annotate roundup/backends/indexer_xapian.py @ 5470:e2baa4e6ed6d

handle words starting with unicode characters
author Christof Meerwald <cmeerw@cmeerw.org>
date Sat, 28 Jul 2018 22:39:24 +0100
parents 93832cec4c31
children e72573996caf
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
1 ''' This implements the full-text indexer using the Xapian indexer.
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
2 '''
5142
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
3 import re, os, time
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
4
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
5 import xapian
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
6
3544
5cd1c83dea50 Features and fixes.
Richard Jones <richard@users.sourceforge.net>
parents: 3295
diff changeset
7 from roundup.backends.indexer_common import Indexer as IndexerBase
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
8
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
9 # TODO: we need to delete documents when a property is *reindexed*
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
10
3544
5cd1c83dea50 Features and fixes.
Richard Jones <richard@users.sourceforge.net>
parents: 3295
diff changeset
11 class Indexer(IndexerBase):
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
12 def __init__(self, db):
3544
5cd1c83dea50 Features and fixes.
Richard Jones <richard@users.sourceforge.net>
parents: 3295
diff changeset
13 IndexerBase.__init__(self, db)
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
14 self.db_path = db.config.DATABASE
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
15 self.reindex = 0
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
16 self.transaction_active = False
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
17
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
18 def _get_database(self):
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
19 index = os.path.join(self.db_path, 'text-index')
5142
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
20 for n in range(10):
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
21 try:
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
22 # if successful return
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
23 return xapian.WritableDatabase(index, xapian.DB_CREATE_OR_OPEN)
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
24 except xapian.DatabaseLockError:
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
25 # adaptive sleep. Get longer as count increases.
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
26 time_to_sleep = 0.01 * (2 << min(5, n))
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
27 time.sleep(time_to_sleep)
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
28 # we are back to the for loop
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
29
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
30 # Get here only if we dropped out of the for loop.
93832cec4c31 issue2550839: Xapian, DatabaseLockError: Unable to get write lock on
John Rouillard <rouilj@ieee.org>
parents: 5108
diff changeset
31 raise xapian.DatabaseLockError("Unable to get lock after 10 retries on %s."%index)
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
32
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
33 def save_index(self):
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
34 '''Save the changes to the index.'''
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
35 if not self.transaction_active:
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
36 return
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
37 database = self._get_database()
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
38 database.commit_transaction()
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
39 self.transaction_active = False
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
40
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
41 def close(self):
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
42 '''close the indexing database'''
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
43 pass
3887
c7363442cdbb change xapian stemmer to use "new" API
Justus Pendleton <jpend@users.sourceforge.net>
parents: 3555
diff changeset
44
3555
91c495476db3 pre-release stuff and test fix
Richard Jones <richard@users.sourceforge.net>
parents: 3547
diff changeset
45 def rollback(self):
91c495476db3 pre-release stuff and test fix
Richard Jones <richard@users.sourceforge.net>
parents: 3547
diff changeset
46 if not self.transaction_active:
91c495476db3 pre-release stuff and test fix
Richard Jones <richard@users.sourceforge.net>
parents: 3547
diff changeset
47 return
91c495476db3 pre-release stuff and test fix
Richard Jones <richard@users.sourceforge.net>
parents: 3547
diff changeset
48 database = self._get_database()
91c495476db3 pre-release stuff and test fix
Richard Jones <richard@users.sourceforge.net>
parents: 3547
diff changeset
49 database.cancel_transaction()
91c495476db3 pre-release stuff and test fix
Richard Jones <richard@users.sourceforge.net>
parents: 3547
diff changeset
50 self.transaction_active = False
91c495476db3 pre-release stuff and test fix
Richard Jones <richard@users.sourceforge.net>
parents: 3547
diff changeset
51
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
52 def force_reindex(self):
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
53 '''Force a reindexing of the database. This essentially
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
54 empties the tables ids and index and sets a flag so
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
55 that the databases are reindexed'''
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
56 self.reindex = 1
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
57
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
58 def should_reindex(self):
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
59 '''returns True if the indexes need to be rebuilt'''
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
60 return self.reindex
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
61
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
62 def add_text(self, identifier, text, mime_type='text/plain'):
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
63 ''' "identifier" is (classname, itemid, property) '''
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
64 if mime_type != 'text/plain':
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
65 return
3547
7728ee93efd2 fix reindexing in Xapian
Richard Jones <richard@users.sourceforge.net>
parents: 3544
diff changeset
66 if not text: text = ''
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
67
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
68 # open the database and start a transaction if needed
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
69 database = self._get_database()
4378
477f2a47cbca - Indexer Xapian, made Xapian 1.2 compatible.
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 4252
diff changeset
70
477f2a47cbca - Indexer Xapian, made Xapian 1.2 compatible.
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 4252
diff changeset
71 # XXX: Xapian now supports transactions,
477f2a47cbca - Indexer Xapian, made Xapian 1.2 compatible.
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 4252
diff changeset
72 # but there is a call to save_index() missing.
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
73 #if not self.transaction_active:
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
74 #database.begin_transaction()
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
75 #self.transaction_active = True
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
76
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
77 # TODO: allow configuration of other languages
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
78 stemmer = xapian.Stem("english")
3547
7728ee93efd2 fix reindexing in Xapian
Richard Jones <richard@users.sourceforge.net>
parents: 3544
diff changeset
79
7728ee93efd2 fix reindexing in Xapian
Richard Jones <richard@users.sourceforge.net>
parents: 3544
diff changeset
80 # We use the identifier twice: once in the actual "text" being
7728ee93efd2 fix reindexing in Xapian
Richard Jones <richard@users.sourceforge.net>
parents: 3544
diff changeset
81 # indexed so we can search on it, and again as the "data" being
7728ee93efd2 fix reindexing in Xapian
Richard Jones <richard@users.sourceforge.net>
parents: 3544
diff changeset
82 # indexed so we know what we're matching when we get results
7728ee93efd2 fix reindexing in Xapian
Richard Jones <richard@users.sourceforge.net>
parents: 3544
diff changeset
83 identifier = '%s:%s:%s'%identifier
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
84
3547
7728ee93efd2 fix reindexing in Xapian
Richard Jones <richard@users.sourceforge.net>
parents: 3544
diff changeset
85 # create the new document
7728ee93efd2 fix reindexing in Xapian
Richard Jones <richard@users.sourceforge.net>
parents: 3544
diff changeset
86 doc = xapian.Document()
7728ee93efd2 fix reindexing in Xapian
Richard Jones <richard@users.sourceforge.net>
parents: 3544
diff changeset
87 doc.set_data(identifier)
4511
931370d96c34 Xapian indexing improved:
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 4470
diff changeset
88 doc.add_term(identifier, 0)
3547
7728ee93efd2 fix reindexing in Xapian
Richard Jones <richard@users.sourceforge.net>
parents: 3544
diff changeset
89
4252
2ff6f39aa391 Indexers behaviour made more consistent regarding length of indexed words...
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 3932
diff changeset
90 for match in re.finditer(r'\b\w{%d,%d}\b'
2ff6f39aa391 Indexers behaviour made more consistent regarding length of indexed words...
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 3932
diff changeset
91 % (self.minlength, self.maxlength),
2ff6f39aa391 Indexers behaviour made more consistent regarding length of indexed words...
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 3932
diff changeset
92 text.upper()):
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
93 word = match.group(0)
3544
5cd1c83dea50 Features and fixes.
Richard Jones <richard@users.sourceforge.net>
parents: 3295
diff changeset
94 if self.is_stopword(word):
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
95 continue
5108
67fad01d2009 issue2550653: xapian search, stemming is not working
John Rouillard <rouilj@ieee.org>
parents: 4841
diff changeset
96 term = stemmer(word.lower())
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
97 doc.add_posting(term, match.start(0))
4511
931370d96c34 Xapian indexing improved:
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 4470
diff changeset
98
931370d96c34 Xapian indexing improved:
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 4470
diff changeset
99 database.replace_document(identifier, doc)
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
100
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
101 def find(self, wordlist):
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
102 '''look up all the words in the wordlist.
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
103 If none are found return an empty dictionary
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
104 * more rules here
3887
c7363442cdbb change xapian stemmer to use "new" API
Justus Pendleton <jpend@users.sourceforge.net>
parents: 3555
diff changeset
105 '''
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
106 if not wordlist:
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
107 return {}
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
108
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
109 database = self._get_database()
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
110
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
111 enquire = xapian.Enquire(database)
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
112 stemmer = xapian.Stem("english")
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
113 terms = []
4252
2ff6f39aa391 Indexers behaviour made more consistent regarding length of indexed words...
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 3932
diff changeset
114 for term in [word.upper() for word in wordlist
2ff6f39aa391 Indexers behaviour made more consistent regarding length of indexed words...
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 3932
diff changeset
115 if self.minlength <= len(word) <= self.maxlength]:
2ff6f39aa391 Indexers behaviour made more consistent regarding length of indexed words...
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 3932
diff changeset
116 if not self.is_stopword(term):
5108
67fad01d2009 issue2550653: xapian search, stemming is not working
John Rouillard <rouilj@ieee.org>
parents: 4841
diff changeset
117 terms.append(stemmer(term.lower()))
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
118 query = xapian.Query(xapian.Query.OP_AND, terms)
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
119
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
120 enquire.set_query(query)
4841
3ff1a288fb9c issue2550583, issue2550635 Do not limit results with Xapian indexer
Thomas Arendsen Hein <thomas@intevation.de>
parents: 4570
diff changeset
121 matches = enquire.get_mset(0, database.get_doccount())
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
122
4470
21a95ba01a42 Fix search for xapian 1.2 issue2550676.
Bernhard Reiter <Bernhard.Reiter@intevation.de>
parents: 4378
diff changeset
123 return [tuple(m.document.get_data().split(':'))
3295
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
124 for m in matches]
a615cc230160 added Xapian indexer; replaces standard indexers if Xapian is available
Richard Jones <richard@users.sourceforge.net>
parents:
diff changeset
125

Roundup Issue Tracker: http://roundup-tracker.org/