Merge branch 'pretty' of github.com:fnp/wolnelektury into pretty

[wolnelektury.git] / apps / search / index.py

diff --git a/apps/search/index.py b/apps/search/index.py
index 77ce877..307376d 100644 (file)
--- a/apps/search/index.py
+++ b/apps/search/index.py
@@ -25,6 +25,7 @@ import re
 import errno
 from librarian import dcparser
 from librarian.parser import WLDocument
 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
 import catalogue.models
 from multiprocessing.pool import ThreadPool
 from threading import current_thread


@@ -54,6 +55,8 @@ class WLAnalyzer(PerFieldAnalyzerWrapper):
         self.addAnalyzer("source_name", simple)
         self.addAnalyzer("publisher", simple)
         self.addAnalyzer("authors", simple)
         self.addAnalyzer("source_name", simple)
         self.addAnalyzer("publisher", simple)
         self.addAnalyzer("authors", simple)

+        self.addAnalyzer("title", simple)
+

         self.addAnalyzer("is_book", keyword)
         # shouldn't the title have two forms? _pl and simple?
 
         self.addAnalyzer("is_book", keyword)
         # shouldn't the title have two forms? _pl and simple?
 


@@ -210,7 +213,7 @@ class Index(BaseIndex):
 
         for tag in catalogue.models.Tag.objects.all():
             doc = Document()
 
         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))
             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))


@@ -221,9 +224,9 @@ class Index(BaseIndex):
         Create a lucene document referring book id.
         """
         doc = Document()
         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:
         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):
         return doc
 
     def remove_book(self, book):


@@ -399,28 +402,28 @@ class Index(BaseIndex):
 
                 if header.tag in self.skip_header_tags:
                     continue
 
                 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):
 
                 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}
                     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(',')))
                     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
                     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.
                         frag = fragments[fid]
                         if frag['themes'] == []:
                             continue  # empty themes list.


@@ -440,12 +443,23 @@ class Index(BaseIndex):
                                        themes=frag['themes'])
 
                         self.index.addDocument(doc)
                                        themes=frag['themes'])
 
                         self.index.addDocument(doc)

+
+                        # Collect content.

                     elif start is not None:
                         for frag in fragments.values():
                             frag['content'].append(start.text)
                     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)
                     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()
 
         finally:
             snippets.close()
 


@@ -540,14 +554,13 @@ class JoinSearch(object):
 
 class SearchResult(object):
     def __init__(self, searcher, scoreDocs, score=None, how_found=None, snippets=None):
 
 class SearchResult(object):
     def __init__(self, searcher, scoreDocs, score=None, how_found=None, snippets=None):

-        self.snippets = []
-

         if score:
             self.score = score
         else:
             self.score = scoreDocs.score
 
         if score:
             self.score = score
         else:
             self.score = scoreDocs.score
 

-        self.hits = []
+        self._hits = []
+        self.hits = None  # processed hits

 
         stored = searcher.doc(scoreDocs.doc)
         self.book_id = int(stored.get("book_id"))
 
         stored = searcher.doc(scoreDocs.doc)
         self.book_id = int(stored.get("book_id"))


@@ -562,14 +575,16 @@ class SearchResult(object):
 
         fragment = stored.get("fragment_anchor")
 
 
         fragment = stored.get("fragment_anchor")
 

-        hit = (sec + (header_span,), fragment, scoreDocs.score, {'how_found': how_found, 'snippets': [snippets]})
+        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)

 
     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))
 
     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))

-        self.hits += other.hits
+        self._hits += other._hits

         if other.score > self.score:
             self.score = other.score
         return self
         if other.score > self.score:
             self.score = other.score
         return self


@@ -580,35 +595,66 @@ class SearchResult(object):
     book = property(get_book)
 
     def process_hits(self):
     book = property(get_book)
 
     def process_hits(self):

-        frags = filter(lambda r: r[1] is not None, self.hits)
-        sect = filter(lambda r: r[1] is None, self.hits)
+        POSITION = 0
+        FRAGMENT = 1
+        POSITION_INDEX = 1
+        POSITION_SPAN = 2
+        SCORE = 2
+        OTHER = 3
+
+        # to sections and fragments
+        frags = filter(lambda r: r[FRAGMENT] is not None, self._hits)
+        sect = filter(lambda r: r[FRAGMENT] is None, self._hits)

         sect = filter(lambda s: 0 == len(filter(
         sect = filter(lambda s: 0 == len(filter(

-            lambda f: s[0][1] >= f[0][1] and s[0][1] < f[0][1] + f[0][2],
+            lambda f: s[POSITION][POSITION_INDEX] >= f[POSITION][POSITION_INDEX]
+            and s[POSITION][POSITION_INDEX] < f[POSITION][POSITION_INDEX] + f[POSITION][POSITION_SPAN],

             frags)), sect)
 
         hits = []
 
             frags)), sect)
 
         hits = []
 

+        # remove duplicate fragments
+        fragments = {}
+        for f in frags:
+            fid = f[FRAGMENT]
+            if fid in fragments:
+                if fragments[fid][SCORE] >= f[SCORE]:
+                    continue
+            fragments[fid] = f
+        frags = fragments.values()
+
+        # remove duplicate sections
+        sections = {}
+

         for s in sect:
         for s in sect:

-            m = {'score': s[2],
-                 'header_index': s[0][1]
+            si = s[POSITION][POSITION_INDEX]
+            # skip existing
+            if si in sections:
+                if sections[si]['score'] >= s[SCORE]:
+                    continue
+
+            m = {'score': s[SCORE],
+                 'section_number': s[POSITION][POSITION_INDEX] + 1,

                  }
                  }

-            m.update(s[3])
-            hits.append(m)
+            m.update(s[OTHER])
+            sections[si] = m
+
+        hits = sections.values()

 
         for f in frags:
 
         for f in frags:

-            frag = catalogue.models.Fragment.objects.get(anchor=f[1])
-            m = {'score': f[2],
+            frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT])
+            m = {'score': f[SCORE],

                  'fragment': frag,
                  'fragment': frag,

+                 'section_number': f[POSITION][POSITION_INDEX] + 1,

                  'themes': frag.tags.filter(category='theme')
                  }
                  'themes': frag.tags.filter(category='theme')
                  }

-            m.update(f[3])
+            m.update(f[OTHER])

             hits.append(m)
 
         hits.sort(lambda a, b: cmp(a['score'], b['score']), reverse=True)
 
             hits.append(m)
 
         hits.sort(lambda a, b: cmp(a['score'], b['score']), reverse=True)
 

-        print("--- %s" % hits)
+        self.hits = hits

 
 

-        return hits
+        return self

 
     def __unicode__(self):
         return u'SearchResult(book_id=%d, score=%d)' % (self.book_id, self.score)
 
     def __unicode__(self):
         return u'SearchResult(book_id=%d, score=%d)' % (self.book_id, self.score)


@@ -660,7 +706,7 @@ class Hint(object):
                 lst = self.book_tags.get(t.category, [])
                 lst.append(t)
                 self.book_tags[t.category] = lst
                 lst = self.book_tags.get(t.category, [])
                 lst.append(t)
                 self.book_tags[t.category] = lst

-            if t.category in ['theme']:
+            if t.category in ['theme', 'theme_pl']:

                 self.part_tags.append(t)
 
     def tag_filter(self, tags, field='tags'):
                 self.part_tags.append(t)
 
     def tag_filter(self, tags, field='tags'):


@@ -853,7 +899,34 @@ class Search(IndexStore):
                 self.chain_filters([only_in, self.term_filter(Term('is_book', 'true'))]),
                 max_results)
             for found in top.scoreDocs:
                 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):
+        fields_to_search = ['tags', 'authors', 'title']
+
+        only_in = None
+        if hint:
+            if not hint.should_search_for_book():
+                return []
+            fields_to_search = hint.just_search_in(fields_to_search)
+            only_in = hint.book_filter()
+
+        tokens = self.get_tokens(searched, field='SIMPLE')
+
+        q = BooleanQuery()
+
+        for fld in fields_to_search:
+            q.add(BooleanClause(self.make_term_query(tokens, field=fld,
+                                fuzzy=fuzzy), BooleanClause.Occur.SHOULD))
+
+        books = []
+        top = self.searcher.search(q,
+                                   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"))
+

         return books
 
     def search_perfect_parts(self, searched, max_results=20, fuzzy=False, hint=None):
         return books
 
     def search_perfect_parts(self, searched, max_results=20, fuzzy=False, hint=None):


@@ -874,7 +947,7 @@ class Search(IndexStore):
                                                            flt]),
                                        max_results)
             for found in top.scoreDocs:
                                                            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
 
 
         return books
 


@@ -893,29 +966,40 @@ class Search(IndexStore):
         # content only query : themes x content
         q = BooleanQuery()
 
         # content only query : themes x content
         q = BooleanQuery()
 

-        tokens = self.get_tokens(searched)
-        if hint is None or hint.just_search_in(['themes_pl']) != []:
-            q.add(BooleanClause(self.make_term_query(tokens, field='themes_pl',
+        tokens_pl = self.get_tokens(searched, field='content')
+        tokens = self.get_tokens(searched, field='SIMPLE')
+
+        # only search in themes when we do not already filter by themes
+        if hint is None or hint.just_search_in(['themes']) != []:
+            q.add(BooleanClause(self.make_term_query(tokens_pl, field='themes_pl',

                                                      fuzzy=fuzzy), BooleanClause.Occur.MUST))
 
                                                      fuzzy=fuzzy), BooleanClause.Occur.MUST))
 

-        q.add(BooleanClause(self.make_term_query(tokens, field='content',
+        q.add(BooleanClause(self.make_term_query(tokens_pl, field='content',

                                                  fuzzy=fuzzy), BooleanClause.Occur.SHOULD))
 
         topDocs = self.searcher.search(q, only_in, max_results)
         for found in topDocs.scoreDocs:
                                                  fuzzy=fuzzy), BooleanClause.Occur.SHOULD))
 
         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
         q = BooleanQuery()
 
         # query themes/content x author/title/tags
         q = BooleanQuery()

-        #        in_meta = BooleanQuery()

         in_content = BooleanQuery()
         in_content = BooleanQuery()

+        in_meta = BooleanQuery()
+
+        for fld in ['themes_pl', 'content']:
+            in_content.add(BooleanClause(self.make_term_query(tokens_pl, field=fld, fuzzy=False), BooleanClause.Occur.SHOULD))

 
 

-        for fld in ['themes', 'content', 'tags', 'authors', 'title']:
-            in_content.add(BooleanClause(self.make_term_query(tokens, field=fld, fuzzy=False), BooleanClause.Occur.SHOULD))
+        for fld in ['tags', 'authors', 'title']:
+            in_meta.add(BooleanClause(self.make_term_query(tokens, field=fld, fuzzy=False), BooleanClause.Occur.SHOULD))
+
+        q.add(BooleanClause(in_content, BooleanClause.Occur.MUST))
+        q.add(BooleanClause(in_meta, BooleanClause.Occur.SHOULD))

 
         topDocs = self.searcher.search(q, only_in, max_results)
         for found in topDocs.scoreDocs:
 
         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
 
 
         return books
 


@@ -961,7 +1045,6 @@ class Search(IndexStore):
 
         # return None
 
 
         # return None
 

-

     def get_snippets(self, scoreDoc, query, field='content'):
         """
         Returns a snippet for found scoreDoc.
     def get_snippets(self, scoreDoc, query, field='content'):
         """
         Returns a snippet for found scoreDoc.


@@ -981,7 +1064,6 @@ class Search(IndexStore):
 
         tokenStream = TokenSources.getAnyTokenStream(self.searcher.getIndexReader(), scoreDoc.doc, field, self.analyzer)
         #  highlighter.getBestTextFragments(tokenStream, text, False, 10)
 
         tokenStream = TokenSources.getAnyTokenStream(self.searcher.getIndexReader(), scoreDoc.doc, field, self.analyzer)
         #  highlighter.getBestTextFragments(tokenStream, text, False, 10)

-        #        import pdb; pdb.set_trace()

         snip = highlighter.getBestFragments(tokenStream, text, 3, "...")
 
         return snip
         snip = highlighter.getBestFragments(tokenStream, text, 3, "...")
 
         return snip