' parent_id="'
275
parent_id = _encode_and_escape(ie.parent_id)
279
if ie.kind == 'file':
281
executable = ' executable="yes"'
285
append('<file%s file_id="%s name="%s%s%s revision="%s '
286
'text_sha1="%s" text_size="%d" />\n' % (
287
executable, _encode_and_escape(ie.file_id),
288
_encode_and_escape(ie.name), parent_str, parent_id,
289
_encode_and_escape(ie.revision), ie.text_sha1,
292
append('<file%s file_id="%s name="%s%s%s />\n' % (
293
executable, _encode_and_escape(ie.file_id),
294
_encode_and_escape(ie.name), parent_str, parent_id))
295
elif ie.kind == 'directory':
297
append('<directory file_id="%s name="%s%s%s revision="%s '
299
_encode_and_escape(ie.file_id),
300
_encode_and_escape(ie.name),
301
parent_str, parent_id,
302
_encode_and_escape(ie.revision)))
304
append('<directory file_id="%s name="%s%s%s />\n' % (
305
_encode_and_escape(ie.file_id),
306
_encode_and_escape(ie.name),
307
parent_str, parent_id))
308
elif ie.kind == 'symlink':
310
append('<symlink file_id="%s name="%s%s%s revision="%s '
311
'symlink_target="%s />\n' % (
312
_encode_and_escape(ie.file_id),
313
_encode_and_escape(ie.name),
314
parent_str, parent_id,
315
_encode_and_escape(ie.revision),
316
_encode_and_escape(ie.symlink_target)))
318
append('<symlink file_id="%s name="%s%s%s />\n' % (
319
_encode_and_escape(ie.file_id),
320
_encode_and_escape(ie.name),
321
parent_str, parent_id))
322
elif ie.kind == 'tree-reference':
323
if ie.kind not in self.supported_kinds:
324
raise errors.UnsupportedInventoryKind(ie.kind)
326
append('<tree-reference file_id="%s name="%s%s%s '
327
'revision="%s reference_revision="%s />\n' % (
328
_encode_and_escape(ie.file_id),
329
_encode_and_escape(ie.name),
330
parent_str, parent_id,
331
_encode_and_escape(ie.revision),
332
_encode_and_escape(ie.reference_revision)))
334
append('<tree-reference file_id="%s name="%s%s%s />\n' % (
335
_encode_and_escape(ie.file_id),
336
_encode_and_escape(ie.name),
337
parent_str, parent_id))
339
raise errors.UnsupportedInventoryKind(ie.kind)
340
append('</inventory>\n')
158
serialize_inventory_flat(inv, append,
159
self.root_id, self.supported_kinds, working)
341
160
if f is not None:
342
161
f.writelines(output)
343
162
# Just to keep the cache from growing without bounds
349
168
"""Append the inventory root to output."""
350
169
if inv.revision_id is not None:
351
170
revid1 = ' revision_id="'
352
revid2 = _encode_and_escape(inv.revision_id)
171
revid2 = encode_and_escape(inv.revision_id)
356
175
append('<inventory format="%s"%s%s>\n' % (
357
176
self.format_num, revid1, revid2))
358
177
append('<directory file_id="%s name="%s revision="%s />\n' % (
359
_encode_and_escape(inv.root.file_id),
360
_encode_and_escape(inv.root.name),
361
_encode_and_escape(inv.root.revision)))
178
encode_and_escape(inv.root.file_id),
179
encode_and_escape(inv.root.name),
180
encode_and_escape(inv.root.revision)))
363
182
def _pack_revision(self, rev):
364
183
"""Revision object -> xml tree"""
408
227
prop_elt.tail = '\n'
409
228
top_elt.tail = '\n'
230
def _unpack_entry(self, elt, entry_cache=None, return_from_cache=False):
231
# This is here because it's overridden by xml7
232
return unpack_inventory_entry(elt, entry_cache,
411
235
def _unpack_inventory(self, elt, revision_id=None, entry_cache=None,
412
236
return_from_cache=False):
413
237
"""Construct from XML Element"""
414
if elt.tag != 'inventory':
415
raise errors.UnexpectedInventoryFormat('Root tag is %r' % elt.tag)
416
format = elt.get('format')
417
if format != self.format_num:
418
raise errors.UnexpectedInventoryFormat('Invalid format version %r'
420
revision_id = elt.get('revision_id')
421
if revision_id is not None:
422
revision_id = cache_utf8.encode(revision_id)
423
inv = inventory.Inventory(root_id=None, revision_id=revision_id)
425
ie = self._unpack_entry(e, entry_cache=entry_cache,
426
return_from_cache=return_from_cache)
238
inv = unpack_inventory_flat(elt, self.format_num, self._unpack_entry,
239
entry_cache, return_from_cache)
428
240
self._check_cache_size(len(inv), entry_cache)
431
def _unpack_entry(self, elt, entry_cache=None, return_from_cache=False):
433
file_id = elt_get('file_id')
434
revision = elt_get('revision')
435
# Check and see if we have already unpacked this exact entry
436
# Some timings for "repo.revision_trees(last_100_revs)"
438
# unmodified 4.1s 40.8s
440
# using fifo 2.83s 29.1s
444
# no_copy 2.00s 20.5s
445
# no_c,dict 1.95s 18.0s
446
# Note that a cache of 10k nodes is more than sufficient to hold all of
447
# the inventory for the last 100 revs for bzr, but not for mysql (20k
448
# is enough for mysql, which saves the same 2s as using a dict)
450
# Breakdown of mysql using time.clock()
451
# 4.1s 2 calls to element.get for file_id, revision_id
452
# 4.5s cache_hit lookup
453
# 7.1s InventoryFile.copy()
454
# 2.4s InventoryDirectory.copy()
455
# 0.4s decoding unique entries
456
# 1.6s decoding entries after FIFO fills up
457
# 0.8s Adding nodes to FIFO (including flushes)
458
# 0.1s cache miss lookups
460
# 4.1s 2 calls to element.get for file_id, revision_id
461
# 9.9s cache_hit lookup
462
# 10.8s InventoryEntry.copy()
463
# 0.3s cache miss lookus
464
# 1.2s decoding entries
465
# 1.0s adding nodes to LRU
466
if entry_cache is not None and revision is not None:
467
key = (file_id, revision)
469
# We copy it, because some operations may mutate it
470
cached_ie = entry_cache[key]
474
# Only copying directory entries drops us 2.85s => 2.35s
475
if return_from_cache:
476
if cached_ie.kind == 'directory':
477
return cached_ie.copy()
479
return cached_ie.copy()
482
if not InventoryEntry.versionable_kind(kind):
483
raise AssertionError('unsupported entry kind %s' % kind)
485
get_cached = _get_utf8_or_ascii
487
file_id = get_cached(file_id)
488
if revision is not None:
489
revision = get_cached(revision)
490
parent_id = elt_get('parent_id')
491
if parent_id is not None:
492
parent_id = get_cached(parent_id)
494
if kind == 'directory':
495
ie = inventory.InventoryDirectory(file_id,
499
ie = inventory.InventoryFile(file_id,
502
ie.text_sha1 = elt_get('text_sha1')
503
if elt_get('executable') == 'yes':
505
v = elt_get('text_size')
506
ie.text_size = v and int(v)
507
elif kind == 'symlink':
508
ie = inventory.InventoryLink(file_id,
511
ie.symlink_target = elt_get('symlink_target')
513
raise errors.UnsupportedInventoryKind(kind)
514
ie.revision = revision
515
if revision is not None and entry_cache is not None:
516
# We cache a copy() because callers like to mutate objects, and
517
# that would cause the item in cache to mutate as well.
518
# This has a small effect on many-inventory performance, because
519
# the majority fraction is spent in cache hits, not misses.
520
entry_cache[key] = ie.copy()
524
243
def _unpack_revision(self, elt):
525
244
"""XML Element -> Revision object"""
526
245
format = elt.get('format')
added
removed
bzrlib/xml8.py
