1
# Copyright (C) 2005-2010 Canonical Ltd
3
1
# This program is free software; you can redistribute it and/or modify
4
2
# it under the terms of the GNU General Public License as published by
5
3
# the Free Software Foundation; either version 2 of the License, or
6
4
# (at your option) any later version.
8
6
# This program is distributed in the hope that it will be useful,
9
7
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
8
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
9
# GNU General Public License for more details.
13
11
# You should have received a copy of the GNU General Public License
14
12
# 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
19
from io import BytesIO
24
revision as _mod_revision,
27
from .xml_serializer import (
34
serialize_inventory_flat,
35
unpack_inventory_entry,
36
unpack_inventory_flat,
38
from ..revision import Revision
39
from ..sixish import unichr
40
from ..errors import BzrError
52
def _unescaper(match, _map=_xml_unescape_map):
57
if not code.startswith(b'#'):
59
return unichr(int(code[1:])).encode('utf8')
62
_unescape_re = lazy_regex.lazy_compile(b'\\&([^;]*);')
65
def _unescape_xml(data):
66
"""Unescape predefined XML entities in a string of data."""
67
return _unescape_re.sub(_unescaper, data)
70
class Serializer_v8(XMLSerializer):
71
"""This serialiser adds rich roots.
73
Its revision format number matches its inventory number.
13
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
16
from bzrlib.xml import ElementTree, SubElement, Element, Serializer
17
from bzrlib.inventory import ROOT_ID, Inventory, InventoryEntry
18
import bzrlib.inventory as inventory
19
from bzrlib.revision import Revision
20
from bzrlib.errors import BzrError
23
class Serializer_v5(Serializer):
24
"""Version 5 serializer
26
Packs objects into XML and vice versa.
79
support_altered_by_hack = True
80
# This format supports the altered-by hack that reads file ids directly out
81
# of the versionedfile, without doing XML parsing.
83
supported_kinds = {'file', 'directory', 'symlink'}
85
revision_format_num = None
87
# The search regex used by xml based repositories to determine what things
88
# where changed in a single commit.
89
_file_ids_altered_regex = lazy_regex.lazy_compile(
90
b'file_id="(?P<file_id>[^"]+)"'
91
b'.* revision="(?P<revision_id>[^"]+)"'
94
def _check_revisions(self, inv):
95
"""Extension point for subclasses to check during serialisation.
97
:param inv: An inventory about to be serialised, to be checked.
98
:raises: AssertionError if an error has occurred.
100
if inv.revision_id is None:
101
raise AssertionError("inv.revision_id is None")
102
if inv.root.revision is None:
103
raise AssertionError("inv.root.revision is None")
105
def _check_cache_size(self, inv_size, entry_cache):
106
"""Check that the entry_cache is large enough.
108
We want the cache to be ~2x the size of an inventory. The reason is
109
because we use a FIFO cache, and how Inventory records are likely to
110
change. In general, you have a small number of records which change
111
often, and a lot of records which do not change at all. So when the
112
cache gets full, you actually flush out a lot of the records you are
113
interested in, which means you need to recreate all of those records.
114
An LRU Cache would be better, but the overhead negates the cache
117
One way to look at it, only the size of the cache > len(inv) is your
118
'working' set. And in general, it shouldn't be a problem to hold 2
119
inventories in memory anyway.
121
:param inv_size: The number of entries in an inventory.
123
if entry_cache is None:
125
# 1.5 times might also be reasonable.
126
recommended_min_cache_size = inv_size * 1.5
127
if entry_cache.cache_size() < recommended_min_cache_size:
128
recommended_cache_size = inv_size * 2
129
trace.mutter('Resizing the inventory entry cache from %d to %d',
130
entry_cache.cache_size(), recommended_cache_size)
131
entry_cache.resize(recommended_cache_size)
133
def write_inventory_to_lines(self, inv):
134
"""Return a list of lines with the encoded inventory."""
135
return self.write_inventory(inv, None)
137
def write_inventory_to_chunks(self, inv):
138
return self.write_inventory(inv, None)
140
def write_inventory(self, inv, f, working=False):
141
"""Write inventory to a file.
143
:param inv: the inventory to write.
144
:param f: the file to write. (May be None if the lines are the desired
146
:param working: If True skip history data - text_sha1, text_size,
147
reference_revision, symlink_target.
148
:return: The inventory as a list of lines.
151
append = output.append
152
self._append_inventory_root(append, inv)
153
serialize_inventory_flat(inv, append,
154
self.root_id, self.supported_kinds, working)
157
# Just to keep the cache from growing without bounds
158
# but we may actually not want to do clear the cache
162
def _append_inventory_root(self, append, inv):
163
"""Append the inventory root to output."""
164
if inv.revision_id is not None:
166
[b' revision_id="', encode_and_escape(inv.revision_id), b'"'])
169
append(b'<inventory format="%s"%s>\n' % (
170
self.format_num, revid1))
171
append(b'<directory file_id="%s" name="%s" revision="%s" />\n' % (
172
encode_and_escape(inv.root.file_id),
173
encode_and_escape(inv.root.name),
174
encode_and_escape(inv.root.revision)))
176
def write_revision_to_lines(self, rev):
31
def _pack_inventory(self, inv):
32
"""Convert to XML Element"""
33
e = Element('inventory',
36
if inv.root.file_id not in (None, ROOT_ID):
37
e.set('file_id', inv.root.file_id)
38
for path, ie in inv.iter_entries():
39
e.append(self._pack_entry(ie))
43
def _pack_entry(self, ie):
44
"""Convert InventoryEntry to XML element"""
45
if not InventoryEntry.versionable_kind(ie.kind):
46
raise AssertionError('unsupported entry kind %s' % ie.kind)
48
e.set('name', ie.name)
49
e.set('file_id', ie.file_id)
51
if ie.text_size != None:
52
e.set('text_size', '%d' % ie.text_size)
54
for f in ['text_sha1', 'revision', 'symlink_target']:
60
e.set('executable', 'yes')
62
# to be conservative, we don't externalize the root pointers
63
# for now, leaving them as null in the xml form. in a future
64
# version it will be implied by nested elements.
65
if ie.parent_id != ROOT_ID:
66
assert isinstance(ie.parent_id, basestring)
67
e.set('parent_id', ie.parent_id)
74
def _pack_revision(self, rev):
177
75
"""Revision object -> xml tree"""
178
# For the XML format, we need to write them as Unicode rather than as
179
# utf-8 strings. So that cElementTree can handle properly escaping
182
el = (b'<revision committer="%s" format="%s" '
183
b'inventory_sha1="%s" revision_id="%s" '
184
b'timestamp="%.3f"' % (
185
encode_and_escape(rev.committer),
186
self.revision_format_num or self.format_num,
188
encode_and_escape(cache_utf8.decode(rev.revision_id)),
190
if rev.timezone is not None:
191
el += b' timezone="%s"' % str(rev.timezone).encode('ascii')
192
lines.append(el + b'>\n')
193
message = encode_and_escape(escape_invalid_chars(rev.message)[0])
194
lines.extend((b'<message>' + message + b'</message>\n').splitlines(True))
76
root = Element('revision',
77
committer = rev.committer,
78
timestamp = '%.9f' % rev.timestamp,
79
revision_id = rev.revision_id,
80
inventory_sha1 = rev.inventory_sha1,
84
root.set('timezone', str(rev.timezone))
86
msg = SubElement(root, 'message')
87
msg.text = rev.message
195
89
if rev.parent_ids:
196
lines.append(b'<parents>\n')
90
pelts = SubElement(root, 'parents')
91
pelts.tail = pelts.text = '\n'
197
92
for parent_id in rev.parent_ids:
198
_mod_revision.check_not_reserved_id(parent_id)
200
b'<revision_ref revision_id="%s" />\n'
201
% encode_and_escape(cache_utf8.decode(parent_id)))
202
lines.append(b'</parents>\n')
93
assert isinstance(parent_id, basestring)
94
p = SubElement(pelts, 'revision_ref')
96
p.set('revision_id', parent_id)
203
97
if rev.properties:
204
preamble = b'<properties>'
205
for prop_name, prop_value in sorted(rev.properties.items()):
207
proplines = (preamble + b'<property name="%s">%s</property>\n' % (
208
encode_and_escape(prop_name),
209
encode_and_escape(escape_invalid_chars(prop_value)[0]))).splitlines(True)
211
proplines = [preamble + b'<property name="%s" />\n' % (encode_and_escape(prop_name), )]
213
lines.extend(proplines)
214
lines.append(b'</properties>\n')
215
lines.append(b'</revision>\n')
218
def _unpack_entry(self, elt, entry_cache=None, return_from_cache=False):
219
# This is here because it's overridden by xml7
220
return unpack_inventory_entry(elt, entry_cache,
223
def _unpack_inventory(self, elt, revision_id=None, entry_cache=None,
224
return_from_cache=False):
225
"""Construct from XML Element"""
226
inv = unpack_inventory_flat(elt, self.format_num, self._unpack_entry,
227
entry_cache, return_from_cache)
228
self._check_cache_size(len(inv), entry_cache)
98
self._pack_revision_properties(rev, root)
102
def _pack_revision_properties(self, rev, under_element):
103
top_elt = SubElement(under_element, 'properties')
104
for prop_name, prop_value in sorted(rev.properties.items()):
105
assert isinstance(prop_name, basestring)
106
assert isinstance(prop_value, basestring)
107
prop_elt = SubElement(top_elt, 'property')
108
prop_elt.set('name', prop_name)
109
prop_elt.text = prop_value
114
def _unpack_inventory(self, elt):
115
"""Construct from XML Element
117
assert elt.tag == 'inventory'
118
root_id = elt.get('file_id') or ROOT_ID
119
format = elt.get('format')
120
if format is not None:
122
raise BzrError("invalid format version %r on inventory"
124
inv = Inventory(root_id)
126
ie = self._unpack_entry(e)
127
if ie.parent_id == ROOT_ID:
128
ie.parent_id = root_id
133
def _unpack_entry(self, elt):
135
if not InventoryEntry.versionable_kind(kind):
136
raise AssertionError('unsupported entry kind %s' % kind)
138
parent_id = elt.get('parent_id')
139
if parent_id == None:
142
if kind == 'directory':
143
ie = inventory.InventoryDirectory(elt.get('file_id'),
147
ie = inventory.InventoryFile(elt.get('file_id'),
150
ie.text_sha1 = elt.get('text_sha1')
151
if elt.get('executable') == 'yes':
153
v = elt.get('text_size')
154
ie.text_size = v and int(v)
155
elif kind == 'symlink':
156
ie = inventory.InventoryLink(elt.get('file_id'),
159
ie.symlink_target = elt.get('symlink_target')
161
raise BzrError("unknown kind %r" % kind)
162
ie.revision = elt.get('revision')
231
167
def _unpack_revision(self, elt):
232
168
"""XML Element -> Revision object"""
169
assert elt.tag == 'revision'
233
170
format = elt.get('format')
234
format_num = self.format_num
235
if self.revision_format_num is not None:
236
format_num = self.revision_format_num
237
171
if format is not None:
238
if format.encode() != format_num:
239
raise BzrError("invalid format version %r on revision"
241
get_cached = get_utf8_or_ascii
242
rev = Revision(committer=elt.get('committer'),
243
timestamp=float(elt.get('timestamp')),
244
revision_id=get_cached(elt.get('revision_id')),
245
inventory_sha1=elt.get('inventory_sha1').encode('ascii')
173
raise BzrError("invalid format version %r on inventory"
175
rev = Revision(committer = elt.get('committer'),
176
timestamp = float(elt.get('timestamp')),
177
revision_id = elt.get('revision_id'),
178
inventory_sha1 = elt.get('inventory_sha1')
247
180
parents = elt.find('parents') or []
248
181
for p in parents:
249
rev.parent_ids.append(get_cached(p.get('revision_id')))
182
assert p.tag == 'revision_ref', \
183
"bad parent node tag %r" % p.tag
184
rev.parent_ids.append(p.get('revision_id'))
250
185
self._unpack_revision_properties(elt, rev)
251
186
v = elt.get('timezone')
255
rev.timezone = int(v)
256
rev.message = elt.findtext('message') # text of <message>
187
rev.timezone = v and int(v)
188
rev.message = elt.findtext('message') # text of <message>
259
192
def _unpack_revision_properties(self, elt, rev):
260
193
"""Unpack properties onto a revision."""
261
194
props_elt = elt.find('properties')
262
if props_elt is None:
195
assert len(rev.properties) == 0
264
198
for prop_elt in props_elt:
265
if prop_elt.tag != 'property':
266
raise AssertionError(
267
"bad tag under properties list: %r" % prop_elt.tag)
199
assert prop_elt.tag == 'property', \
200
"bad tag under properties list: %r" % p.tag
268
201
name = prop_elt.get('name')
269
202
value = prop_elt.text
270
# If a property had an empty value ('') cElementTree reads
271
# that back as None, convert it back to '', so that all
272
# properties have string values
275
if name in rev.properties:
276
raise AssertionError("repeated property %r" % name)
203
assert name not in rev.properties, \
204
"repeated property %r" % p.name
277
205
rev.properties[name] = value
279
def _find_text_key_references(self, line_iterator):
280
"""Core routine for extracting references to texts from inventories.
282
This performs the translation of xml lines to revision ids.
284
:param line_iterator: An iterator of lines, origin_version_id
285
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
286
to whether they were referred to by the inventory of the
287
revision_id that they contain. Note that if that revision_id was
288
not part of the line_iterator's output then False will be given -
289
even though it may actually refer to that key.
291
if not self.support_altered_by_hack:
292
raise AssertionError(
293
"_find_text_key_references only "
294
"supported for branches which store inventory as unnested xml"
295
", not on %r" % self)
298
# this code needs to read every new line in every inventory for the
299
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
300
# not present in one of those inventories is unnecessary but not
301
# harmful because we are filtering by the revision id marker in the
302
# inventory lines : we only select file ids altered in one of those
303
# revisions. We don't need to see all lines in the inventory because
304
# only those added in an inventory in rev X can contain a revision=X
306
unescape_revid_cache = {}
307
unescape_fileid_cache = {}
309
# jam 20061218 In a big fetch, this handles hundreds of thousands
310
# of lines, so it has had a lot of inlining and optimizing done.
311
# Sorry that it is a little bit messy.
312
# Move several functions to be local variables, since this is a long
314
search = self._file_ids_altered_regex.search
315
unescape = _unescape_xml
316
setdefault = result.setdefault
317
for line, line_key in line_iterator:
321
# One call to match.group() returning multiple items is quite a
322
# bit faster than 2 calls to match.group() each returning 1
323
file_id, revision_id = match.group('file_id', 'revision_id')
325
# Inlining the cache lookups helps a lot when you make 170,000
326
# lines and 350k ids, versus 8.4 unique ids.
327
# Using a cache helps in 2 ways:
328
# 1) Avoids unnecessary decoding calls
329
# 2) Re-uses cached strings, which helps in future set and
331
# (2) is enough that removing encoding entirely along with
332
# the cache (so we are using plain strings) results in no
333
# performance improvement.
335
revision_id = unescape_revid_cache[revision_id]
337
unescaped = unescape(revision_id)
338
unescape_revid_cache[revision_id] = unescaped
339
revision_id = unescaped
341
# Note that unconditionally unescaping means that we deserialise
342
# every fileid, which for general 'pull' is not great, but we don't
343
# really want to have some many fulltexts that this matters anyway.
346
file_id = unescape_fileid_cache[file_id]
348
unescaped = unescape(file_id)
349
unescape_fileid_cache[file_id] = unescaped
352
key = (file_id, revision_id)
353
setdefault(key, False)
354
if revision_id == line_key[-1]:
359
serializer_v8 = Serializer_v8()
208
serializer_v5 = Serializer_v5()
added
removed
bzrlib/xml5.py
