pass context
[wolnelektury.git] / apps / search / index.py
index c2dfdee..df44888 100644 (file)
@@ -25,6 +25,7 @@ import re
 import errno
 from librarian import dcparser
 from librarian.parser import WLDocument
+from lxml import etree
 import catalogue.models
 from multiprocessing.pool import ThreadPool
 from threading import current_thread
@@ -212,7 +213,7 @@ class Index(BaseIndex):
 
         for tag in catalogue.models.Tag.objects.all():
             doc = Document()
-            doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(tag.id))
+            doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(int(tag.id)))
             doc.add(Field("tag_name", tag.name, Field.Store.NO, Field.Index.ANALYZED))
             doc.add(Field("tag_name_pl", tag.name, Field.Store.NO, Field.Index.ANALYZED))
             doc.add(Field("tag_category", tag.category, Field.Store.NO, Field.Index.NOT_ANALYZED))
@@ -223,9 +224,9 @@ class Index(BaseIndex):
         Create a lucene document referring book id.
         """
         doc = Document()
-        doc.add(NumericField("book_id", Field.Store.YES, True).setIntValue(book.id))
+        doc.add(NumericField("book_id", Field.Store.YES, True).setIntValue(int(book.id)))
         if book.parent is not None:
-            doc.add(NumericField("parent_id", Field.Store.YES, True).setIntValue(book.parent.id))
+            doc.add(NumericField("parent_id", Field.Store.YES, True).setIntValue(int(book.parent.id)))
         return doc
 
     def remove_book(self, book):
@@ -401,28 +402,28 @@ class Index(BaseIndex):
 
                 if header.tag in self.skip_header_tags:
                     continue
+                if header.tag is etree.Comment:
+                    continue
 
-                content = u' '.join([t for t in header.itertext()])
-                content = fix_format(content)
-
-                doc = add_part(snippets, header_index=position, header_type=header.tag, content=content)
-
-                self.index.addDocument(doc)
+                # section content
+                content = []
 
                 for start, end in walker(header):
+                        # 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}
-                        fragments[fid]['content'].append(start.tail)
+
                     elif start is not None and start.tag == 'motyw':
                         fid = start.attrib['id'][1:]
                         if start.text is not None:
                             fragments[fid]['themes'] += map(str.strip, map(give_me_utf8, start.text.split(',')))
-                        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 <end> node, skip it
+                                      #                        import pdb; pdb.set_trace()
                         frag = fragments[fid]
                         if frag['themes'] == []:
                             continue  # empty themes list.
@@ -442,12 +443,23 @@ class Index(BaseIndex):
                                        themes=frag['themes'])
 
                         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)
+
+                        # 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))))
+
+                self.index.addDocument(doc)
+
         finally:
             snippets.close()
 
@@ -541,11 +553,15 @@ class JoinSearch(object):
 
 
 class SearchResult(object):
-    def __init__(self, searcher, scoreDocs, score=None, how_found=None, snippets=None):
+    def __init__(self, searcher, 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
+            self._score = score
         else:
-            self.score = scoreDocs.score
+            self._score = scoreDocs.score
+            
+        self.boost = 1.0
 
         self._hits = []
         self.hits = None  # processed hits
@@ -563,10 +579,20 @@ class SearchResult(object):
 
         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 []})
 
         self._hits.append(hit)
 
+        self.searcher = searcher
+        self.searched = searched
+        self.tokens_cache = tokens_cache
+
+    @property
+    def score(self):
+        return self._score * self.boost
+
     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))
@@ -610,27 +636,43 @@ class SearchResult(object):
 
         # remove duplicate sections
         sections = {}
-        
+
         for s in sect:
             si = s[POSITION][POSITION_INDEX]
             # skip existing
             if si in sections:
                 if sections[si]['score'] >= s[SCORE]:
                     continue
-                
+
             m = {'score': s[SCORE],
-                 'header_index': s[POSITION][POSITION_INDEX]
+                 'section_number': s[POSITION][POSITION_INDEX] + 1,
                  }
             m.update(s[OTHER])
             sections[si] = m
-            
+
         hits = sections.values()
 
         for f in frags:
             frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT])
+
+            # Figure out if we were searching for a token matching some word in theme name.
+            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)
+                for theme in themes:
+                    name_tokens = self.searcher.get_tokens(theme.name, 'POLISH')
+                    for t in tokens:
+                        if name_tokens.index(t):
+                            if not theme in themes_hit:
+                                themes_hit.append(theme)
+                            break
+
             m = {'score': f[SCORE],
                  'fragment': frag,
-                 'themes': frag.tags.filter(category='theme')
+                 'section_number': f[POSITION][POSITION_INDEX] + 1,
+                 'themes': themes,
+                 'themes_hit': themes_hit
                  }
             m.update(f[OTHER])
             hits.append(m)
@@ -787,11 +829,14 @@ class Search(IndexStore):
             bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id")))
         return (bks, tops.totalHits)
 
-    def get_tokens(self, searched, field='content'):
+    def get_tokens(self, searched, field='content', cached=None):
         """returns tokens analyzed by a proper (for a field) analyzer
         argument can be: StringReader, string/unicode, or tokens. In the last case
         they will just be returned (so we can reuse tokens, if we don't change the analyzer)
         """
+        if cached is not None and field in cached:
+            return cached[field]
+
         if isinstance(searched, str) or isinstance(searched, unicode):
             searched = StringReader(searched)
         elif isinstance(searched, list):
@@ -803,6 +848,10 @@ class Search(IndexStore):
         while tokens.incrementToken():
             cta = tokens.getAttribute(CharTermAttribute.class_)
             toks.append(cta.toString())
+
+        if cached is not None:
+            cached[field] = toks
+
         return toks
 
     def fuzziness(self, fuzzy):
@@ -859,9 +908,39 @@ class Search(IndexStore):
             q.add(BooleanClause(term, modal))
         return q
 
-    # def content_query(self, query):
-    #     return BlockJoinQuery(query, self.parent_filter,
-    #                           BlockJoinQuery.ScoreMode.Total)
+    def search_phrase(self, searched, field, book=True, max_results=20, fuzzy=False,
+                      filters=None, tokens_cache=None, boost=None):
+        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)
+        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) 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):
+        if filters is None: filters = []
+        if tokens_cache is None: tokens_cache = {}
+
+        if book:
+            filters.append(self.term_filter(Term('is_book', 'true')))
+
+        query = BooleanQuery()
+
+        for fld in fields:
+            tokens = self.get_tokens(searched, fld, cached=tokens_cache)
+
+            query.add(BooleanClause(self.make_term_query(tokens, field=fld,
+                                fuzzy=fuzzy), BooleanClause.Occur.SHOULD))
+
+        top = self.searcher.search(query, self.chain_filters(filters), max_results)
+
+        return [SearchResult(self.searcher, found, searched=searched, tokens_cache=tokens_cache) for found in top.scoreDocs]
 
     def search_perfect_book(self, searched, max_results=20, fuzzy=False, hint=None):
         """
@@ -884,7 +963,7 @@ 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))
+                books.append(SearchResult(self.searcher, found, how_found="search_perfect_book"))
         return books
 
     def search_book(self, searched, max_results=20, fuzzy=False, hint=None):
@@ -910,13 +989,13 @@ 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))
+            books.append(SearchResult(self.searcher, found, how_found="search_book"))
 
         return books
 
     def search_perfect_parts(self, searched, max_results=20, fuzzy=False, hint=None):
         """
-        Search for book parts which containt a phrase perfectly matching (with a slop of 2, default for make_phrase())
+        Search for book parts which contains a phrase perfectly matching (with a slop of 2, default for make_phrase())
         some part/fragment of the book.
         """
         qrys = [self.make_phrase(self.get_tokens(searched), field=fld, fuzzy=fuzzy) for fld in ['content']]
@@ -932,16 +1011,17 @@ class Search(IndexStore):
                                                            flt]),
                                        max_results)
             for found in top.scoreDocs:
-                books.append(SearchResult(self.searcher, found, snippets=self.get_snippets(found, q)))
+                books.append(SearchResult(self.searcher, found, snippets=self.get_snippets(found, q), how_found='search_perfect_parts'))
 
         return books
 
-    def search_everywhere(self, searched, max_results=20, fuzzy=False, hint=None):
+    def search_everywhere(self, searched, max_results=20, fuzzy=False, hint=None, tokens_cache=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.
         """
+        if tokens_cache is None: tokens_cache = {}
         books = []
         only_in = None
 
@@ -951,8 +1031,8 @@ class Search(IndexStore):
         # content only query : themes x content
         q = BooleanQuery()
 
-        tokens_pl = self.get_tokens(searched, field='content')
-        tokens = self.get_tokens(searched, field='SIMPLE')
+        tokens_pl = self.get_tokens(searched, field='content', cached=tokens_cache)
+        tokens = self.get_tokens(searched, field='SIMPLE', cached=tokens_cache)
 
         # only search in themes when we do not already filter by themes
         if hint is None or hint.just_search_in(['themes']) != []:
@@ -964,7 +1044,7 @@ class Search(IndexStore):
 
         topDocs = self.searcher.search(q, only_in, max_results)
         for found in topDocs.scoreDocs:
-            books.append(SearchResult(self.searcher, found))
+            books.append(SearchResult(self.searcher, found, how_found='search_everywhere_themesXcontent'))
             print "* %s theme x content: %s" % (searched, books[-1]._hits)
 
         # query themes/content x author/title/tags
@@ -983,7 +1063,7 @@ class Search(IndexStore):
 
         topDocs = self.searcher.search(q, only_in, max_results)
         for found in topDocs.scoreDocs:
-            books.append(SearchResult(self.searcher, found))
+            books.append(SearchResult(self.searcher, found, how_found='search_everywhere'))
             print "* %s scatter search: %s" % (searched, books[-1]._hits)
 
         return books
@@ -1030,7 +1110,6 @@ class Search(IndexStore):
 
         # return None
 
-
     def get_snippets(self, scoreDoc, query, field='content'):
         """
         Returns a snippet for found scoreDoc.
@@ -1157,7 +1236,7 @@ class Search(IndexStore):
         Chains a filter list together
         """
         filters = filter(lambda x: x is not None, filters)
-        if not filters:
+        if not filters or filters is []:
             return None
         chf = ChainedFilter(JArray('object')(filters, Filter), op)
         return chf