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
22
revision as _mod_revision,
28
from .xml_serializer import (
35
serialize_inventory_flat,
36
unpack_inventory_entry,
37
unpack_inventory_flat,
39
from .revision import Revision
40
from .errors import BzrError
52
def _unescaper(match, _map=_xml_unescape_map):
57
if not code.startswith('#'):
59
return unichr(int(code[1:])).encode('utf8')
62
_unescape_re = lazy_regex.lazy_compile('\&([^;]*);')
64
def _unescape_xml(data):
65
"""Unescape predefined XML entities in a string of data."""
66
return _unescape_re.sub(_unescaper, data)
69
class Serializer_v8(XMLSerializer):
70
"""This serialiser adds rich roots.
72
Its revision format number matches its inventory number.
78
support_altered_by_hack = True
79
# This format supports the altered-by hack that reads file ids directly out
80
# of the versionedfile, without doing XML parsing.
82
supported_kinds = {'file', 'directory', 'symlink'}
84
revision_format_num = None
86
# The search regex used by xml based repositories to determine what things
87
# where changed in a single commit.
88
_file_ids_altered_regex = lazy_regex.lazy_compile(
89
r'file_id="(?P<file_id>[^"]+)"'
90
r'.* revision="(?P<revision_id>[^"]+)"'
93
def _check_revisions(self, inv):
94
"""Extension point for subclasses to check during serialisation.
96
:param inv: An inventory about to be serialised, to be checked.
97
:raises: AssertionError if an error has occurred.
99
if inv.revision_id is None:
100
raise AssertionError("inv.revision_id is None")
101
if inv.root.revision is None:
102
raise AssertionError("inv.root.revision is None")
104
def _check_cache_size(self, inv_size, entry_cache):
105
"""Check that the entry_cache is large enough.
107
We want the cache to be ~2x the size of an inventory. The reason is
108
because we use a FIFO cache, and how Inventory records are likely to
109
change. In general, you have a small number of records which change
110
often, and a lot of records which do not change at all. So when the
111
cache gets full, you actually flush out a lot of the records you are
112
interested in, which means you need to recreate all of those records.
113
An LRU Cache would be better, but the overhead negates the cache
116
One way to look at it, only the size of the cache > len(inv) is your
117
'working' set. And in general, it shouldn't be a problem to hold 2
118
inventories in memory anyway.
120
:param inv_size: The number of entries in an inventory.
122
if entry_cache is None:
124
# 1.5 times might also be reasonable.
125
recommended_min_cache_size = inv_size * 1.5
126
if entry_cache.cache_size() < recommended_min_cache_size:
127
recommended_cache_size = inv_size * 2
128
trace.mutter('Resizing the inventory entry cache from %d to %d',
129
entry_cache.cache_size(), recommended_cache_size)
130
entry_cache.resize(recommended_cache_size)
132
def write_inventory_to_lines(self, inv):
133
"""Return a list of lines with the encoded inventory."""
134
return self.write_inventory(inv, None)
136
def write_inventory_to_string(self, inv, working=False):
137
"""Just call write_inventory with a BytesIO and return the value.
139
:param working: If True skip history data - text_sha1, text_size,
140
reference_revision, symlink_target.
143
self.write_inventory(inv, sio, working)
144
return sio.getvalue()
146
def write_inventory(self, inv, f, working=False):
147
"""Write inventory to a file.
149
:param inv: the inventory to write.
150
:param f: the file to write. (May be None if the lines are the desired
152
:param working: If True skip history data - text_sha1, text_size,
153
reference_revision, symlink_target.
154
:return: The inventory as a list of lines.
157
append = output.append
158
self._append_inventory_root(append, inv)
159
serialize_inventory_flat(inv, append,
160
self.root_id, self.supported_kinds, working)
163
# Just to keep the cache from growing without bounds
164
# but we may actually not want to do clear the cache
168
def _append_inventory_root(self, append, inv):
169
"""Append the inventory root to output."""
170
if inv.revision_id is not None:
171
revid1 = ' revision_id="'
172
revid2 = encode_and_escape(inv.revision_id)
176
append('<inventory format="%s"%s%s>\n' % (
177
self.format_num, revid1, revid2))
178
append('<directory file_id="%s name="%s revision="%s />\n' % (
179
encode_and_escape(inv.root.file_id),
180
encode_and_escape(inv.root.name),
181
encode_and_escape(inv.root.revision)))
183
def _pack_revision(self, rev):
184
"""Revision object -> xml tree"""
185
# For the XML format, we need to write them as Unicode rather than as
186
# utf-8 strings. So that cElementTree can handle properly escaping
188
decode_utf8 = cache_utf8.decode
189
revision_id = rev.revision_id
190
if isinstance(revision_id, str):
191
revision_id = decode_utf8(revision_id)
192
format_num = self.format_num
193
if self.revision_format_num is not None:
194
format_num = self.revision_format_num
195
root = Element('revision',
196
committer = rev.committer,
197
timestamp = '%.3f' % rev.timestamp,
198
revision_id = revision_id,
199
inventory_sha1 = rev.inventory_sha1,
202
if rev.timezone is not None:
203
root.set('timezone', str(rev.timezone))
205
msg = SubElement(root, 'message')
206
msg.text = escape_invalid_chars(rev.message)[0]
209
pelts = SubElement(root, 'parents')
210
pelts.tail = pelts.text = '\n'
211
for parent_id in rev.parent_ids:
212
_mod_revision.check_not_reserved_id(parent_id)
213
p = SubElement(pelts, 'revision_ref')
215
if isinstance(parent_id, str):
216
parent_id = decode_utf8(parent_id)
217
p.set('revision_id', parent_id)
219
self._pack_revision_properties(rev, root)
222
def _pack_revision_properties(self, rev, under_element):
223
top_elt = SubElement(under_element, 'properties')
224
for prop_name, prop_value in sorted(rev.properties.items()):
225
prop_elt = SubElement(top_elt, 'property')
226
prop_elt.set('name', prop_name)
227
prop_elt.text = prop_value
231
def _unpack_entry(self, elt, entry_cache=None, return_from_cache=False):
232
# This is here because it's overridden by xml7
233
return unpack_inventory_entry(elt, entry_cache,
236
def _unpack_inventory(self, elt, revision_id=None, entry_cache=None,
237
return_from_cache=False):
238
"""Construct from XML Element"""
239
inv = unpack_inventory_flat(elt, self.format_num, self._unpack_entry,
240
entry_cache, return_from_cache)
241
self._check_cache_size(len(inv), entry_cache)
244
def _unpack_revision(self, elt):
245
"""XML Element -> Revision object"""
246
format = elt.get('format')
247
format_num = self.format_num
248
if self.revision_format_num is not None:
249
format_num = self.revision_format_num
250
if format is not None:
251
if format != format_num:
252
raise BzrError("invalid format version %r on revision"
254
get_cached = get_utf8_or_ascii
255
rev = Revision(committer = elt.get('committer'),
256
timestamp = float(elt.get('timestamp')),
257
revision_id = get_cached(elt.get('revision_id')),
258
inventory_sha1 = elt.get('inventory_sha1')
260
parents = elt.find('parents') or []
262
rev.parent_ids.append(get_cached(p.get('revision_id')))
263
self._unpack_revision_properties(elt, rev)
264
v = elt.get('timezone')
268
rev.timezone = int(v)
269
rev.message = elt.findtext('message') # text of <message>
272
def _unpack_revision_properties(self, elt, rev):
273
"""Unpack properties onto a revision."""
274
props_elt = elt.find('properties')
277
for prop_elt in props_elt:
278
if prop_elt.tag != 'property':
279
raise AssertionError(
280
"bad tag under properties list: %r" % prop_elt.tag)
281
name = prop_elt.get('name')
282
value = prop_elt.text
283
# If a property had an empty value ('') cElementTree reads
284
# that back as None, convert it back to '', so that all
285
# properties have string values
288
if name in rev.properties:
289
raise AssertionError("repeated property %r" % name)
290
rev.properties[name] = value
292
def _find_text_key_references(self, line_iterator):
293
"""Core routine for extracting references to texts from inventories.
295
This performs the translation of xml lines to revision ids.
297
:param line_iterator: An iterator of lines, origin_version_id
298
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
299
to whether they were referred to by the inventory of the
300
revision_id that they contain. Note that if that revision_id was
301
not part of the line_iterator's output then False will be given -
302
even though it may actually refer to that key.
304
if not self.support_altered_by_hack:
305
raise AssertionError(
306
"_find_text_key_references only "
307
"supported for branches which store inventory as unnested xml"
308
", not on %r" % self)
311
# this code needs to read every new line in every inventory for the
312
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
313
# not present in one of those inventories is unnecessary but not
314
# harmful because we are filtering by the revision id marker in the
315
# inventory lines : we only select file ids altered in one of those
316
# revisions. We don't need to see all lines in the inventory because
317
# only those added in an inventory in rev X can contain a revision=X
319
unescape_revid_cache = {}
320
unescape_fileid_cache = {}
322
# jam 20061218 In a big fetch, this handles hundreds of thousands
323
# of lines, so it has had a lot of inlining and optimizing done.
324
# Sorry that it is a little bit messy.
325
# Move several functions to be local variables, since this is a long
327
search = self._file_ids_altered_regex.search
328
unescape = _unescape_xml
329
setdefault = result.setdefault
330
for line, line_key in line_iterator:
334
# One call to match.group() returning multiple items is quite a
335
# bit faster than 2 calls to match.group() each returning 1
336
file_id, revision_id = match.group('file_id', 'revision_id')
338
# Inlining the cache lookups helps a lot when you make 170,000
339
# lines and 350k ids, versus 8.4 unique ids.
340
# Using a cache helps in 2 ways:
341
# 1) Avoids unnecessary decoding calls
342
# 2) Re-uses cached strings, which helps in future set and
344
# (2) is enough that removing encoding entirely along with
345
# the cache (so we are using plain strings) results in no
346
# performance improvement.
348
revision_id = unescape_revid_cache[revision_id]
350
unescaped = unescape(revision_id)
351
unescape_revid_cache[revision_id] = unescaped
352
revision_id = unescaped
354
# Note that unconditionally unescaping means that we deserialise
355
# every fileid, which for general 'pull' is not great, but we don't
356
# really want to have some many fulltexts that this matters anyway.
359
file_id = unescape_fileid_cache[file_id]
361
unescaped = unescape(file_id)
362
unescape_fileid_cache[file_id] = unescaped
365
key = (file_id, revision_id)
366
setdefault(key, False)
367
if revision_id == line_key[-1]:
372
serializer_v8 = Serializer_v8()