~bzr-pqm/bzr/bzr.dev

1773.4.1 by Martin Pool
Add pyflakes makefile target; fix many warnings
1
# Copyright (C) 2005, 2006 Canonical Ltd
2
#
1189 by Martin Pool
- BROKEN: partial support for commit into weave
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.
1887.1.1 by Adeodato Simó
Do not separate paragraphs in the copyright statement with blank lines,
7
#
1189 by Martin Pool
- BROKEN: partial support for commit into weave
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.
1887.1.1 by Adeodato Simó
Do not separate paragraphs in the copyright statement with blank lines,
12
#
1189 by Martin Pool
- BROKEN: partial support for commit into weave
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
16
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
17
import cStringIO
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
18
import re
1189 by Martin Pool
- BROKEN: partial support for commit into weave
19
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
20
from bzrlib import (
21
    cache_utf8,
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
22
    inventory,
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
23
    )
1773.4.1 by Martin Pool
Add pyflakes makefile target; fix many warnings
24
from bzrlib.xml_serializer import SubElement, Element, Serializer
1189 by Martin Pool
- BROKEN: partial support for commit into weave
25
from bzrlib.inventory import ROOT_ID, Inventory, InventoryEntry
1773.4.1 by Martin Pool
Add pyflakes makefile target; fix many warnings
26
from bzrlib.revision import Revision
1189 by Martin Pool
- BROKEN: partial support for commit into weave
27
from bzrlib.errors import BzrError
28
29
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
30
_utf8_re = None
31
_utf8_escape_map = {
32
    "&":'&',
33
    "'":"'", # FIXME: overkill
34
    "\"":""",
35
    "<":"&lt;",
36
    ">":"&gt;",
37
    }
38
39
40
def _ensure_utf8_re():
41
    """Make sure the _utf8_re regex has been compiled"""
42
    global _utf8_re
43
    if _utf8_re is not None:
44
        return
1934.1.12 by John Arbash Meinel
Switch back to using Entity serializer, since performance is equivalent, yet still compatible
45
    _utf8_re = re.compile(u'[&<>\'\"\u0080-\uffff]')
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
46
47
48
def _utf8_escape_replace(match, _map=_utf8_escape_map):
49
    """Replace a string of non-ascii, non XML safe characters with their escape
50
51
    This will escape both Standard XML escapes, like <>"', etc.
52
    As well as escaping non ascii characters, because ElementTree did.
53
    This helps us remain compatible to older versions of bzr. We may change
54
    our policy in the future, though.
55
    """
1934.1.7 by John Arbash Meinel
Comment why we do caching the way we do
56
    # jam 20060816 Benchmarks show that try/KeyError is faster if you
57
    # expect the entity to rarely miss. There is about a 10% difference
58
    # in overall time. But if you miss frequently, then if None is much
59
    # faster. For our use case, we *rarely* have a revision id, file id
60
    # or path name that is unicode. So use try/KeyError.
1934.1.12 by John Arbash Meinel
Switch back to using Entity serializer, since performance is equivalent, yet still compatible
61
    try:
62
        return _map[match.group()]
63
    except KeyError:
64
        return "&#%d;" % ord(match.group())
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
65
66
1934.1.5 by John Arbash Meinel
Cache the entity escaping cuts us down to 450ms
67
_unicode_to_escaped_map = {}
68
1934.1.12 by John Arbash Meinel
Switch back to using Entity serializer, since performance is equivalent, yet still compatible
69
def _encode_and_escape(unicode_str, _map=_unicode_to_escaped_map):
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
70
    """Encode the string into utf8, and escape invalid XML characters"""
1934.1.7 by John Arbash Meinel
Comment why we do caching the way we do
71
    # We frequently get entities we have not seen before, so it is better
72
    # to check if None, rather than try/KeyError
1934.1.5 by John Arbash Meinel
Cache the entity escaping cuts us down to 450ms
73
    text = _map.get(unicode_str)
74
    if text is None:
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
75
        # The alternative policy is to do a regular UTF8 encoding
1934.1.12 by John Arbash Meinel
Switch back to using Entity serializer, since performance is equivalent, yet still compatible
76
        # and then escape only XML meta characters.
77
        # Performance is equivalent once you use cache_utf8. *However*
78
        # this makes the serialized texts incompatible with old versions
79
        # of bzr. So no net gain. (Perhaps the read code would handle utf8
80
        # better than entity escapes, but cElementTree seems to do just fine
81
        # either way)
82
        text = str(_utf8_re.sub(_utf8_escape_replace, unicode_str)) + '"'
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
83
        _map[unicode_str] = text
1934.1.5 by John Arbash Meinel
Cache the entity escaping cuts us down to 450ms
84
    return text
85
86
87
def _clear_cache():
88
    """Clean out the unicode => escaped map"""
89
    _unicode_to_escaped_map.clear()
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
90
91
1189 by Martin Pool
- BROKEN: partial support for commit into weave
92
class Serializer_v5(Serializer):
93
    """Version 5 serializer
94
95
    Packs objects into XML and vice versa.
96
    """
97
    
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
98
    __slots__ = []
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
99
100
    def write_inventory_to_string(self, inv):
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
101
        """Just call write_inventory with a StringIO and return the value"""
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
102
        sio = cStringIO.StringIO()
103
        self.write_inventory(inv, sio)
104
        return sio.getvalue()
105
106
    def write_inventory(self, inv, f):
107
        """Write inventory to a file.
108
        
109
        :param inv: the inventory to write.
110
        :param f: the file to write.
111
        """
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
112
        _ensure_utf8_re()
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
113
        output = []
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
114
        append = output.append
115
        self._append_inventory_root(append, inv)
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
116
        entries = inv.iter_entries()
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
117
        # Skip the root
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
118
        root_path, root_ie = entries.next()
119
        for path, ie in entries:
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
120
            self._append_entry(append, ie)
121
        append('</inventory>\n')
1934.1.4 by John Arbash Meinel
rewrite escaper to use xml numerical entities, rather than using encode('utf8')
122
        f.writelines(output)
1934.1.5 by John Arbash Meinel
Cache the entity escaping cuts us down to 450ms
123
        # Just to keep the cache from growing without bounds
124
        # but we may actually not want to do clear the cache
1934.1.6 by John Arbash Meinel
With a full cache the time is down to 381 ms
125
        #_clear_cache()
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
126
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
127
    def _append_inventory_root(self, append, inv):
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
128
        """Append the inventory root to output."""
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
129
        append('<inventory')
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
130
        if inv.root.file_id not in (None, ROOT_ID):
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
131
            append(' file_id="')
1934.1.9 by John Arbash Meinel
Caching the trailing " and directly appending, down another 50ms, 275ms cached, 354ms uncached
132
            append(_encode_and_escape(inv.root.file_id))
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
133
        append(' format="5"')
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
134
        if inv.revision_id is not None:
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
135
            append(' revision_id="')
1934.1.9 by John Arbash Meinel
Caching the trailing " and directly appending, down another 50ms, 275ms cached, 354ms uncached
136
            append(_encode_and_escape(inv.revision_id))
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
137
        append('>\n')
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
138
        
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
139
    def _append_entry(self, append, ie):
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
140
        """Convert InventoryEntry to XML element and append to output."""
141
        # TODO: should just be a plain assertion
142
        assert InventoryEntry.versionable_kind(ie.kind), \
143
            'unsupported entry kind %s' % ie.kind
144
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
145
        append("<")
146
        append(ie.kind)
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
147
        if ie.executable:
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
148
            append(' executable="yes"')
149
        append(' file_id="')
1934.1.9 by John Arbash Meinel
Caching the trailing " and directly appending, down another 50ms, 275ms cached, 354ms uncached
150
        append(_encode_and_escape(ie.file_id))
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
151
        append(' name="')
1934.1.9 by John Arbash Meinel
Caching the trailing " and directly appending, down another 50ms, 275ms cached, 354ms uncached
152
        append(_encode_and_escape(ie.name))
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
153
        if ie.parent_id != ROOT_ID:
154
            assert isinstance(ie.parent_id, basestring)
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
155
            append(' parent_id="')
1934.1.9 by John Arbash Meinel
Caching the trailing " and directly appending, down another 50ms, 275ms cached, 354ms uncached
156
            append(_encode_and_escape(ie.parent_id))
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
157
        if ie.revision is not None:
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
158
            append(' revision="')
1934.1.9 by John Arbash Meinel
Caching the trailing " and directly appending, down another 50ms, 275ms cached, 354ms uncached
159
            append(_encode_and_escape(ie.revision))
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
160
        if ie.symlink_target is not None:
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
161
            append(' symlink_target="')
1934.1.9 by John Arbash Meinel
Caching the trailing " and directly appending, down another 50ms, 275ms cached, 354ms uncached
162
            append(_encode_and_escape(ie.symlink_target))
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
163
        if ie.text_sha1 is not None:
1934.1.15 by John Arbash Meinel
One field was incorrect, need text_sha1 not text_size
164
            append(' text_sha1="')
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
165
            append(ie.text_sha1)
166
            append('"')
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
167
        if ie.text_size is not None:
1934.1.8 by John Arbash Meinel
Passing around the append function rather than the list shaves off another 10%, down to 400ms
168
            append(' text_size="%d"' % ie.text_size)
169
        append(" />\n")
1934.1.3 by John Arbash Meinel
[merge] robert's custom XML serializer, and cleanup for benchmarks and iter_entries() differences
170
        return
171
1189 by Martin Pool
- BROKEN: partial support for commit into weave
172
    def _pack_revision(self, rev):
173
        """Revision object -> xml tree"""
174
        root = Element('revision',
175
                       committer = rev.committer,
176
                       timestamp = '%.9f' % rev.timestamp,
177
                       revision_id = rev.revision_id,
178
                       inventory_sha1 = rev.inventory_sha1,
1393.1.59 by Martin Pool
- put 'format=5' on inventory and revision xml
179
                       format='5',
1189 by Martin Pool
- BROKEN: partial support for commit into weave
180
                       )
1913.1.1 by John Arbash Meinel
Fix bug #55783
181
        if rev.timezone is not None:
1189 by Martin Pool
- BROKEN: partial support for commit into weave
182
            root.set('timezone', str(rev.timezone))
183
        root.text = '\n'
184
        msg = SubElement(root, 'message')
185
        msg.text = rev.message
186
        msg.tail = '\n'
1313 by Martin Pool
- rename to Revision.parent_ids to avoid confusion with old usage
187
        if rev.parent_ids:
1189 by Martin Pool
- BROKEN: partial support for commit into weave
188
            pelts = SubElement(root, 'parents')
189
            pelts.tail = pelts.text = '\n'
1313 by Martin Pool
- rename to Revision.parent_ids to avoid confusion with old usage
190
            for parent_id in rev.parent_ids:
1311 by Martin Pool
- remove RevisionReference; just hold parent ids directly
191
                assert isinstance(parent_id, basestring)
1189 by Martin Pool
- BROKEN: partial support for commit into weave
192
                p = SubElement(pelts, 'revision_ref')
193
                p.tail = '\n'
1311 by Martin Pool
- remove RevisionReference; just hold parent ids directly
194
                p.set('revision_id', parent_id)
1185.16.36 by Martin Pool
- store revision properties in revision xml
195
        if rev.properties:
196
            self._pack_revision_properties(rev, root)
1189 by Martin Pool
- BROKEN: partial support for commit into weave
197
        return root
1185.16.36 by Martin Pool
- store revision properties in revision xml
198
199
    def _pack_revision_properties(self, rev, under_element):
200
        top_elt = SubElement(under_element, 'properties')
201
        for prop_name, prop_value in sorted(rev.properties.items()):
202
            assert isinstance(prop_name, basestring) 
203
            assert isinstance(prop_value, basestring) 
204
            prop_elt = SubElement(top_elt, 'property')
205
            prop_elt.set('name', prop_name)
206
            prop_elt.text = prop_value
207
            prop_elt.tail = '\n'
208
        top_elt.tail = '\n'
209
1189 by Martin Pool
- BROKEN: partial support for commit into weave
210
    def _unpack_inventory(self, elt):
211
        """Construct from XML Element
212
        """
213
        assert elt.tag == 'inventory'
214
        root_id = elt.get('file_id') or ROOT_ID
1393.1.59 by Martin Pool
- put 'format=5' on inventory and revision xml
215
        format = elt.get('format')
216
        if format is not None:
217
            if format != '5':
218
                raise BzrError("invalid format version %r on inventory"
219
                                % format)
1638.1.2 by Robert Collins
Change the basis-inventory file to not have the revision-id in the file name.
220
        revision_id = elt.get('revision_id')
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
221
        if revision_id is not None:
222
            revision_id = cache_utf8.get_cached_unicode(revision_id)
1638.1.2 by Robert Collins
Change the basis-inventory file to not have the revision-id in the file name.
223
        inv = Inventory(root_id, revision_id=revision_id)
1189 by Martin Pool
- BROKEN: partial support for commit into weave
224
        for e in elt:
225
            ie = self._unpack_entry(e)
226
            if ie.parent_id == ROOT_ID:
227
                ie.parent_id = root_id
228
            inv.add(ie)
229
        return inv
230
231
    def _unpack_entry(self, elt):
232
        kind = elt.tag
1399.1.6 by Robert Collins
move exporting functionality into inventory.py - uncovers bug in symlink support
233
        if not InventoryEntry.versionable_kind(kind):
1092.2.20 by Robert Collins
symlink and weaves, whaddya know
234
            raise AssertionError('unsupported entry kind %s' % kind)
1189 by Martin Pool
- BROKEN: partial support for commit into weave
235
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
236
        get_cached = cache_utf8.get_cached_unicode
237
1189 by Martin Pool
- BROKEN: partial support for commit into weave
238
        parent_id = elt.get('parent_id')
1963.2.6 by Robey Pointer
pychecker is on crack; go back to using 'is None'.
239
        if parent_id is None:
1189 by Martin Pool
- BROKEN: partial support for commit into weave
240
            parent_id = ROOT_ID
1934.1.20 by John Arbash Meinel
Document why we aren't caching file ids at the moment
241
        # TODO: jam 20060817 At present, caching file ids costs us too 
242
        #       much time. It slows down overall read performances from
243
        #       approx 500ms to 700ms. And doesn't improve future reads.
244
        #       it might be because revision ids and file ids are mixing.
245
        #       Consider caching *just* the file ids, for a limited period
246
        #       of time.
1934.1.18 by John Arbash Meinel
Caching revision_ids costs us a little (512 vs 475), but caching file-ids costs us a lot (780 vs 512), so disabling caching of file ids for now
247
        #parent_id = get_cached(parent_id)
248
        #file_id = get_cached(elt.get('file_id'))
249
        file_id = elt.get('file_id')
1189 by Martin Pool
- BROKEN: partial support for commit into weave
250
1399.1.8 by Robert Collins
factor out inventory directory logic into 'InventoryDirectory' class
251
        if kind == 'directory':
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
252
            ie = inventory.InventoryDirectory(file_id,
1399.1.8 by Robert Collins
factor out inventory directory logic into 'InventoryDirectory' class
253
                                              elt.get('name'),
254
                                              parent_id)
1399.1.9 by Robert Collins
factor out file related logic from InventoryEntry to InventoryFile
255
        elif kind == 'file':
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
256
            ie = inventory.InventoryFile(file_id,
1399.1.9 by Robert Collins
factor out file related logic from InventoryEntry to InventoryFile
257
                                         elt.get('name'),
258
                                         parent_id)
259
            ie.text_sha1 = elt.get('text_sha1')
260
            if elt.get('executable') == 'yes':
261
                ie.executable = True
262
            v = elt.get('text_size')
263
            ie.text_size = v and int(v)
1399.1.10 by Robert Collins
remove kind from the InventoryEntry constructor - only child classes should be created now
264
        elif kind == 'symlink':
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
265
            ie = inventory.InventoryLink(file_id,
1399.1.10 by Robert Collins
remove kind from the InventoryEntry constructor - only child classes should be created now
266
                                         elt.get('name'),
267
                                         parent_id)
268
            ie.symlink_target = elt.get('symlink_target')
1399.1.8 by Robert Collins
factor out inventory directory logic into 'InventoryDirectory' class
269
        else:
1399.1.10 by Robert Collins
remove kind from the InventoryEntry constructor - only child classes should be created now
270
            raise BzrError("unknown kind %r" % kind)
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
271
        revision = elt.get('revision')
272
        if revision is not None:
273
            revision = get_cached(revision)
274
        ie.revision = revision
1189 by Martin Pool
- BROKEN: partial support for commit into weave
275
276
        return ie
277
278
    def _unpack_revision(self, elt):
279
        """XML Element -> Revision object"""
280
        assert elt.tag == 'revision'
1393.1.59 by Martin Pool
- put 'format=5' on inventory and revision xml
281
        format = elt.get('format')
282
        if format is not None:
283
            if format != '5':
284
                raise BzrError("invalid format version %r on inventory"
285
                                % format)
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
286
        get_cached = cache_utf8.get_cached_unicode
1189 by Martin Pool
- BROKEN: partial support for commit into weave
287
        rev = Revision(committer = elt.get('committer'),
288
                       timestamp = float(elt.get('timestamp')),
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
289
                       revision_id = get_cached(elt.get('revision_id')),
1189 by Martin Pool
- BROKEN: partial support for commit into weave
290
                       inventory_sha1 = elt.get('inventory_sha1')
291
                       )
292
        parents = elt.find('parents') or []
293
        for p in parents:
294
            assert p.tag == 'revision_ref', \
295
                   "bad parent node tag %r" % p.tag
1911.2.6 by John Arbash Meinel
Cache revision ids and file ids as part of xml processing. A custom xml parser could just call decode/encode directly.
296
            rev.parent_ids.append(get_cached(p.get('revision_id')))
1185.16.37 by Martin Pool
- properties are retrieved when revisions are loaded
297
        self._unpack_revision_properties(elt, rev)
1189 by Martin Pool
- BROKEN: partial support for commit into weave
298
        v = elt.get('timezone')
1913.1.1 by John Arbash Meinel
Fix bug #55783
299
        if v is None:
300
            rev.timezone = 0
301
        else:
302
            rev.timezone = int(v)
1189 by Martin Pool
- BROKEN: partial support for commit into weave
303
        rev.message = elt.findtext('message') # text of <message>
304
        return rev
305
1185.16.37 by Martin Pool
- properties are retrieved when revisions are loaded
306
    def _unpack_revision_properties(self, elt, rev):
307
        """Unpack properties onto a revision."""
308
        props_elt = elt.find('properties')
309
        assert len(rev.properties) == 0
310
        if not props_elt:
311
            return
312
        for prop_elt in props_elt:
313
            assert prop_elt.tag == 'property', \
1773.4.1 by Martin Pool
Add pyflakes makefile target; fix many warnings
314
                "bad tag under properties list: %r" % prop_elt.tag
1185.16.37 by Martin Pool
- properties are retrieved when revisions are loaded
315
            name = prop_elt.get('name')
316
            value = prop_elt.text
1886.1.1 by John Arbash Meinel
Fix bug #47782,
317
            # If a property had an empty value ('') cElementTree reads
318
            # that back as None, convert it back to '', so that all
319
            # properties have string values
320
            if value is None:
321
                value = ''
1185.16.37 by Martin Pool
- properties are retrieved when revisions are loaded
322
            assert name not in rev.properties, \
1773.4.1 by Martin Pool
Add pyflakes makefile target; fix many warnings
323
                "repeated property %r" % name
1185.16.37 by Martin Pool
- properties are retrieved when revisions are loaded
324
            rev.properties[name] = value
325
326
1189 by Martin Pool
- BROKEN: partial support for commit into weave
327
serializer_v5 = Serializer_v5()