130
135
return re.subn(u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
131
136
lambda match: match.group(0).encode('unicode_escape'),
140
def get_utf8_or_ascii(a_str, _encode_utf8=cache_utf8.encode):
141
"""Return a cached version of the string.
143
cElementTree will return a plain string if the XML is plain ascii. It only
144
returns Unicode when it needs to. We want to work in utf-8 strings. So if
145
cElementTree returns a plain string, we can just return the cached version.
146
If it is Unicode, then we need to encode it.
148
:param a_str: An 8-bit string or Unicode as returned by
149
cElementTree.Element.get()
150
:return: A utf-8 encoded 8-bit string.
152
# This is fairly optimized because we know what cElementTree does, this is
153
# not meant as a generic function for all cases. Because it is possible for
154
# an 8-bit string to not be ascii or valid utf8.
155
if a_str.__class__ is unicode:
156
return _encode_utf8(a_str)
161
_utf8_re = lazy_regex.lazy_compile('[&<>\'\"]|[\x80-\xff]+')
162
_unicode_re = lazy_regex.lazy_compile(u'[&<>\'\"\u0080-\uffff]')
167
"'":"'", # FIXME: overkill
174
def _unicode_escape_replace(match, _map=_xml_escape_map):
175
"""Replace a string of non-ascii, non XML safe characters with their escape
177
This will escape both Standard XML escapes, like <>"', etc.
178
As well as escaping non ascii characters, because ElementTree did.
179
This helps us remain compatible to older versions of bzr. We may change
180
our policy in the future, though.
182
# jam 20060816 Benchmarks show that try/KeyError is faster if you
183
# expect the entity to rarely miss. There is about a 10% difference
184
# in overall time. But if you miss frequently, then if None is much
185
# faster. For our use case, we *rarely* have a revision id, file id
186
# or path name that is unicode. So use try/KeyError.
188
return _map[match.group()]
190
return "&#%d;" % ord(match.group())
193
def _utf8_escape_replace(match, _map=_xml_escape_map):
194
"""Escape utf8 characters into XML safe ones.
196
This uses 2 tricks. It is either escaping "standard" characters, like "&<>,
197
or it is handling characters with the high-bit set. For ascii characters,
198
we just lookup the replacement in the dictionary. For everything else, we
199
decode back into Unicode, and then use the XML escape code.
202
return _map[match.group()]
204
return ''.join('&#%d;' % ord(uni_chr)
205
for uni_chr in match.group().decode('utf8'))
210
def encode_and_escape(unicode_or_utf8_str, _map=_to_escaped_map):
211
"""Encode the string into utf8, and escape invalid XML characters"""
212
# We frequently get entities we have not seen before, so it is better
213
# to check if None, rather than try/KeyError
214
text = _map.get(unicode_or_utf8_str)
216
if unicode_or_utf8_str.__class__ is unicode:
217
# The alternative policy is to do a regular UTF8 encoding
218
# and then escape only XML meta characters.
219
# Performance is equivalent once you use cache_utf8. *However*
220
# this makes the serialized texts incompatible with old versions
221
# of bzr. So no net gain. (Perhaps the read code would handle utf8
222
# better than entity escapes, but cElementTree seems to do just fine
224
text = str(_unicode_re.sub(_unicode_escape_replace,
225
unicode_or_utf8_str)) + '"'
227
# Plain strings are considered to already be in utf-8 so we do a
228
# slightly different method for escaping.
229
text = _utf8_re.sub(_utf8_escape_replace,
230
unicode_or_utf8_str) + '"'
231
_map[unicode_or_utf8_str] = text
236
"""Clean out the unicode => escaped map"""
237
_to_escaped_map.clear()
240
def unpack_inventory_entry(elt, entry_cache=None, return_from_cache=False):
242
file_id = elt_get('file_id')
243
revision = elt_get('revision')
244
# Check and see if we have already unpacked this exact entry
245
# Some timings for "repo.revision_trees(last_100_revs)"
247
# unmodified 4.1s 40.8s
249
# using fifo 2.83s 29.1s
253
# no_copy 2.00s 20.5s
254
# no_c,dict 1.95s 18.0s
255
# Note that a cache of 10k nodes is more than sufficient to hold all of
256
# the inventory for the last 100 revs for bzr, but not for mysql (20k
257
# is enough for mysql, which saves the same 2s as using a dict)
259
# Breakdown of mysql using time.clock()
260
# 4.1s 2 calls to element.get for file_id, revision_id
261
# 4.5s cache_hit lookup
262
# 7.1s InventoryFile.copy()
263
# 2.4s InventoryDirectory.copy()
264
# 0.4s decoding unique entries
265
# 1.6s decoding entries after FIFO fills up
266
# 0.8s Adding nodes to FIFO (including flushes)
267
# 0.1s cache miss lookups
269
# 4.1s 2 calls to element.get for file_id, revision_id
270
# 9.9s cache_hit lookup
271
# 10.8s InventoryEntry.copy()
272
# 0.3s cache miss lookus
273
# 1.2s decoding entries
274
# 1.0s adding nodes to LRU
275
if entry_cache is not None and revision is not None:
276
key = (file_id, revision)
278
# We copy it, because some operations may mutate it
279
cached_ie = entry_cache[key]
283
# Only copying directory entries drops us 2.85s => 2.35s
284
if return_from_cache:
285
if cached_ie.kind == 'directory':
286
return cached_ie.copy()
288
return cached_ie.copy()
291
if not inventory.InventoryEntry.versionable_kind(kind):
292
raise AssertionError('unsupported entry kind %s' % kind)
294
file_id = get_utf8_or_ascii(file_id)
295
if revision is not None:
296
revision = get_utf8_or_ascii(revision)
297
parent_id = elt_get('parent_id')
298
if parent_id is not None:
299
parent_id = get_utf8_or_ascii(parent_id)
301
if kind == 'directory':
302
ie = inventory.InventoryDirectory(file_id,
306
ie = inventory.InventoryFile(file_id,
309
ie.text_sha1 = elt_get('text_sha1')
310
if elt_get('executable') == 'yes':
312
v = elt_get('text_size')
313
ie.text_size = v and int(v)
314
elif kind == 'symlink':
315
ie = inventory.InventoryLink(file_id,
318
ie.symlink_target = elt_get('symlink_target')
320
raise errors.UnsupportedInventoryKind(kind)
321
ie.revision = revision
322
if revision is not None and entry_cache is not None:
323
# We cache a copy() because callers like to mutate objects, and
324
# that would cause the item in cache to mutate as well.
325
# This has a small effect on many-inventory performance, because
326
# the majority fraction is spent in cache hits, not misses.
327
entry_cache[key] = ie.copy()
332
def unpack_inventory_flat(elt, format_num, unpack_entry,
333
entry_cache=None, return_from_cache=False):
334
"""Unpack a flat XML inventory.
336
:param elt: XML element for the inventory
337
:param format_num: Expected format number
338
:param unpack_entry: Function for unpacking inventory entries
339
:return: An inventory
340
:raise UnexpectedInventoryFormat: When unexpected elements or data is
343
if elt.tag != 'inventory':
344
raise errors.UnexpectedInventoryFormat('Root tag is %r' % elt.tag)
345
format = elt.get('format')
346
if format != format_num:
347
raise errors.UnexpectedInventoryFormat('Invalid format version %r'
349
revision_id = elt.get('revision_id')
350
if revision_id is not None:
351
revision_id = cache_utf8.encode(revision_id)
352
inv = inventory.Inventory(root_id=None, revision_id=revision_id)
354
ie = unpack_entry(e, entry_cache, return_from_cache)
359
def serialize_inventory_flat(inv, append, root_id, supported_kinds, working):
360
"""Serialize an inventory to a flat XML file.
362
:param inv: Inventory to serialize
363
:param append: Function for writing a line of output
364
:param working: If True skip history data - text_sha1, text_size,
365
reference_revision, symlink_target. self._check_revisions(inv)
367
entries = inv.iter_entries()
369
root_path, root_ie = entries.next()
370
for path, ie in entries:
371
if ie.parent_id != root_id:
372
parent_str = ' parent_id="'
373
parent_id = encode_and_escape(ie.parent_id)
377
if ie.kind == 'file':
379
executable = ' executable="yes"'
383
append('<file%s file_id="%s name="%s%s%s revision="%s '
384
'text_sha1="%s" text_size="%d" />\n' % (
385
executable, encode_and_escape(ie.file_id),
386
encode_and_escape(ie.name), parent_str, parent_id,
387
encode_and_escape(ie.revision), ie.text_sha1,
390
append('<file%s file_id="%s name="%s%s%s />\n' % (
391
executable, encode_and_escape(ie.file_id),
392
encode_and_escape(ie.name), parent_str, parent_id))
393
elif ie.kind == 'directory':
395
append('<directory file_id="%s name="%s%s%s revision="%s '
397
encode_and_escape(ie.file_id),
398
encode_and_escape(ie.name),
399
parent_str, parent_id,
400
encode_and_escape(ie.revision)))
402
append('<directory file_id="%s name="%s%s%s />\n' % (
403
encode_and_escape(ie.file_id),
404
encode_and_escape(ie.name),
405
parent_str, parent_id))
406
elif ie.kind == 'symlink':
408
append('<symlink file_id="%s name="%s%s%s revision="%s '
409
'symlink_target="%s />\n' % (
410
encode_and_escape(ie.file_id),
411
encode_and_escape(ie.name),
412
parent_str, parent_id,
413
encode_and_escape(ie.revision),
414
encode_and_escape(ie.symlink_target)))
416
append('<symlink file_id="%s name="%s%s%s />\n' % (
417
encode_and_escape(ie.file_id),
418
encode_and_escape(ie.name),
419
parent_str, parent_id))
420
elif ie.kind == 'tree-reference':
421
if ie.kind not in supported_kinds:
422
raise errors.UnsupportedInventoryKind(ie.kind)
424
append('<tree-reference file_id="%s name="%s%s%s '
425
'revision="%s reference_revision="%s />\n' % (
426
encode_and_escape(ie.file_id),
427
encode_and_escape(ie.name),
428
parent_str, parent_id,
429
encode_and_escape(ie.revision),
430
encode_and_escape(ie.reference_revision)))
432
append('<tree-reference file_id="%s name="%s%s%s />\n' % (
433
encode_and_escape(ie.file_id),
434
encode_and_escape(ie.name),
435
parent_str, parent_id))
437
raise errors.UnsupportedInventoryKind(ie.kind)
438
append('</inventory>\n')