1 ''' This implements the full-text indexer over two RDBMS tables. The first
2 is a mapping of words to occurance IDs. The second maps the IDs to (Class,
3 propname, itemid) instances.
4 '''
5 import re
7 from indexer_dbm import Indexer
9 class Indexer(Indexer):
10 disallows = {'THE':1, 'THIS':1, 'ZZZ':1, 'THAT':1, 'WITH':1}
11 def __init__(self, db):
12 self.db = db
13 self.reindex = 0
15 def close(self):
16 '''close the indexing database'''
17 # just nuke the circular reference
18 self.db = None
20 def force_reindex(self):
21 '''Force a reindexing of the database. This essentially
22 empties the tables ids and index and sets a flag so
23 that the databases are reindexed'''
24 self.reindex = 1
26 def should_reindex(self):
27 '''returns True if the indexes need to be rebuilt'''
28 return self.reindex
30 def add_text(self, identifier, text, mime_type='text/plain'):
31 ''' "identifier" is (classname, itemid, property) '''
32 if mime_type != 'text/plain':
33 return
35 # first, find the id of the (classname, itemid, property)
36 a = self.db.arg
37 sql = 'select _textid from __textids where _class=%s and '\
38 '_itemid=%s and _prop=%s'%(a, a, a)
39 self.db.cursor.execute(sql, identifier)
40 r = self.db.cursor.fetchone()
41 if not r:
42 id = self.db.newid('__textids')
43 sql = 'insert into __textids (_textid, _class, _itemid, _prop)'\
44 ' values (%s, %s, %s, %s)'%(a, a, a, a)
45 self.db.cursor.execute(sql, (id, ) + identifier)
46 self.db.cursor.execute('select max(_textid) from __textids')
47 id = self.db.cursor.fetchone()[0]
48 else:
49 id = int(r[0])
50 # clear out any existing indexed values
51 sql = 'delete from __words where _textid=%s'%a
52 self.db.cursor.execute(sql, (id, ))
54 # ok, find all the words in the text
55 wordlist = re.findall(r'\b\w{2,25}\b', str(text).upper())
56 words = {}
57 for word in wordlist:
58 if not self.disallows.has_key(word):
59 words[word] = 1
60 words = words.keys()
62 # for each word, add an entry in the db
63 for word in words:
64 # don't dupe
65 sql = 'select * from __words where _word=%s and _textid=%s'%(a, a)
66 self.db.cursor.execute(sql, (word, id))
67 if self.db.cursor.fetchall():
68 continue
69 sql = 'insert into __words (_word, _textid) values (%s, %s)'%(a, a)
70 self.db.cursor.execute(sql, (word, id))
72 def find(self, wordlist):
73 '''look up all the words in the wordlist.
74 If none are found return an empty dictionary
75 * more rules here
76 '''
77 l = [word.upper() for word in wordlist if 26 > len(word) > 2]
79 a = ','.join([self.db.arg] * len(l))
80 sql = 'select distinct(_textid) from __words where _word in (%s)'%a
81 self.db.cursor.execute(sql, tuple(l))
82 r = self.db.cursor.fetchall()
83 if not r:
84 return {}
85 a = ','.join([self.db.arg] * len(r))
86 sql = 'select _class, _itemid, _prop from __textids '\
87 'where _textid in (%s)'%a
88 self.db.cursor.execute(sql, tuple([int(id) for (id,) in r]))
89 # self.search_index has the results as {some id: identifier} ...
90 # sigh
91 r = {}
92 k = 0
93 for c,n,p in self.db.cursor.fetchall():
94 key = (str(c), str(n), str(p))
95 r[k] = key
96 k += 1
97 return r
99 def save_index(self):
100 # the normal RDBMS backend transaction mechanisms will handle this
101 pass
103 def rollback(self):
104 # the normal RDBMS backend transaction mechanisms will handle this
105 pass