basic fragment choosing

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

diff --git a/apps/search/index.py b/apps/search/index.py
index 10069e4..9972c2c 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?
 


@@ -71,6 +74,11 @@ class WLAnalyzer(PerFieldAnalyzerWrapper):
 
 
 class IndexStore(object):
 
 
 class IndexStore(object):

+    """
+    Provides access to search index.
+
+    self.store - lucene index directory
+    """

     def __init__(self):
         self.make_index_dir()
         self.store = SimpleFSDirectory(File(settings.SEARCH_INDEX))
     def __init__(self):
         self.make_index_dir()
         self.store = SimpleFSDirectory(File(settings.SEARCH_INDEX))


@@ -95,6 +103,11 @@ class IndexChecker(IndexStore):
 
 
 class Snippets(object):
 
 
 class Snippets(object):

+    """
+    This class manages snippet files for indexed object (book)
+    the snippets are concatenated together, and their positions and
+    lengths are kept in lucene index fields.
+    """

     SNIPPET_DIR = "snippets"
 
     def __init__(self, book_id):
     SNIPPET_DIR = "snippets"
 
     def __init__(self, book_id):


@@ -108,6 +121,9 @@ class Snippets(object):
         self.file = None
 
     def open(self, mode='r'):
         self.file = None
 
     def open(self, mode='r'):

+        """
+        Open the snippet file. Call .close() afterwards.
+        """

         if not 'b' in mode:
             mode += 'b'
         self.file = open(os.path.join(settings.SEARCH_INDEX, self.SNIPPET_DIR, str(self.book_id)), mode)
         if not 'b' in mode:
             mode += 'b'
         self.file = open(os.path.join(settings.SEARCH_INDEX, self.SNIPPET_DIR, str(self.book_id)), mode)


@@ -115,6 +131,10 @@ class Snippets(object):
         return self
 
     def add(self, snippet):
         return self
 
     def add(self, snippet):

+        """
+        Append a snippet (unicode) to the snippet file.
+        Return a (position, length) tuple
+        """

         txt = snippet.encode('utf-8')
         l = len(txt)
         self.file.write(txt)
         txt = snippet.encode('utf-8')
         l = len(txt)
         self.file.write(txt)


@@ -123,17 +143,26 @@ class Snippets(object):
         return pos
 
     def get(self, pos):
         return pos
 
     def get(self, pos):

+        """
+        Given a tuple of (position, length) return an unicode
+        of the snippet stored there.
+        """

         self.file.seek(pos[0], 0)
         txt = self.file.read(pos[1]).decode('utf-8')
         return txt
 
     def close(self):
         self.file.seek(pos[0], 0)
         txt = self.file.read(pos[1]).decode('utf-8')
         return txt
 
     def close(self):

+        """Close snippet file"""

         self.file.close()
 
 
         self.file.close()
 
 

-class Index(IndexStore):
+class BaseIndex(IndexStore):
+    """
+    Base index class.
+    Provides basic operations on index: opening, closing, optimizing.
+    """

     def __init__(self, analyzer=None):
     def __init__(self, analyzer=None):

-        IndexStore.__init__(self)
+        super(BaseIndex, self).__init__()

         self.index = None
         if not analyzer:
             analyzer = WLAnalyzer()
         self.index = None
         if not analyzer:
             analyzer = WLAnalyzer()


@@ -158,23 +187,60 @@ class Index(IndexStore):
         self.index.close()
         self.index = None
 
         self.index.close()
         self.index = None
 

+    def __enter__(self):
+        self.open()
+        return self
+
+    def __exit__(self, type, value, tb):
+        self.close()
+
+
+class Index(BaseIndex):
+    """
+    Class indexing books.
+    """
+    def __init__(self, analyzer=None):
+        super(Index, self).__init__(analyzer)
+

     def index_tags(self):
     def index_tags(self):

+        """
+        Re-index global tag list.
+        Removes all tags from index, then index them again.
+        Indexed fields include: id, name (with and without polish stems), category
+        """

         q = NumericRangeQuery.newIntRange("tag_id", 0, Integer.MAX_VALUE, True, True)
         self.index.deleteDocuments(q)
 
         for tag in catalogue.models.Tag.objects.all():
             doc = Document()
         q = NumericRangeQuery.newIntRange("tag_id", 0, Integer.MAX_VALUE, True, True)
         self.index.deleteDocuments(q)
 
         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))
             self.index.addDocument(doc)
 
             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))
             self.index.addDocument(doc)
 

+    def create_book_doc(self, book):
+        """
+        Create a lucene document referring book id.
+        """
+        doc = Document()
+        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(int(book.parent.id)))
+        return doc
+

     def remove_book(self, book):
     def remove_book(self, book):

+        """Removes a book from search index.
+        book - Book instance."""

         q = NumericRangeQuery.newIntRange("book_id", book.id, book.id, True, True)
         self.index.deleteDocuments(q)
 
     def index_book(self, book, book_info=None, overwrite=True):
         q = NumericRangeQuery.newIntRange("book_id", book.id, book.id, True, True)
         self.index.deleteDocuments(q)
 
     def index_book(self, book, book_info=None, overwrite=True):

+        """
+        Indexes the book.
+        Creates a lucene document for extracted metadata
+        and calls self.index_content() to index the contents of the book.
+        """

         if overwrite:
             self.remove_book(book)
 
         if overwrite:
             self.remove_book(book)
 


@@ -203,17 +269,10 @@ class Index(IndexStore):
 
     skip_header_tags = ['autor_utworu', 'nazwa_utworu', 'dzielo_nadrzedne']
 
 
     skip_header_tags = ['autor_utworu', 'nazwa_utworu', 'dzielo_nadrzedne']
 

-    def create_book_doc(self, book):
+    def extract_metadata(self, book, book_info=None):

         """
         """

-        Create a lucene document connected to the book
+        Extract metadata from book and returns a map of fields keyed by fieldname

         """
         """

-        doc = Document()
-        doc.add(NumericField("book_id", Field.Store.YES, True).setIntValue(book.id))
-        if book.parent is not None:
-            doc.add(NumericField("parent_id", Field.Store.YES, True).setIntValue(book.parent.id))
-        return doc
-
-    def extract_metadata(self, book, book_info=None):

         fields = {}
 
         if book_info is None:
         fields = {}
 
         if book_info is None:


@@ -252,18 +311,29 @@ class Index(IndexStore):
 
         return fields
 
 
         return fields
 

-    def get_master(self, root):
-        for master in root.iter():
-            if master.tag in self.master_tags:
-                return master
-

     def add_gaps(self, fields, fieldname):
     def add_gaps(self, fields, fieldname):

+        """
+        Interposes a list of fields with gap-fields, which are indexed spaces and returns it.
+        This allows for doing phrase queries which do not overlap the gaps (when slop is 0).
+        """

         def gap():
             while True:
                 yield Field(fieldname, ' ', Field.Store.NO, Field.Index.NOT_ANALYZED)
         return reduce(lambda a, b: a + b, zip(fields, gap()))[0:-1]
 
         def gap():
             while True:
                 yield Field(fieldname, ' ', Field.Store.NO, Field.Index.NOT_ANALYZED)
         return reduce(lambda a, b: a + b, zip(fields, gap()))[0:-1]
 

+    def get_master(self, root):
+        """
+        Returns the first master tag from an etree.
+        """
+        for master in root.iter():
+            if master.tag in self.master_tags:
+                return master
+

     def index_content(self, book, book_fields=[]):
     def index_content(self, book, book_fields=[]):

+        """
+        Walks the book XML and extract content from it.
+        Adds parts for each header tag and for each fragment.
+        """

         wld = WLDocument.from_file(book.xml_file.path, parse_dublincore=False)
         root = wld.edoc.getroot()
 
         wld = WLDocument.from_file(book.xml_file.path, parse_dublincore=False)
         root = wld.edoc.getroot()
 


@@ -280,7 +350,7 @@ class Index(IndexStore):
             return
 
         def fix_format(text):
             return
 
         def fix_format(text):

-            return re.sub("/$", "", text, flags=re.M)
+            return re.sub("(?m)/$", "", text)

 
         def add_part(snippets, **fields):
             doc = self.create_book_doc(book)
 
         def add_part(snippets, **fields):
             doc = self.create_book_doc(book)


@@ -325,7 +395,6 @@ class Index(IndexStore):
             else:
                 return s
 
             else:
                 return s
 

-

         fragments = {}
         snippets = Snippets(book.id).open('w')
         try:
         fragments = {}
         snippets = Snippets(book.id).open('w')
         try:


@@ -333,28 +402,28 @@ class Index(IndexStore):
 
                 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.


@@ -374,22 +443,25 @@ class Index(IndexStore):
                                        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)

-        finally:
-            snippets.close()
+                        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))))

 
 

-    def __enter__(self):
-        self.open()
-        return self
+                self.index.addDocument(doc)

 
 

-    def __exit__(self, type, value, tb):
-        self.close()
+        finally:
+            snippets.close()

 
 
 def log_exception_wrapper(f):
 
 
 def log_exception_wrapper(f):


@@ -437,20 +509,13 @@ class ReusableIndex(Index):
         pass
 
 
         pass
 
 

-class Search(IndexStore):
-    def __init__(self, default_field="content"):
-        IndexStore.__init__(self)
-        self.analyzer = WLAnalyzer() #PolishAnalyzer(Version.LUCENE_34)
-        ## self.analyzer = WLAnalyzer()
-        self.searcher = IndexSearcher(self.store, True)
-        self.parser = QueryParser(Version.LUCENE_34, default_field,
-                                  self.analyzer)
-
-        self.parent_filter = TermsFilter()
-        self.parent_filter.addTerm(Term("is_book", "true"))
-
-    def query(self, query):
-        return self.parser.parse(query)
+class JoinSearch(object):
+    """
+    This mixin could be used to handle block join queries.
+    (currently unused)
+    """
+    def __init__(self, *args, **kw):
+        super(JoinSearch, self).__init__(*args, **kw)

 
     def wrapjoins(self, query, fields=[]):
         """
 
     def wrapjoins(self, query, fields=[]):
         """


@@ -475,29 +540,6 @@ class Search(IndexStore):
             return BlockJoinQuery(query, self.parent_filter,
                                   BlockJoinQuery.ScoreMode.Total)
 
             return BlockJoinQuery(query, self.parent_filter,
                                   BlockJoinQuery.ScoreMode.Total)
 

-    def simple_search(self, query, max_results=50):
-        """Returns (books, total_hits)
-        """
-
-        tops = self.searcher.search(self.query(query), 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)
-
-
-    def search(self, query, max_results=50):
-        query = self.query(query)
-        query = self.wrapjoins(query, ["content", "themes"])
-
-        tops = self.searcher.search(query, 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)
-

     def bsearch(self, query, max_results=50):
         q = self.query(query)
         bjq = BlockJoinQuery(q, self.parent_filter, BlockJoinQuery.ScoreMode.Avg)
     def bsearch(self, query, max_results=50):
         q = self.query(query)
         bjq = BlockJoinQuery(q, self.parent_filter, BlockJoinQuery.ScoreMode.Avg)


@@ -509,27 +551,20 @@ class Search(IndexStore):
             bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id")))
         return (bks, tops.totalHits)
 
             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 SearchResult(object):
 
 class SearchResult(object):

-    def __init__(self, searcher, scoreDocs, score=None, how_found=None, snippets=None):
-        self.snippets = []
-
+    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:
         if score:

-            self.score = score
+            self._score = score

         else:
         else:

-            self.score = scoreDocs.score
+            self._score = scoreDocs.score
+            
+        self.boost = 1.0

 
 

-        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"))


@@ -544,14 +579,24 @@ 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)
+
+        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))
 
     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


@@ -562,35 +607,85 @@ 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],
+            try:
+                frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT])
+            except catalogue.models.Fragment.DoesNotExist:
+                # stale index
+                continue
+
+            # 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,
                  'fragment': frag,

-                 'themes': frag.tags.filter(category='theme')
+                 'section_number': f[POSITION][POSITION_INDEX] + 1,
+                 'themes': themes,
+                 'themes_hit': themes_hit

                  }
                  }

-            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)


@@ -612,25 +707,44 @@ class SearchResult(object):
 
 
 class Hint(object):
 
 
 class Hint(object):

+    """
+    Given some hint information (information we already know about)
+    our search target - like author, title (specific book), epoch, genre, kind
+    we can narrow down search using filters.
+    """

     def __init__(self, search):
     def __init__(self, search):

+        """
+        Accepts a Searcher instance.
+        """

         self.search = search
         self.book_tags = {}
         self.part_tags = []
         self._books = []
 
     def books(self, *books):
         self.search = search
         self.book_tags = {}
         self.part_tags = []
         self._books = []
 
     def books(self, *books):

+        """
+        Give a hint that we search these books.
+        """

         self._books = books
 
     def tags(self, tags):
         self._books = books
 
     def tags(self, tags):

+        """
+        Give a hint that these Tag objects (a list of)
+        is necessary.
+        """

         for t in tags:
             if t.category in ['author', 'title', 'epoch', 'genre', 'kind']:
                 lst = self.book_tags.get(t.category, [])
                 lst.append(t)
                 self.book_tags[t.category] = lst
         for t in tags:
             if t.category in ['author', 'title', 'epoch', 'genre', 'kind']:
                 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'):

+        """
+        Given a lsit of tags and an optional field (but they are normally in tags field)
+        returns a filter accepting only books with specific tags.
+        """

         q = BooleanQuery()
 
         for tag in tags:
         q = BooleanQuery()
 
         for tag in tags:


@@ -643,6 +757,9 @@ class Hint(object):
         return QueryWrapperFilter(q)
 
     def book_filter(self):
         return QueryWrapperFilter(q)
 
     def book_filter(self):

+        """
+        Filters using book tags (all tag kinds except a theme)
+        """

         tags = reduce(lambda a, b: a + b, self.book_tags.values(), [])
         if tags:
             return self.tag_filter(tags)
         tags = reduce(lambda a, b: a + b, self.book_tags.values(), [])
         if tags:
             return self.tag_filter(tags)


@@ -650,6 +767,10 @@ class Hint(object):
             return None
 
     def part_filter(self):
             return None
 
     def part_filter(self):

+        """
+        This filter can be used to look for book parts.
+        It filters on book id and/or themes.
+        """

         fs = []
         if self.part_tags:
             fs.append(self.tag_filter(self.part_tags, field='themes'))
         fs = []
         if self.part_tags:
             fs.append(self.tag_filter(self.part_tags, field='themes'))


@@ -661,7 +782,7 @@ class Hint(object):
                 bf.add(FilterClause(id_filter, BooleanClause.Occur.SHOULD))
             fs.append(bf)
 
                 bf.add(FilterClause(id_filter, BooleanClause.Occur.SHOULD))
             fs.append(bf)
 

-        return MultiSearch.chain_filters(fs)
+        return Search.chain_filters(fs)

 
     def should_search_for_book(self):
         return self._books == []
 
     def should_search_for_book(self):
         return self._books == []


@@ -680,13 +801,46 @@ class Hint(object):
         return some
 
 
         return some
 
 

-class MultiSearch(Search):
-    """Class capable of IMDb-like searching"""
-    def get_tokens(self, searched, field='content'):
+class Search(IndexStore):
+    """
+    Search facilities.
+    """
+    def __init__(self, default_field="content"):
+        IndexStore.__init__(self)
+        self.analyzer = WLAnalyzer()  # PolishAnalyzer(Version.LUCENE_34)
+        # self.analyzer = WLAnalyzer()
+        self.searcher = IndexSearcher(self.store, True)
+        self.parser = QueryParser(Version.LUCENE_34, default_field,
+                                  self.analyzer)
+
+        self.parent_filter = TermsFilter()
+        self.parent_filter.addTerm(Term("is_book", "true"))
+
+    def query(self, query):
+        """Parse query in default Lucene Syntax. (for humans)
+        """
+        return self.parser.parse(query)
+
+    def simple_search(self, query, max_results=50):
+        """Runs a query for books using lucene syntax. (for humans)
+        Returns (books, total_hits)
+        """
+
+        tops = self.searcher.search(self.query(query), 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)
+
+    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)
         """
         """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):
         if isinstance(searched, str) or isinstance(searched, unicode):
             searched = StringReader(searched)
         elif isinstance(searched, list):


@@ -698,9 +852,14 @@ class MultiSearch(Search):
         while tokens.incrementToken():
             cta = tokens.getAttribute(CharTermAttribute.class_)
             toks.append(cta.toString())
         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):
         return toks
 
     def fuzziness(self, fuzzy):

+        """Helper method to sanitize fuzziness"""

         if not fuzzy:
             return None
         if isinstance(fuzzy, float) and fuzzy > 0.0 and fuzzy <= 1.0:
         if not fuzzy:
             return None
         if isinstance(fuzzy, float) and fuzzy > 0.0 and fuzzy <= 1.0:


@@ -709,6 +868,9 @@ class MultiSearch(Search):
             return 0.5
 
     def make_phrase(self, tokens, field='content', slop=2, fuzzy=False):
             return 0.5
 
     def make_phrase(self, tokens, field='content', slop=2, fuzzy=False):

+        """
+        Return a PhraseQuery with a series of tokens.
+        """

         if fuzzy:
             phrase = MultiPhraseQuery()
             for t in tokens:
         if fuzzy:
             phrase = MultiPhraseQuery()
             for t in tokens:


@@ -735,6 +897,11 @@ class MultiSearch(Search):
         return phrase
 
     def make_term_query(self, tokens, field='content', modal=BooleanClause.Occur.SHOULD, fuzzy=False):
         return phrase
 
     def make_term_query(self, tokens, field='content', modal=BooleanClause.Occur.SHOULD, fuzzy=False):

+        """
+        Returns term queries joined by boolean query.
+        modal - applies to boolean query
+        fuzzy - should the query by fuzzy.
+        """

         q = BooleanQuery()
         for t in tokens:
             term = Term(field, t)
         q = BooleanQuery()
         for t in tokens:
             term = Term(field, t)


@@ -745,11 +912,45 @@ class MultiSearch(Search):
             q.add(BooleanClause(term, modal))
         return q
 
             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):
 
     def search_perfect_book(self, searched, max_results=20, fuzzy=False, hint=None):

+        """
+        Search for perfect book matches. Just see if the query matches with some author or title,
+        taking hints into account.
+        """

         fields_to_search = ['authors', 'title']
         only_in = None
         if hint:
         fields_to_search = ['authors', 'title']
         only_in = None
         if hint:


@@ -763,13 +964,44 @@ class MultiSearch(Search):
         books = []
         for q in qrys:
             top = self.searcher.search(q,
         books = []
         for q in qrys:
             top = self.searcher.search(q,

-                                       self.chain_filters([only_in, self.term_filter(Term('is_book', 'true'))]),
+                self.chain_filters([only_in, self.term_filter(Term('is_book', 'true'))]),

                 max_results)
             for found in top.scoreDocs:
                 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):

+        """
+        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']]
 
         flt = None
         qrys = [self.make_phrase(self.get_tokens(searched), field=fld, fuzzy=fuzzy) for fld in ['content']]
 
         flt = None


@@ -780,15 +1012,20 @@ class MultiSearch(Search):
         for q in qrys:
             top = self.searcher.search(q,
                                        self.chain_filters([self.term_filter(Term('is_book', 'true'), inverse=True),
         for q in qrys:
             top = self.searcher.search(q,
                                        self.chain_filters([self.term_filter(Term('is_book', 'true'), inverse=True),

-                                                           flt
-                                                          ]),
+                                                           flt]),

                                        max_results)
             for found in top.scoreDocs:
                                        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
 

-    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
 
         books = []
         only_in = None
 


@@ -798,43 +1035,54 @@ class MultiSearch(Search):
         # 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', 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']) != []:
+            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', 'content', 'tags', 'authors', 'title']:
-            in_content.add(BooleanClause(self.make_term_query(tokens, field=fld, fuzzy=False), BooleanClause.Occur.SHOULD))
+        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 ['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
 

-    def multisearch(self, query, max_results=50):
-        """
-        Search strategy:
-        - (phrase) OR -> content
-                      -> title
-                      -> authors
-        - (keywords)  -> authors
-                      -> motyw
-                      -> tags
-                      -> content
-        """
+    # def multisearch(self, query, max_results=50):
+    #     """
+    #     Search strategy:
+    #     - (phrase) OR -> content
+    #                   -> title
+    #                   -> authors
+    #     - (keywords)  -> authors
+    #                   -> motyw
+    #                   -> tags
+    #                   -> content
+    #     """

         # queryreader = StringReader(query)
         # tokens = self.get_tokens(queryreader)
 
         # queryreader = StringReader(query)
         # tokens = self.get_tokens(queryreader)
 


@@ -864,21 +1112,12 @@ class MultiSearch(Search):
         # top_level.add(BooleanClause(phrase_level, Should))
         # top_level.add(BooleanClause(kw_level, Should))
 
         # top_level.add(BooleanClause(phrase_level, Should))
         # top_level.add(BooleanClause(kw_level, Should))
 

-        return None
-
-    def book_search(self, query, filter=None, max_results=50, collector=None):
-        tops = self.searcher.search(query, filter, max_results)
-        #tops = self.searcher.search(p_content, max_results)
-
-        bks = []
-        for found in tops.scoreDocs:
-            doc = self.searcher.doc(found.doc)
-            b = catalogue.models.Book.objects.get(id=doc.get("book_id"))
-            bks.append(b)
-            print "%s (%d) -> %f" % (b, b.id, found.score)
-        return bks
+        # return None

 
     def get_snippets(self, scoreDoc, query, field='content'):
 
     def get_snippets(self, scoreDoc, query, field='content'):

+        """
+        Returns a snippet for found scoreDoc.
+        """

         htmlFormatter = SimpleHTMLFormatter()
         highlighter = Highlighter(htmlFormatter, QueryScorer(query))
 
         htmlFormatter = SimpleHTMLFormatter()
         highlighter = Highlighter(htmlFormatter, QueryScorer(query))
 


@@ -894,10 +1133,9 @@ class MultiSearch(Search):
 
         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, "...")
 
         snip = highlighter.getBestFragments(tokenStream, text, 3, "...")
 

-        return [snip]
+        return snip

 
     @staticmethod
     def enum_to_array(enum):
 
     @staticmethod
     def enum_to_array(enum):


@@ -917,6 +1155,9 @@ class MultiSearch(Search):
             return JArray('object')(terms, Term)
 
     def search_tags(self, query, filter=None, max_results=40):
             return JArray('object')(terms, Term)
 
     def search_tags(self, query, filter=None, max_results=40):

+        """
+        Search for Tag objects using query.
+        """

         tops = self.searcher.search(query, filter, max_results)
 
         tags = []
         tops = self.searcher.search(query, filter, max_results)
 
         tags = []


@@ -929,6 +1170,9 @@ class MultiSearch(Search):
         return tags
 
     def search_books(self, query, filter=None, max_results=10):
         return tags
 
     def search_books(self, query, filter=None, max_results=10):

+        """
+        Searches for Book objects using query
+        """

         bks = []
         tops = self.searcher.search(query, filter, max_results)
         for found in tops.scoreDocs:
         bks = []
         tops = self.searcher.search(query, filter, max_results)
         for found in tops.scoreDocs:


@@ -941,7 +1185,7 @@ class MultiSearch(Search):
         for i in range(len(toks)):
             t = Term(field, toks[i])
             if i == len(toks) - 1:
         for i in range(len(toks)):
             t = Term(field, toks[i])
             if i == len(toks) - 1:

-                pterms = MultiSearch.enum_to_array(PrefixTermEnum(self.searcher.getIndexReader(), t))
+                pterms = Search.enum_to_array(PrefixTermEnum(self.searcher.getIndexReader(), t))

                 if pterms:
                     q.add(pterms)
                 else:
                 if pterms:
                     q.add(pterms)
                 else:


@@ -963,6 +1207,10 @@ class MultiSearch(Search):
         return only_term
 
     def hint_tags(self, string, max_results=50):
         return only_term
 
     def hint_tags(self, string, max_results=50):

+        """
+        Return auto-complete hints for tags
+        using prefix search.
+        """

         toks = self.get_tokens(string, field='SIMPLE')
         top = BooleanQuery()
 
         toks = self.get_tokens(string, field='SIMPLE')
         top = BooleanQuery()
 


@@ -975,21 +1223,32 @@ class MultiSearch(Search):
         return self.search_tags(top, no_book_cat, max_results=max_results)
 
     def hint_books(self, string, max_results=50):
         return self.search_tags(top, no_book_cat, max_results=max_results)
 
     def hint_books(self, string, max_results=50):

+        """
+        Returns auto-complete hints for book titles
+        Because we do not index 'pseudo' title-tags.
+        Prefix search.
+        """

         toks = self.get_tokens(string, field='SIMPLE')
 
         q = self.create_prefix_phrase(toks, 'title')
 
         toks = self.get_tokens(string, field='SIMPLE')
 
         q = self.create_prefix_phrase(toks, 'title')
 

-        return self.book_search(q, self.term_filter(Term("is_book", "true")), max_results=max_results)
+        return self.search_books(q, self.term_filter(Term("is_book", "true")), max_results=max_results)

 
     @staticmethod
     def chain_filters(filters, op=ChainedFilter.AND):
 
     @staticmethod
     def chain_filters(filters, op=ChainedFilter.AND):

+        """
+        Chains a filter list together
+        """

         filters = filter(lambda x: x is not None, filters)
         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
 
     def filtered_categories(self, tags):
             return None
         chf = ChainedFilter(JArray('object')(filters, Filter), op)
         return chf
 
     def filtered_categories(self, tags):

+        """
+        Return a list of tag categories, present in tags list.
+        """

         cats = {}
         for t in tags:
             cats[t.category] = True
         cats = {}
         for t in tags:
             cats[t.category] = True