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 io import BytesIO
22
revision as _mod_revision,
25
from .xml_serializer import (
32
serialize_inventory_flat,
33
unpack_inventory_entry,
34
unpack_inventory_flat,
36
from ..revision import Revision
37
from ..errors import BzrError
49
def _unescaper(match, _map=_xml_unescape_map):
54
if not code.startswith(b'#'):
56
return chr(int(code[1:])).encode('utf8')
59
_unescape_re = lazy_regex.lazy_compile(b'\\&([^;]*);')
62
def _unescape_xml(data):
63
"""Unescape predefined XML entities in a string of data."""
64
return _unescape_re.sub(_unescaper, data)
67
class Serializer_v8(XMLSerializer):
68
"""This serialiser adds rich roots.
70
Its revision format number matches its inventory number.
76
support_altered_by_hack = True
77
# This format supports the altered-by hack that reads file ids directly out
78
# of the versionedfile, without doing XML parsing.
80
supported_kinds = {'file', 'directory', 'symlink'}
82
revision_format_num = None
84
# The search regex used by xml based repositories to determine what things
85
# where changed in a single commit.
86
_file_ids_altered_regex = lazy_regex.lazy_compile(
87
b'file_id="(?P<file_id>[^"]+)"'
88
b'.* revision="(?P<revision_id>[^"]+)"'
91
def _check_revisions(self, inv):
92
"""Extension point for subclasses to check during serialisation.
94
:param inv: An inventory about to be serialised, to be checked.
95
:raises: AssertionError if an error has occurred.
97
if inv.revision_id is None:
98
raise AssertionError("inv.revision_id is None")
99
if inv.root.revision is None:
100
raise AssertionError("inv.root.revision is None")
102
def _check_cache_size(self, inv_size, entry_cache):
103
"""Check that the entry_cache is large enough.
105
We want the cache to be ~2x the size of an inventory. The reason is
106
because we use a FIFO cache, and how Inventory records are likely to
107
change. In general, you have a small number of records which change
108
often, and a lot of records which do not change at all. So when the
109
cache gets full, you actually flush out a lot of the records you are
110
interested in, which means you need to recreate all of those records.
111
An LRU Cache would be better, but the overhead negates the cache
114
One way to look at it, only the size of the cache > len(inv) is your
115
'working' set. And in general, it shouldn't be a problem to hold 2
116
inventories in memory anyway.
118
:param inv_size: The number of entries in an inventory.
120
if entry_cache is None:
122
# 1.5 times might also be reasonable.
123
recommended_min_cache_size = inv_size * 1.5
124
if entry_cache.cache_size() < recommended_min_cache_size:
125
recommended_cache_size = inv_size * 2
126
trace.mutter('Resizing the inventory entry cache from %d to %d',
127
entry_cache.cache_size(), recommended_cache_size)
128
entry_cache.resize(recommended_cache_size)
130
def write_inventory_to_lines(self, inv):
131
"""Return a list of lines with the encoded inventory."""
132
return self.write_inventory(inv, None)
134
def write_inventory_to_chunks(self, inv):
135
return self.write_inventory(inv, None)
137
def write_inventory(self, inv, f, working=False):
138
"""Write inventory to a file.
140
:param inv: the inventory to write.
141
:param f: the file to write. (May be None if the lines are the desired
143
:param working: If True skip history data - text_sha1, text_size,
144
reference_revision, symlink_target.
145
:return: The inventory as a list of lines.
148
append = output.append
149
self._append_inventory_root(append, inv)
150
serialize_inventory_flat(inv, append,
151
self.root_id, self.supported_kinds, working)
154
# Just to keep the cache from growing without bounds
155
# but we may actually not want to do clear the cache
159
def _append_inventory_root(self, append, inv):
160
"""Append the inventory root to output."""
161
if inv.revision_id is not None:
163
[b' revision_id="', encode_and_escape(inv.revision_id), b'"'])
166
append(b'<inventory format="%s"%s>\n' % (
167
self.format_num, revid1))
168
append(b'<directory file_id="%s" name="%s" revision="%s" />\n' % (
169
encode_and_escape(inv.root.file_id),
170
encode_and_escape(inv.root.name),
171
encode_and_escape(inv.root.revision)))
173
def write_revision_to_lines(self, rev):
174
"""Revision object -> xml tree"""
175
# For the XML format, we need to write them as Unicode rather than as
176
# utf-8 strings. So that cElementTree can handle properly escaping
179
el = (b'<revision committer="%s" format="%s" '
180
b'inventory_sha1="%s" revision_id="%s" '
181
b'timestamp="%.3f"' % (
182
encode_and_escape(rev.committer),
183
self.revision_format_num or self.format_num,
185
encode_and_escape(cache_utf8.decode(rev.revision_id)),
187
if rev.timezone is not None:
188
el += b' timezone="%s"' % str(rev.timezone).encode('ascii')
189
lines.append(el + b'>\n')
190
message = encode_and_escape(escape_invalid_chars(rev.message)[0])
191
lines.extend((b'<message>' + message + b'</message>\n').splitlines(True))
193
lines.append(b'<parents>\n')
194
for parent_id in rev.parent_ids:
195
_mod_revision.check_not_reserved_id(parent_id)
197
b'<revision_ref revision_id="%s" />\n'
198
% encode_and_escape(cache_utf8.decode(parent_id)))
199
lines.append(b'</parents>\n')
201
preamble = b'<properties>'
202
for prop_name, prop_value in sorted(rev.properties.items()):
204
proplines = (preamble + b'<property name="%s">%s</property>\n' % (
205
encode_and_escape(prop_name),
206
encode_and_escape(escape_invalid_chars(prop_value)[0]))).splitlines(True)
208
proplines = [preamble + b'<property name="%s" />\n' % (encode_and_escape(prop_name), )]
210
lines.extend(proplines)
211
lines.append(b'</properties>\n')
212
lines.append(b'</revision>\n')
215
def _unpack_entry(self, elt, entry_cache=None, return_from_cache=False):
216
# This is here because it's overridden by xml7
217
return unpack_inventory_entry(elt, entry_cache,
220
def _unpack_inventory(self, elt, revision_id=None, entry_cache=None,
221
return_from_cache=False):
222
"""Construct from XML Element"""
223
inv = unpack_inventory_flat(elt, self.format_num, self._unpack_entry,
224
entry_cache, return_from_cache)
225
self._check_cache_size(len(inv), entry_cache)
228
def _unpack_revision(self, elt):
229
"""XML Element -> Revision object"""
230
format = elt.get('format')
231
format_num = self.format_num
232
if self.revision_format_num is not None:
233
format_num = self.revision_format_num
234
if format is not None:
235
if format.encode() != format_num:
236
raise BzrError("invalid format version %r on revision"
238
get_cached = get_utf8_or_ascii
239
rev = Revision(committer=elt.get('committer'),
240
timestamp=float(elt.get('timestamp')),
241
revision_id=get_cached(elt.get('revision_id')),
242
inventory_sha1=elt.get('inventory_sha1').encode('ascii')
244
parents = elt.find('parents')
245
if parents is not None:
247
rev.parent_ids.append(get_cached(p.get('revision_id')))
248
self._unpack_revision_properties(elt, rev)
249
v = elt.get('timezone')
253
rev.timezone = int(v)
254
rev.message = elt.findtext('message') # text of <message>
257
def _unpack_revision_properties(self, elt, rev):
258
"""Unpack properties onto a revision."""
259
props_elt = elt.find('properties')
260
if props_elt is None:
262
for prop_elt in props_elt:
263
if prop_elt.tag != 'property':
264
raise AssertionError(
265
"bad tag under properties list: %r" % prop_elt.tag)
266
name = prop_elt.get('name')
267
value = prop_elt.text
268
# If a property had an empty value ('') cElementTree reads
269
# that back as None, convert it back to '', so that all
270
# properties have string values
273
if name in rev.properties:
274
raise AssertionError("repeated property %r" % name)
275
rev.properties[name] = value
277
def _find_text_key_references(self, line_iterator):
278
"""Core routine for extracting references to texts from inventories.
280
This performs the translation of xml lines to revision ids.
282
:param line_iterator: An iterator of lines, origin_version_id
283
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
284
to whether they were referred to by the inventory of the
285
revision_id that they contain. Note that if that revision_id was
286
not part of the line_iterator's output then False will be given -
287
even though it may actually refer to that key.
289
if not self.support_altered_by_hack:
290
raise AssertionError(
291
"_find_text_key_references only "
292
"supported for branches which store inventory as unnested xml"
293
", not on %r" % self)
296
# this code needs to read every new line in every inventory for the
297
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
298
# not present in one of those inventories is unnecessary but not
299
# harmful because we are filtering by the revision id marker in the
300
# inventory lines : we only select file ids altered in one of those
301
# revisions. We don't need to see all lines in the inventory because
302
# only those added in an inventory in rev X can contain a revision=X
304
unescape_revid_cache = {}
305
unescape_fileid_cache = {}
307
# jam 20061218 In a big fetch, this handles hundreds of thousands
308
# of lines, so it has had a lot of inlining and optimizing done.
309
# Sorry that it is a little bit messy.
310
# Move several functions to be local variables, since this is a long
312
search = self._file_ids_altered_regex.search
313
unescape = _unescape_xml
314
setdefault = result.setdefault
315
for line, line_key in line_iterator:
319
# One call to match.group() returning multiple items is quite a
320
# bit faster than 2 calls to match.group() each returning 1
321
file_id, revision_id = match.group('file_id', 'revision_id')
323
# Inlining the cache lookups helps a lot when you make 170,000
324
# lines and 350k ids, versus 8.4 unique ids.
325
# Using a cache helps in 2 ways:
326
# 1) Avoids unnecessary decoding calls
327
# 2) Re-uses cached strings, which helps in future set and
329
# (2) is enough that removing encoding entirely along with
330
# the cache (so we are using plain strings) results in no
331
# performance improvement.
333
revision_id = unescape_revid_cache[revision_id]
335
unescaped = unescape(revision_id)
336
unescape_revid_cache[revision_id] = unescaped
337
revision_id = unescaped
339
# Note that unconditionally unescaping means that we deserialise
340
# every fileid, which for general 'pull' is not great, but we don't
341
# really want to have some many fulltexts that this matters anyway.
344
file_id = unescape_fileid_cache[file_id]
346
unescaped = unescape(file_id)
347
unescape_fileid_cache[file_id] = unescaped
350
key = (file_id, revision_id)
351
setdefault(key, False)
352
if revision_id == line_key[-1]:
357
serializer_v8 = Serializer_v8()
added
removed
breezy/bzr/xml8.py
