X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/05e5b36046d980e84776978660f420b008db4879..154870f0416b6b387637d6035c96321410512e95:/src/search/index.py diff --git a/src/search/index.py b/src/search/index.py index 0bd5c0530..a712b0702 100644 --- a/src/search/index.py +++ b/src/search/index.py @@ -1,26 +1,34 @@ -# -*- coding: utf-8 -*- # This file is part of Wolnelektury, licensed under GNU Affero GPLv3 or later. # Copyright © Fundacja Nowoczesna Polska. See NOTICE for more information. # -from django.conf import settings - +from functools import total_ordering +from itertools import chain +import logging +import operator import os import re +from django.conf import settings from librarian import dcparser from librarian.parser import WLDocument from lxml import etree +import scorched import catalogue.models +import picture.models from pdcounter.models import Author as PDCounterAuthor, BookStub as PDCounterBook -from itertools import chain -import sunburnt -import custom -import operator -import logging from wolnelektury.utils import makedirs +from . import custom log = logging.getLogger('search') +if os.path.isfile(settings.SOLR_STOPWORDS): + stopwords = set( + line.strip() + for line in open(settings.SOLR_STOPWORDS) if not line.startswith('#')) +else: + stopwords = set() + + class SolrIndex(object): def __init__(self, mode=None): self.index = custom.CustomSolrInterface(settings.SOLR, mode=mode) @@ -121,7 +129,7 @@ class Index(SolrIndex): """ uids = set() for q in queries: - if isinstance(q, sunburnt.search.LuceneQuery): + if isinstance(q, scorched.search.LuceneQuery): q = self.index.query(q) q.field_limiter.update(['uid']) st = 0 @@ -271,14 +279,14 @@ class Index(SolrIndex): 'dramat_wierszowany_lp', 'dramat_wspolczesny', 'liryka_l', 'liryka_lp', 'wywiad', - ] + ] ignore_content_tags = [ - 'uwaga', 'extra', 'nota_red', + 'uwaga', 'extra', 'nota_red', 'abstrakt', 'zastepnik_tekstu', 'sekcja_asterysk', 'separator_linia', 'zastepnik_wersu', 'didaskalia', 'naglowek_aktu', 'naglowek_sceny', 'naglowek_czesc', - ] + ] footnote_tags = ['pa', 'pt', 'pr', 'pe'] @@ -297,7 +305,6 @@ class Index(SolrIndex): book_info = dcparser.parse(open(book.xml_file.path)) fields['slug'] = book.slug - fields['tags'] = [t.name for t in book.tags] fields['is_book'] = True # validator, name @@ -317,9 +324,9 @@ class Index(SolrIndex): elif type_indicator == dcparser.as_person: p = getattr(book_info, field.name) if isinstance(p, dcparser.Person): - persons = unicode(p) + persons = str(p) else: - persons = ', '.join(map(unicode, p)) + persons = ', '.join(map(str, p)) fields[field.name] = persons elif type_indicator == dcparser.as_date: dt = getattr(book_info, field.name) @@ -471,7 +478,7 @@ class Index(SolrIndex): fid = start.attrib['id'][1:] handle_text.append(lambda text: None) if start.text is not None: - fragments[fid]['themes'] += map(unicode.strip, map(unicode, (start.text.split(',')))) + fragments[fid]['themes'] += map(str.strip, map(str, (start.text.split(',')))) elif end is not None and end.tag == 'motyw': handle_text.pop() @@ -508,7 +515,49 @@ class Index(SolrIndex): finally: snippets.close() + def remove_picture(self, picture_or_id): + """Removes a picture from search index.""" + if isinstance(picture_or_id, picture.models.Picture): + picture_id = picture_or_id.id + else: + picture_id = picture_or_id + self.delete_query(self.index.Q(picture_id=picture_id)) + + def index_picture(self, picture, picture_info=None, overwrite=True): + """ + Indexes the picture. + Creates a lucene document for extracted metadata + and calls self.index_area() to index the contents of the picture. + """ + if overwrite: + # we don't remove snippets, since they might be still needed by + # threads using not reopened index + self.remove_picture(picture) + + picture_doc = {'picture_id': int(picture.id)} + meta_fields = self.extract_metadata(picture, picture_info, dc_only=[ + 'authors', 'title', 'epochs', 'kinds', 'genres']) + + picture_doc.update(meta_fields) + + picture_doc['uid'] = "picture%s" % picture_doc['picture_id'] + self.index.add(picture_doc) + del picture_doc['is_book'] + for area in picture.areas.all(): + self.index_area(area, picture_fields=picture_doc) + + def index_area(self, area, picture_fields): + """ + Indexes themes and objects on the area. + """ + doc = dict(picture_fields) + doc['area_id'] = area.id + doc['themes'] = list(area.tags.filter(category__in=('theme', 'thing')).values_list('name', flat=True)) + doc['uid'] = 'area%s' % area.id + self.index.add(doc) + +@total_ordering class SearchResult(object): def __init__(self, doc, how_found=None, query_terms=None): self.boost = 1.0 @@ -551,14 +600,25 @@ class SearchResult(object): self._hits.append(hit) - def __unicode__(self): + @classmethod + def from_book(cls, book, how_found=None, query_terms=None): + doc = { + 'score': book.popularity.count, + 'book_id': book.id, + 'published_date': 0, + } + result = cls(doc, how_found=how_found, query_terms=query_terms) + result._book = book + return result + + def __str__(self): return u"" % \ (self.book_id, len(self._hits), len(self._processed_hits) if self._processed_hits else -1, self._score, len(self.snippets)) - def __str__(self): - return unicode(self).encode('utf-8') + def __bytes__(self): + return str(self).encode('utf-8') @property def score(self): @@ -566,10 +626,9 @@ class SearchResult(object): 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)) + raise ValueError("this search result is for book %d; tried to merge with %d" % (self.book_id, other.book_id)) self._hits += other._hits - if other.score > self.score: - self._score = other._score + self._score += max(other._score, 0) return self def get_book(self): @@ -595,12 +654,12 @@ class SearchResult(object): # to sections and fragments frags = filter(lambda r: r[self.FRAGMENT] is not None, self._hits) - sect = filter(lambda r: r[self.FRAGMENT] is None, self._hits) + sect = [hit for hit in self._hits if hit[self.FRAGMENT] is None] # sections not covered by fragments - sect = filter(lambda s: 0 == len(filter( + sect = filter(lambda s: 0 == len(list(filter( lambda f: f[self.POSITION][self.POSITION_INDEX] <= s[self.POSITION][self.POSITION_INDEX] < - f[self.POSITION][self.POSITION_INDEX] + f[self.POSITION][self.POSITION_SPAN], frags)), sect) + f[self.POSITION][self.POSITION_INDEX] + f[self.POSITION][self.POSITION_SPAN], frags))), sect) def remove_duplicates(lst, keyfn, compare): els = {} @@ -633,7 +692,7 @@ class SearchResult(object): m.update(s[self.OTHER]) sections[si] = m - hits = sections.values() + hits = list(sections.values()) for f in frags: try: @@ -647,7 +706,7 @@ class SearchResult(object): 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) + tms = map(str.lower, tms) for qt in self.query_terms: if qt in tms: themes_hit.add(f[self.OTHER]['themes'][i]) @@ -687,13 +746,13 @@ class SearchResult(object): books[r.book_id] = r return books.values() - def __cmp__(self, other): - c = cmp(self.score, other.score) - if c == 0: - # this is inverted, because earlier date is better - return cmp(other.published_date, self.published_date) - else: - return c + def __lt__(self, other): + return (-self.score, self.published_date, self.book.sort_key_author, self.book.sort_key) > \ + (-other.score, other.published_date, other.book.sort_key_author, other.book.sort_key) + + def __eq__(self, other): + return (self.score, self.published_date, self.book.sort_key_author, self.book.sort_key) == \ + (other.score, other.published_date, other.book.sort_key_author, other.book.sort_key) def __len__(self): return len(self.hits) @@ -708,6 +767,114 @@ class SearchResult(object): return None +@total_ordering +class PictureResult(object): + def __init__(self, doc, how_found=None, query_terms=None): + self.boost = 1.0 + self.query_terms = query_terms + self._picture = None + self._hits = [] + self._processed_hits = None + + if 'score' in doc: + self._score = doc['score'] + else: + self._score = 0 + + self.picture_id = int(doc["picture_id"]) + + if doc.get('area_id'): + hit = (self._score, { + 'how_found': how_found, + 'area_id': doc['area_id'], + 'themes': doc.get('themes', []), + 'themes_pl': doc.get('themes_pl', []), + }) + + self._hits.append(hit) + + def __str__(self): + return u"" % (self.picture_id, self._score) + + def __repr__(self): + return str(self) + + @property + def score(self): + return self._score * self.boost + + def merge(self, other): + if self.picture_id != other.picture_id: + raise ValueError( + "this search result is for picture %d; tried to merge with %d" % (self.picture_id, other.picture_id)) + self._hits += other._hits + self._score += max(other._score, 0) + return self + + SCORE = 0 + OTHER = 1 + + @property + def hits(self): + if self._processed_hits is not None: + return self._processed_hits + + hits = [] + for hit in self._hits: + try: + area = picture.models.PictureArea.objects.get(id=hit[self.OTHER]['area_id']) + except picture.models.PictureArea.DoesNotExist: + # stale index + continue + # Figure out if we were searching for a token matching some word in theme name. + themes_hit = set() + if self.query_terms is not None: + for i in range(0, len(hit[self.OTHER]['themes'])): + tms = hit[self.OTHER]['themes'][i].split(r' +') + hit[self.OTHER]['themes_pl'][i].split(' ') + tms = map(str.lower, tms) + for qt in self.query_terms: + if qt in tms: + themes_hit.add(hit[self.OTHER]['themes'][i]) + break + + m = { + 'score': hit[self.SCORE], + 'area': area, + 'themes_hit': themes_hit, + } + m.update(hit[self.OTHER]) + hits.append(m) + + hits.sort(key=lambda h: h['score'], reverse=True) + hits = hits[:1] + self._processed_hits = hits + return hits + + def get_picture(self): + if self._picture is None: + self._picture = picture.models.Picture.objects.get(id=self.picture_id) + return self._picture + + picture = property(get_picture) + + @staticmethod + def aggregate(*result_lists): + books = {} + for rl in result_lists: + for r in rl: + if r.picture_id in books: + books[r.picture_id].merge(r) + else: + books[r.picture_id] = r + return books.values() + + def __lt__(self, other): + return self.score < other.score + + def __eq__(self, other): + return self.score == other.score + + class Search(SolrIndex): """ Search facilities. @@ -728,23 +895,51 @@ class Search(SolrIndex): return q - def search_words(self, words, fields, book=True): + def search_by_author(self, words): + from catalogue.models import Book + books = Book.objects.filter(parent=None).order_by('-popularity__count') + for word in words: + books = books.filter(cached_author__iregex='\m%s\M' % word).select_related('popularity__count') + return [SearchResult.from_book(book, how_found='search_by_author', query_terms=words) for book in books[:30]] + + def search_words(self, words, fields, required=None, book=True, picture=False): + if book and not picture and fields == ['authors']: + return self.search_by_author(words) filters = [] for word in words: - word_filter = None - for field in fields: - q = self.index.Q(**{field: word}) - if word_filter is None: - word_filter = q - else: - word_filter |= q - filters.append(word_filter) + if book or picture or (word not in stopwords): + word_filter = None + for field in fields: + q = self.index.Q(**{field: word}) + if word_filter is None: + word_filter = q + else: + word_filter |= q + filters.append(word_filter) + if required: + required_filter = None + for field in required: + for word in words: + if book or picture or (word not in stopwords): + q = self.index.Q(**{field: word}) + if required_filter is None: + required_filter = q + else: + required_filter |= q + filters.append(required_filter) + if not filters: + return [] + params = {} if book: - query = self.index.query(is_book=True) + params['is_book'] = True + if picture: + params['picture_id__gt'] = 0 else: - query = self.index.query() + params['book_id__gt'] = 0 + query = self.index.query(**params) query = self.apply_filters(query, filters).field_limit(score=True, all_fields=True) - return [SearchResult(found, how_found='search_words', query_terms=words) for found in query.execute()] + result_class = PictureResult if picture else SearchResult + return [result_class(found, how_found='search_words', query_terms=words) for found in query.execute()] def get_snippets(self, searchresult, query, field='text', num=1): """ @@ -767,13 +962,15 @@ class Search(SolrIndex): text = snippets.get((int(position), int(length))) snip = self.index.highlight(text=text, field=field, q=query) + if not snip and field == 'text': + snip = self.index.highlight(text=text, field='text_nonstem', q=query) if snip not in snips: snips[idx] = snip if snip: num -= 1 idx += 1 - except IOError, e: + except IOError as e: book = catalogue.models.Book.objects.filter(id=book_id) if not book: log.error("Book does not exist for book id = %d" % book_id) @@ -783,107 +980,13 @@ class Search(SolrIndex): finally: snippets.close() - # remove verse end markers.. - snips = map(lambda s: s and s.replace("/\n", "\n"), snips) + # remove verse end markers.. + snips = [s.replace("/\n", "\n") if s else s for s in snips] searchresult.snippets = snips return snips - def hint_tags(self, query, pdcounter=True, prefix=True): - """ - Return auto-complete hints for tags - using prefix search. - """ - 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) - - return self.search_tags(qu, pdcounter=pdcounter) - - def search_tags(self, query, filters=None, pdcounter=False): - """ - Search for Tag objects using query. - """ - if not filters: - filters = [] - if not pdcounter: - filters.append(~self.index.Q(is_pdcounter=True)) - res = self.apply_filters(query, filters).execute() - - tags = [] - pd_tags = [] - - for doc in res: - is_pdcounter = doc.get('is_pdcounter', False) - category = doc.get('tag_category') - try: - if is_pdcounter: - if category == 'pd_author': - tag = PDCounterAuthor.objects.get(id=doc.get('tag_id')) - else: # category == 'pd_book': - tag = PDCounterBook.objects.get(id=doc.get('tag_id')) - tag.category = 'pd_book' # make it look more lik a tag. - pd_tags.append(tag) - else: - tag = catalogue.models.Tag.objects.get(id=doc.get("tag_id")) - 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: t.slug not in tags_slugs, pd_tags) - - log.debug('search_tags: %s' % tags) - - return tags - - def hint_books(self, query, prefix=True): - """ - Returns auto-complete hints for book titles - Because we do not index 'pseudo' title-tags. - Prefix search. - """ - q = self.index.Q() - query = query.strip() - if prefix: - q |= self.index.Q(title=query + "*") - q |= self.index.Q(title_orig=query + "*") - else: - q |= self.make_term_query(query, field='title') - q |= self.make_term_query(query, field='title_orig') - qu = self.index.query(q) - only_books = self.index.Q(is_book=True) - return self.search_books(qu, [only_books]) - - def search_books(self, query, filters=None, max_results=10): - """ - Searches for Book objects using query - """ - 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 bid not in bks_found: - bks.append(catalogue.models.Book.objects.get(id=bid)) - bks_found.add(bid) - except catalogue.models.Book.DoesNotExist: - pass - return bks - @staticmethod def apply_filters(query, filters): """