X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/d285fd31ba8f1aaf3539606e98894f4eb396d41c..462d84ea8718cd3ca1112fe11c2120a0f20f475f:/apps/search/index.py diff --git a/apps/search/index.py b/apps/search/index.py index 307376de9..df44888df 100644 --- a/apps/search/index.py +++ b/apps/search/index.py @@ -553,11 +553,15 @@ class JoinSearch(object): class SearchResult(object): - def __init__(self, searcher, scoreDocs, score=None, how_found=None, snippets=None): + def __init__(self, searcher, scoreDocs, score=None, how_found=None, snippets=None, searched=None, tokens_cache=None): + if tokens_cache is None: tokens_cache = {} + if score: - self.score = score + self._score = score else: - self.score = scoreDocs.score + self._score = scoreDocs.score + + self.boost = 1.0 self._hits = [] self.hits = None # processed hits @@ -581,6 +585,14 @@ class SearchResult(object): self._hits.append(hit) + self.searcher = searcher + self.searched = searched + self.tokens_cache = tokens_cache + + @property + def score(self): + return self._score * self.boost + def merge(self, other): if self.book_id != other.book_id: raise ValueError("this search result is or book %d; tried to merge with %d" % (self.book_id, other.book_id)) @@ -642,10 +654,25 @@ class SearchResult(object): for f in frags: frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT]) + + # 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.searcher.get_tokens(self.searched, 'POLISH', tokens_cache=self.tokens_cache) + for theme in themes: + name_tokens = self.searcher.get_tokens(theme.name, 'POLISH') + for t in tokens: + if name_tokens.index(t): + if not theme in themes_hit: + themes_hit.append(theme) + break + m = {'score': f[SCORE], 'fragment': frag, 'section_number': f[POSITION][POSITION_INDEX] + 1, - 'themes': frag.tags.filter(category='theme') + 'themes': themes, + 'themes_hit': themes_hit } m.update(f[OTHER]) hits.append(m) @@ -802,11 +829,14 @@ class Search(IndexStore): bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) return (bks, tops.totalHits) - def get_tokens(self, searched, field='content'): + def get_tokens(self, searched, field='content', 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): @@ -818,6 +848,10 @@ class Search(IndexStore): while tokens.incrementToken(): cta = tokens.getAttribute(CharTermAttribute.class_) toks.append(cta.toString()) + + if cached is not None: + cached[field] = toks + return toks def fuzziness(self, fuzzy): @@ -874,9 +908,39 @@ class Search(IndexStore): q.add(BooleanClause(term, modal)) return q - # def content_query(self, query): - # return BlockJoinQuery(query, self.parent_filter, - # BlockJoinQuery.ScoreMode.Total) + def search_phrase(self, searched, field, book=True, max_results=20, fuzzy=False, + filters=None, tokens_cache=None, boost=None): + if filters is None: filters = [] + if tokens_cache is None: tokens_cache = {} + + tokens = self.get_tokens(searched, field, cached=tokens_cache) + + query = self.make_phrase(tokens, field=field, fuzzy=fuzzy) + 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.searcher, found) 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): + if filters is None: filters = [] + if tokens_cache is None: tokens_cache = {} + + if book: + filters.append(self.term_filter(Term('is_book', 'true'))) + + query = BooleanQuery() + + 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.searcher, found, searched=searched, tokens_cache=tokens_cache) for found in top.scoreDocs] def search_perfect_book(self, searched, max_results=20, fuzzy=False, hint=None): """ @@ -931,7 +995,7 @@ class Search(IndexStore): def search_perfect_parts(self, searched, max_results=20, fuzzy=False, hint=None): """ - Search for book parts which containt a phrase perfectly matching (with a slop of 2, default for make_phrase()) + 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 ['content']] @@ -951,12 +1015,13 @@ class Search(IndexStore): return books - def search_everywhere(self, searched, max_results=20, fuzzy=False, hint=None): + def search_everywhere(self, searched, max_results=20, fuzzy=False, hint=None, tokens_cache=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 @@ -966,8 +1031,8 @@ class Search(IndexStore): # content only query : themes x content q = BooleanQuery() - tokens_pl = self.get_tokens(searched, field='content') - tokens = self.get_tokens(searched, field='SIMPLE') + tokens_pl = self.get_tokens(searched, field='content', 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']) != []: @@ -1171,7 +1236,7 @@ class Search(IndexStore): Chains a filter list together """ filters = filter(lambda x: x is not None, filters) - if not filters: + if not filters or filters is []: return None chf = ChainedFilter(JArray('object')(filters, Filter), op) return chf