X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/18a5524b295619382c9438b8de44627e3acc28b0..7e9be4e3962e1cccc43798d0507b2c82cd1d275c:/src/search/index.py diff --git a/src/search/index.py b/src/search/index.py index c84841b19..2be60fdf7 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 reduce, 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) @@ -87,7 +95,10 @@ class Snippets(object): of the snippet stored there. """ self.file.seek(pos[0], 0) - txt = self.file.read(pos[1]).decode('utf-8') + try: + txt = self.file.read(pos[1]).decode('utf-8') + except: + return '' return txt def close(self): @@ -121,7 +132,7 @@ 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 @@ -134,7 +145,8 @@ class Index(SolrIndex): uids.add(res['uid']) st += rows if uids: - self.index.delete(uids) + # FIXME: With Solr API change, this doesn't work. + #self.index.delete(uids) return True else: return False @@ -240,7 +252,8 @@ 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', 'translators', 'title']) + meta_fields = self.extract_metadata(book, book_info, dc_only=[ + 'source_name', 'authors', 'translators', 'title', 'epochs', 'kinds', 'genres']) # let's not index it - it's only used for extracting publish date if 'source_name' in meta_fields: del meta_fields['source_name'] @@ -257,8 +270,9 @@ class Index(SolrIndex): 'published_date': meta_fields['published_date'] } - if 'translators' in meta_fields: - book_fields['translators'] = meta_fields['translators'] + for tag_name in ('translators', 'epochs', 'kinds', 'genres'): + if tag_name in meta_fields: + book_fields[tag_name] = meta_fields[tag_name] self.index_content(book, book_fields=book_fields) @@ -269,14 +283,14 @@ class Index(SolrIndex): 'dramat_wierszowany_lp', 'dramat_wspolczesny', 'liryka_l', 'liryka_lp', 'wywiad', - ] + ] ignore_content_tags = [ - 'uwaga', 'extra', + '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'] @@ -295,7 +309,6 @@ 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 @@ -315,9 +328,9 @@ 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) @@ -365,8 +378,8 @@ class Index(SolrIndex): if master is None: return [] - def walker(node, ignore_tags=()): - if node.tag not in ignore_tags: + def walker(node): + if node.tag not in self.ignore_content_tags: yield node, None, None if node.text is not None: yield None, node.text, None @@ -380,16 +393,16 @@ class Index(SolrIndex): return def fix_format(text): - # separator = [u" ", u"\t", u".", u";", u","] + # separator = [" ", "\t", ".", ";", ","] if isinstance(text, list): # need to join it first text = filter(lambda s: s is not None, content) - text = u' '.join(text) + text = ' '.join(text) # for i in range(len(text)): # if i > 0: # if text[i][0] not in separator\ # and text[i - 1][-1] not in separator: - # text.insert(i, u" ") + # text.insert(i, " ") return re.sub("(?m)/$", "", text) @@ -417,17 +430,10 @@ class Index(SolrIndex): if 'themes' in fields: doc['themes'] = fields['themes'] - doc['uid'] = "part%s%s%s" % (doc['header_index'], - doc['header_span'], - doc.get('fragment_anchor', '')) + doc['uid'] = "part%s-%s-%s-%s" % ( + book.id, doc['header_index'], doc['header_span'], doc.get('fragment_anchor', '')) return doc - def give_me_utf8(s): - if isinstance(s, unicode): - return s.encode('utf-8') - else: - return s - fragments = {} snippets = Snippets(book.id).open('w') try: @@ -448,7 +454,7 @@ class Index(SolrIndex): content.append(text) handle_text = [all_content] - for start, text, end in walker(header, ignore_tags=self.ignore_content_tags): + for start, text, end in walker(header): # handle footnotes if start is not None and start.tag in self.footnote_tags: footnote = [] @@ -460,7 +466,7 @@ class Index(SolrIndex): elif end is not None and footnote is not [] and end.tag in self.footnote_tags: handle_text.pop() doc = add_part(snippets, header_index=position, header_type=header.tag, - text=u''.join(footnote), + text=''.join(footnote), is_footnote=True) self.index.add(doc) footnote = [] @@ -476,7 +482,7 @@ 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() @@ -513,10 +519,51 @@ 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=None, query_terms=None): - # self.search = search + def __init__(self, doc, how_found=None, query_terms=None): self.boost = 1.0 self._hits = [] self._processed_hits = None # processed hits @@ -557,14 +604,25 @@ class SearchResult(object): self._hits.append(hit) - def __unicode__(self): - return u"" % \ + @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 "" % \ (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): @@ -572,16 +630,18 @@ 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): if self._book is not None: return self._book - self._book = catalogue.models.Book.objects.get(id=self.book_id) + try: + self._book = catalogue.models.Book.objects.get(id=self.book_id, findable=True) + except catalogue.models.Book.DoesNotExist: + self._book = None return self._book book = property(get_book) @@ -601,27 +661,25 @@ 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): + def remove_duplicates(lst, keyfn, larger): els = {} for e in lst: eif = keyfn(e) if eif in els: - if compare(els[eif], e) >= 1: + if larger(els[eif], e): continue els[eif] = e return els.values() # remove fragments with duplicated fid's and duplicated snippets - 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])) + frags = remove_duplicates(frags, lambda f: f[self.FRAGMENT], lambda a, b: a[self.SCORE] > b[self.SCORE]) # remove duplicate sections sections = {} @@ -639,7 +697,7 @@ class SearchResult(object): m.update(s[self.OTHER]) sections[si] = m - hits = sections.values() + hits = list(sections.values()) for f in frags: try: @@ -653,19 +711,19 @@ 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]) break def theme_by_name(n): - th = filter(lambda t: t.name == n, themes) + th = list(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)) + themes_hit = list(filter(lambda a: a is not None, map(theme_by_name, themes_hit))) m = {'score': f[self.SCORE], 'fragment': frag, @@ -676,7 +734,7 @@ class SearchResult(object): m.update(f[self.OTHER]) hits.append(m) - hits.sort(lambda a, b: cmp(a['score'], b['score']), reverse=True) + hits.sort(key=lambda h: h['score'], reverse=True) self._processed_hits = hits @@ -693,13 +751,17 @@ 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 get_sort_key(self): + return (-self.score, + self.published_date, + self.book.sort_key_author if self.book else '', + self.book.sort_key if self.book else '') + + def __lt__(self, other): + return self.get_sort_key() > other.get_sort_key() + + def __eq__(self, other): + return self.get_sort_key() == other.get_sort_key() def __len__(self): return len(self.hits) @@ -714,6 +776,114 @@ 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 "" % (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. @@ -734,71 +904,51 @@ class Search(SolrIndex): return q - def search_phrase(self, searched, field='text', book=False, - filters=None, - snippets=False): - if filters is None: - filters = [] - if book: - filters.append(self.index.Q(is_book=True)) - - 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] - - def search_some(self, searched, fields, book=True, - filters=None, snippets=True, query_terms=None): - assert isinstance(fields, list) - if filters is None: - filters = [] + def search_by_author(self, words): + from catalogue.models import Book + books = Book.objects.filter(parent=None, findable=True).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: + 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: - filters.append(self.index.Q(is_book=True)) - - query = self.index.Q() - - for fld in fields: - query = self.index.Q(query | self.make_term_query(searched, fld)) - - query = self.index.query(query) + params['is_book'] = True + if picture: + params['picture_id__gt'] = 0 + else: + params['book_id__gt'] = 0 + query = self.index.query(**params) 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] - - 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. - """ - books = [] - # content only query : themes x content - q = self.make_term_query(searched, 'text') - q_themes = self.make_term_query(searched, 'themes_pl') - - query = self.index.query(q).query(q_themes).field_limit(score=True, all_fields=True) - res = query.execute() - - for found in res: - books.append(SearchResult(found, how_found='search_everywhere_themesXcontent', query_terms=query_terms)) - - # query themes/content x author/title/tags - in_content = self.index.Q() - in_meta = self.index.Q() - - for fld in ['themes_pl', 'text']: - in_content |= self.make_term_query(searched, field=fld) - - for fld in ['tags', 'authors', 'title']: - in_meta |= self.make_term_query(searched, field=fld) - - q = in_content & in_meta - res = self.index.query(q).field_limit(score=True, all_fields=True).execute() - - for found in res: - books.append(SearchResult(found, how_found='search_everywhere', query_terms=query_terms)) - - return books + 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): """ @@ -821,122 +971,31 @@ class Search(SolrIndex): text = snippets.get((int(position), int(length))) snip = self.index.highlight(text=text, field=field, q=query) - snips[idx] = snip - if snip: - num -= 1 + 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: - book = catalogue.models.Book.objects.get(id=book_id) - if not book.children.exists(): + except IOError as e: + book = catalogue.models.Book.objects.filter(id=book_id, findable=True) + if not book: + log.error("Book does not exist for book id = %d" % book_id) + elif not book.get().children.exists(): log.error("Cannot open snippet file for book id = %d [rev=%s], %s" % (book_id, revision, e)) return [] 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')) - 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: - # WTF - 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")) - 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 + "*") - else: - 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]) - - 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): """