some speedups for batch indexing
[wolnelektury.git] / apps / search / index.py
index 33836ad..12554d2 100644 (file)
@@ -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
 from multiprocessing.pool import ThreadPool
 from threading import current_thread
 import atexit
@@ -219,6 +220,15 @@ 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", 'pdcounter', Field.Store.NO, 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.
@@ -276,7 +286,7 @@ class Index(BaseIndex):
 
     footnote_tags = ['pa', 'pt', 'pr', 'pe']
 
-    skip_header_tags = ['autor_utworu', 'nazwa_utworu', 'dzielo_nadrzedne']
+    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]+)[\]. ]*$")
 
@@ -321,11 +331,13 @@ class Index(BaseIndex):
                                                (dt.year, dt.month, dt.day), Field.Store.NO, Field.Index.NOT_ANALYZED)
 
         # get published date
-        source = book_info.source_name
-        match = self.published_date_re.search(source)
-        print("published date is %s %s" % (match, match is not None and match.groups()))
-        if match is not None:
-            fields["published_date"] = Field("published_date", str(match.groups()[0]), Field.Store.YES, Field.Index.NOT_ANALYZED)
+        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
 
@@ -360,11 +372,18 @@ class Index(BaseIndex):
             return []
 
         def walker(node, ignore_tags=[]):
-            yield node, None
-            for child in filter(lambda n: n.tag not in ignore_tags, list(node)):
-                for b, e in walker(child):
-                    yield b, e
-            yield None, node
+
+            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):
@@ -426,9 +445,8 @@ 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
@@ -437,35 +455,50 @@ class Index(BaseIndex):
 
                 # section content
                 content = []
-                footnote = None
+                footnote = []
+
+                def all_content(text):
+                    for frag in fragments.values():
+                        frag['content'].append(text)
+                    content.append(text)
+                handle_text = [all_content]
+
 
-                for start, end in walker(header, ignore_tags=self.ignore_content_tags):
+                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 = ' '.join(start.itertext())
-                    elif end is not None and footnote is not None and end.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=footnote)
-
+                                       content=u''.join(footnote),
+                                       is_footnote=Field("is_footnote", 'true', Field.Store.NO, Field.Index.NOT_ANALYZED))
+                
                         self.index.addDocument(doc)
-
-                        footnote = None
-
+                        #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 <end> node, skip it
-                                      #                        import pdb; pdb.set_trace()
                         frag = fragments[fid]
                         if frag['themes'] == []:
                             continue  # empty themes list.
@@ -478,25 +511,22 @@ class Index(BaseIndex):
                                        fragment_anchor=fid,
                                        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(content))
+                #print '@ CONTENT: %s' % fix_format(content)
 
                 self.index.addDocument(doc)
-                position += 1
 
         finally:
             snippets.close()
@@ -607,27 +637,25 @@ class SearchResult(object):
         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
-
-        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")
-
         pd = stored.get("published_date")
         if pd is None:
-            print "published_date is none for book %d" % self.book_id
             pd = 0
         self.published_date = int(pd)
 
-        if snippets:
-            snippets = snippets.replace("/\n", "\n")
-        hit = (sec + (header_span,), fragment, scoreDocs.score, {'how_found': how_found, 'snippets': snippets and [snippets] or []})
+        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")
 
-        self._hits.append(hit)
+            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.search = search
         self.searched = searched
@@ -642,7 +670,7 @@ 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):
@@ -714,7 +742,6 @@ class SearchResult(object):
                 tokens = self.search.get_tokens(self.searched, 'POLISH', cached=self.tokens_cache)
                 for theme in themes:
                     name_tokens = self.search.get_tokens(theme.name, 'POLISH')
-                    print "THEME HIT: %s in %s" % (tokens, name_tokens)
                     for t in tokens:
                         if t in name_tokens:
                             if not theme in themes_hit:
@@ -967,13 +994,13 @@ 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)
@@ -981,7 +1008,7 @@ class Search(IndexStore):
         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 = {}
 
@@ -999,7 +1026,7 @@ class Search(IndexStore):
         top = self.searcher.search(query, self.chain_filters(filters), max_results)
 
         return [SearchResult(self, found, searched=searched, tokens_cache=tokens_cache,
-                             snippets=self.get_snippets(found, query)) for found in top.scoreDocs]
+                             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):
         """
@@ -1213,19 +1240,27 @@ 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')
+            if is_pdcounter:
+                tag = PDCounterAuthor.objects.get(id=doc.get('tag_id'))
+            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):
@@ -1239,7 +1274,7 @@ 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])
@@ -1265,7 +1300,7 @@ 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):
         """
         Return auto-complete hints for tags
         using prefix search.
@@ -1274,14 +1309,17 @@ 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)
             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):
         """
         Returns auto-complete hints for book titles
         Because we do not index 'pseudo' title-tags.
@@ -1289,7 +1327,10 @@ 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')
 
         return self.search_books(q, self.term_filter(Term("is_book", "true")), max_results=max_results)