38a7f2ee9d87cbd6075dcfa52522eb281a0a5442
[auf_roundup.git] / build / lib / roundup / backends / indexer_xapian.py
1 #$Id: indexer_xapian.py,v 1.6 2007-10-25 07:02:42 richard Exp $
2 ''' This implements the full-text indexer using the Xapian indexer.
3 '''
4 import re, os
5
6 import xapian
7
8 from roundup.backends.indexer_common import Indexer as IndexerBase
9
10 # TODO: we need to delete documents when a property is *reindexed*
11
12 class Indexer(IndexerBase):
13 def __init__(self, db):
14 IndexerBase.__init__(self, db)
15 self.db_path = db.config.DATABASE
16 self.reindex = 0
17 self.transaction_active = False
18
19 def _get_database(self):
20 index = os.path.join(self.db_path, 'text-index')
21 return xapian.WritableDatabase(index, xapian.DB_CREATE_OR_OPEN)
22
23 def save_index(self):
24 '''Save the changes to the index.'''
25 if not self.transaction_active:
26 return
27 # XXX: Xapian databases don't actually implement transactions yet
28 database = self._get_database()
29 database.commit_transaction()
30 self.transaction_active = False
31
32 def close(self):
33 '''close the indexing database'''
34 pass
35
36 def rollback(self):
37 if not self.transaction_active:
38 return
39 # XXX: Xapian databases don't actually implement transactions yet
40 database = self._get_database()
41 database.cancel_transaction()
42 self.transaction_active = False
43
44 def force_reindex(self):
45 '''Force a reindexing of the database. This essentially
46 empties the tables ids and index and sets a flag so
47 that the databases are reindexed'''
48 self.reindex = 1
49
50 def should_reindex(self):
51 '''returns True if the indexes need to be rebuilt'''
52 return self.reindex
53
54 def add_text(self, identifier, text, mime_type='text/plain'):
55 ''' "identifier" is (classname, itemid, property) '''
56 if mime_type != 'text/plain':
57 return
58 if not text: text = ''
59
60 # open the database and start a transaction if needed
61 database = self._get_database()
62 # XXX: Xapian databases don't actually implement transactions yet
63 #if not self.transaction_active:
64 #database.begin_transaction()
65 #self.transaction_active = True
66
67 # TODO: allow configuration of other languages
68 stemmer = xapian.Stem("english")
69
70 # We use the identifier twice: once in the actual "text" being
71 # indexed so we can search on it, and again as the "data" being
72 # indexed so we know what we're matching when we get results
73 identifier = '%s:%s:%s'%identifier
74
75 # see if the id is in the database
76 enquire = xapian.Enquire(database)
77 query = xapian.Query(xapian.Query.OP_AND, [identifier])
78 enquire.set_query(query)
79 matches = enquire.get_mset(0, 10)
80 if matches.size(): # would it killya to implement __len__()??
81 b = matches.begin()
82 docid = b.get_docid()
83 else:
84 docid = None
85
86 # create the new document
87 doc = xapian.Document()
88 doc.set_data(identifier)
89 doc.add_posting(identifier, 0)
90
91 for match in re.finditer(r'\b\w{%d,%d}\b'
92 % (self.minlength, self.maxlength),
93 text.upper()):
94 word = match.group(0)
95 if self.is_stopword(word):
96 continue
97 term = stemmer(word)
98 doc.add_posting(term, match.start(0))
99 if docid:
100 database.replace_document(docid, doc)
101 else:
102 database.add_document(doc)
103
104 def find(self, wordlist):
105 '''look up all the words in the wordlist.
106 If none are found return an empty dictionary
107 * more rules here
108 '''
109 if not wordlist:
110 return {}
111
112 database = self._get_database()
113
114 enquire = xapian.Enquire(database)
115 stemmer = xapian.Stem("english")
116 terms = []
117 for term in [word.upper() for word in wordlist
118 if self.minlength <= len(word) <= self.maxlength]:
119 if not self.is_stopword(term):
120 terms.append(stemmer(term))
121 query = xapian.Query(xapian.Query.OP_AND, terms)
122
123 enquire.set_query(query)
124 matches = enquire.get_mset(0, 10)
125
126 return [tuple(m[xapian.MSET_DOCUMENT].get_data().split(':'))
127 for m in matches]
128