1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
|
#
# Copyright (C) 2006 Robey Pointer <robey@lag.net>
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
#
"""
indexing of the comment text of revisions, for fast searching.
two separate database files are created:
- recorded: revid -> 1 (if the revid is indexed)
- index: 3-letter substring -> list(revids)
"""
import os
import re
import time
from loggerhead import util
from loggerhead.lockfile import LockFile
from loggerhead.changecache import FakeShelf
# if any substring index reaches this many revids, replace the entry with
# an ALL marker -- it's not worth an explicit index.
ALL_THRESHOLD = 1000
ALL = 'ALL'
with_lock = util.with_lock('_lock')
def normalize_string(s):
"""
remove any punctuation and normalize all whitespace to a single space.
"""
s = util.to_utf8(s).lower()
# remove apostrophes completely.
s = re.sub(r"'", '', s)
# convert other garbage into space
s = re.sub(r'[^\w\d]', ' ', s)
# compress multiple spaces into one.
s = re.sub(r'\s{2,}', ' ', s)
# and finally remove leading/trailing whitespace
s = s.strip()
return s
class TextIndex (object):
def __init__(self, history, cache_path):
self.history = history
self.log = history.log
if not os.path.exists(cache_path):
os.mkdir(cache_path)
self._recorded_filename = os.path.join(cache_path, 'textindex-recorded.sql')
self._index_filename = os.path.join(cache_path, 'textindex.sql')
# use a lockfile since the cache folder could be shared across different processes.
self._lock = LockFile(os.path.join(cache_path, 'index-lock'))
self._closed = False
self.log.info('Using search index; %d entries.', len(self))
def _index(self):
return FakeShelf(self._index_filename)
def _recorded(self):
return FakeShelf(self._recorded_filename)
def _is_indexed(self, revid, recorded):
return recorded.get(util.to_utf8(revid)) is not None
@with_lock
def is_indexed(self, revid):
recorded = self._recorded()
try:
return self._is_indexed(revid, recorded)
finally:
recorded.close()
@with_lock
def __len__(self):
recorded = self._recorded()
try:
return recorded.count()
finally:
recorded.close()
@with_lock
def close(self):
self._closed = True
@with_lock
def closed(self):
return self._closed
@with_lock
def flush(self):
pass
@with_lock
def full(self):
recorded = self._recorded()
last_revid = util.to_utf8(self.history.last_revid)
try:
return (recorded.count() >= len(self.history.get_revision_history())
and recorded.get(last_revid) is not None)
finally:
recorded.close()
def _index_change(self, change, recorded, index):
"""
currently, only indexes the 'comment' field.
"""
comment = normalize_string(change.comment)
if len(comment) < 3:
return
for i in xrange(len(comment) - 2):
sub = comment[i:i + 3]
orig = revid_set = index.get(sub)
if revid_set is None:
revid_set = set()
elif revid_set == ALL:
# this entry got too big
continue
revid_set.add(change.revid)
if len(revid_set) > ALL_THRESHOLD:
revid_set = ALL
if orig is not None:
index.update([(sub, revid_set)], commit=False)
else:
index.add([(sub, revid_set)], commit=False)
recorded.add([(util.to_utf8(change.revid), True)], commit=False)
@with_lock
def index_changes(self, revid_list):
recorded = self._recorded()
index = self._index()
try:
revid_list = [r for r in revid_list if not self._is_indexed(r, recorded)]
change_list = self.history.get_changes(revid_list)
for change in change_list:
self._index_change(change, recorded, index)
finally:
index.close(commit=True)
recorded.close(commit=True)
@with_lock
def find(self, text, revid_list=None):
index = self._index()
try:
text = normalize_string(text)
if len(text) < 3:
return []
total_set = None
if revid_list is not None:
total_set = set(revid_list)
seen_all = False
for i in xrange(len(text) - 2):
sub = text[i:i + 3]
revid_set = index.get(sub)
if revid_set is None:
# zero matches, stop here.
return []
if revid_set == ALL:
# skip
seen_all = True
continue
if total_set is None:
total_set = revid_set
else:
total_set.intersection_update(revid_set)
if len(total_set) == 0:
return []
finally:
index.close()
# tricky: if seen_all is True, one of the substring indices was ALL
# (in other words, unindexed), so our results are actually a superset
# of the exact answer.
#
# if we cared, we could do a direct match on the result set and cull
# out any that aren't actually matches. for now, i'm gonna say that
# we DON'T care, and if one of the substrings hit ALL, there's a small
# chance that we'll give a few false positives.
return total_set
def check_rebuild(self, max_time=3600):
"""
check if there are any un-indexed revisions, and if so, index them.
but don't spend longer than C{max_time} on it.
"""
if self.closed() or self.full():
# all done
return
self.log.info('Building search index...')
work = list(self.history.get_revision_history())
start_time = time.time()
last_update = time.time()
count = 0
jump = 100
for i in xrange(0, len(work), jump):
r = work[i:i + jump]
self.index_changes(r)
if self.closed():
return
count += jump
now = time.time()
if now - start_time > 3600:
# there's no point working for hours. eventually we might even
# hit the next re-index interval, which would suck mightily.
self.log.info('Search indexing has worked for an hour; giving up for now.')
return
if now - last_update > 60:
self.log.info('Search indexing continues: %d/%d' % (min(count, len(work)), len(work)))
last_update = time.time()
# give someone else a chance at the lock
time.sleep(1)
self.log.info('Search index completed.')
self.flush()
|