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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
24
revision as _mod_revision,
27
from bzrlib.xml_serializer import SubElement, Element, Serializer
28
from bzrlib.inventory import ROOT_ID, Inventory, InventoryEntry
29
from bzrlib.revision import Revision
30
from bzrlib.errors import BzrError
37
"'":"'", # FIXME: overkill
44
def _ensure_utf8_re():
45
"""Make sure the _utf8_re and _unicode_re regexes have been compiled."""
46
global _utf8_re, _unicode_re
48
_utf8_re = re.compile('[&<>\'\"]|[\x80-\xff]+')
49
if _unicode_re is None:
50
_unicode_re = re.compile(u'[&<>\'\"\u0080-\uffff]')
53
def _unicode_escape_replace(match, _map=_xml_escape_map):
54
"""Replace a string of non-ascii, non XML safe characters with their escape
56
This will escape both Standard XML escapes, like <>"', etc.
57
As well as escaping non ascii characters, because ElementTree did.
58
This helps us remain compatible to older versions of bzr. We may change
59
our policy in the future, though.
61
# jam 20060816 Benchmarks show that try/KeyError is faster if you
62
# expect the entity to rarely miss. There is about a 10% difference
63
# in overall time. But if you miss frequently, then if None is much
64
# faster. For our use case, we *rarely* have a revision id, file id
65
# or path name that is unicode. So use try/KeyError.
67
return _map[match.group()]
69
return "&#%d;" % ord(match.group())
72
def _utf8_escape_replace(match, _map=_xml_escape_map):
73
"""Escape utf8 characters into XML safe ones.
75
This uses 2 tricks. It is either escaping "standard" characters, like "&<>,
76
or it is handling characters with the high-bit set. For ascii characters,
77
we just lookup the replacement in the dictionary. For everything else, we
78
decode back into Unicode, and then use the XML escape code.
81
return _map[match.group()]
83
return ''.join('&#%d;' % ord(uni_chr)
84
for uni_chr in match.group().decode('utf8'))
89
def _encode_and_escape(unicode_or_utf8_str, _map=_to_escaped_map):
90
"""Encode the string into utf8, and escape invalid XML characters"""
91
# We frequently get entities we have not seen before, so it is better
92
# to check if None, rather than try/KeyError
93
text = _map.get(unicode_or_utf8_str)
95
if unicode_or_utf8_str.__class__ == unicode:
96
# The alternative policy is to do a regular UTF8 encoding
97
# and then escape only XML meta characters.
98
# Performance is equivalent once you use cache_utf8. *However*
99
# this makes the serialized texts incompatible with old versions
100
# of bzr. So no net gain. (Perhaps the read code would handle utf8
101
# better than entity escapes, but cElementTree seems to do just fine
103
text = str(_unicode_re.sub(_unicode_escape_replace,
104
unicode_or_utf8_str)) + '"'
106
# Plain strings are considered to already be in utf-8 so we do a
107
# slightly different method for escaping.
108
text = _utf8_re.sub(_utf8_escape_replace,
109
unicode_or_utf8_str) + '"'
110
_map[unicode_or_utf8_str] = text
114
def _get_utf8_or_ascii(a_str,
115
_encode_utf8=cache_utf8.encode,
116
_get_cached_ascii=cache_utf8.get_cached_ascii):
117
"""Return a cached version of the string.
119
cElementTree will return a plain string if the XML is plain ascii. It only
120
returns Unicode when it needs to. We want to work in utf-8 strings. So if
121
cElementTree returns a plain string, we can just return the cached version.
122
If it is Unicode, then we need to encode it.
124
:param a_str: An 8-bit string or Unicode as returned by
125
cElementTree.Element.get()
126
:return: A utf-8 encoded 8-bit string.
128
# This is fairly optimized because we know what cElementTree does, this is
129
# not meant as a generic function for all cases. Because it is possible for
130
# an 8-bit string to not be ascii or valid utf8.
131
if a_str.__class__ == unicode:
132
return _encode_utf8(a_str)
134
return _get_cached_ascii(a_str)
138
"""Clean out the unicode => escaped map"""
139
_to_escaped_map.clear()
142
class Serializer_v8(Serializer):
143
"""This serialiser adds rich roots.
145
Its revision format number matches its inventory number.
151
support_altered_by_hack = True
152
# This format supports the altered-by hack that reads file ids directly out
153
# of the versionedfile, without doing XML parsing.
155
supported_kinds = set(['file', 'directory', 'symlink'])
157
revision_format_num = None
159
def _check_revisions(self, inv):
160
"""Extension point for subclasses to check during serialisation.
162
:param inv: An inventory about to be serialised, to be checked.
163
:raises: AssertionError if an error has occured.
165
if inv.revision_id is None:
166
raise AssertionError()
167
if inv.root.revision is None:
168
raise AssertionError()
170
def _check_cache_size(self, inv_size, entry_cache):
171
"""Check that the entry_cache is large enough.
173
We want the cache to be ~2x the size of an inventory. The reason is
174
because we use a FIFO cache, and how Inventory records are likely to
175
change. In general, you have a small number of records which change
176
often, and a lot of records which do not change at all. So when the
177
cache gets full, you actually flush out a lot of the records you are
178
interested in, which means you need to recreate all of those records.
179
An LRU Cache would be better, but the overhead negates the cache
182
One way to look at it, only the size of the cache > len(inv) is your
183
'working' set. And in general, it shouldn't be a problem to hold 2
184
inventories in memory anyway.
186
:param inv_size: The number of entries in an inventory.
188
if entry_cache is None:
190
# 1.5 times might also be reasonable.
191
recommended_min_cache_size = inv_size * 1.5
192
if entry_cache.cache_size() < recommended_min_cache_size:
193
recommended_cache_size = inv_size * 2
194
trace.mutter('Resizing the inventory entry cache from %d to %d',
195
entry_cache.cache_size(), recommended_cache_size)
196
entry_cache.resize(recommended_cache_size)
198
def write_inventory_to_lines(self, inv):
199
"""Return a list of lines with the encoded inventory."""
200
return self.write_inventory(inv, None)
202
def write_inventory_to_string(self, inv, working=False):
203
"""Just call write_inventory with a StringIO and return the value.
205
:param working: If True skip history data - text_sha1, text_size,
206
reference_revision, symlink_target.
208
sio = cStringIO.StringIO()
209
self.write_inventory(inv, sio, working)
210
return sio.getvalue()
212
def write_inventory(self, inv, f, working=False):
213
"""Write inventory to a file.
215
:param inv: the inventory to write.
216
:param f: the file to write. (May be None if the lines are the desired
218
:param working: If True skip history data - text_sha1, text_size,
219
reference_revision, symlink_target.
220
:return: The inventory as a list of lines.
223
self._check_revisions(inv)
225
append = output.append
226
self._append_inventory_root(append, inv)
227
entries = inv.iter_entries()
229
root_path, root_ie = entries.next()
230
for path, ie in entries:
231
if ie.parent_id != self.root_id:
232
parent_str = ' parent_id="'
233
parent_id = _encode_and_escape(ie.parent_id)
237
if ie.kind == 'file':
239
executable = ' executable="yes"'
243
append('<file%s file_id="%s name="%s%s%s revision="%s '
244
'text_sha1="%s" text_size="%d" />\n' % (
245
executable, _encode_and_escape(ie.file_id),
246
_encode_and_escape(ie.name), parent_str, parent_id,
247
_encode_and_escape(ie.revision), ie.text_sha1,
250
append('<file%s file_id="%s name="%s%s%s />\n' % (
251
executable, _encode_and_escape(ie.file_id),
252
_encode_and_escape(ie.name), parent_str, parent_id))
253
elif ie.kind == 'directory':
255
append('<directory file_id="%s name="%s%s%s revision="%s '
257
_encode_and_escape(ie.file_id),
258
_encode_and_escape(ie.name),
259
parent_str, parent_id,
260
_encode_and_escape(ie.revision)))
262
append('<directory file_id="%s name="%s%s%s />\n' % (
263
_encode_and_escape(ie.file_id),
264
_encode_and_escape(ie.name),
265
parent_str, parent_id))
266
elif ie.kind == 'symlink':
268
append('<symlink file_id="%s name="%s%s%s revision="%s '
269
'symlink_target="%s />\n' % (
270
_encode_and_escape(ie.file_id),
271
_encode_and_escape(ie.name),
272
parent_str, parent_id,
273
_encode_and_escape(ie.revision),
274
_encode_and_escape(ie.symlink_target)))
276
append('<symlink file_id="%s name="%s%s%s />\n' % (
277
_encode_and_escape(ie.file_id),
278
_encode_and_escape(ie.name),
279
parent_str, parent_id))
280
elif ie.kind == 'tree-reference':
281
if ie.kind not in self.supported_kinds:
282
raise errors.UnsupportedInventoryKind(ie.kind)
284
append('<tree-reference file_id="%s name="%s%s%s '
285
'revision="%s reference_revision="%s />\n' % (
286
_encode_and_escape(ie.file_id),
287
_encode_and_escape(ie.name),
288
parent_str, parent_id,
289
_encode_and_escape(ie.revision),
290
_encode_and_escape(ie.reference_revision)))
292
append('<tree-reference file_id="%s name="%s%s%s />\n' % (
293
_encode_and_escape(ie.file_id),
294
_encode_and_escape(ie.name),
295
parent_str, parent_id))
297
raise errors.UnsupportedInventoryKind(ie.kind)
298
append('</inventory>\n')
301
# Just to keep the cache from growing without bounds
302
# but we may actually not want to do clear the cache
306
def _append_inventory_root(self, append, inv):
307
"""Append the inventory root to output."""
308
if inv.revision_id is not None:
309
revid1 = ' revision_id="'
310
revid2 = _encode_and_escape(inv.revision_id)
314
append('<inventory format="%s"%s%s>\n' % (
315
self.format_num, revid1, revid2))
316
append('<directory file_id="%s name="%s revision="%s />\n' % (
317
_encode_and_escape(inv.root.file_id),
318
_encode_and_escape(inv.root.name),
319
_encode_and_escape(inv.root.revision)))
321
def _pack_revision(self, rev):
322
"""Revision object -> xml tree"""
323
# For the XML format, we need to write them as Unicode rather than as
324
# utf-8 strings. So that cElementTree can handle properly escaping
326
decode_utf8 = cache_utf8.decode
327
revision_id = rev.revision_id
328
if isinstance(revision_id, str):
329
revision_id = decode_utf8(revision_id)
330
format_num = self.format_num
331
if self.revision_format_num is not None:
332
format_num = self.revision_format_num
333
root = Element('revision',
334
committer = rev.committer,
335
timestamp = '%.3f' % rev.timestamp,
336
revision_id = revision_id,
337
inventory_sha1 = rev.inventory_sha1,
340
if rev.timezone is not None:
341
root.set('timezone', str(rev.timezone))
343
msg = SubElement(root, 'message')
344
msg.text = rev.message
347
pelts = SubElement(root, 'parents')
348
pelts.tail = pelts.text = '\n'
349
for parent_id in rev.parent_ids:
350
_mod_revision.check_not_reserved_id(parent_id)
351
p = SubElement(pelts, 'revision_ref')
353
if isinstance(parent_id, str):
354
parent_id = decode_utf8(parent_id)
355
p.set('revision_id', parent_id)
357
self._pack_revision_properties(rev, root)
360
def _pack_revision_properties(self, rev, under_element):
361
top_elt = SubElement(under_element, 'properties')
362
for prop_name, prop_value in sorted(rev.properties.items()):
363
prop_elt = SubElement(top_elt, 'property')
364
prop_elt.set('name', prop_name)
365
prop_elt.text = prop_value
369
def _unpack_inventory(self, elt, revision_id=None, entry_cache=None):
370
"""Construct from XML Element"""
371
if elt.tag != 'inventory':
372
raise errors.UnexpectedInventoryFormat('Root tag is %r' % elt.tag)
373
format = elt.get('format')
374
if format != self.format_num:
375
raise errors.UnexpectedInventoryFormat('Invalid format version %r'
377
revision_id = elt.get('revision_id')
378
if revision_id is not None:
379
revision_id = cache_utf8.encode(revision_id)
380
inv = inventory.Inventory(root_id=None, revision_id=revision_id)
382
ie = self._unpack_entry(e, entry_cache=entry_cache)
384
self._check_cache_size(len(inv), entry_cache)
387
def _unpack_entry(self, elt, entry_cache=None):
389
file_id = elt_get('file_id')
390
revision = elt_get('revision')
391
# Check and see if we have already unpacked this exact entry
392
# Some timings for "repo.revision_trees(last_100_revs)"
394
# unmodified 4.1s 40.8s
396
# using fifo 2.83s 29.1s
400
# no_copy 2.00s 20.5s
401
# no_c,dict 1.95s 18.0s
402
# Note that a cache of 10k nodes is more than sufficient to hold all of
403
# the inventory for the last 100 revs for bzr, but not for mysql (20k
404
# is enough for mysql, which saves the same 2s as using a dict)
406
# Breakdown of mysql using time.clock()
407
# 4.1s 2 calls to element.get for file_id, revision_id
408
# 4.5s cache_hit lookup
409
# 7.1s InventoryFile.copy()
410
# 2.4s InventoryDirectory.copy()
411
# 0.4s decoding unique entries
412
# 1.6s decoding entries after FIFO fills up
413
# 0.8s Adding nodes to FIFO (including flushes)
414
# 0.1s cache miss lookups
416
# 4.1s 2 calls to element.get for file_id, revision_id
417
# 9.9s cache_hit lookup
418
# 10.8s InventoryEntry.copy()
419
# 0.3s cache miss lookus
420
# 1.2s decoding entries
421
# 1.0s adding nodes to LRU
422
if entry_cache is not None and revision is not None:
423
key = (file_id, revision)
425
# We copy it, because some operatations may mutate it
426
cached_ie = entry_cache[key]
430
# Only copying directory entries drops us 2.85s => 2.35s
431
# if cached_ie.kind == 'directory':
432
# return cached_ie.copy()
434
return cached_ie.copy()
437
if not InventoryEntry.versionable_kind(kind):
438
raise AssertionError('unsupported entry kind %s' % kind)
440
get_cached = _get_utf8_or_ascii
442
file_id = get_cached(file_id)
443
if revision is not None:
444
revision = get_cached(revision)
445
parent_id = elt_get('parent_id')
446
if parent_id is not None:
447
parent_id = get_cached(parent_id)
449
if kind == 'directory':
450
ie = inventory.InventoryDirectory(file_id,
454
ie = inventory.InventoryFile(file_id,
457
ie.text_sha1 = elt_get('text_sha1')
458
if elt_get('executable') == 'yes':
460
v = elt_get('text_size')
461
ie.text_size = v and int(v)
462
elif kind == 'symlink':
463
ie = inventory.InventoryLink(file_id,
466
ie.symlink_target = elt_get('symlink_target')
468
raise errors.UnsupportedInventoryKind(kind)
469
ie.revision = revision
470
if revision is not None and entry_cache is not None:
471
# We cache a copy() because callers like to mutate objects, and
472
# that would cause the item in cache to mutate as well.
473
# This has a small effect on many-inventory performance, because
474
# the majority fraction is spent in cache hits, not misses.
475
entry_cache[key] = ie.copy()
479
def _unpack_revision(self, elt):
480
"""XML Element -> Revision object"""
481
format = elt.get('format')
482
format_num = self.format_num
483
if self.revision_format_num is not None:
484
format_num = self.revision_format_num
485
if format is not None:
486
if format != format_num:
487
raise BzrError("invalid format version %r on revision"
489
get_cached = _get_utf8_or_ascii
490
rev = Revision(committer = elt.get('committer'),
491
timestamp = float(elt.get('timestamp')),
492
revision_id = get_cached(elt.get('revision_id')),
493
inventory_sha1 = elt.get('inventory_sha1')
495
parents = elt.find('parents') or []
497
rev.parent_ids.append(get_cached(p.get('revision_id')))
498
self._unpack_revision_properties(elt, rev)
499
v = elt.get('timezone')
503
rev.timezone = int(v)
504
rev.message = elt.findtext('message') # text of <message>
507
def _unpack_revision_properties(self, elt, rev):
508
"""Unpack properties onto a revision."""
509
props_elt = elt.find('properties')
512
for prop_elt in props_elt:
513
if prop_elt.tag != 'property':
514
raise AssertionError(
515
"bad tag under properties list: %r" % prop_elt.tag)
516
name = prop_elt.get('name')
517
value = prop_elt.text
518
# If a property had an empty value ('') cElementTree reads
519
# that back as None, convert it back to '', so that all
520
# properties have string values
523
if name in rev.properties:
524
raise AssertionError("repeated property %r" % name)
525
rev.properties[name] = value
528
serializer_v8 = Serializer_v8()
added
removed
bzrlib/xml8.py
