68
64
return "&#%d;" % ord(match.group())
71
def _utf8_escape_replace(match, _map=_xml_escape_map):
72
"""Escape utf8 characters into XML safe ones.
74
This uses 2 tricks. It is either escaping "standard" characters, like "&<>,
75
or it is handling characters with the high-bit set. For ascii characters,
76
we just lookup the replacement in the dictionary. For everything else, we
77
decode back into Unicode, and then use the XML escape code.
80
return _map[match.group()]
82
return ''.join('&#%d;' % ord(uni_chr)
83
for uni_chr in match.group().decode('utf8'))
88
def _encode_and_escape(unicode_or_utf8_str, _map=_to_escaped_map):
67
_unicode_to_escaped_map = {}
69
def _encode_and_escape(unicode_str, _map=_unicode_to_escaped_map):
89
70
"""Encode the string into utf8, and escape invalid XML characters"""
90
71
# We frequently get entities we have not seen before, so it is better
91
72
# to check if None, rather than try/KeyError
92
text = _map.get(unicode_or_utf8_str)
73
text = _map.get(unicode_str)
94
if unicode_or_utf8_str.__class__ == unicode:
95
# The alternative policy is to do a regular UTF8 encoding
96
# and then escape only XML meta characters.
97
# Performance is equivalent once you use cache_utf8. *However*
98
# this makes the serialized texts incompatible with old versions
99
# of bzr. So no net gain. (Perhaps the read code would handle utf8
100
# better than entity escapes, but cElementTree seems to do just fine
102
text = str(_unicode_re.sub(_unicode_escape_replace,
103
unicode_or_utf8_str)) + '"'
105
# Plain strings are considered to already be in utf-8 so we do a
106
# slightly different method for escaping.
107
text = _utf8_re.sub(_utf8_escape_replace,
108
unicode_or_utf8_str) + '"'
109
_map[unicode_or_utf8_str] = text
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
113
def _get_utf8_or_ascii(a_str,
114
_encode_utf8=cache_utf8.encode,
115
_get_cached_ascii=cache_utf8.get_cached_ascii):
116
"""Return a cached version of the string.
118
cElementTree will return a plain string if the XML is plain ascii. It only
119
returns Unicode when it needs to. We want to work in utf-8 strings. So if
120
cElementTree returns a plain string, we can just return the cached version.
121
If it is Unicode, then we need to encode it.
123
:param a_str: An 8-bit string or Unicode as returned by
124
cElementTree.Element.get()
125
:return: A utf-8 encoded 8-bit string.
127
# This is fairly optimized because we know what cElementTree does, this is
128
# not meant as a generic function for all cases. Because it is possible for
129
# an 8-bit string to not be ascii or valid utf8.
130
if a_str.__class__ == unicode:
131
return _encode_utf8(a_str)
133
return _get_cached_ascii(a_str)
136
87
def _clear_cache():
137
88
"""Clean out the unicode => escaped map"""
138
_to_escaped_map.clear()
89
_unicode_to_escaped_map.clear()
141
92
class Serializer_v5(Serializer):
150
100
support_altered_by_hack = True
151
101
# This format supports the altered-by hack that reads file ids directly out
152
102
# of the versionedfile, without doing XML parsing.
154
supported_kinds = set(['file', 'directory', 'symlink'])
157
def _check_revisions(self, inv):
158
"""Extension point for subclasses to check during serialisation.
160
By default no checking is done.
162
:param inv: An inventory about to be serialised, to be checked.
163
:raises: AssertionError if an error has occured.
166
def write_inventory_to_lines(self, inv):
167
"""Return a list of lines with the encoded inventory."""
168
return self.write_inventory(inv, None)
170
def write_inventory_to_string(self, inv, working=False):
171
"""Just call write_inventory with a StringIO and return the value.
173
:param working: If True skip history data - text_sha1, text_size,
174
reference_revision, symlink_target.
104
def write_inventory_to_string(self, inv):
105
"""Just call write_inventory with a StringIO and return the value"""
176
106
sio = cStringIO.StringIO()
177
self.write_inventory(inv, sio, working)
107
self.write_inventory(inv, sio)
178
108
return sio.getvalue()
180
def write_inventory(self, inv, f, working=False):
110
def write_inventory(self, inv, f):
181
111
"""Write inventory to a file.
183
113
:param inv: the inventory to write.
184
:param f: the file to write. (May be None if the lines are the desired
186
:param working: If True skip history data - text_sha1, text_size,
187
reference_revision, symlink_target.
188
:return: The inventory as a list of lines.
114
:param f: the file to write.
190
116
_ensure_utf8_re()
191
self._check_revisions(inv)
193
118
append = output.append
194
119
self._append_inventory_root(append, inv)
197
122
root_path, root_ie = entries.next()
198
123
for path, ie in entries:
199
if ie.parent_id != self.root_id:
200
parent_str = ' parent_id="'
201
parent_id = _encode_and_escape(ie.parent_id)
205
if ie.kind == 'file':
207
executable = ' executable="yes"'
211
append('<file%s file_id="%s name="%s%s%s revision="%s '
212
'text_sha1="%s" text_size="%d" />\n' % (
213
executable, _encode_and_escape(ie.file_id),
214
_encode_and_escape(ie.name), parent_str, parent_id,
215
_encode_and_escape(ie.revision), ie.text_sha1,
218
append('<file%s file_id="%s name="%s%s%s />\n' % (
219
executable, _encode_and_escape(ie.file_id),
220
_encode_and_escape(ie.name), parent_str, parent_id))
221
elif ie.kind == 'directory':
223
append('<directory file_id="%s name="%s%s%s revision="%s '
225
_encode_and_escape(ie.file_id),
226
_encode_and_escape(ie.name),
227
parent_str, parent_id,
228
_encode_and_escape(ie.revision)))
230
append('<directory file_id="%s name="%s%s%s />\n' % (
231
_encode_and_escape(ie.file_id),
232
_encode_and_escape(ie.name),
233
parent_str, parent_id))
234
elif ie.kind == 'symlink':
236
append('<symlink file_id="%s name="%s%s%s revision="%s '
237
'symlink_target="%s />\n' % (
238
_encode_and_escape(ie.file_id),
239
_encode_and_escape(ie.name),
240
parent_str, parent_id,
241
_encode_and_escape(ie.revision),
242
_encode_and_escape(ie.symlink_target)))
244
append('<symlink file_id="%s name="%s%s%s />\n' % (
245
_encode_and_escape(ie.file_id),
246
_encode_and_escape(ie.name),
247
parent_str, parent_id))
248
elif ie.kind == 'tree-reference':
249
if ie.kind not in self.supported_kinds:
250
raise errors.UnsupportedInventoryKind(ie.kind)
252
append('<tree-reference file_id="%s name="%s%s%s '
253
'revision="%s reference_revision="%s />\n' % (
254
_encode_and_escape(ie.file_id),
255
_encode_and_escape(ie.name),
256
parent_str, parent_id,
257
_encode_and_escape(ie.revision),
258
_encode_and_escape(ie.reference_revision)))
260
append('<tree-reference file_id="%s name="%s%s%s />\n' % (
261
_encode_and_escape(ie.file_id),
262
_encode_and_escape(ie.name),
263
parent_str, parent_id))
265
raise errors.UnsupportedInventoryKind(ie.kind)
124
self._append_entry(append, ie)
266
125
append('</inventory>\n')
269
127
# Just to keep the cache from growing without bounds
270
128
# but we may actually not want to do clear the cache
274
131
def _append_inventory_root(self, append, inv):
275
132
"""Append the inventory root to output."""
276
134
if inv.root.file_id not in (None, ROOT_ID):
277
fileid1 = ' file_id="'
278
fileid2 = _encode_and_escape(inv.root.file_id)
136
append(_encode_and_escape(inv.root.file_id))
137
append(' format="5"')
282
138
if inv.revision_id is not None:
283
revid1 = ' revision_id="'
284
revid2 = _encode_and_escape(inv.revision_id)
288
append('<inventory%s%s format="5"%s%s>\n' % (
289
fileid1, fileid2, revid1, revid2))
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
291
179
def _pack_revision(self, rev):
292
180
"""Revision object -> xml tree"""
293
# For the XML format, we need to write them as Unicode rather than as
294
# utf-8 strings. So that cElementTree can handle properly escaping
296
decode_utf8 = cache_utf8.decode
297
revision_id = rev.revision_id
298
if isinstance(revision_id, str):
299
revision_id = decode_utf8(revision_id)
300
181
root = Element('revision',
301
182
committer = rev.committer,
302
183
timestamp = '%.3f' % rev.timestamp,
303
revision_id = revision_id,
184
revision_id = rev.revision_id,
304
185
inventory_sha1 = rev.inventory_sha1,
336
214
prop_elt.tail = '\n'
337
215
top_elt.tail = '\n'
339
def _unpack_inventory(self, elt, revision_id):
217
def _unpack_inventory(self, elt):
340
218
"""Construct from XML Element
342
220
assert elt.tag == 'inventory'
343
221
root_id = elt.get('file_id') or ROOT_ID
344
root_id = _get_utf8_or_ascii(root_id)
346
222
format = elt.get('format')
347
223
if format is not None:
348
224
if format != '5':
349
225
raise BzrError("invalid format version %r on inventory"
351
data_revision_id = elt.get('revision_id')
352
if data_revision_id is not None:
353
revision_id = cache_utf8.encode(data_revision_id)
227
revision_id = elt.get('revision_id')
228
if revision_id is not None:
229
revision_id = cache_utf8.get_cached_unicode(revision_id)
354
230
inv = Inventory(root_id, revision_id=revision_id)
356
232
ie = self._unpack_entry(e)
357
if ie.parent_id is None:
233
if ie.parent_id == ROOT_ID:
358
234
ie.parent_id = root_id
360
if revision_id is not None:
361
inv.root.revision = revision_id
364
def _unpack_entry(self, elt):
238
def _unpack_entry(self, elt, none_parents=False):
366
240
if not InventoryEntry.versionable_kind(kind):
367
241
raise AssertionError('unsupported entry kind %s' % kind)
369
get_cached = _get_utf8_or_ascii
243
get_cached = cache_utf8.get_cached_unicode
371
245
parent_id = elt.get('parent_id')
372
if parent_id is not None:
373
parent_id = get_cached(parent_id)
374
file_id = get_cached(elt.get('file_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')
376
258
if kind == 'directory':
377
259
ie = inventory.InventoryDirectory(file_id,
added
removed
bzrlib/xml5.py
