14
14
# along with this program; if not, write to the Free Software
15
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20
17
from bzrlib import (
24
from bzrlib.xml_serializer import SubElement, Element, Serializer
25
from bzrlib.inventory import ROOT_ID, Inventory, InventoryEntry
26
from bzrlib.revision import Revision
27
from bzrlib.errors import BzrError
33
"'":"'", # FIXME: overkill
40
def _ensure_utf8_re():
41
"""Make sure the _utf8_re regex has been compiled"""
43
if _utf8_re is not None:
45
_utf8_re = re.compile(u'[&<>\'\"\u0080-\uffff]')
48
def _utf8_escape_replace(match, _map=_utf8_escape_map):
49
"""Replace a string of non-ascii, non XML safe characters with their escape
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.
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.
62
return _map[match.group()]
64
return "&#%d;" % ord(match.group())
67
_unicode_to_escaped_map = {}
69
def _encode_and_escape(unicode_str, _map=_unicode_to_escaped_map):
70
"""Encode the string into utf8, and escape invalid XML characters"""
71
# We frequently get entities we have not seen before, so it is better
72
# to check if None, rather than try/KeyError
73
text = _map.get(unicode_str)
75
# The alternative policy is to do a regular UTF8 encoding
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
82
text = str(_utf8_re.sub(_utf8_escape_replace, unicode_str)) + '"'
83
_map[unicode_str] = text
88
"""Clean out the unicode => escaped map"""
89
_unicode_to_escaped_map.clear()
92
class Serializer_v5(Serializer):
24
class Serializer_v5(xml6.Serializer_v6):
93
25
"""Version 5 serializer
95
27
Packs objects into XML and vice versa.
100
support_altered_by_hack = True
101
# This format supports the altered-by hack that reads file ids directly out
102
# of the versionedfile, without doing XML parsing.
104
def write_inventory_to_string(self, inv):
105
"""Just call write_inventory with a StringIO and return the value"""
106
sio = cStringIO.StringIO()
107
self.write_inventory(inv, sio)
108
return sio.getvalue()
110
def write_inventory(self, inv, f):
111
"""Write inventory to a file.
113
:param inv: the inventory to write.
114
:param f: the file to write.
118
append = output.append
119
self._append_inventory_root(append, inv)
120
entries = inv.iter_entries()
122
root_path, root_ie = entries.next()
123
for path, ie in entries:
124
self._append_entry(append, ie)
125
append('</inventory>\n')
127
# Just to keep the cache from growing without bounds
128
# but we may actually not want to do clear the cache
131
def _append_inventory_root(self, append, inv):
132
"""Append the inventory root to output."""
134
if inv.root.file_id not in (None, ROOT_ID):
136
append(_encode_and_escape(inv.root.file_id))
137
append(' format="5"')
138
if inv.revision_id is not None:
139
append(' revision_id="')
140
append(_encode_and_escape(inv.revision_id))
143
def _append_entry(self, append, ie):
144
"""Convert InventoryEntry to XML element and append to output."""
145
# TODO: should just be a plain assertion
146
assert InventoryEntry.versionable_kind(ie.kind), \
147
'unsupported entry kind %s' % ie.kind
152
append(' executable="yes"')
154
append(_encode_and_escape(ie.file_id))
156
append(_encode_and_escape(ie.name))
157
if self._parent_condition(ie):
158
assert isinstance(ie.parent_id, basestring)
159
append(' parent_id="')
160
append(_encode_and_escape(ie.parent_id))
161
if ie.revision is not None:
162
append(' revision="')
163
append(_encode_and_escape(ie.revision))
164
if ie.symlink_target is not None:
165
append(' symlink_target="')
166
append(_encode_and_escape(ie.symlink_target))
167
if ie.text_sha1 is not None:
168
append(' text_sha1="')
171
if ie.text_size is not None:
172
append(' text_size="%d"' % ie.text_size)
176
def _parent_condition(self, ie):
177
return ie.parent_id != ROOT_ID
179
def _pack_revision(self, rev):
180
"""Revision object -> xml tree"""
181
root = Element('revision',
182
committer = rev.committer,
183
timestamp = '%.9f' % rev.timestamp,
184
revision_id = rev.revision_id,
185
inventory_sha1 = rev.inventory_sha1,
188
if rev.timezone is not None:
189
root.set('timezone', str(rev.timezone))
191
msg = SubElement(root, 'message')
192
msg.text = rev.message
195
pelts = SubElement(root, 'parents')
196
pelts.tail = pelts.text = '\n'
197
for parent_id in rev.parent_ids:
198
assert isinstance(parent_id, basestring)
199
p = SubElement(pelts, 'revision_ref')
201
p.set('revision_id', parent_id)
203
self._pack_revision_properties(rev, root)
206
def _pack_revision_properties(self, rev, under_element):
207
top_elt = SubElement(under_element, 'properties')
208
for prop_name, prop_value in sorted(rev.properties.items()):
209
assert isinstance(prop_name, basestring)
210
assert isinstance(prop_value, basestring)
211
prop_elt = SubElement(top_elt, 'property')
212
prop_elt.set('name', prop_name)
213
prop_elt.text = prop_value
217
def _unpack_inventory(self, elt):
30
root_id = inventory.ROOT_ID
32
def _unpack_inventory(self, elt, revision_id):
218
33
"""Construct from XML Element
220
assert elt.tag == 'inventory'
221
root_id = elt.get('file_id') or ROOT_ID
35
root_id = elt.get('file_id') or inventory.ROOT_ID
36
root_id = xml8._get_utf8_or_ascii(root_id)
222
38
format = elt.get('format')
223
39
if format is not None:
225
41
raise BzrError("invalid format version %r on inventory"
227
revision_id = elt.get('revision_id')
228
if revision_id is not None:
229
revision_id = cache_utf8.get_cached_unicode(revision_id)
230
inv = Inventory(root_id, revision_id=revision_id)
43
data_revision_id = elt.get('revision_id')
44
if data_revision_id is not None:
45
revision_id = cache_utf8.encode(data_revision_id)
46
inv = inventory.Inventory(root_id, revision_id=revision_id)
232
48
ie = self._unpack_entry(e)
233
if ie.parent_id == ROOT_ID:
49
if ie.parent_id is None:
234
50
ie.parent_id = root_id
52
if revision_id is not None:
53
inv.root.revision = revision_id
238
def _unpack_entry(self, elt, none_parents=False):
240
if not InventoryEntry.versionable_kind(kind):
241
raise AssertionError('unsupported entry kind %s' % kind)
243
get_cached = cache_utf8.get_cached_unicode
245
parent_id = elt.get('parent_id')
246
if parent_id is None and not none_parents:
248
# TODO: jam 20060817 At present, caching file ids costs us too
249
# much time. It slows down overall read performances from
250
# approx 500ms to 700ms. And doesn't improve future reads.
251
# it might be because revision ids and file ids are mixing.
252
# Consider caching *just* the file ids, for a limited period
254
#parent_id = get_cached(parent_id)
255
#file_id = get_cached(elt.get('file_id'))
256
file_id = elt.get('file_id')
258
if kind == 'directory':
259
ie = inventory.InventoryDirectory(file_id,
263
ie = inventory.InventoryFile(file_id,
266
ie.text_sha1 = elt.get('text_sha1')
267
if elt.get('executable') == 'yes':
269
v = elt.get('text_size')
270
ie.text_size = v and int(v)
271
elif kind == 'symlink':
272
ie = inventory.InventoryLink(file_id,
275
ie.symlink_target = elt.get('symlink_target')
277
raise BzrError("unknown kind %r" % kind)
278
revision = elt.get('revision')
279
if revision is not None:
280
revision = get_cached(revision)
281
ie.revision = revision
285
def _unpack_revision(self, elt):
286
"""XML Element -> Revision object"""
287
assert elt.tag == 'revision'
288
format = elt.get('format')
289
if format is not None:
291
raise BzrError("invalid format version %r on inventory"
293
get_cached = cache_utf8.get_cached_unicode
294
rev = Revision(committer = elt.get('committer'),
295
timestamp = float(elt.get('timestamp')),
296
revision_id = get_cached(elt.get('revision_id')),
297
inventory_sha1 = elt.get('inventory_sha1')
299
parents = elt.find('parents') or []
301
assert p.tag == 'revision_ref', \
302
"bad parent node tag %r" % p.tag
303
rev.parent_ids.append(get_cached(p.get('revision_id')))
304
self._unpack_revision_properties(elt, rev)
305
v = elt.get('timezone')
309
rev.timezone = int(v)
310
rev.message = elt.findtext('message') # text of <message>
313
def _unpack_revision_properties(self, elt, rev):
314
"""Unpack properties onto a revision."""
315
props_elt = elt.find('properties')
316
assert len(rev.properties) == 0
319
for prop_elt in props_elt:
320
assert prop_elt.tag == 'property', \
321
"bad tag under properties list: %r" % prop_elt.tag
322
name = prop_elt.get('name')
323
value = prop_elt.text
324
# If a property had an empty value ('') cElementTree reads
325
# that back as None, convert it back to '', so that all
326
# properties have string values
329
assert name not in rev.properties, \
330
"repeated property %r" % name
331
rev.properties[name] = value
56
def _check_revisions(self, inv):
57
"""Extension point for subclasses to check during serialisation.
59
In this version, no checking is done.
61
:param inv: An inventory about to be serialised, to be checked.
62
:raises: AssertionError if an error has occured.
65
def _append_inventory_root(self, append, inv):
66
"""Append the inventory root to output."""
67
if inv.root.file_id not in (None, inventory.ROOT_ID):
68
fileid1 = ' file_id="'
69
fileid2 = xml8._encode_and_escape(inv.root.file_id)
73
if inv.revision_id is not None:
74
revid1 = ' revision_id="'
75
revid2 = xml8._encode_and_escape(inv.revision_id)
79
append('<inventory%s%s format="5"%s%s>\n' % (
80
fileid1, fileid2, revid1, revid2))
334
83
serializer_v5 = Serializer_v5()
added
removed
bzrlib/xml5.py
