X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/e0e33e24cf53aa1b65e08bcabcc674529d09741a..6f8ce976c498b1df0768c0965c0c9dc2e84bda1c:/apps/search/index.py diff --git a/apps/search/index.py b/apps/search/index.py index adb76797e..dc33575bc 100644 --- a/apps/search/index.py +++ b/apps/search/index.py @@ -1,7 +1,7 @@ # -*- coding: utf-8 -*- from django.conf import settings -from lucene import SimpleFSDirectory, IndexWriter, CheckIndex, \ +from lucene import SimpleFSDirectory, IndexWriter, IndexWriterConfig, CheckIndex, \ File, Field, Integer, \ NumericField, Version, Document, JavaError, IndexSearcher, \ QueryParser, PerFieldAnalyzerWrapper, \ @@ -27,6 +27,7 @@ from librarian import dcparser from librarian.parser import WLDocument from lxml import etree import catalogue.models +from pdcounter.models import Author as PDCounterAuthor, BookStub as PDCounterBook from multiprocessing.pool import ThreadPool from threading import current_thread import atexit @@ -168,11 +169,13 @@ class BaseIndex(IndexStore): analyzer = WLAnalyzer() self.analyzer = analyzer - def open(self, analyzer=None): + def open(self, timeout=None): if self.index: raise Exception("Index is already opened") - self.index = IndexWriter(self.store, self.analyzer,\ - IndexWriter.MaxFieldLength.LIMITED) + conf = IndexWriterConfig(Version.LUCENE_34, self.analyzer) + if timeout: + conf.setWriteLockTimeout(long(timeout)) + self.index = IndexWriter(self.store, conf) return self.index def optimize(self): @@ -219,6 +222,24 @@ class Index(BaseIndex): doc.add(Field("tag_category", tag.category, Field.Store.NO, Field.Index.NOT_ANALYZED)) self.index.addDocument(doc) + for pdtag in PDCounterAuthor.objects.all(): + doc = Document() + doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(int(pdtag.id))) + doc.add(Field("tag_name", pdtag.name, Field.Store.NO, Field.Index.ANALYZED)) + doc.add(Field("tag_name_pl", pdtag.name, Field.Store.NO, Field.Index.ANALYZED)) + doc.add(Field("tag_category", 'pd_author', Field.Store.YES, Field.Index.NOT_ANALYZED)) + doc.add(Field("is_pdcounter", 'true', Field.Store.YES, Field.Index.NOT_ANALYZED)) + self.index.addDocument(doc) + + for pdtag in PDCounterBook.objects.all(): + doc = Document() + doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(int(pdtag.id))) + doc.add(Field("tag_name", pdtag.title, Field.Store.NO, Field.Index.ANALYZED)) + doc.add(Field("tag_name_pl", pdtag.title, Field.Store.NO, Field.Index.ANALYZED)) + doc.add(Field("tag_category", 'pd_book', Field.Store.YES, Field.Index.NOT_ANALYZED)) + doc.add(Field("is_pdcounter", 'true', Field.Store.YES, Field.Index.NOT_ANALYZED)) + self.index.addDocument(doc) + def create_book_doc(self, book): """ Create a lucene document referring book id. @@ -252,11 +273,10 @@ class Index(BaseIndex): book_doc.add(elem) else: book_doc.add(f) - self.index.addDocument(book_doc) del book_doc - self.index_content(book, book_fields=[meta_fields['title'], meta_fields['authors']]) + self.index_content(book, book_fields=[meta_fields['title'], meta_fields['authors'], meta_fields['published_date']]) master_tags = [ 'opowiadanie', @@ -264,10 +284,21 @@ class Index(BaseIndex): 'dramat_wierszowany_l', 'dramat_wierszowany_lp', 'dramat_wspolczesny', 'liryka_l', 'liryka_lp', - 'wywiad' + 'wywiad', ] - skip_header_tags = ['autor_utworu', 'nazwa_utworu', 'dzielo_nadrzedne'] + ignore_content_tags = [ + 'uwaga', 'extra', + 'zastepnik_tekstu', 'sekcja_asterysk', 'separator_linia', 'zastepnik_wersu', + 'didaskalia', + 'naglowek_aktu', 'naglowek_sceny', 'naglowek_czesc', + ] + + footnote_tags = ['pa', 'pt', 'pr', 'pe'] + + skip_header_tags = ['autor_utworu', 'nazwa_utworu', 'dzielo_nadrzedne', '{http://www.w3.org/1999/02/22-rdf-syntax-ns#}RDF'] + + published_date_re = re.compile("([0-9]+)[\]. ]*$") def extract_metadata(self, book, book_info=None): """ @@ -309,6 +340,15 @@ class Index(BaseIndex): fields[field.name] = Field(field.name, "%04d%02d%02d" %\ (dt.year, dt.month, dt.day), Field.Store.NO, Field.Index.NOT_ANALYZED) + # get published date + pd = None + if hasattr(book_info, 'source_name') and book_info.source_name: + match = self.published_date_re.search(book_info.source_name) + if match is not None: + pd = str(match.groups()[0]) + if not pd: pd = "" + fields["published_date"] = Field("published_date", pd, Field.Store.YES, Field.Index.NOT_ANALYZED) + return fields def add_gaps(self, fields, fieldname): @@ -341,15 +381,33 @@ class Index(BaseIndex): if master is None: return [] - def walker(node): - yield node, None - for child in list(node): - for b, e in walker(child): - yield b, e - yield None, node + def walker(node, ignore_tags=[]): + + if node.tag not in ignore_tags: + yield node, None, None + if node.text is not None: + yield None, node.text, None + for child in list(node): + for b, t, e in walker(child): + yield b, t, e + yield None, None, node + + if node.tail is not None: + yield None, node.tail, None return def fix_format(text): + # separator = [u" ", u"\t", u".", u";", u","] + if isinstance(text, list): + # need to join it first + text = filter(lambda s: s is not None, content) + text = u' '.join(text) + # for i in range(len(text)): + # if i > 0: + # if text[i][0] not in separator\ + # and text[i - 1][-1] not in separator: + # text.insert(i, u" ") + return re.sub("(?m)/$", "", text) def add_part(snippets, **fields): @@ -397,9 +455,8 @@ class Index(BaseIndex): fragments = {} snippets = Snippets(book.id).open('w') - position = 0 try: - for header in list(master): + for header, position in zip(list(master), range(len(master))): if header.tag in self.skip_header_tags: continue @@ -408,59 +465,78 @@ class Index(BaseIndex): # section content content = [] - - for start, end in walker(header): - # handle fragments and themes. + footnote = [] + + def all_content(text): + for frag in fragments.values(): + frag['content'].append(text) + content.append(text) + handle_text = [all_content] + + + for start, text, end in walker(header, ignore_tags=self.ignore_content_tags): + # handle footnotes + if start is not None and start.tag in self.footnote_tags: + footnote = [] + def collect_footnote(t): + footnote.append(t) + handle_text.append(collect_footnote) + 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, + content=u''.join(footnote), + is_footnote=Field("is_footnote", 'true', Field.Store.NO, Field.Index.NOT_ANALYZED)) + + self.index.addDocument(doc) + #print "@ footnote text: %s" % footnote + footnote = [] + + # handle fragments and themes. if start is not None and start.tag == 'begin': fid = start.attrib['id'][1:] fragments[fid] = {'content': [], 'themes': [], 'start_section': position, 'start_header': header.tag} + # themes for this fragment elif start is not None and start.tag == 'motyw': fid = start.attrib['id'][1:] + handle_text.append(None) if start.text is not None: fragments[fid]['themes'] += map(str.strip, map(give_me_utf8, start.text.split(','))) + elif end is not None and end.tag == 'motyw': + handle_text.pop() elif start is not None and start.tag == 'end': fid = start.attrib['id'][1:] if fid not in fragments: continue # a broken node, skip it - # import pdb; pdb.set_trace() frag = fragments[fid] if frag['themes'] == []: continue # empty themes list. del fragments[fid] - def jstr(l): - return u' '.join(map( - lambda x: x == None and u'(none)' or unicode(x), - l)) - doc = add_part(snippets, header_type=frag['start_header'], header_index=frag['start_section'], header_span=position - frag['start_section'] + 1, fragment_anchor=fid, - content=u' '.join(filter(lambda s: s is not None, frag['content'])), + content=fix_format(frag['content']), themes=frag['themes']) - + #print '@ FRAG %s' % frag['content'] self.index.addDocument(doc) # Collect content. - elif start is not None: - for frag in fragments.values(): - frag['content'].append(start.text) - content.append(start.text) - elif end is not None: - for frag in fragments.values(): - frag['content'].append(end.tail) - content.append(end.tail) + + if text is not None and handle_text is not []: + hdl = handle_text[-1] + if hdl is not None: + hdl(text) # in the end, add a section text. doc = add_part(snippets, header_index=position, header_type=header.tag, - content=fix_format(u' '.join(filter(lambda s: s is not None, content)))) + content=fix_format(content)) + #print '@ CONTENT: %s' % fix_format(content) self.index.addDocument(doc) - position += 1 finally: snippets.close() @@ -487,12 +563,12 @@ class ReusableIndex(Index): """ index = None - def open(self, analyzer=None, threads=4): - if ReusableIndex.index is not None: + def open(self, analyzer=None, **kw): + if ReusableIndex.index: self.index = ReusableIndex.index else: print("opening index") - Index.open(self, analyzer) + Index.open(self, analyzer, **kw) ReusableIndex.index = self.index atexit.register(ReusableIndex.close_reusable) @@ -502,13 +578,15 @@ class ReusableIndex(Index): @staticmethod def close_reusable(): - if ReusableIndex.index is not None: + if ReusableIndex.index: + print("closing index") ReusableIndex.index.optimize() ReusableIndex.index.close() ReusableIndex.index = None def close(self): - pass + if ReusableIndex.index: + ReusableIndex.index.commit() class JoinSearch(object): @@ -555,39 +633,44 @@ class JoinSearch(object): class SearchResult(object): - def __init__(self, searcher, scoreDocs, score=None, how_found=None, snippets=None, searched=None, tokens_cache=None): + def __init__(self, search, 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 else: self._score = scoreDocs.score - + self.boost = 1.0 self._hits = [] self._processed_hits = None # processed hits - stored = searcher.doc(scoreDocs.doc) + stored = search.searcher.doc(scoreDocs.doc) self.book_id = int(stored.get("book_id")) - header_type = stored.get("header_type") - if not header_type: - return + pd = stored.get("published_date") + try: + self.published_date = int(pd) + except ValueError: + self.published_date = 0 - sec = (header_type, int(stored.get("header_index"))) - header_span = stored.get('header_span') - header_span = header_span is not None and int(header_span) or 1 + header_type = stored.get("header_type") + # we have a content hit in some header of fragment + if header_type is not None: + sec = (header_type, int(stored.get("header_index"))) + header_span = stored.get('header_span') + header_span = header_span is not None and int(header_span) or 1 - fragment = stored.get("fragment_anchor") + fragment = stored.get("fragment_anchor") - if snippets: - snippets = snippets.replace("/\n", "\n") - hit = (sec + (header_span,), fragment, scoreDocs.score, {'how_found': how_found, 'snippets': snippets and [snippets] or []}) + if snippets: + snippets = snippets.replace("/\n", "\n") + hit = (sec + (header_span,), fragment, scoreDocs.score, {'how_found': how_found, 'snippets': snippets and [snippets] or []}) - self._hits.append(hit) + self._hits.append(hit) - self.searcher = searcher + self.search = search self.searched = searched self.tokens_cache = tokens_cache @@ -600,7 +683,7 @@ class SearchResult(object): raise ValueError("this search result is or 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 = other._score return self def get_book(self): @@ -660,7 +743,7 @@ class SearchResult(object): for f in frags: try: - frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT]) + frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT], book__id=self.book_id) except catalogue.models.Fragment.DoesNotExist: # stale index continue @@ -669,11 +752,11 @@ class SearchResult(object): 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) + tokens = self.search.get_tokens(self.searched, 'POLISH', cached=self.tokens_cache) for theme in themes: - name_tokens = self.searcher.get_tokens(theme.name, 'POLISH') + name_tokens = self.search.get_tokens(theme.name, 'POLISH') for t in tokens: - if name_tokens.index(t): + if t in name_tokens: if not theme in themes_hit: themes_hit.append(theme) break @@ -709,7 +792,12 @@ class SearchResult(object): return books.values() def __cmp__(self, other): - return cmp(self.score, other.score) + 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 class Hint(object): @@ -919,21 +1007,21 @@ class Search(IndexStore): return q def search_phrase(self, searched, field, book=True, max_results=20, fuzzy=False, - filters=None, tokens_cache=None, boost=None, snippets=False): + filters=None, tokens_cache=None, boost=None, snippets=False, slop=2): 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) + 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.searcher, found, snippets=(snippets and self.get_snippets(found, query) or None)) for found in top.scoreDocs] + 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): + filters=None, tokens_cache=None, boost=None, snippets=True): if filters is None: filters = [] if tokens_cache is None: tokens_cache = {} @@ -950,7 +1038,8 @@ class Search(IndexStore): top = self.searcher.search(query, self.chain_filters(filters), max_results) - return [SearchResult(self.searcher, found, searched=searched, tokens_cache=tokens_cache, snippets=self.get_snippets(found, query)) for found in top.scoreDocs] + 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): """ @@ -973,7 +1062,7 @@ class Search(IndexStore): self.chain_filters([only_in, self.term_filter(Term('is_book', 'true'))]), max_results) for found in top.scoreDocs: - books.append(SearchResult(self.searcher, found, how_found="search_perfect_book")) + books.append(SearchResult(self, found, how_found="search_perfect_book")) return books def search_book(self, searched, max_results=20, fuzzy=False, hint=None): @@ -999,7 +1088,7 @@ class Search(IndexStore): self.chain_filters([only_in, self.term_filter(Term('is_book', 'true'))]), max_results) for found in top.scoreDocs: - books.append(SearchResult(self.searcher, found, how_found="search_book")) + books.append(SearchResult(self, found, how_found="search_book")) return books @@ -1021,7 +1110,7 @@ class Search(IndexStore): flt]), max_results) for found in top.scoreDocs: - books.append(SearchResult(self.searcher, found, snippets=self.get_snippets(found, q), how_found='search_perfect_parts')) + books.append(SearchResult(self, found, snippets=self.get_snippets(found, q), how_found='search_perfect_parts')) return books @@ -1054,7 +1143,7 @@ class Search(IndexStore): topDocs = self.searcher.search(q, only_in, max_results) for found in topDocs.scoreDocs: - books.append(SearchResult(self.searcher, found, how_found='search_everywhere_themesXcontent')) + books.append(SearchResult(self, found, how_found='search_everywhere_themesXcontent', searched=searched)) print "* %s theme x content: %s" % (searched, books[-1]._hits) # query themes/content x author/title/tags @@ -1073,7 +1162,7 @@ class Search(IndexStore): topDocs = self.searcher.search(q, only_in, max_results) for found in topDocs.scoreDocs: - books.append(SearchResult(self.searcher, found, how_found='search_everywhere')) + books.append(SearchResult(self, found, how_found='search_everywhere', searched=searched)) print "* %s scatter search: %s" % (searched, books[-1]._hits) return books @@ -1134,17 +1223,25 @@ class Search(IndexStore): if position is None or length is None: return None # locate content. - snippets = Snippets(stored.get('book_id')).open() + book_id = int(stored.get('book_id')) + snippets = Snippets(book_id).open() try: - text = snippets.get((int(position), - int(length))) - finally: - snippets.close() + try: + text = snippets.get((int(position), + int(length))) + finally: + snippets.close() - tokenStream = TokenSources.getAnyTokenStream(self.searcher.getIndexReader(), scoreDoc.doc, field, self.analyzer) - # highlighter.getBestTextFragments(tokenStream, text, False, 10) - snip = highlighter.getBestFragments(tokenStream, text, 3, "...") + tokenStream = TokenSources.getAnyTokenStream(self.searcher.getIndexReader(), scoreDoc.doc, field, self.analyzer) + # highlighter.getBestTextFragments(tokenStream, text, False, 10) + snip = highlighter.getBestFragments(tokenStream, text, 3, "...") + 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 @staticmethod @@ -1164,19 +1261,34 @@ class Search(IndexStore): if terms: return JArray('object')(terms, Term) - def search_tags(self, query, filter=None, max_results=40): + def search_tags(self, query, filters=None, max_results=40, pdcounter=False): """ Search for Tag objects using query. """ - tops = self.searcher.search(query, filter, max_results) + if not pdcounter: + filters = self.chain_filters([filter, self.term_filter(Term('is_pdcounter', 'true'), inverse=True)]) + tops = self.searcher.search(query, filters, max_results) tags = [] for found in tops.scoreDocs: doc = self.searcher.doc(found.doc) - tag = catalogue.models.Tag.objects.get(id=doc.get("tag_id")) - tags.append(tag) - print "%s (%d) -> %f" % (tag, tag.id, found.score) - + is_pdcounter = doc.get('is_pdcounter') + category = doc.get('tag_category') + 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) + else: + tag = catalogue.models.Tag.objects.get(id=doc.get("tag_id")) + # don't add the pdcounter tag if same tag already exists + if not (is_pdcounter and filter(lambda t: tag.slug == t.slug, tags)): + tags.append(tag) + # print "%s (%d) -> %f" % (tag, tag.id, found.score) + print 'returning %s' % tags return tags def search_books(self, query, filter=None, max_results=10): @@ -1190,7 +1302,7 @@ class Search(IndexStore): bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) return bks - def create_prefix_phrase(self, toks, field): + def make_prefix_phrase(self, toks, field): q = MultiPhraseQuery() for i in range(len(toks)): t = Term(field, toks[i]) @@ -1216,7 +1328,7 @@ class Search(IndexStore): return only_term - def hint_tags(self, string, max_results=50): + def hint_tags(self, string, max_results=50, pdcounter=True, prefix=True, fuzzy=False): """ Return auto-complete hints for tags using prefix search. @@ -1225,14 +1337,17 @@ class Search(IndexStore): top = BooleanQuery() for field in ['tag_name', 'tag_name_pl']: - q = self.create_prefix_phrase(toks, field) + 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) + return self.search_tags(top, no_book_cat, max_results=max_results, pdcounter=pdcounter) - def hint_books(self, string, max_results=50): + def hint_books(self, string, max_results=50, prefix=True, fuzzy=False): """ Returns auto-complete hints for book titles Because we do not index 'pseudo' title-tags. @@ -1240,7 +1355,10 @@ class Search(IndexStore): """ toks = self.get_tokens(string, field='SIMPLE') - q = self.create_prefix_phrase(toks, 'title') + if prefix: + q = self.make_prefix_phrase(toks, 'title') + 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)