« back to all changes in this revision
Viewing changes to bzrlib/xml5.py
- browse files at revision 1942
- compare with another revision
- compare with revision 1910.2.18
- download diff
- download tarball
- view history from revision 1942
- Committer: Canonical.com Patch Queue Manager
- Date: 2006-08-17 14:18:30 UTC
- mfrom: (1934.1.21 xml_writer)
- Revision ID: pqm@pqm.ubuntu.com-20060817141830-383cad75e9090732
(robertc,jam) Custom xml serializer saves a lot of time when serializing inventories
added
removed
bzrlib/xml5.py
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
16
16
17
import cStringIO
18
import re
17
19
18
20
from bzrlib import (
19
21
cache_utf8,
22
inventory,
20
23
)
21
24
from bzrlib.xml_serializer import SubElement, Element, Serializer
22
25
from bzrlib.inventory import ROOT_ID, Inventory, InventoryEntry
23
import bzrlib.inventory as inventory
24
26
from bzrlib.revision import Revision
25
27
from bzrlib.errors import BzrError
26
28
27
29
30
_utf8_re = None
31
_utf8_escape_map = {
32
"&":'&',
33
"'":"'", # FIXME: overkill
34
"\"":""",
35
"<":"<",
36
">":">",
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
45
_utf8_re = re.compile(u'[&<>\'\"\u0080-\uffff]')
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
"""
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.
61
try:
62
return _map[match.group()]
63
except KeyError:
64
return "&#%d;" % ord(match.group())
65
66
67
_unicode_to_escaped_map = {}
68
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)
74
if text is None:
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
81
# either way)
82
text = str(_utf8_re.sub(_utf8_escape_replace, unicode_str)) + '"'
83
_map[unicode_str] = text
84
return text
85
86
87
def _clear_cache():
88
"""Clean out the unicode => escaped map"""
89
_unicode_to_escaped_map.clear()
90
91
28
92
class Serializer_v5(Serializer):
29
93
"""Version 5 serializer
30
94
32
96
"""
33
97
34
98
__slots__ = []
35
36
def _pack_inventory(self, inv):
37
"""Convert to XML Element"""
99
100
def write_inventory_to_string(self, inv):
101
"""Just call write_inventory with a StringIO and return the value"""
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
"""
112
_ensure_utf8_re()
113
output = []
114
append = output.append
115
self._append_inventory_root(append, inv)
38
116
entries = inv.iter_entries()
39
e = Element('inventory',
40
format='5')
41
e.text = '\n'
42
path, root = entries.next()
43
if root.file_id not in (None, ROOT_ID):
44
e.set('file_id', root.file_id)
117
# Skip the root
118
root_path, root_ie = entries.next()
119
for path, ie in entries:
120
self._append_entry(append, ie)
121
append('</inventory>\n')
122
f.writelines(output)
123
# Just to keep the cache from growing without bounds
124
# but we may actually not want to do clear the cache
125
#_clear_cache()
126
127
def _append_inventory_root(self, append, inv):
128
"""Append the inventory root to output."""
129
append('<inventory')
130
if inv.root.file_id not in (None, ROOT_ID):
131
append(' file_id="')
132
append(_encode_and_escape(inv.root.file_id))
133
append(' format="5"')
45
134
if inv.revision_id is not None:
46
e.set('revision_id', inv.revision_id)
47
for path, ie in entries:
48
e.append(self._pack_entry(ie))
49
return e
50
51
def _pack_entry(self, ie):
52
"""Convert InventoryEntry to XML element"""
135
append(' revision_id="')
136
append(_encode_and_escape(inv.revision_id))
137
append('>\n')
138
139
def _append_entry(self, append, ie):
140
"""Convert InventoryEntry to XML element and append to output."""
53
141
# TODO: should just be a plain assertion
54
if not InventoryEntry.versionable_kind(ie.kind):
55
raise AssertionError('unsupported entry kind %s' % ie.kind)
56
e = Element(ie.kind)
57
e.set('name', ie.name)
58
e.set('file_id', ie.file_id)
59
60
if ie.text_size != None:
61
e.set('text_size', '%d' % ie.text_size)
62
63
for f in ['text_sha1', 'revision', 'symlink_target']:
64
v = getattr(ie, f)
65
if v != None:
66
e.set(f, v)
67
142
assert InventoryEntry.versionable_kind(ie.kind), \
143
'unsupported entry kind %s' % ie.kind
144
145
append("<")
146
append(ie.kind)
68
147
if ie.executable:
69
e.set('executable', 'yes')
70
71
# to be conservative, we don't externalize the root pointers
72
# for now, leaving them as null in the xml form. in a future
73
# version it will be implied by nested elements.
148
append(' executable="yes"')
149
append(' file_id="')
150
append(_encode_and_escape(ie.file_id))
151
append(' name="')
152
append(_encode_and_escape(ie.name))
74
153
if ie.parent_id != ROOT_ID:
75
154
assert isinstance(ie.parent_id, basestring)
76
e.set('parent_id', ie.parent_id)
77
e.tail = '\n'
78
return e
155
append(' parent_id="')
156
append(_encode_and_escape(ie.parent_id))
157
if ie.revision is not None:
158
append(' revision="')
159
append(_encode_and_escape(ie.revision))
160
if ie.symlink_target is not None:
161
append(' symlink_target="')
162
append(_encode_and_escape(ie.symlink_target))
163
if ie.text_sha1 is not None:
164
append(' text_sha1="')
165
append(ie.text_sha1)
166
append('"')
167
if ie.text_size is not None:
168
append(' text_size="%d"' % ie.text_size)
169
append(" />\n")
170
return
79
171
80
172
def _pack_revision(self, rev):
81
173
"""Revision object -> xml tree"""
104
196
self._pack_revision_properties(rev, root)
105
197
return root
106
198
107
108
199
def _pack_revision_properties(self, rev, under_element):
109
200
top_elt = SubElement(under_element, 'properties')
110
201
for prop_name, prop_value in sorted(rev.properties.items()):
116
207
prop_elt.tail = '\n'
117
208
top_elt.tail = '\n'
118
209
119
120
210
def _unpack_inventory(self, elt):
121
211
"""Construct from XML Element
122
212
"""
138
228
inv.add(ie)
139
229
return inv
140
230
141
142
231
def _unpack_entry(self, elt):
143
232
kind = elt.tag
144
233
if not InventoryEntry.versionable_kind(kind):
149
238
parent_id = elt.get('parent_id')
150
239
if parent_id == None:
151
240
parent_id = ROOT_ID
152
parent_id = get_cached(parent_id)
153
file_id = get_cached(elt.get('file_id'))
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.
247
#parent_id = get_cached(parent_id)
248
#file_id = get_cached(elt.get('file_id'))
249
file_id = elt.get('file_id')
154
250
155
251
if kind == 'directory':
156
252
ie = inventory.InventoryDirectory(file_id,
179
275
180
276
return ie
181
277
182
183
278
def _unpack_revision(self, elt):
184
279
"""XML Element -> Revision object"""
185
280
assert elt.tag == 'revision'
208
303
rev.message = elt.findtext('message') # text of <message>
209
304
return rev
210
305
211
212
306
def _unpack_revision_properties(self, elt, rev):
213
307
"""Unpack properties onto a revision."""
214
308
props_elt = elt.find('properties')
Loggerhead is a web-based interface for Breezy
Version: 2.0.1