X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/480da6ff4ed8c04b205710c318b2fcfcb961bbc7..4a4c7cd36d9f6899b37b6d749d4af626cf133e21:/apps/search/index.py?ds=inline diff --git a/apps/search/index.py b/apps/search/index.py index e7f28c9da..85bcb616d 100644 --- a/apps/search/index.py +++ b/apps/search/index.py @@ -15,12 +15,14 @@ import traceback import logging log = logging.getLogger('search') import sunburnt -import highlight +import custom +import operator +log = logging.getLogger('search') class SolrIndex(object): def __init__(self, mode=None): - self.index = highlight.HLSolrInterface(settings.SOLR, mode=mode) + self.index = custom.CustomSolrInterface(settings.SOLR, mode=mode) class Snippets(object): @@ -91,7 +93,8 @@ class Snippets(object): def close(self): """Close snippet file""" - self.file.close() + if self.file: + self.file.close() def remove(self): self.revision = None @@ -110,7 +113,7 @@ class Index(SolrIndex): Class indexing books. """ def __init__(self): - super(Index, self).__init__() + super(Index, self).__init__(mode='rw') def delete_query(self, *queries): """ @@ -130,8 +133,7 @@ class Index(SolrIndex): break for res in ids: uids.add(res['uid']) - st+=rows - # print "Will delete %s" % ','.join([x for x in uids]) + st += rows if uids: self.index.delete(uids) return True @@ -144,6 +146,7 @@ class Index(SolrIndex): Removes all tags from index, then index them again. Indexed fields include: id, name (with and without polish stems), category """ + log.debug("Indexing tags") remove_only = kw.get('remove_only', False) # first, remove tags from index. if tags: @@ -179,7 +182,8 @@ class Index(SolrIndex): "tag_name": tag.name, "tag_name_pl": tag.name, "tag_category": 'pd_author', - "is_pdcounter": True + "is_pdcounter": True, + "uid": "tag%d_pd_a" % tag.id } elif isinstance(tag, PDCounterBook): doc = { @@ -187,7 +191,8 @@ class Index(SolrIndex): "tag_name": tag.title, "tag_name_pl": tag.title, "tag_category": 'pd_book', - "is_pdcounter": True + "is_pdcounter": True, + "uid": "tag%d_pd_b" % tag.id } else: doc = { @@ -195,9 +200,9 @@ class Index(SolrIndex): "tag_name": tag.name, "tag_name_pl": tag.name, "tag_category": tag.category, - "is_pdcounter": False + "is_pdcounter": False, + "uid": "tag%d" % tag.id } - doc['uid'] = "tag%d" % tag.id self.index.add(doc) def create_book_doc(self, book): @@ -237,7 +242,7 @@ class Index(SolrIndex): self.remove_book(book, remove_snippets=False) book_doc = self.create_book_doc(book) - meta_fields = self.extract_metadata(book, book_info, dc_only=['source_name', 'authors', 'title']) + meta_fields = self.extract_metadata(book, book_info, dc_only=['source_name', 'authors', 'translators', 'title']) # let's not index it - it's only used for extracting publish date if 'source_name' in meta_fields: del meta_fields['source_name'] @@ -248,11 +253,16 @@ class Index(SolrIndex): book_doc['uid'] = "book%s" % book_doc['book_id'] self.index.add(book_doc) del book_doc - - self.index_content(book, book_fields={ + book_fields = { 'title': meta_fields['title'], 'authors': meta_fields['authors'], - 'published_date': meta_fields['published_date']}) + 'published_date': meta_fields['published_date'] + } + + if 'translators' in meta_fields: + book_fields['translators'] = meta_fields['translators'] + + self.index_content(book, book_fields=book_fields) master_tags = [ 'opowiadanie', @@ -410,7 +420,7 @@ class Index(SolrIndex): doc['themes'] = fields['themes'] doc['uid'] = "part%s%s%s" % (doc['header_index'], doc['header_span'], - doc.get('fragment_anchor','')) + doc.get('fragment_anchor', '')) return doc def give_me_utf8(s): @@ -453,9 +463,7 @@ class Index(SolrIndex): doc = add_part(snippets, header_index=position, header_type=header.tag, text=u''.join(footnote), is_footnote=True) - self.index.add(doc) - #print "@ footnote text: %s" % footnote footnote = [] # handle fragments and themes. @@ -488,7 +496,6 @@ class Index(SolrIndex): fragment_anchor=fid, text=fix_format(frag['text']), themes=frag['themes']) - #print '@ FRAG %s' % frag['content'] self.index.add(doc) # Collect content. @@ -501,7 +508,6 @@ class Index(SolrIndex): # in the end, add a section text. doc = add_part(snippets, header_index=position, header_type=header.tag, text=fix_format(content)) - #print '@ CONTENT: %s' % fix_format(content) self.index.add(doc) @@ -509,47 +515,54 @@ class Index(SolrIndex): snippets.close() - class SearchResult(object): - def __init__(self, search, doc, how_found=None, snippets=None, searched=None, tokens_cache=None): - if tokens_cache is None: tokens_cache = {} + def __init__(self, doc, how_found=None, query=None, query_terms=None): + # self.search = search + self.boost = 1.0 + self._hits = [] + self._processed_hits = None # processed hits + self.snippets = [] + self.query_terms = query_terms if 'score' in doc: self._score = doc['score'] else: self._score = 0 - self.boost = 1.0 - - self._hits = [] - self._processed_hits = None # processed hits - self.book_id = int(doc["book_id"]) - pd = doc["published_date"] try: - self.published_date = int(pd) + self.published_date = int(doc.get("published_date")) except ValueError: self.published_date = 0 + # content hits header_type = doc.get("header_type", None) # we have a content hit in some header of fragment if header_type is not None: sec = (header_type, int(doc["header_index"])) header_span = doc['header_span'] header_span = header_span is not None and int(header_span) or 1 - fragment = doc.get("fragment_anchor", None) + snippets_pos = (doc['snippets_position'], doc['snippets_length']) + snippets_rev = doc.get('snippets_revision', None) - if snippets: - snippets = snippets.replace("/\n", "\n") - hit = (sec + (header_span,), fragment, self._score, {'how_found': how_found, 'snippets': snippets and [snippets] or []}) + hit = (sec + (header_span,), fragment, self._score, { + 'how_found': how_found, + 'snippets_pos': snippets_pos, + 'snippets_revision': snippets_rev, + 'themes': doc.get('themes', []), + 'themes_pl': doc.get('themes_pl', []) + }) self._hits.append(hit) - self.search = search - self.searched = searched - self.tokens_cache = tokens_cache + def __unicode__(self): + return u"" % \ + (self.book_id, len(self._hits), self._processed_hits and len(self._processed_hits) or -1, self._score, len(self.snippets)) + + def __str__(self): + return unicode(self).encode('utf-8') @property def score(self): @@ -566,31 +579,32 @@ class SearchResult(object): def get_book(self): if hasattr(self, '_book'): return self._book - return catalogue.models.Book.objects.get(id=self.book_id) + self._book = catalogue.models.Book.objects.get(id=self.book_id) + return self._book book = property(get_book) + POSITION = 0 + FRAGMENT = 1 + POSITION_INDEX = 1 + POSITION_SPAN = 2 + SCORE = 2 + OTHER = 3 + @property def hits(self): if self._processed_hits is not None: return self._processed_hits - POSITION = 0 - FRAGMENT = 1 - POSITION_INDEX = 1 - POSITION_SPAN = 2 - SCORE = 2 - OTHER = 3 - # to sections and fragments - frags = filter(lambda r: r[FRAGMENT] is not None, self._hits) + frags = filter(lambda r: r[self.FRAGMENT] is not None, self._hits) - sect = filter(lambda r: r[FRAGMENT] is None, self._hits) + sect = filter(lambda r: r[self.FRAGMENT] is None, self._hits) # sections not covered by fragments sect = filter(lambda s: 0 == len(filter( - lambda f: s[POSITION][POSITION_INDEX] >= f[POSITION][POSITION_INDEX] - and s[POSITION][POSITION_INDEX] < f[POSITION][POSITION_INDEX] + f[POSITION][POSITION_SPAN], + lambda f: s[self.POSITION][self.POSITION_INDEX] >= f[self.POSITION][self.POSITION_INDEX] + and s[self.POSITION][self.POSITION_INDEX] < f[self.POSITION][self.POSITION_INDEX] + f[self.POSITION][self.POSITION_SPAN], frags)), sect) hits = [] @@ -606,55 +620,61 @@ class SearchResult(object): return els.values() # remove fragments with duplicated fid's and duplicated snippets - frags = remove_duplicates(frags, lambda f: f[FRAGMENT], lambda a, b: cmp(a[SCORE], b[SCORE])) - frags = remove_duplicates(frags, lambda f: f[OTHER]['snippets'] and f[OTHER]['snippets'][0] or f[FRAGMENT], - lambda a, b: cmp(a[SCORE], b[SCORE])) + frags = remove_duplicates(frags, lambda f: f[self.FRAGMENT], lambda a, b: cmp(a[self.SCORE], b[self.SCORE])) + # frags = remove_duplicates(frags, lambda f: f[OTHER]['snippet_pos'] and f[OTHER]['snippet_pos'] or f[FRAGMENT], + # lambda a, b: cmp(a[SCORE], b[SCORE])) # remove duplicate sections sections = {} for s in sect: - si = s[POSITION][POSITION_INDEX] + si = s[self.POSITION][self.POSITION_INDEX] # skip existing if si in sections: - if sections[si]['score'] >= s[SCORE]: + if sections[si]['score'] >= s[self.SCORE]: continue - m = {'score': s[SCORE], - 'section_number': s[POSITION][POSITION_INDEX] + 1, + m = {'score': s[self.SCORE], + 'section_number': s[self.POSITION][self.POSITION_INDEX] + 1, } - m.update(s[OTHER]) + m.update(s[self.OTHER]) sections[si] = m hits = sections.values() for f in frags: try: - frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT], book__id=self.book_id) + frag = catalogue.models.Fragment.objects.get(anchor=f[self.FRAGMENT], book__id=self.book_id) except catalogue.models.Fragment.DoesNotExist: # stale index continue - # Figure out if we were searching for a token matching some word in theme name. themes = frag.tags.filter(category='theme') - themes_hit = [] - if self.searched is not None: - tokens = self.search.get_tokens(self.searched, 'POLISH', cached=self.tokens_cache) - for theme in themes: - name_tokens = self.search.get_tokens(theme.name, 'POLISH') - for t in tokens: - if t in name_tokens: - if not theme in themes_hit: - themes_hit.append(theme) + themes_hit = set() + if self.query_terms is not None: + for i in range(0, len(f[self.OTHER]['themes'])): + tms = f[self.OTHER]['themes'][i].split(r' +') + f[self.OTHER]['themes_pl'][i].split(' ') + tms = map(unicode.lower, tms) + for qt in self.query_terms: + if qt in tms: + themes_hit.add(f[self.OTHER]['themes'][i]) break - m = {'score': f[SCORE], + def theme_by_name(n): + th = filter(lambda t: t.name == n, themes) + if th: + return th[0] + else: + return None + themes_hit = filter(lambda a: a is not None, map(theme_by_name, themes_hit)) + + m = {'score': f[self.SCORE], 'fragment': frag, - 'section_number': f[POSITION][POSITION_INDEX] + 1, + 'section_number': f[self.POSITION][self.POSITION_INDEX] + 1, 'themes': themes, 'themes_hit': themes_hit } - m.update(f[OTHER]) + m.update(f[self.OTHER]) hits.append(m) hits.sort(lambda a, b: cmp(a['score'], b['score']), reverse=True) @@ -663,9 +683,6 @@ class SearchResult(object): return hits - def __unicode__(self): - return u'SearchResult(book_id=%d, score=%d)' % (self.book_id, self.score) - @staticmethod def aggregate(*result_lists): books = {} @@ -685,624 +702,239 @@ class SearchResult(object): else: return c + def __len__(self): + return len(self.hits) -class Hint(object): - """ - Given some hint information (information we already know about) - our search target - like author, title (specific book), epoch, genre, kind - we can narrow down search using filters. - """ - def __init__(self, search): - """ - Accepts a Searcher instance. - """ - self.search = search - self.book_tags = {} - self.part_tags = [] - self._books = [] - - def books(self, *books): - """ - Give a hint that we search these books. - """ - self._books = books + def snippet_pos(self, idx=0): + return self.hits[idx]['snippets_pos'] - def tags(self, tags): - """ - Give a hint that these Tag objects (a list of) - is necessary. - """ - for t in tags: - if t.category in ['author', 'title', 'epoch', 'genre', 'kind']: - lst = self.book_tags.get(t.category, []) - lst.append(t) - self.book_tags[t.category] = lst - if t.category in ['theme', 'theme_pl']: - self.part_tags.append(t) - - def tag_filter(self, tags, field='tags'): - """ - Given a lsit of tags and an optional field (but they are normally in tags field) - returns a filter accepting only books with specific tags. - """ - q = BooleanQuery() - - for tag in tags: - toks = self.search.get_tokens(tag.name, field=field) - tag_phrase = PhraseQuery() - for tok in toks: - tag_phrase.add(Term(field, tok)) - q.add(BooleanClause(tag_phrase, BooleanClause.Occur.MUST)) - - return QueryWrapperFilter(q) - - def book_filter(self): - """ - Filters using book tags (all tag kinds except a theme) - """ - tags = reduce(lambda a, b: a + b, self.book_tags.values(), []) - if tags: - return self.tag_filter(tags) - else: + def snippet_revision(self, idx=0): + try: + return self.hits[idx]['snippets_revision'] + except: return None - def part_filter(self): - """ - This filter can be used to look for book parts. - It filters on book id and/or themes. - """ - fs = [] - if self.part_tags: - fs.append(self.tag_filter(self.part_tags, field='themes')) - - if self._books != []: - bf = BooleanFilter() - for b in self._books: - id_filter = NumericRangeFilter.newIntRange('book_id', b.id, b.id, True, True) - bf.add(FilterClause(id_filter, BooleanClause.Occur.SHOULD)) - fs.append(bf) - - return Search.chain_filters(fs) - - def should_search_for_book(self): - return self._books == [] - - def just_search_in(self, all): - """Holds logic to figure out which indexes should be search, when we have some hinst already""" - some = [] - for field in all: - if field == 'authors' and 'author' in self.book_tags: - continue - if field == 'title' and self._books != []: - continue - if (field == 'themes' or field == 'themes_pl') and self.part_tags: - continue - some.append(field) - return some - class Search(SolrIndex): """ Search facilities. """ def __init__(self, default_field="text"): - IndexStore.__init__(self) - self.analyzer = WLAnalyzer() # PolishAnalyzer(Version.LUCENE_34) - # self.analyzer = WLAnalyzer() - reader = IndexReader.open(self.store, True) - self.searcher = IndexSearcher(reader) - self.parser = QueryParser(Version.LUCENE_34, default_field, - self.analyzer) - - self.parent_filter = TermsFilter() - self.parent_filter.addTerm(Term("is_book", "true")) - index_changed.connect(self.reopen) - - def close(self): - reader = self.searcher.getIndexReader() - self.searcher.close() - reader.close() - super(Search, self).close() - index_changed.disconnect(self.reopen) - - def reopen(self, **unused): - reader = self.searcher.getIndexReader() - rdr = reader.reopen() - if not rdr.equals(reader): - log.debug('Reopening index') - oldsearch = self.searcher - self.searcher = IndexSearcher(rdr) - oldsearch.close() - reader.close() - - def query(self, query): - """Parse query in default Lucene Syntax. (for humans) - """ - return self.parser.parse(query) - - def simple_search(self, query, max_results=50): - """Runs a query for books using lucene syntax. (for humans) - Returns (books, total_hits) - """ - - tops = self.searcher.search(self.query(query), max_results) - bks = [] - for found in tops.scoreDocs: - doc = self.searcher.doc(found.doc) - bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) - return (bks, tops.totalHits) - - def get_tokens(self, searched, field='text', cached=None): - """returns tokens analyzed by a proper (for a field) analyzer - argument can be: StringReader, string/unicode, or tokens. In the last case - they will just be returned (so we can reuse tokens, if we don't change the analyzer) - """ - if cached is not None and field in cached: - return cached[field] - - if isinstance(searched, str) or isinstance(searched, unicode): - searched = StringReader(searched) - elif isinstance(searched, list): - return searched - - searched.reset() - tokens = self.analyzer.reusableTokenStream(field, searched) - toks = [] - while tokens.incrementToken(): - cta = tokens.getAttribute(CharTermAttribute.class_) - toks.append(cta.toString()) - - if cached is not None: - cached[field] = toks - - return toks - - @staticmethod - def fuzziness(fuzzy): - """Helper method to sanitize fuzziness""" - if not fuzzy: - return None - if isinstance(fuzzy, float) and fuzzy > 0.0 and fuzzy <= 1.0: - return fuzzy - else: - return 0.5 - - def make_phrase(self, tokens, field='text', slop=2, fuzzy=False): - """ - Return a PhraseQuery with a series of tokens. - """ - if fuzzy: - phrase = MultiPhraseQuery() - for t in tokens: - term = Term(field, t) - fuzzterm = FuzzyTermEnum(self.searcher.getIndexReader(), term, self.fuzziness(fuzzy)) - fuzzterms = [] + super(Search, self).__init__(mode='r') - while True: - ft = fuzzterm.term() - if ft: - fuzzterms.append(ft) - if not fuzzterm.next(): break - if fuzzterms: - phrase.add(JArray('object')(fuzzterms, Term)) - else: - phrase.add(term) - else: - phrase = PhraseQuery() - phrase.setSlop(slop) - for t in tokens: - term = Term(field, t) - phrase.add(term) - return phrase - @staticmethod - def make_term_query(tokens, field='text', modal='BooleanClause.Occur.SHOULD XXX', fuzzy=False): + def make_term_query(self, query, field='text', modal=operator.or_): """ Returns term queries joined by boolean query. modal - applies to boolean query fuzzy - should the query by fuzzy. """ - q = BooleanQuery() - for t in tokens: - term = Term(field, t) - if fuzzy: - term = FuzzyQuery(term, self.fuzziness(fuzzy)) - else: - term = TermQuery(term) - q.add(BooleanClause(term, modal)) + if query is None: query = '' + q = self.index.Q() + q = reduce(modal, map(lambda s: self.index.Q(**{field: s}), + query.split(r" ")), q) + return q - def search_phrase(self, searched, field, book=True, max_results=20, fuzzy=False, - filters=None, tokens_cache=None, boost=None, snippets=False, slop=2): + def search_phrase(self, searched, field='text', book=False, + filters=None, + snippets=False): if filters is None: filters = [] - if tokens_cache is None: tokens_cache = {} + if book: filters.append(self.index.Q(is_book=True)) - tokens = self.get_tokens(searched, field, cached=tokens_cache) + q = self.index.query(**{field: searched}) + q = self.apply_filters(q, filters).field_limit(score=True, all_fields=True) + res = q.execute() + return [SearchResult(found, how_found=u'search_phrase') for found in res] - query = self.make_phrase(tokens, field=field, fuzzy=fuzzy, slop=slop) - if book: - filters.append(self.term_filter(Term('is_book', 'true'))) - top = self.searcher.search(query, self.chain_filters(filters), max_results) - - return [SearchResult(self, found, snippets=(snippets and self.get_snippets(found, query) or None), searched=searched) for found in top.scoreDocs] - - def search_some(self, searched, fields, book=True, max_results=20, fuzzy=False, - filters=None, tokens_cache=None, boost=None, snippets=True): + def search_some(self, searched, fields, book=True, + filters=None, snippets=True, query_terms=None): + assert isinstance(fields, list) if filters is None: filters = [] - if tokens_cache is None: tokens_cache = {} - - if book: - filters.append(self.term_filter(Term('is_book', 'true'))) + if book: filters.append(self.index.Q(is_book=True)) - query = BooleanQuery() + query = self.index.Q() for fld in fields: - tokens = self.get_tokens(searched, fld, cached=tokens_cache) - - query.add(BooleanClause(self.make_term_query(tokens, field=fld, - fuzzy=fuzzy), BooleanClause.Occur.SHOULD)) - - top = self.searcher.search(query, self.chain_filters(filters), max_results) - - return [SearchResult(self, found, searched=searched, tokens_cache=tokens_cache, - snippets=(snippets and self.get_snippets(found, query) or None)) for found in top.scoreDocs] - - def search_perfect_book(self, searched, max_results=20, fuzzy=False, hint=None): - """ - Search for perfect book matches. Just see if the query matches with some author or title, - taking hints into account. - """ - fields_to_search = ['authors', 'title'] - only_in = None - if hint: - if not hint.should_search_for_book(): - return [] - fields_to_search = hint.just_search_in(fields_to_search) - only_in = hint.book_filter() - - qrys = [self.make_phrase(self.get_tokens(searched, field=fld), field=fld, fuzzy=fuzzy) for fld in fields_to_search] - - books = [] - for q in qrys: - top = self.searcher.search(q, - self.chain_filters([only_in, self.term_filter(Term('is_book', 'true'))]), - max_results) - for found in top.scoreDocs: - books.append(SearchResult(self, found, how_found="search_perfect_book")) - return books - - def search_book(self, searched, max_results=20, fuzzy=False, hint=None): - fields_to_search = ['tags', 'authors', 'title'] - - only_in = None - if hint: - if not hint.should_search_for_book(): - return [] - fields_to_search = hint.just_search_in(fields_to_search) - only_in = hint.book_filter() - - tokens = self.get_tokens(searched, field='SIMPLE') - - q = BooleanQuery() + query = self.index.Q(query | self.make_term_query(searched, fld)) - for fld in fields_to_search: - q.add(BooleanClause(self.make_term_query(tokens, field=fld, - fuzzy=fuzzy), BooleanClause.Occur.SHOULD)) + query = self.index.query(query) + query = self.apply_filters(query, filters).field_limit(score=True, all_fields=True) + res = query.execute() + return [SearchResult(found, how_found='search_some', query_terms=query_terms) for found in res] - books = [] - top = self.searcher.search(q, - self.chain_filters([only_in, self.term_filter(Term('is_book', 'true'))]), - max_results) - for found in top.scoreDocs: - books.append(SearchResult(self, found, how_found="search_book")) - - return books - - def search_perfect_parts(self, searched, max_results=20, fuzzy=False, hint=None): - """ - Search for book parts which contains a phrase perfectly matching (with a slop of 2, default for make_phrase()) - some part/fragment of the book. - """ - qrys = [self.make_phrase(self.get_tokens(searched), field=fld, fuzzy=fuzzy) for fld in ['text']] - flt = None - if hint: - flt = hint.part_filter() - - books = [] - for q in qrys: - top = self.searcher.search(q, - self.chain_filters([self.term_filter(Term('is_book', 'true'), inverse=True), - flt]), - max_results) - for found in top.scoreDocs: - books.append(SearchResult(self, found, snippets=self.get_snippets(found, q), how_found='search_perfect_parts')) - - return books - - def search_everywhere(self, searched, max_results=20, fuzzy=False, hint=None, tokens_cache=None): + def search_everywhere(self, searched, query_terms=None): """ Tries to use search terms to match different fields of book (or its parts). E.g. one word can be an author survey, another be a part of the title, and the rest are some words from third chapter. """ - if tokens_cache is None: tokens_cache = {} books = [] - only_in = None - - if hint: - only_in = hint.part_filter() - # content only query : themes x content - q = BooleanQuery() - - tokens_pl = self.get_tokens(searched, field='text', cached=tokens_cache) - tokens = self.get_tokens(searched, field='SIMPLE', cached=tokens_cache) - - # only search in themes when we do not already filter by themes - if hint is None or hint.just_search_in(['themes']) != []: - q.add(BooleanClause(self.make_term_query(tokens_pl, field='themes_pl', - fuzzy=fuzzy), BooleanClause.Occur.MUST)) + q = self.make_term_query(searched, 'text') + q_themes = self.make_term_query(searched, 'themes_pl') - q.add(BooleanClause(self.make_term_query(tokens_pl, field='text', - fuzzy=fuzzy), BooleanClause.Occur.SHOULD)) + query = self.index.query(q).query(q_themes).field_limit(score=True, all_fields=True) + res = query.execute() - topDocs = self.searcher.search(q, only_in, max_results) - for found in topDocs.scoreDocs: - books.append(SearchResult(self, found, how_found='search_everywhere_themesXcontent', searched=searched)) + for found in res: + books.append(SearchResult(found, how_found='search_everywhere_themesXcontent', query_terms=query_terms)) # query themes/content x author/title/tags - q = BooleanQuery() - in_content = BooleanQuery() - in_meta = BooleanQuery() + in_content = self.index.Q() + in_meta = self.index.Q() for fld in ['themes_pl', 'text']: - in_content.add(BooleanClause(self.make_term_query(tokens_pl, field=fld, fuzzy=False), BooleanClause.Occur.SHOULD)) + in_content |= self.make_term_query(searched, field=fld) for fld in ['tags', 'authors', 'title']: - in_meta.add(BooleanClause(self.make_term_query(tokens, field=fld, fuzzy=False), BooleanClause.Occur.SHOULD)) + in_meta |= self.make_term_query(searched, field=fld) - q.add(BooleanClause(in_content, BooleanClause.Occur.MUST)) - q.add(BooleanClause(in_meta, BooleanClause.Occur.SHOULD)) + q = in_content & in_meta + res = self.index.query(q).field_limit(score=True, all_fields=True).execute() - topDocs = self.searcher.search(q, only_in, max_results) - for found in topDocs.scoreDocs: - books.append(SearchResult(self, found, how_found='search_everywhere', searched=searched)) + for found in res: + books.append(SearchResult(found, how_found='search_everywhere', query_terms=query_terms)) return books - # def multisearch(self, query, max_results=50): - # """ - # Search strategy: - # - (phrase) OR -> content - # -> title - # -> authors - # - (keywords) -> authors - # -> motyw - # -> tags - # -> content - # """ - # queryreader = StringReader(query) - # tokens = self.get_tokens(queryreader) - - # top_level = BooleanQuery() - # Should = BooleanClause.Occur.SHOULD - - # phrase_level = BooleanQuery() - # phrase_level.setBoost(1.3) - - # p_content = self.make_phrase(tokens, joined=True) - # p_title = self.make_phrase(tokens, 'title') - # p_author = self.make_phrase(tokens, 'author') - - # phrase_level.add(BooleanClause(p_content, Should)) - # phrase_level.add(BooleanClause(p_title, Should)) - # phrase_level.add(BooleanClause(p_author, Should)) - - # kw_level = BooleanQuery() - - # kw_level.add(self.make_term_query(tokens, 'author'), Should) - # j_themes = self.make_term_query(tokens, 'themes', joined=True) - # kw_level.add(j_themes, Should) - # kw_level.add(self.make_term_query(tokens, 'tags'), Should) - # j_con = self.make_term_query(tokens, joined=True) - # kw_level.add(j_con, Should) - - # top_level.add(BooleanClause(phrase_level, Should)) - # top_level.add(BooleanClause(kw_level, Should)) - - # return None - - def get_snippets(self, scoreDoc, query, field='text'): + def get_snippets(self, searchresult, query, field='text', num=1): """ Returns a snippet for found scoreDoc. """ - htmlFormatter = SimpleHTMLFormatter() - highlighter = Highlighter(htmlFormatter, QueryScorer(query)) - - stored = self.searcher.doc(scoreDoc.doc) - - position = stored.get('snippets_position') - length = stored.get('snippets_length') - if position is None or length is None: - return None - revision = stored.get('snippets_revision') - if revision: revision = int(revision) - - # locate content. - book_id = int(stored.get('book_id')) + maxnum = len(searchresult) + if num is None or num < 0 or num > maxnum: + num = maxnum + book_id = searchresult.book_id + revision = searchresult.snippet_revision() snippets = Snippets(book_id, revision=revision) - + snips = [None] * maxnum try: snippets.open() + idx = 0 + while idx < maxnum and num > 0: + position, length = searchresult.snippet_pos(idx) + if position is None or length is None: + continue + text = snippets.get((int(position), + int(length))) + snip = self.index.highlight(text=text, field=field, q=query) + snips[idx] = snip + if snip: + num -= 1 + idx += 1 + except IOError, e: - log.error("Cannot open snippet file for book id = %d [rev=%d], %s" % (book_id, revision, e)) + log.error("Cannot open snippet file for book id = %d [rev=%s], %s" % (book_id, revision, e)) return [] + finally: + snippets.close() - try: - try: - text = snippets.get((int(position), - int(length))) - finally: - snippets.close() + # remove verse end markers.. + snips = map(lambda s: s and s.replace("/\n", "\n"), snips) - tokenStream = TokenSources.getAnyTokenStream(self.searcher.getIndexReader(), scoreDoc.doc, field, self.analyzer) - # highlighter.getBestTextFragments(tokenStream, text, False, 10) - snip = highlighter.getBestFragments(tokenStream, text, 3, "...") + searchresult.snippets = snips - except Exception, e: - e2 = e - if hasattr(e, 'getJavaException'): - e2 = unicode(e.getJavaException()) - raise Exception("Problem fetching snippets for book %d, @%d len=%d" % (book_id, int(position), int(length)), - e2) - return snip + return snips - @staticmethod - def enum_to_array(enum): + def hint_tags(self, query, pdcounter=True, prefix=True): """ - Converts a lucene TermEnum to array of Terms, suitable for - addition to queries + Return auto-complete hints for tags + using prefix search. """ - terms = [] - - while True: - t = enum.term() - if t: - terms.append(t) - if not enum.next(): break + q = self.index.Q() + query = query.strip() + for field in ['tag_name', 'tag_name_pl']: + if prefix: + q |= self.index.Q(**{field: query + "*"}) + else: + q |= self.make_term_query(query, field=field) + qu = self.index.query(q).exclude(tag_category="book") - if terms: - return JArray('object')(terms, Term) + return self.search_tags(qu, pdcounter=pdcounter) - def search_tags(self, query, filt=None, max_results=40, pdcounter=False): + def search_tags(self, query, filters=None, pdcounter=False): """ Search for Tag objects using query. """ + if not filters: filters = [] if not pdcounter: - filters = self.chain_filters([filt, self.term_filter(Term('is_pdcounter', 'true'), inverse=True)]) - tops = self.searcher.search(query, filt, max_results) + filters.append(~self.index.Q(is_pdcounter=True)) + res = self.apply_filters(query, filters).execute() tags = [] - for found in tops.scoreDocs: - doc = self.searcher.doc(found.doc) - is_pdcounter = doc.get('is_pdcounter') + pd_tags = [] + + for doc in res: + is_pdcounter = doc.get('is_pdcounter', False) category = doc.get('tag_category') try: - if is_pdcounter == 'true': + if is_pdcounter == True: if category == 'pd_author': tag = PDCounterAuthor.objects.get(id=doc.get('tag_id')) elif category == 'pd_book': tag = PDCounterBook.objects.get(id=doc.get('tag_id')) tag.category = 'pd_book' # make it look more lik a tag. else: - print "Warning. cannot get pdcounter tag_id=%d from db; cat=%s" % (int(doc.get('tag_id')), category) + print ("Warning. cannot get pdcounter tag_id=%d from db; cat=%s" % (int(doc.get('tag_id')), category)).encode('utf-8') + pd_tags.append(tag) else: tag = catalogue.models.Tag.objects.get(id=doc.get("tag_id")) - # don't add the pdcounter tag if same tag already exists - - tags.append(tag) + tags.append(tag) except catalogue.models.Tag.DoesNotExist: pass except PDCounterAuthor.DoesNotExist: pass except PDCounterBook.DoesNotExist: pass + tags_slugs = set(map(lambda t: t.slug, tags)) + tags = tags + filter(lambda t: not t.slug in tags_slugs, pd_tags) + log.debug('search_tags: %s' % tags) return tags - def search_books(self, query, filt=None, max_results=10): - """ - Searches for Book objects using query - """ - bks = [] - tops = self.searcher.search(query, filt, max_results) - for found in tops.scoreDocs: - doc = self.searcher.doc(found.doc) - try: - bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) - except catalogue.models.Book.DoesNotExist: pass - return bks - - def make_prefix_phrase(self, toks, field): - q = MultiPhraseQuery() - for i in range(len(toks)): - t = Term(field, toks[i]) - if i == len(toks) - 1: - pterms = Search.enum_to_array(PrefixTermEnum(self.searcher.getIndexReader(), t)) - if pterms: - q.add(pterms) - else: - q.add(t) - else: - q.add(t) - return q - - @staticmethod - def term_filter(term, inverse=False): - only_term = TermsFilter() - only_term.addTerm(term) - - if inverse: - neg = BooleanFilter() - neg.add(FilterClause(only_term, BooleanClause.Occur.MUST_NOT)) - only_term = neg - - return only_term - - def hint_tags(self, string, max_results=50, pdcounter=True, prefix=True, fuzzy=False): - """ - Return auto-complete hints for tags - using prefix search. - """ - toks = self.get_tokens(string, field='SIMPLE') - top = BooleanQuery() - - for field in ['tag_name', 'tag_name_pl']: - if prefix: - q = self.make_prefix_phrase(toks, field) - else: - q = self.make_term_query(toks, field, fuzzy=fuzzy) - top.add(BooleanClause(q, BooleanClause.Occur.SHOULD)) - - no_book_cat = self.term_filter(Term("tag_category", "book"), inverse=True) - - return self.search_tags(top, no_book_cat, max_results=max_results, pdcounter=pdcounter) - - def hint_books(self, string, max_results=50, prefix=True, fuzzy=False): + def hint_books(self, query, prefix=True): """ Returns auto-complete hints for book titles Because we do not index 'pseudo' title-tags. Prefix search. """ - toks = self.get_tokens(string, field='SIMPLE') - + q = self.index.Q() + query = query.strip() if prefix: - q = self.make_prefix_phrase(toks, 'title') + q |= self.index.Q(title=query + "*") else: - q = self.make_term_query(toks, 'title', fuzzy=fuzzy) - - return self.search_books(q, self.term_filter(Term("is_book", "true")), max_results=max_results) + q |= self.make_term_query(query, field='title') + qu = self.index.query(q) + only_books = self.index.Q(is_book=True) + return self.search_books(qu, [only_books]) - @staticmethod - def chain_filters(filters, op='XXXChainedFilter.AND'): + def search_books(self, query, filters=None, max_results=10): """ - Chains a filter list together + Searches for Book objects using query """ - filters = filter(lambda x: x is not None, filters) - if not filters or filters is []: - return None - chf = ChainedFilter(JArray('object')(filters, Filter), op) - return chf + bks = [] + bks_found = set() + query = query.query(is_book=True) + res = self.apply_filters(query, filters).field_limit(['book_id']) + for r in res: + try: + bid = r['book_id'] + if not bid in bks_found: + bks.append(catalogue.models.Book.objects.get(id=bid)) + bks_found.add(bid) + except catalogue.models.Book.DoesNotExist: pass + return bks + - def filtered_categories(self, tags): + @staticmethod + def apply_filters(query, filters): """ - Return a list of tag categories, present in tags list. + Apply filters to a query """ - cats = {} - for t in tags: - cats[t.category] = True - return cats.keys() - - def hint(self): - return Hint(self) + if filters is None: filters = [] + filters = filter(lambda x: x is not None, filters) + for f in filters: + query = query.query(f) + return query