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
24
18
from bzrlib.xml_serializer import SubElement, Element, Serializer
25
19
from bzrlib.inventory import ROOT_ID, Inventory, InventoryEntry
20
import bzrlib.inventory as inventory
26
21
from bzrlib.revision import Revision
27
22
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
25
class Serializer_v5(Serializer):
93
26
"""Version 5 serializer
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)
33
def _pack_inventory(self, inv):
34
"""Convert to XML Element"""
120
35
entries = inv.iter_entries()
122
root_path, root_ie = entries.next()
36
e = Element('inventory',
39
path, root = entries.next()
40
if root.file_id not in (None, ROOT_ID):
41
e.set('file_id', root.file_id)
42
if inv.revision_id is not None:
43
e.set('revision_id', inv.revision_id)
123
44
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
45
e.append(self._pack_entry(ie))
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."""
48
def _pack_entry(self, ie):
49
"""Convert InventoryEntry to XML element"""
145
50
# TODO: should just be a plain assertion
146
assert InventoryEntry.versionable_kind(ie.kind), \
147
'unsupported entry kind %s' % ie.kind
51
if not InventoryEntry.versionable_kind(ie.kind):
52
raise AssertionError('unsupported entry kind %s' % ie.kind)
54
e.set('name', ie.name)
55
e.set('file_id', ie.file_id)
57
if ie.text_size != None:
58
e.set('text_size', '%d' % ie.text_size)
60
for f in ['text_sha1', 'revision', 'symlink_target']:
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):
66
e.set('executable', 'yes')
68
# to be conservative, we don't externalize the root pointers
69
# for now, leaving them as null in the xml form. in a future
70
# version it will be implied by nested elements.
71
if ie.parent_id != ROOT_ID:
158
72
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
73
e.set('parent_id', ie.parent_id)
179
77
def _pack_revision(self, rev):
180
78
"""Revision object -> xml tree"""
238
def _unpack_entry(self, elt, none_parents=False):
137
def _unpack_entry(self, elt):
240
139
if not InventoryEntry.versionable_kind(kind):
241
140
raise AssertionError('unsupported entry kind %s' % kind)
243
get_cached = cache_utf8.get_cached_unicode
245
142
parent_id = elt.get('parent_id')
246
if parent_id is None and not none_parents:
143
if parent_id == None:
247
144
parent_id = ROOT_ID
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
146
if kind == 'directory':
259
ie = inventory.InventoryDirectory(file_id,
147
ie = inventory.InventoryDirectory(elt.get('file_id'),
262
150
elif kind == 'file':
263
ie = inventory.InventoryFile(file_id,
151
ie = inventory.InventoryFile(elt.get('file_id'),
266
154
ie.text_sha1 = elt.get('text_sha1')
290
176
if format != '5':
291
177
raise BzrError("invalid format version %r on inventory"
293
get_cached = cache_utf8.get_cached_unicode
294
179
rev = Revision(committer = elt.get('committer'),
295
180
timestamp = float(elt.get('timestamp')),
296
revision_id = get_cached(elt.get('revision_id')),
181
revision_id = elt.get('revision_id'),
297
182
inventory_sha1 = elt.get('inventory_sha1')
299
184
parents = elt.find('parents') or []
300
185
for p in parents:
301
186
assert p.tag == 'revision_ref', \
302
187
"bad parent node tag %r" % p.tag
303
rev.parent_ids.append(get_cached(p.get('revision_id')))
188
rev.parent_ids.append(p.get('revision_id'))
304
189
self._unpack_revision_properties(elt, rev)
305
190
v = elt.get('timezone')
309
rev.timezone = int(v)
191
rev.timezone = v and int(v)
310
192
rev.message = elt.findtext('message') # text of <message>
313
196
def _unpack_revision_properties(self, elt, rev):
314
197
"""Unpack properties onto a revision."""
315
198
props_elt = elt.find('properties')
added
removed
bzrlib/xml5.py
