1
# Copyright (C) 2005-2010 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
17
from __future__ import absolute_import
24
revision as _mod_revision,
27
from bzrlib.xml_serializer import (
34
serialize_inventory_flat,
35
unpack_inventory_entry,
36
unpack_inventory_flat,
38
from bzrlib.revision import Revision
39
from bzrlib.errors import BzrError
51
def _unescaper(match, _map=_xml_unescape_map):
56
if not code.startswith('#'):
58
return unichr(int(code[1:])).encode('utf8')
61
_unescape_re = lazy_regex.lazy_compile('\&([^;]*);')
63
def _unescape_xml(data):
64
"""Unescape predefined XML entities in a string of data."""
65
return _unescape_re.sub(_unescaper, data)
68
class Serializer_v8(XMLSerializer):
69
"""This serialiser adds rich roots.
71
Its revision format number matches its inventory number.
77
support_altered_by_hack = True
78
# This format supports the altered-by hack that reads file ids directly out
79
# of the versionedfile, without doing XML parsing.
81
supported_kinds = set(['file', 'directory', 'symlink'])
83
revision_format_num = None
85
# The search regex used by xml based repositories to determine what things
86
# where changed in a single commit.
87
_file_ids_altered_regex = lazy_regex.lazy_compile(
88
r'file_id="(?P<file_id>[^"]+)"'
89
r'.* revision="(?P<revision_id>[^"]+)"'
92
def _check_revisions(self, inv):
93
"""Extension point for subclasses to check during serialisation.
95
:param inv: An inventory about to be serialised, to be checked.
96
:raises: AssertionError if an error has occurred.
98
if inv.revision_id is None:
99
raise AssertionError("inv.revision_id is None")
100
if inv.root.revision is None:
101
raise AssertionError("inv.root.revision is None")
103
def _check_cache_size(self, inv_size, entry_cache):
104
"""Check that the entry_cache is large enough.
106
We want the cache to be ~2x the size of an inventory. The reason is
107
because we use a FIFO cache, and how Inventory records are likely to
108
change. In general, you have a small number of records which change
109
often, and a lot of records which do not change at all. So when the
110
cache gets full, you actually flush out a lot of the records you are
111
interested in, which means you need to recreate all of those records.
112
An LRU Cache would be better, but the overhead negates the cache
115
One way to look at it, only the size of the cache > len(inv) is your
116
'working' set. And in general, it shouldn't be a problem to hold 2
117
inventories in memory anyway.
119
:param inv_size: The number of entries in an inventory.
121
if entry_cache is None:
123
# 1.5 times might also be reasonable.
124
recommended_min_cache_size = inv_size * 1.5
125
if entry_cache.cache_size() < recommended_min_cache_size:
126
recommended_cache_size = inv_size * 2
127
trace.mutter('Resizing the inventory entry cache from %d to %d',
128
entry_cache.cache_size(), recommended_cache_size)
129
entry_cache.resize(recommended_cache_size)
131
def write_inventory_to_lines(self, inv):
132
"""Return a list of lines with the encoded inventory."""
133
return self.write_inventory(inv, None)
135
def write_inventory_to_string(self, inv, working=False):
136
"""Just call write_inventory with a StringIO and return the value.
138
:param working: If True skip history data - text_sha1, text_size,
139
reference_revision, symlink_target.
141
sio = cStringIO.StringIO()
142
self.write_inventory(inv, sio, working)
143
return sio.getvalue()
145
def write_inventory(self, inv, f, working=False):
146
"""Write inventory to a file.
148
:param inv: the inventory to write.
149
:param f: the file to write. (May be None if the lines are the desired
151
:param working: If True skip history data - text_sha1, text_size,
152
reference_revision, symlink_target.
153
:return: The inventory as a list of lines.
156
append = output.append
157
self._append_inventory_root(append, inv)
158
serialize_inventory_flat(inv, append,
159
self.root_id, self.supported_kinds, working)
162
# Just to keep the cache from growing without bounds
163
# but we may actually not want to do clear the cache
167
def _append_inventory_root(self, append, inv):
168
"""Append the inventory root to output."""
169
if inv.revision_id is not None:
170
revid1 = ' revision_id="'
171
revid2 = encode_and_escape(inv.revision_id)
175
append('<inventory format="%s"%s%s>\n' % (
176
self.format_num, revid1, revid2))
177
append('<directory file_id="%s name="%s revision="%s />\n' % (
178
encode_and_escape(inv.root.file_id),
179
encode_and_escape(inv.root.name),
180
encode_and_escape(inv.root.revision)))
182
def _pack_revision(self, rev):
183
"""Revision object -> xml tree"""
184
# For the XML format, we need to write them as Unicode rather than as
185
# utf-8 strings. So that cElementTree can handle properly escaping
187
decode_utf8 = cache_utf8.decode
188
revision_id = rev.revision_id
189
if isinstance(revision_id, str):
190
revision_id = decode_utf8(revision_id)
191
format_num = self.format_num
192
if self.revision_format_num is not None:
193
format_num = self.revision_format_num
194
root = Element('revision',
195
committer = rev.committer,
196
timestamp = '%.3f' % rev.timestamp,
197
revision_id = revision_id,
198
inventory_sha1 = rev.inventory_sha1,
201
if rev.timezone is not None:
202
root.set('timezone', str(rev.timezone))
204
msg = SubElement(root, 'message')
205
msg.text = escape_invalid_chars(rev.message)[0]
208
pelts = SubElement(root, 'parents')
209
pelts.tail = pelts.text = '\n'
210
for parent_id in rev.parent_ids:
211
_mod_revision.check_not_reserved_id(parent_id)
212
p = SubElement(pelts, 'revision_ref')
214
if isinstance(parent_id, str):
215
parent_id = decode_utf8(parent_id)
216
p.set('revision_id', parent_id)
218
self._pack_revision_properties(rev, root)
221
def _pack_revision_properties(self, rev, under_element):
222
top_elt = SubElement(under_element, 'properties')
223
for prop_name, prop_value in sorted(rev.properties.items()):
224
prop_elt = SubElement(top_elt, 'property')
225
prop_elt.set('name', prop_name)
226
prop_elt.text = prop_value
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,
235
def _unpack_inventory(self, elt, revision_id=None, entry_cache=None,
236
return_from_cache=False):
237
"""Construct from XML Element"""
238
inv = unpack_inventory_flat(elt, self.format_num, self._unpack_entry,
239
entry_cache, return_from_cache)
240
self._check_cache_size(len(inv), entry_cache)
243
def _unpack_revision(self, elt):
244
"""XML Element -> Revision object"""
245
format = elt.get('format')
246
format_num = self.format_num
247
if self.revision_format_num is not None:
248
format_num = self.revision_format_num
249
if format is not None:
250
if format != format_num:
251
raise BzrError("invalid format version %r on revision"
253
get_cached = get_utf8_or_ascii
254
rev = Revision(committer = elt.get('committer'),
255
timestamp = float(elt.get('timestamp')),
256
revision_id = get_cached(elt.get('revision_id')),
257
inventory_sha1 = elt.get('inventory_sha1')
259
parents = elt.find('parents') or []
261
rev.parent_ids.append(get_cached(p.get('revision_id')))
262
self._unpack_revision_properties(elt, rev)
263
v = elt.get('timezone')
267
rev.timezone = int(v)
268
rev.message = elt.findtext('message') # text of <message>
271
def _unpack_revision_properties(self, elt, rev):
272
"""Unpack properties onto a revision."""
273
props_elt = elt.find('properties')
276
for prop_elt in props_elt:
277
if prop_elt.tag != 'property':
278
raise AssertionError(
279
"bad tag under properties list: %r" % prop_elt.tag)
280
name = prop_elt.get('name')
281
value = prop_elt.text
282
# If a property had an empty value ('') cElementTree reads
283
# that back as None, convert it back to '', so that all
284
# properties have string values
287
if name in rev.properties:
288
raise AssertionError("repeated property %r" % name)
289
rev.properties[name] = value
291
def _find_text_key_references(self, line_iterator):
292
"""Core routine for extracting references to texts from inventories.
294
This performs the translation of xml lines to revision ids.
296
:param line_iterator: An iterator of lines, origin_version_id
297
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
298
to whether they were referred to by the inventory of the
299
revision_id that they contain. Note that if that revision_id was
300
not part of the line_iterator's output then False will be given -
301
even though it may actually refer to that key.
303
if not self.support_altered_by_hack:
304
raise AssertionError(
305
"_find_text_key_references only "
306
"supported for branches which store inventory as unnested xml"
307
", not on %r" % self)
310
# this code needs to read every new line in every inventory for the
311
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
312
# not present in one of those inventories is unnecessary but not
313
# harmful because we are filtering by the revision id marker in the
314
# inventory lines : we only select file ids altered in one of those
315
# revisions. We don't need to see all lines in the inventory because
316
# only those added in an inventory in rev X can contain a revision=X
318
unescape_revid_cache = {}
319
unescape_fileid_cache = {}
321
# jam 20061218 In a big fetch, this handles hundreds of thousands
322
# of lines, so it has had a lot of inlining and optimizing done.
323
# Sorry that it is a little bit messy.
324
# Move several functions to be local variables, since this is a long
326
search = self._file_ids_altered_regex.search
327
unescape = _unescape_xml
328
setdefault = result.setdefault
329
for line, line_key in line_iterator:
333
# One call to match.group() returning multiple items is quite a
334
# bit faster than 2 calls to match.group() each returning 1
335
file_id, revision_id = match.group('file_id', 'revision_id')
337
# Inlining the cache lookups helps a lot when you make 170,000
338
# lines and 350k ids, versus 8.4 unique ids.
339
# Using a cache helps in 2 ways:
340
# 1) Avoids unnecessary decoding calls
341
# 2) Re-uses cached strings, which helps in future set and
343
# (2) is enough that removing encoding entirely along with
344
# the cache (so we are using plain strings) results in no
345
# performance improvement.
347
revision_id = unescape_revid_cache[revision_id]
349
unescaped = unescape(revision_id)
350
unescape_revid_cache[revision_id] = unescaped
351
revision_id = unescaped
353
# Note that unconditionally unescaping means that we deserialise
354
# every fileid, which for general 'pull' is not great, but we don't
355
# really want to have some many fulltexts that this matters anyway.
358
file_id = unescape_fileid_cache[file_id]
360
unescaped = unescape(file_id)
361
unescape_fileid_cache[file_id] = unescaped
364
key = (file_id, revision_id)
365
setdefault(key, False)
366
if revision_id == line_key[-1]:
371
serializer_v8 = Serializer_v8()
added
removed
bzrlib/xml8.py
