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
|
#!/usr/bin/python3
import os
import os.path
import sqlite3
from argparse import ArgumentParser
import utils
from myconfig import MyConfig
from dirwalk import walkIndexFast
INSERT_NGRAM_DML = '''
INSERT INTO ngram(words, freq) VALUES(?, 1);
'''
UPDATE_NGRAM_DML = '''
UPDATE ngram SET freq = freq + 1 WHERE words = ?;
'''
PRUNE_NGRAM_DML = '''
DELETE FROM ngram WHERE freq <= ?;
'''
config = MyConfig()
#maximum combine number
N = config.getMaximumCombineNumber()
#change cwd to the word recognizer directory
words_dir = config.getWordRecognizerDir()
os.chdir(words_dir)
#chdir done
def handleOneDocument(infile, cur, length):
print(infile, length)
infilestatuspath = infile + config.getStatusPostfix()
infilestatus = utils.load_status(infilestatuspath)
if not utils.check_epoch(infilestatus, 'Segment'):
raise utils.EpochError('Please segment first.\n')
if utils.check_epoch(infilestatus, 'Populate'):
return False
sep = config.getWordSep()
#train
docfile = open(infile + config.getSegmentPostfix(), 'r')
words = []
for oneline in docfile.readlines():
oneline = oneline.rstrip(os.linesep)
if len(oneline) == 0:
continue
(token, word) = oneline.split(" ", 1)
token = int(token)
if 0 == token:
words = []
else:
words.append(word)
if len(words) < length:
continue
if len(words) > length:
words.pop(0)
assert len(words) == length
#do sqlite training
words_str = sep + sep.join(words) + sep
#print(words_str)
rowcount = cur.execute(UPDATE_NGRAM_DML, (words_str,)).rowcount
#print(rowcount)
assert rowcount <= 1
if 0 == rowcount:
cur.execute(INSERT_NGRAM_DML, (words_str,))
docfile.close()
#sign epoch only after last pass
if N == length:
utils.sign_epoch(infilestatus, 'Populate')
utils.store_status(infilestatuspath, infilestatus)
return True
def handleOnePass(indexpath, workdir, length):
print(indexpath, workdir, length)
filename = config.getNgramFileName(length)
filepath = workdir + os.sep + filename
conn = sqlite3.connect(filepath)
cur = conn.cursor()
#begin processing
indexfile = open(indexpath, 'r')
for oneline in indexfile.readlines():
oneline = oneline.rstrip(os.linesep)
(title, textpath) = oneline.split('#')
infile = config.getTextDir() + textpath
infilesize = utils.get_file_length(infile + config.getSegmentPostfix())
if infilesize < config.getMinimumFileSize():
print("Skipping " + title + '#' + textpath)
continue
#process one document
handleOneDocument(infile, cur, length)
indexfile.close()
conn.commit()
if conn:
conn.close()
def pruneNgramTable(indexpath, workdir, length):
print(indexpath, workdir, length, 'prune')
threshold = config.getPruneMinimumOccurrence()
filename = config.getNgramFileName(length)
filepath = workdir + os.sep + filename
conn = sqlite3.connect(filepath)
cur = conn.cursor()
rowcount = cur.execute(PRUNE_NGRAM_DML, (threshold, )).rowcount
#print(rowcount)
conn.commit()
if conn:
conn.close()
def handleOneIndex(indexpath, subdir, indexname, fast):
print(indexpath, subdir, indexname)
indexstatuspath = indexpath + config.getStatusPostfix()
indexstatus = utils.load_status(indexstatuspath)
if not utils.check_epoch(indexstatus, 'Prepare'):
raise utils.EpochError('Please prepare first.\n')
if utils.check_epoch(indexstatus, 'Populate'):
return
workdir = config.getWordRecognizerDir() + os.sep + \
subdir + os.sep + indexname
print(workdir)
shmdir = config.getInMemoryFileSystem()
for i in range(1, N + 1):
if fast:
#copy file
filename = config.getNgramFileName(i)
filepath = workdir + os.sep + filename
shmfilepath = shmdir + os.sep + filename
utils.copyfile(filepath, shmfilepath)
handleOnePass(indexpath, shmdir, i)
pruneNgramTable(indexpath, shmdir, i)
utils.copyfile(shmfilepath, filepath)
os.unlink(shmfilepath)
else:
handleOnePass(indexpath, workdir, i)
pruneNgramTable(indexpath, workdir, i)
#sign epoch
utils.sign_epoch(indexstatus, 'Populate')
utils.store_status(indexstatuspath, indexstatus)
if __name__ == '__main__':
parser = ArgumentParser(description='Populate n-gram.')
parser.add_argument('--indexdir', action='store', \
help='index directory', \
default=config.getTextIndexDir())
parser.add_argument('--fast', action='store_const', \
help='Use /dev/shm to speed up populate', \
const=True, default=False)
args = parser.parse_args()
print(args)
walkIndexFast(handleOneIndex, args.indexdir, args.fast)
print('done')
|