1
# Copyright (C) 2005, 2006, 2007, 2008 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24
revision as _mod_revision,
27
from bzrlib.xml_serializer import (
32
from bzrlib.inventory import ROOT_ID, Inventory, InventoryEntry
33
from bzrlib.revision import Revision
34
from bzrlib.errors import BzrError
41
"'":"'", # FIXME: overkill
48
def _ensure_utf8_re():
49
"""Make sure the _utf8_re and _unicode_re regexes have been compiled."""
50
global _utf8_re, _unicode_re
52
_utf8_re = re.compile('[&<>\'\"]|[\x80-\xff]+')
53
if _unicode_re is None:
54
_unicode_re = re.compile(u'[&<>\'\"\u0080-\uffff]')
57
def _unicode_escape_replace(match, _map=_xml_escape_map):
58
"""Replace a string of non-ascii, non XML safe characters with their escape
60
This will escape both Standard XML escapes, like <>"', etc.
61
As well as escaping non ascii characters, because ElementTree did.
62
This helps us remain compatible to older versions of bzr. We may change
63
our policy in the future, though.
65
# jam 20060816 Benchmarks show that try/KeyError is faster if you
66
# expect the entity to rarely miss. There is about a 10% difference
67
# in overall time. But if you miss frequently, then if None is much
68
# faster. For our use case, we *rarely* have a revision id, file id
69
# or path name that is unicode. So use try/KeyError.
71
return _map[match.group()]
73
return "&#%d;" % ord(match.group())
76
def _utf8_escape_replace(match, _map=_xml_escape_map):
77
"""Escape utf8 characters into XML safe ones.
79
This uses 2 tricks. It is either escaping "standard" characters, like "&<>,
80
or it is handling characters with the high-bit set. For ascii characters,
81
we just lookup the replacement in the dictionary. For everything else, we
82
decode back into Unicode, and then use the XML escape code.
85
return _map[match.group()]
87
return ''.join('&#%d;' % ord(uni_chr)
88
for uni_chr in match.group().decode('utf8'))
93
def _encode_and_escape(unicode_or_utf8_str, _map=_to_escaped_map):
94
"""Encode the string into utf8, and escape invalid XML characters"""
95
# We frequently get entities we have not seen before, so it is better
96
# to check if None, rather than try/KeyError
97
text = _map.get(unicode_or_utf8_str)
99
if unicode_or_utf8_str.__class__ is unicode:
100
# The alternative policy is to do a regular UTF8 encoding
101
# and then escape only XML meta characters.
102
# Performance is equivalent once you use cache_utf8. *However*
103
# this makes the serialized texts incompatible with old versions
104
# of bzr. So no net gain. (Perhaps the read code would handle utf8
105
# better than entity escapes, but cElementTree seems to do just fine
107
text = str(_unicode_re.sub(_unicode_escape_replace,
108
unicode_or_utf8_str)) + '"'
110
# Plain strings are considered to already be in utf-8 so we do a
111
# slightly different method for escaping.
112
text = _utf8_re.sub(_utf8_escape_replace,
113
unicode_or_utf8_str) + '"'
114
_map[unicode_or_utf8_str] = text
118
def _get_utf8_or_ascii(a_str,
119
_encode_utf8=cache_utf8.encode,
120
_get_cached_ascii=cache_utf8.get_cached_ascii):
121
"""Return a cached version of the string.
123
cElementTree will return a plain string if the XML is plain ascii. It only
124
returns Unicode when it needs to. We want to work in utf-8 strings. So if
125
cElementTree returns a plain string, we can just return the cached version.
126
If it is Unicode, then we need to encode it.
128
:param a_str: An 8-bit string or Unicode as returned by
129
cElementTree.Element.get()
130
:return: A utf-8 encoded 8-bit string.
132
# This is fairly optimized because we know what cElementTree does, this is
133
# not meant as a generic function for all cases. Because it is possible for
134
# an 8-bit string to not be ascii or valid utf8.
135
if a_str.__class__ is unicode:
136
return _encode_utf8(a_str)
142
"""Clean out the unicode => escaped map"""
143
_to_escaped_map.clear()
146
class Serializer_v8(XMLSerializer):
147
"""This serialiser adds rich roots.
149
Its revision format number matches its inventory number.
155
support_altered_by_hack = True
156
# This format supports the altered-by hack that reads file ids directly out
157
# of the versionedfile, without doing XML parsing.
159
supported_kinds = set(['file', 'directory', 'symlink'])
161
revision_format_num = None
163
def _check_revisions(self, inv):
164
"""Extension point for subclasses to check during serialisation.
166
:param inv: An inventory about to be serialised, to be checked.
167
:raises: AssertionError if an error has occurred.
169
if inv.revision_id is None:
170
raise AssertionError()
171
if inv.root.revision is None:
172
raise AssertionError()
174
def _check_cache_size(self, inv_size, entry_cache):
175
"""Check that the entry_cache is large enough.
177
We want the cache to be ~2x the size of an inventory. The reason is
178
because we use a FIFO cache, and how Inventory records are likely to
179
change. In general, you have a small number of records which change
180
often, and a lot of records which do not change at all. So when the
181
cache gets full, you actually flush out a lot of the records you are
182
interested in, which means you need to recreate all of those records.
183
An LRU Cache would be better, but the overhead negates the cache
186
One way to look at it, only the size of the cache > len(inv) is your
187
'working' set. And in general, it shouldn't be a problem to hold 2
188
inventories in memory anyway.
190
:param inv_size: The number of entries in an inventory.
192
if entry_cache is None:
194
# 1.5 times might also be reasonable.
195
recommended_min_cache_size = inv_size * 1.5
196
if entry_cache.cache_size() < recommended_min_cache_size:
197
recommended_cache_size = inv_size * 2
198
trace.mutter('Resizing the inventory entry cache from %d to %d',
199
entry_cache.cache_size(), recommended_cache_size)
200
entry_cache.resize(recommended_cache_size)
202
def write_inventory_to_lines(self, inv):
203
"""Return a list of lines with the encoded inventory."""
204
return self.write_inventory(inv, None)
206
def write_inventory_to_string(self, inv, working=False):
207
"""Just call write_inventory with a StringIO and return the value.
209
:param working: If True skip history data - text_sha1, text_size,
210
reference_revision, symlink_target.
212
sio = cStringIO.StringIO()
213
self.write_inventory(inv, sio, working)
214
return sio.getvalue()
216
def write_inventory(self, inv, f, working=False):
217
"""Write inventory to a file.
219
:param inv: the inventory to write.
220
:param f: the file to write. (May be None if the lines are the desired
222
:param working: If True skip history data - text_sha1, text_size,
223
reference_revision, symlink_target.
224
:return: The inventory as a list of lines.
227
self._check_revisions(inv)
229
append = output.append
230
self._append_inventory_root(append, inv)
231
entries = inv.iter_entries()
233
root_path, root_ie = entries.next()
234
for path, ie in entries:
235
if ie.parent_id != self.root_id:
236
parent_str = ' parent_id="'
237
parent_id = _encode_and_escape(ie.parent_id)
241
if ie.kind == 'file':
243
executable = ' executable="yes"'
247
append('<file%s file_id="%s name="%s%s%s revision="%s '
248
'text_sha1="%s" text_size="%d" />\n' % (
249
executable, _encode_and_escape(ie.file_id),
250
_encode_and_escape(ie.name), parent_str, parent_id,
251
_encode_and_escape(ie.revision), ie.text_sha1,
254
append('<file%s file_id="%s name="%s%s%s />\n' % (
255
executable, _encode_and_escape(ie.file_id),
256
_encode_and_escape(ie.name), parent_str, parent_id))
257
elif ie.kind == 'directory':
259
append('<directory file_id="%s name="%s%s%s revision="%s '
261
_encode_and_escape(ie.file_id),
262
_encode_and_escape(ie.name),
263
parent_str, parent_id,
264
_encode_and_escape(ie.revision)))
266
append('<directory file_id="%s name="%s%s%s />\n' % (
267
_encode_and_escape(ie.file_id),
268
_encode_and_escape(ie.name),
269
parent_str, parent_id))
270
elif ie.kind == 'symlink':
272
append('<symlink file_id="%s name="%s%s%s revision="%s '
273
'symlink_target="%s />\n' % (
274
_encode_and_escape(ie.file_id),
275
_encode_and_escape(ie.name),
276
parent_str, parent_id,
277
_encode_and_escape(ie.revision),
278
_encode_and_escape(ie.symlink_target)))
280
append('<symlink file_id="%s name="%s%s%s />\n' % (
281
_encode_and_escape(ie.file_id),
282
_encode_and_escape(ie.name),
283
parent_str, parent_id))
284
elif ie.kind == 'tree-reference':
285
if ie.kind not in self.supported_kinds:
286
raise errors.UnsupportedInventoryKind(ie.kind)
288
append('<tree-reference file_id="%s name="%s%s%s '
289
'revision="%s reference_revision="%s />\n' % (
290
_encode_and_escape(ie.file_id),
291
_encode_and_escape(ie.name),
292
parent_str, parent_id,
293
_encode_and_escape(ie.revision),
294
_encode_and_escape(ie.reference_revision)))
296
append('<tree-reference file_id="%s name="%s%s%s />\n' % (
297
_encode_and_escape(ie.file_id),
298
_encode_and_escape(ie.name),
299
parent_str, parent_id))
301
raise errors.UnsupportedInventoryKind(ie.kind)
302
append('</inventory>\n')
305
# Just to keep the cache from growing without bounds
306
# but we may actually not want to do clear the cache
310
def _append_inventory_root(self, append, inv):
311
"""Append the inventory root to output."""
312
if inv.revision_id is not None:
313
revid1 = ' revision_id="'
314
revid2 = _encode_and_escape(inv.revision_id)
318
append('<inventory format="%s"%s%s>\n' % (
319
self.format_num, revid1, revid2))
320
append('<directory file_id="%s name="%s revision="%s />\n' % (
321
_encode_and_escape(inv.root.file_id),
322
_encode_and_escape(inv.root.name),
323
_encode_and_escape(inv.root.revision)))
325
def _pack_revision(self, rev):
326
"""Revision object -> xml tree"""
327
# For the XML format, we need to write them as Unicode rather than as
328
# utf-8 strings. So that cElementTree can handle properly escaping
330
decode_utf8 = cache_utf8.decode
331
revision_id = rev.revision_id
332
if isinstance(revision_id, str):
333
revision_id = decode_utf8(revision_id)
334
format_num = self.format_num
335
if self.revision_format_num is not None:
336
format_num = self.revision_format_num
337
root = Element('revision',
338
committer = rev.committer,
339
timestamp = '%.3f' % rev.timestamp,
340
revision_id = revision_id,
341
inventory_sha1 = rev.inventory_sha1,
344
if rev.timezone is not None:
345
root.set('timezone', str(rev.timezone))
347
msg = SubElement(root, 'message')
348
msg.text = rev.message
351
pelts = SubElement(root, 'parents')
352
pelts.tail = pelts.text = '\n'
353
for parent_id in rev.parent_ids:
354
_mod_revision.check_not_reserved_id(parent_id)
355
p = SubElement(pelts, 'revision_ref')
357
if isinstance(parent_id, str):
358
parent_id = decode_utf8(parent_id)
359
p.set('revision_id', parent_id)
361
self._pack_revision_properties(rev, root)
364
def _pack_revision_properties(self, rev, under_element):
365
top_elt = SubElement(under_element, 'properties')
366
for prop_name, prop_value in sorted(rev.properties.items()):
367
prop_elt = SubElement(top_elt, 'property')
368
prop_elt.set('name', prop_name)
369
prop_elt.text = prop_value
373
def _unpack_inventory(self, elt, revision_id=None, entry_cache=None):
374
"""Construct from XML Element"""
375
if elt.tag != 'inventory':
376
raise errors.UnexpectedInventoryFormat('Root tag is %r' % elt.tag)
377
format = elt.get('format')
378
if format != self.format_num:
379
raise errors.UnexpectedInventoryFormat('Invalid format version %r'
381
revision_id = elt.get('revision_id')
382
if revision_id is not None:
383
revision_id = cache_utf8.encode(revision_id)
384
inv = inventory.Inventory(root_id=None, revision_id=revision_id)
386
ie = self._unpack_entry(e, entry_cache=entry_cache)
388
self._check_cache_size(len(inv), entry_cache)
391
def _unpack_entry(self, elt, entry_cache=None):
393
file_id = elt_get('file_id')
394
revision = elt_get('revision')
395
# Check and see if we have already unpacked this exact entry
396
# Some timings for "repo.revision_trees(last_100_revs)"
398
# unmodified 4.1s 40.8s
400
# using fifo 2.83s 29.1s
404
# no_copy 2.00s 20.5s
405
# no_c,dict 1.95s 18.0s
406
# Note that a cache of 10k nodes is more than sufficient to hold all of
407
# the inventory for the last 100 revs for bzr, but not for mysql (20k
408
# is enough for mysql, which saves the same 2s as using a dict)
410
# Breakdown of mysql using time.clock()
411
# 4.1s 2 calls to element.get for file_id, revision_id
412
# 4.5s cache_hit lookup
413
# 7.1s InventoryFile.copy()
414
# 2.4s InventoryDirectory.copy()
415
# 0.4s decoding unique entries
416
# 1.6s decoding entries after FIFO fills up
417
# 0.8s Adding nodes to FIFO (including flushes)
418
# 0.1s cache miss lookups
420
# 4.1s 2 calls to element.get for file_id, revision_id
421
# 9.9s cache_hit lookup
422
# 10.8s InventoryEntry.copy()
423
# 0.3s cache miss lookus
424
# 1.2s decoding entries
425
# 1.0s adding nodes to LRU
426
if entry_cache is not None and revision is not None:
427
key = (file_id, revision)
429
# We copy it, because some operations may mutate it
430
cached_ie = entry_cache[key]
434
# Only copying directory entries drops us 2.85s => 2.35s
435
# if cached_ie.kind == 'directory':
436
# return cached_ie.copy()
438
return cached_ie.copy()
441
if not InventoryEntry.versionable_kind(kind):
442
raise AssertionError('unsupported entry kind %s' % kind)
444
get_cached = _get_utf8_or_ascii
446
file_id = get_cached(file_id)
447
if revision is not None:
448
revision = get_cached(revision)
449
parent_id = elt_get('parent_id')
450
if parent_id is not None:
451
parent_id = get_cached(parent_id)
453
if kind == 'directory':
454
ie = inventory.InventoryDirectory(file_id,
458
ie = inventory.InventoryFile(file_id,
461
ie.text_sha1 = elt_get('text_sha1')
462
if elt_get('executable') == 'yes':
464
v = elt_get('text_size')
465
ie.text_size = v and int(v)
466
elif kind == 'symlink':
467
ie = inventory.InventoryLink(file_id,
470
ie.symlink_target = elt_get('symlink_target')
472
raise errors.UnsupportedInventoryKind(kind)
473
ie.revision = revision
474
if revision is not None and entry_cache is not None:
475
# We cache a copy() because callers like to mutate objects, and
476
# that would cause the item in cache to mutate as well.
477
# This has a small effect on many-inventory performance, because
478
# the majority fraction is spent in cache hits, not misses.
479
entry_cache[key] = ie.copy()
483
def _unpack_revision(self, elt):
484
"""XML Element -> Revision object"""
485
format = elt.get('format')
486
format_num = self.format_num
487
if self.revision_format_num is not None:
488
format_num = self.revision_format_num
489
if format is not None:
490
if format != format_num:
491
raise BzrError("invalid format version %r on revision"
493
get_cached = _get_utf8_or_ascii
494
rev = Revision(committer = elt.get('committer'),
495
timestamp = float(elt.get('timestamp')),
496
revision_id = get_cached(elt.get('revision_id')),
497
inventory_sha1 = elt.get('inventory_sha1')
499
parents = elt.find('parents') or []
501
rev.parent_ids.append(get_cached(p.get('revision_id')))
502
self._unpack_revision_properties(elt, rev)
503
v = elt.get('timezone')
507
rev.timezone = int(v)
508
rev.message = elt.findtext('message') # text of <message>
511
def _unpack_revision_properties(self, elt, rev):
512
"""Unpack properties onto a revision."""
513
props_elt = elt.find('properties')
516
for prop_elt in props_elt:
517
if prop_elt.tag != 'property':
518
raise AssertionError(
519
"bad tag under properties list: %r" % prop_elt.tag)
520
name = prop_elt.get('name')
521
value = prop_elt.text
522
# If a property had an empty value ('') cElementTree reads
523
# that back as None, convert it back to '', so that all
524
# properties have string values
527
if name in rev.properties:
528
raise AssertionError("repeated property %r" % name)
529
rev.properties[name] = value
532
serializer_v8 = Serializer_v8()
added
removed
bzrlib/xml8.py
