From 5a950930667055e1757bd24b408a48ddbfc8e4c4 Mon Sep 17 00:00:00 2001 From: Sascha Silbe Date: Mon, 17 Aug 2009 19:41:04 +0200 Subject: [PATCH] IndexStore refactoring and prefix term support --- src/carquinyol/indexstore.py | 287 +++++++++++++++++++++++++---------------- 1 files changed, 175 insertions(+), 112 deletions(-) diff --git a/src/carquinyol/indexstore.py b/src/carquinyol/indexstore.py index 42c3132..7b620b5 100644 --- a/src/carquinyol/indexstore.py +++ b/src/carquinyol/indexstore.py @@ -16,10 +16,11 @@ import logging import os +import sys import gobject import xapian -from xapian import WritableDatabase, Document, Enquire, Query, QueryParser +from xapian import WritableDatabase, Document, Enquire, Query from carquinyol import layoutmanager from carquinyol.layoutmanager import MAX_QUERY_LIMIT @@ -28,6 +29,8 @@ _VALUE_UID = 0 _VALUE_TIMESTAMP = 1 _VALUE_TITLE = 2 +_PREFIX_NONE = 'N' +_PREFIX_FULL_VALUE = 'F' _PREFIX_UID = 'Q' _PREFIX_ACTIVITY = 'A' _PREFIX_ACTIVITY_ID = 'I' @@ -40,10 +43,164 @@ _FLUSH_THRESHOLD = 20 # Force a flush after _n_ seconds since the last change to the db _FLUSH_TIMEOUT = 60 -_PROPERTIES_NOT_TO_INDEX = ['timestamp', 'activity_id', 'keep', 'preview'] +_PROPERTIES_NOT_TO_INDEX = ['timestamp', 'preview'] _MAX_RESULTS = int(2 ** 31 - 1) +_QUERY_TERM_MAP = { + 'uid': _PREFIX_UID, + 'activity': _PREFIX_ACTIVITY, + 'activity_id': _PREFIX_ACTIVITY_ID, + 'mime_type': _PREFIX_MIME_TYPE, + 'keep': _PREFIX_KEEP, +} + +_QUERY_VALUE_MAP = { + 'timestamp': _VALUE_TIMESTAMP, +} + + +class TermGenerator (xapian.TermGenerator): + + def index_document(self, document, properties): + document.add_value(_VALUE_TIMESTAMP, str(properties['timestamp'])) + document.add_value(_VALUE_TITLE, properties.get('title', '').strip()) + + xapian.TermGenerator.set_document(self, document) + + properties = dict(properties) + self._index_known(document, properties) + self._index_unknown(document, properties) + + def _index_known(self, document, properties): + for name, prefix in _QUERY_TERM_MAP.items(): + if (name not in properties): + continue + + self._index_property(document, name, properties.pop(name), prefix) + + def _index_unknown(self, document, properties): + for name, value in properties.items(): + self._index_property(document, name, value) + + def _index_property(self, doc, name, value, prefix=''): + if name in _PROPERTIES_NOT_TO_INDEX or not value: + return + + if isinstance(value, unicode): + value = value.encode('utf-8') + elif not isinstance(value, basestring): + value = str(value) + + # We need to add the full value (i.e. not split into words) so + # we can enumerate unique values. It also simplifies setting up + # dictionary-based queries. + if prefix: + doc.add_term(_PREFIX_FULL_VALUE + prefix + value) + + self.index_text(value, 1, prefix or _PREFIX_NONE) + self.increase_termpos() + + +class QueryParser (xapian.QueryParser): + """QueryParser that understands dictionaries and Xapian query strings. + + The dictionary contains metadata names as keys and either basic types + (exact match), 2-tuples (range, only valid for value-stored metadata) + or a list (multiple exact matches joined with OR) as values. + An empty dictionary matches everything. Queries from different keys + (i.e. different metadata names) are joined with AND. + """ + + def __init__(self): + xapian.QueryParser.__init__(self) + + for name, prefix in _QUERY_TERM_MAP.items(): + self.add_prefix(name, prefix) + self.add_prefix('', prefix) + + self.add_prefix('', _PREFIX_NONE) + + def _parse_query_term(self, name, prefix, value): + if isinstance(value, list): + subqueries = [self._parse_query_term(name, prefix, word) + for word in value] + return Query(Query.OP_OR, subqueries) + + elif prefix: + return Query(_PREFIX_FULL_VALUE + prefix + str(value)) + else: + return Query(_PREFIX_NONE + str(value)) + + def _parse_query_value_range(self, name, value, value_no): + if len(value) != 2: + raise TypeError( + 'Only tuples of size 2 have a defined meaning. ' + 'Did you mean to pass a list instead?') + + start, end = value + return Query(Query.OP_VALUE_RANGE, value_no, str(start), str(end)) + + def _parse_query_value(self, name, value_no, value): + if isinstance(value, list): + subqueries = [self._parse_query_value(name, value_no, word) + for word in value] + return Query(Query.OP_OR, subqueries) + + elif isinstance(value, tuple): + return self._parse_query_value_range(name, value, value_no) + + elif isinstance(value, dict): + # compatibility option for timestamp: {'start': 0, 'end': 1} + start = value.get('start', 0) + end = value.get('end', sys.maxint) + return self._parse_query_value_range(name, (start, end), value_no) + + else: + return Query(Query.OP_VALUE_RANGE, + _QUERY_VALUE_MAP[name], str(value), str(value)) + + def _parse_query_xapian(self, query_str): + try: + return xapian.QueryParser.parse_query( + self, query_str, + QueryParser.FLAG_PHRASE | + QueryParser.FLAG_BOOLEAN | + QueryParser.FLAG_LOVEHATE | + QueryParser.FLAG_WILDCARD, + '') + + except xapian.QueryParserError, exception: + logging.warning('Invalid query string: '+exception.get_msg()) + return Query() + + def parse_query(self, query_dict, query_string): + logging.debug('parse_query %r %r', query_dict, query_string) + queries = [] + query_dict = dict(query_dict) + + if query_string: + queries.append(self._parse_query_xapian(str(query_string))) + + for name, value in query_dict.items(): + if name in _QUERY_TERM_MAP: + queries.append(self._parse_query_term(name, + _QUERY_TERM_MAP[name], value)) + elif name in _QUERY_VALUE_MAP: + queries.append(self._parse_query_value(name, + _QUERY_VALUE_MAP[name], value)) + else: + logging.warning('Unknown term: %r=%r', name, value) + + if not queries: + queries.append(Query('')) + + if query_dict: + logging.warning('Unknown term(s): %r', query_dict) + + logging.debug('queries: %r', [str(q) for q in queries]) + return Query(Query.OP_AND, queries) + class IndexStore(object): """Index metadata and provide rich query facilities on it. @@ -70,7 +227,8 @@ class IndexStore(object): os.remove(os.path.join(index_path, f)) def contains(self, uid): - postings = self._database.postlist(_PREFIX_UID + uid) + postings = self._database.postlist(_PREFIX_FULL_VALUE + \ + _PREFIX_UID + uid) try: postlist_item = postings.next() except StopIteration: @@ -79,60 +237,28 @@ class IndexStore(object): def store(self, uid, properties): document = Document() - document.add_term(_PREFIX_UID + uid) - document.add_term(_PREFIX_ACTIVITY + properties.get('activity', '')) - document.add_term(_PREFIX_MIME_TYPE + properties.get('mime_type', '')) - document.add_term(_PREFIX_ACTIVITY_ID + - properties.get('activity_id', '')) - document.add_term(_PREFIX_KEEP + str(properties.get('keep', 0))) - document.add_value(_VALUE_UID, uid) - document.add_value(_VALUE_TIMESTAMP, str(properties['timestamp'])) - document.add_value(_VALUE_TITLE, properties.get('title', '').strip()) - - term_generator = xapian.TermGenerator() - - # TODO: we should do stemming, but in which language? - #if language is not None: - # term_generator.set_stemmer(_xapian.Stem(language)) - - # TODO: we should use a stopper - #if stop is not None: - # stopper = _xapian.SimpleStopper() - # for term in stop: - # stopper.add (term) - # term_generator.set_stopper (stopper) - - term_generator.set_document(document) - term_generator.index_text_without_positions( - self._extract_text(properties), 1, '') + term_generator = TermGenerator() + term_generator.index_document(document, properties) if not self.contains(uid): self._database.add_document(document) else: - self._database.replace_document(_PREFIX_UID + uid, document) - self._flush() + self._database.replace_document(_PREFIX_FULL_VALUE + \ + _PREFIX_UID + uid, document) - def _extract_text(self, properties): - text = '' - for key, value in properties.items(): - if key not in _PROPERTIES_NOT_TO_INDEX: - if text: - text += ' ' - if isinstance(value, unicode): - value = value.encode('utf-8') - elif not isinstance(value, basestring): - value = str(value) - text += value - return text + self._flush() def find(self, query): offset = query.pop('offset', 0) limit = query.pop('limit', MAX_QUERY_LIMIT) order_by = query.pop('order_by', []) + query_string = query.pop('query', None) + query_parser = QueryParser() + query_parser.set_database(self._database) enquire = Enquire(self._database) - enquire.set_query(self._parse_query(query)) + enquire.set_query(query_parser.parse_query(query, query_string)) # This will assure that the results count is exact. check_at_least = offset + limit + 1 @@ -151,7 +277,7 @@ class IndexStore(object): elif order_by == '-title': enquire.set_sort_by_value(_VALUE_TITLE, False) else: - logging.warning('Unsupported property for sorting: %s' % order_by) + logging.warning('Unsupported property for sorting: %s', order_by) query_result = enquire.get_mset(offset, limit, check_at_least) total_count = query_result.get_matches_estimated() @@ -162,77 +288,14 @@ class IndexStore(object): return (uids, total_count) - def _parse_query(self, query_dict): - logging.debug('_parse_query %r' % query_dict) - queries = [] - - query_str = query_dict.pop('query', None) - if query_str is not None: - query_parser = QueryParser() - query_parser.set_database(self._database) - #query_parser.set_default_op(Query.OP_AND) - - # TODO: we should do stemming, but in which language? - #query_parser.set_stemmer(_xapian.Stem(lang)) - #query_parser.set_stemming_strategy(qp.STEM_SOME) - - query = query_parser.parse_query( - query_str, - QueryParser.FLAG_PHRASE | - QueryParser.FLAG_BOOLEAN | - QueryParser.FLAG_LOVEHATE | - QueryParser.FLAG_WILDCARD, - '') - - queries.append(query) - - timestamp = query_dict.pop('timestamp', None) - if timestamp is not None: - start = str(timestamp.pop('start', 0)) - end = str(timestamp.pop('end', _MAX_RESULTS)) - query = Query(Query.OP_VALUE_RANGE, _VALUE_TIMESTAMP, start, end) - queries.append(query) - - uid = query_dict.pop('uid', None) - if uid is not None: - queries.append(Query(_PREFIX_UID + uid)) - - activity = query_dict.pop('activity', None) - if activity is not None: - queries.append(Query(_PREFIX_ACTIVITY + activity)) - - activity_id = query_dict.pop('activity_id', None) - if activity_id is not None: - query = Query(_PREFIX_ACTIVITY_ID + activity_id) - queries.append(query) - - keep = query_dict.pop('keep', None) - if keep is not None: - query = Query(_PREFIX_KEEP + str(keep)) - queries.append(query) - - mime_type = query_dict.pop('mime_type', None) - if mime_type is not None: - mime_queries = [] - for mime_type in mime_type: - mime_queries.append(Query(_PREFIX_MIME_TYPE + mime_type)) - queries.append(Query(Query.OP_OR, mime_queries)) - - if not queries: - queries.append(Query('')) - - if query_dict: - logging.warning('Unknown term(s): %r' % query_dict) - - return Query(Query.OP_AND, queries) - def delete(self, uid): - self._database.delete_document(_PREFIX_UID + uid) + self._database.delete_document(_PREFIX_FULL_VALUE + _PREFIX_UID + uid) def get_activities(self): activities = [] - for term in self._database.allterms(_PREFIX_ACTIVITY): - activities.append(term.term[len(_PREFIX_ACTIVITY):]) + prefix = _PREFIX_FULL_VALUE + _PREFIX_ACTIVITY + for term in self._database.allterms(prefix): + activities.append(term.term[len(prefix):]) return activities def _flush_timeout_cb(self): -- 1.6.3.3