X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/04b00eea4d3f41a4094151b0a5eb3416e25d2bc6..b31cc21d48f349a62a6dc74aef9d46a347fb4f0e:/apps/search/index.py diff --git a/apps/search/index.py b/apps/search/index.py index d5d696d34..60e76a263 100644 --- a/apps/search/index.py +++ b/apps/search/index.py @@ -6,13 +6,17 @@ from lucene import SimpleFSDirectory, IndexWriter, File, Field, \ SimpleAnalyzer, PolishAnalyzer, ArrayList, \ KeywordAnalyzer, NumericRangeQuery, BooleanQuery, \ BlockJoinQuery, BlockJoinCollector, TermsFilter, \ - HashSet, BooleanClause, Term + HashSet, BooleanClause, Term, CharTermAttribute, \ + PhraseQuery, StringReader # KeywordAnalyzer +import sys import os import errno from librarian import dcparser from librarian.parser import WLDocument -from catalogue.models import Book +import catalogue.models +from multiprocessing.pool import ThreadPool +import atexit class WLAnalyzer(PerFieldAnalyzerWrapper): @@ -78,6 +82,7 @@ class Index(IndexStore): def index_book(self, book, overwrite=True): if overwrite: self.remove_book(book) + doc = self.extract_metadata(book) parts = self.extract_content(book) @@ -97,6 +102,8 @@ class Index(IndexStore): 'wywiad' ] + skip_header_tags = ['autor_utworu', 'nazwa_utworu'] + def create_book_doc(self, book): """ Create a lucene document connected to the book @@ -147,6 +154,7 @@ class Index(IndexStore): if master.tag in self.master_tags: return master + def extract_content(self, book): wld = WLDocument.from_file(book.xml_file.path) root = wld.edoc.getroot() @@ -156,9 +164,13 @@ class Index(IndexStore): # header_index - the 0-indexed position of header element. # content master = self.get_master(root) + if master is None: + return [] + header_docs = [] for header, position in zip(list(master), range(len(master))): - print("header %s @%d" % (header, position)) + if header.tag in self.skip_header_tags: + continue doc = self.create_book_doc(book) doc.add(NumericField("header_index", Field.Store.YES, True).setIntValue(position)) doc.add(Field("header_type", header.tag, Field.Store.YES, Field.Index.NOT_ANALYZED)) @@ -181,7 +193,6 @@ class Index(IndexStore): # will contain (framgent id -> { content: [], themes: [] } fragments = {} for start, end in walker(master): - print("%s %s" % (start, end)) if start is not None and start.tag == 'begin': fid = start.attrib['id'][1:] fragments[fid] = {'content': [], 'themes': []} @@ -192,14 +203,26 @@ class Index(IndexStore): fragments[fid]['content'].append(start.tail) 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 frag = fragments[fid] del fragments[fid] - print("Fragment %s complete, themes: %s contents: %s" % (fid, frag['themes'], frag['content'])) + + def jstr(l): + return u' '.join(map( + lambda x: x == None and u'(none)' or unicode(x), + l)) doc = self.create_book_doc(book) - doc.add(Field("fragment_anchor", fid, Field.Store.YES, Field.Index.NOT_ANALYZED)) - doc.add(Field("content", u' '.join(filter(lambda s: s is not None, frag['content'])), Field.Store.NO, Field.Index.ANALYZED)) - doc.add(Field("themes", u' '.join(frag['themes']), Field.Store.NO, Field.Index.ANALYZED)) + doc.add(Field("fragment_anchor", fid, + Field.Store.YES, Field.Index.NOT_ANALYZED)) + doc.add(Field("content", + u' '.join(filter(lambda s: s is not None, frag['content'])), + Field.Store.NO, Field.Index.ANALYZED)) + doc.add(Field("themes", + u' '.join(filter(lambda s: s is not None, frag['themes'])), + Field.Store.NO, Field.Index.ANALYZED)) + fragment_docs.append(doc) elif start is not None: for frag in fragments.values(): @@ -218,6 +241,48 @@ class Index(IndexStore): self.close() +class ReusableIndex(Index): + """ + Works like index, but does not close/optimize Lucene index + until program exit (uses atexit hook). + This is usefull for importbooks command. + + if you cannot rely on atexit, use ReusableIndex.close_reusable() yourself. + """ + index = None + pool = None + pool_jobs = None + + def open(self, analyzer=None, threads=4): + if ReusableIndex.index is not None: + self.index = ReusableIndex.index + else: + ReusableIndex.pool = ThreadPool(threads) + ReusableIndex.pool_jobs = [] + Index.open(self, analyzer) + ReusableIndex.index = self.index + atexit.register(ReusableIndex.close_reusable) + + def index_book(self, *args, **kw): + job = ReusableIndex.pool.apply_async(Index.index_book, (self,)+ args, kw) + ReusableIndex.pool_jobs.append(job) + + @staticmethod + def close_reusable(): + import pdb; pdb.set_trace() + if ReusableIndex.index is not None: + for job in ReusableIndex.pool_jobs: + job.wait() + ReusableIndex.pool.close() + + ReusableIndex.index.optimize() + ReusableIndex.index.close() + ReusableIndex.index = None + + def close(self): + pass + + class Search(IndexStore): def __init__(self, default_field="content"): IndexStore.__init__(self) @@ -264,7 +329,7 @@ class Search(IndexStore): bks = [] for found in tops.scoreDocs: doc = self.searcher.doc(found.doc) - bks.append(Book.objects.get(id=doc.get("book_id"))) + bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) return (bks, tops.totalHits) def search(self, query, max_results=50): @@ -275,18 +340,106 @@ class Search(IndexStore): bks = [] for found in tops.scoreDocs: doc = self.searcher.doc(found.doc) - bks.append(Book.objects.get(id=doc.get("book_id"))) + bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) return (bks, tops.totalHits) def bsearch(self, query, max_results=50): q = self.query(query) - f = TermsFilter() - f.addTerm(Term("is_book", "true")) - bjq = BlockJoinQuery(q, f, BlockJoinQuery.ScoreMode.Avg) + bjq = BlockJoinQuery(q, self.parent_filter, BlockJoinQuery.ScoreMode.Avg) tops = self.searcher.search(bjq, max_results) bks = [] for found in tops.scoreDocs: doc = self.searcher.doc(found.doc) - bks.append(Book.objects.get(id=doc.get("book_id"))) + bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id"))) + return (bks, tops.totalHits) + +# TokenStream tokenStream = analyzer.tokenStream(fieldName, reader); +# OffsetAttribute offsetAttribute = tokenStream.getAttribute(OffsetAttribute.class); +# CharTermAttribute charTermAttribute = tokenStream.getAttribute(CharTermAttribute.class); + +# while (tokenStream.incrementToken()) { +# int startOffset = offsetAttribute.startOffset(); +# int endOffset = offsetAttribute.endOffset(); +# String term = charTermAttribute.toString(); +# } + + +class MultiSearch(Search): + """Class capable of IMDb-like searching""" + def get_tokens(self, queryreader): + if isinstance(queryreader, str): + queryreader = StringReader(queryreader) + queryreader.reset() + tokens = self.analyzer.reusableTokenStream('content', queryreader) + toks = [] + while tokens.incrementToken(): + cta = tokens.getAttribute(CharTermAttribute.class_) + toks.append(cta) + return toks + + def make_phrase(self, tokens, field='content', joined=False): + phrase = PhraseQuery() + for t in tokens: + term = Term(field, t) + phrase.add(term) + if joined: + phrase = self.content_query(phrase) + return phrase + + def make_term_query(self, tokens, field='content', modal=BooleanClause.Occur.SHOULD, joined=False): + q = BooleanQuery() + for t in tokens: + term = Term(field, t) + q.add(BooleanClause(term, modal)) + if joined: + self.content_query(q) + return q + + def content_query(self, query): + return BlockJoinQuery(query, self.parent_filter, + BlockJoinQuery.ScoreMode.Total) + + def multiseach(self, query, max_results=50): + """ + Search strategy: + - (phrase) OR -> content + -> title + -> author + - (keywords) -> author + -> motyw + -> tags + -> content + """ + queryreader = StringReader(query) + tokens = self.get_tokens(queryreader) + + top_level = BooleanQuery() + Should = BooleanClause.Occur.SHOULD + + phrase_level = BooleanQuery() + + 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) + kw_level.add(self.make_term_query(tokens, 'themes', joined=True), Should) + kw_level.add(self.make_term_query(tokens, 'tags'), Should) + kw_level.add(self.make_term_query(tokens, joined=True), Should) + + top_level.add(BooleanClause(phrase_level, Should)) + top_level.add(BooleanClause(kw_level, Should)) + + tops = self.searcher.search(top_level, 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)