link to polski.

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

diff --git a/apps/search/index.py b/apps/search/index.py
index 9617077..a0bf715 100644 (file)
--- a/apps/search/index.py
+++ b/apps/search/index.py
@@ -1,7 +1,8 @@
 # -*- coding: utf-8 -*-
 
 from django.conf import settings
 # -*- coding: utf-8 -*-
 
 from django.conf import settings

-from lucene import SimpleFSDirectory, IndexWriter, CheckIndex, \
+from django.dispatch import Signal
+from lucene import SimpleFSDirectory, NIOFSDirectory, IndexWriter, IndexReader, IndexWriterConfig, CheckIndex, \

     File, Field, Integer, \
     NumericField, Version, Document, JavaError, IndexSearcher, \
     QueryParser, PerFieldAnalyzerWrapper, \
     File, Field, Integer, \
     NumericField, Version, Document, JavaError, IndexSearcher, \
     QueryParser, PerFieldAnalyzerWrapper, \


@@ -17,7 +18,7 @@ from lucene import SimpleFSDirectory, IndexWriter, CheckIndex, \
     # KeywordAnalyzer
 
 # Initialize jvm
     # KeywordAnalyzer
 
 # Initialize jvm

-JVM = initVM(CLASSPATH)
+JVM = initVM(CLASSPATH, maxheap=settings.JVM_MAXHEAP)

 
 import sys
 import os
 
 import sys
 import os


@@ -27,11 +28,14 @@ from librarian import dcparser
 from librarian.parser import WLDocument
 from lxml import etree
 import catalogue.models
 from librarian.parser import WLDocument
 from lxml import etree
 import catalogue.models

+from pdcounter.models import Author as PDCounterAuthor, BookStub as PDCounterBook

 from multiprocessing.pool import ThreadPool
 from threading import current_thread
 from multiprocessing.pool import ThreadPool
 from threading import current_thread

+from itertools import chain

 import atexit
 import traceback
 import atexit
 import traceback

-
+import logging
+log = logging.getLogger('search')

 
 class WLAnalyzer(PerFieldAnalyzerWrapper):
     def __init__(self):
 
 class WLAnalyzer(PerFieldAnalyzerWrapper):
     def __init__(self):


@@ -81,7 +85,7 @@ class IndexStore(object):
     """
     def __init__(self):
         self.make_index_dir()
     """
     def __init__(self):
         self.make_index_dir()

-        self.store = SimpleFSDirectory(File(settings.SEARCH_INDEX))
+        self.store = NIOFSDirectory(File(settings.SEARCH_INDEX))

 
     def make_index_dir(self):
         try:
 
     def make_index_dir(self):
         try:


@@ -91,6 +95,9 @@ class IndexStore(object):
                 pass
             else: raise
 
                 pass
             else: raise
 

+    def close(self):
+        self.store.close()
+

 
 class IndexChecker(IndexStore):
     def __init__(self):
 
 class IndexChecker(IndexStore):
     def __init__(self):


@@ -110,7 +117,7 @@ class Snippets(object):
     """
     SNIPPET_DIR = "snippets"
 
     """
     SNIPPET_DIR = "snippets"
 

-    def __init__(self, book_id):
+    def __init__(self, book_id, revision=None):

         try:
             os.makedirs(os.path.join(settings.SEARCH_INDEX, self.SNIPPET_DIR))
         except OSError as exc:
         try:
             os.makedirs(os.path.join(settings.SEARCH_INDEX, self.SNIPPET_DIR))
         except OSError as exc:


@@ -118,15 +125,32 @@ class Snippets(object):
                 pass
             else: raise
         self.book_id = book_id
                 pass
             else: raise
         self.book_id = book_id

+        self.revision = revision

         self.file = None
 
         self.file = None
 

+    @property
+    def path(self):
+        if self.revision: fn = "%d.%d" % (self.book_id, self.revision)
+        else: fn = "%d" % self.book_id
+
+        return os.path.join(settings.SEARCH_INDEX, self.SNIPPET_DIR, fn)
+

     def open(self, mode='r'):
         """
         Open the snippet file. Call .close() afterwards.
         """
         if not 'b' in mode:
             mode += 'b'
     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 'w' in mode:
+            if os.path.exists(self.path):
+                self.revision = 1
+                while True:
+                    if not os.path.exists(self.path):
+                        break
+                    self.revision += 1
+
+        self.file = open(self.path, mode)

         self.position = 0
         return self
 
         self.position = 0
         return self
 


@@ -155,6 +179,17 @@ class Snippets(object):
         """Close snippet file"""
         self.file.close()
 
         """Close snippet file"""
         self.file.close()
 

+    def remove(self):
+        self.revision = None
+        try:
+            os.unlink(self.path)
+            self.revision = 0
+            while True:
+                self.revision += 1
+                os.unlink(self.path)
+        except OSError:
+            pass
+

 
 class BaseIndex(IndexStore):
     """
 
 class BaseIndex(IndexStore):
     """


@@ -168,11 +203,13 @@ class BaseIndex(IndexStore):
             analyzer = WLAnalyzer()
         self.analyzer = analyzer
 
             analyzer = WLAnalyzer()
         self.analyzer = analyzer
 

-    def open(self, analyzer=None):
+    def open(self, timeout=None):

         if self.index:
             raise Exception("Index is already opened")
         if self.index:
             raise Exception("Index is already opened")

-        self.index = IndexWriter(self.store, self.analyzer,\
-                                 IndexWriter.MaxFieldLength.LIMITED)
+        conf = IndexWriterConfig(Version.LUCENE_34, self.analyzer)
+        if timeout:
+            conf.setWriteLockTimeout(long(timeout))
+        self.index = IndexWriter(self.store, conf)

         return self.index
 
     def optimize(self):
         return self.index
 
     def optimize(self):


@@ -182,11 +219,15 @@ class BaseIndex(IndexStore):
         try:
             self.index.optimize()
         except JavaError, je:
         try:
             self.index.optimize()
         except JavaError, je:

-            print "Error during optimize phase, check index: %s" % je
+            log.error("Error during optimize phase, check index: %s" % je)

 
         self.index.close()
         self.index = None
 
 
         self.index.close()
         self.index = None
 

+        index_changed.send_robust(self)
+
+        super(BaseIndex, self).close()
+

     def __enter__(self):
         self.open()
         return self
     def __enter__(self):
         self.open()
         return self


@@ -195,6 +236,9 @@ class BaseIndex(IndexStore):
         self.close()
 
 
         self.close()
 
 

+index_changed = Signal()
+
+

 class Index(BaseIndex):
     """
     Class indexing books.
 class Index(BaseIndex):
     """
     Class indexing books.


@@ -202,22 +246,66 @@ class Index(BaseIndex):
     def __init__(self, analyzer=None):
         super(Index, self).__init__(analyzer)
 
     def __init__(self, analyzer=None):
         super(Index, self).__init__(analyzer)
 

-    def index_tags(self):
+    def index_tags(self, *tags, **kw):

         """
         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
         """
         """
         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)
+        remove_only = kw.get('remove_only', False)
+        # first, remove tags from index.
+        if tags:
+            q = BooleanQuery()
+            for tag in tags:
+                b_id_cat = BooleanQuery()

 
 

-        for tag in catalogue.models.Tag.objects.all():
-            doc = Document()
-            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)
+                q_id = NumericRangeQuery.newIntRange("tag_id", tag.id, tag.id, True, True)
+                b_id_cat.add(q_id, BooleanClause.Occur.MUST)
+
+                if isinstance(tag, PDCounterAuthor):
+                    q_cat = TermQuery(Term('tag_category', 'pd_author'))
+                elif isinstance(tag, PDCounterBook):
+                    q_cat = TermQuery(Term('tag_category', 'pd_book'))
+                else:
+                    q_cat = TermQuery(Term('tag_category', tag.category))
+                b_id_cat.add(q_cat, BooleanClause.Occur.MUST)
+
+                q.add(b_id_cat, BooleanClause.Occur.SHOULD)
+        else:  # all
+            q = NumericRangeQuery.newIntRange("tag_id", 0, Integer.MAX_VALUE, True, True)
+            self.index.deleteDocuments(q)
+
+        if not remove_only:
+            # then add them [all or just one passed]
+            if not tags:
+                tags = chain(catalogue.models.Tag.objects.exclude(category='set'), \
+                    PDCounterAuthor.objects.all(), \
+                    PDCounterBook.objects.all())
+
+            for tag in tags:
+                if isinstance(tag, PDCounterAuthor):
+                    doc = Document()
+                    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", 'pd_author', Field.Store.YES, Field.Index.NOT_ANALYZED))
+                    doc.add(Field("is_pdcounter", 'true', Field.Store.YES, Field.Index.NOT_ANALYZED))
+                    self.index.addDocument(doc)
+                elif isinstance(tag, PDCounterBook):
+                    doc = Document()
+                    doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(int(tag.id)))
+                    doc.add(Field("tag_name", tag.title, Field.Store.NO, Field.Index.ANALYZED))
+                    doc.add(Field("tag_name_pl", tag.title, Field.Store.NO, Field.Index.ANALYZED))
+                    doc.add(Field("tag_category", 'pd_book', Field.Store.YES, Field.Index.NOT_ANALYZED))
+                    doc.add(Field("is_pdcounter", 'true', Field.Store.YES, Field.Index.NOT_ANALYZED))
+                    self.index.addDocument(doc)
+                else:
+                    doc = Document()
+                    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)

 
     def create_book_doc(self, book):
         """
 
     def create_book_doc(self, book):
         """


@@ -229,12 +317,21 @@ class Index(BaseIndex):
             doc.add(NumericField("parent_id", Field.Store.YES, True).setIntValue(int(book.parent.id)))
         return doc
 
             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_or_id, remove_snippets=True):

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

-        q = NumericRangeQuery.newIntRange("book_id", book.id, book.id, True, True)
+        if isinstance(book_or_id, catalogue.models.Book):
+            book_id = book_or_id.id
+        else:
+            book_id = book_or_id
+
+        q = NumericRangeQuery.newIntRange("book_id", book_id, book_id, True, True)

         self.index.deleteDocuments(q)
 
         self.index.deleteDocuments(q)
 

+        if remove_snippets:
+            snippets = Snippets(book_id)
+            snippets.remove()
+

     def index_book(self, book, book_info=None, overwrite=True):
         """
         Indexes the book.
     def index_book(self, book, book_info=None, overwrite=True):
         """
         Indexes the book.


@@ -242,17 +339,22 @@ class Index(BaseIndex):
         and calls self.index_content() to index the contents of the book.
         """
         if overwrite:
         and calls self.index_content() to index the contents of the book.
         """
         if overwrite:

-            self.remove_book(book)
+            # we don't remove snippets, since they might be still needed by
+            # threads using not reopened index
+            self.remove_book(book, remove_snippets=False)

 
         book_doc = self.create_book_doc(book)
 
         book_doc = self.create_book_doc(book)

-        meta_fields = self.extract_metadata(book, book_info)
+        meta_fields = self.extract_metadata(book, book_info, dc_only=['source_name', 'authors', 'title'])
+        # let's not index it - it's only used for extracting publish date
+        if 'source_name' in meta_fields:
+            del meta_fields['source_name']
+        

         for f in meta_fields.values():
             if isinstance(f, list) or isinstance(f, tuple):
                 for elem in f:
                     book_doc.add(elem)
             else:
                 book_doc.add(f)
         for f in meta_fields.values():
             if isinstance(f, list) or isinstance(f, tuple):
                 for elem in f:
                     book_doc.add(elem)
             else:
                 book_doc.add(f)

-

         self.index.addDocument(book_doc)
         del book_doc
 
         self.index.addDocument(book_doc)
         del book_doc
 


@@ -276,11 +378,11 @@ class Index(BaseIndex):
 
     footnote_tags = ['pa', 'pt', 'pr', 'pe']
 
 
     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]+)[\]. ]*$")
 
 
     published_date_re = re.compile("([0-9]+)[\]. ]*$")
 

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

         """
         Extract metadata from book and returns a map of fields keyed by fieldname
         """
         """
         Extract metadata from book and returns a map of fields keyed by fieldname
         """


@@ -295,6 +397,8 @@ class Index(BaseIndex):
 
         # validator, name
         for field in dcparser.BookInfo.FIELDS:
 
         # validator, name
         for field in dcparser.BookInfo.FIELDS:

+            if dc_only and field.name not in dc_only:
+                continue

             if hasattr(book_info, field.name):
                 if not getattr(book_info, field.name):
                     continue
             if hasattr(book_info, field.name):
                 if not getattr(book_info, field.name):
                     continue


@@ -321,11 +425,13 @@ class Index(BaseIndex):
                                                (dt.year, dt.month, dt.day), Field.Store.NO, Field.Index.NOT_ANALYZED)
 
         # get published date
                                                (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
 
 
         return fields
 


@@ -360,11 +466,18 @@ class Index(BaseIndex):
             return []
 
         def walker(node, ignore_tags=[]):
             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):
             return
 
         def fix_format(text):


@@ -397,6 +510,8 @@ class Index(BaseIndex):
             snip_pos = snippets.add(fields["content"])
             doc.add(NumericField("snippets_position", Field.Store.YES, True).setIntValue(snip_pos[0]))
             doc.add(NumericField("snippets_length", Field.Store.YES, True).setIntValue(snip_pos[1]))
             snip_pos = snippets.add(fields["content"])
             doc.add(NumericField("snippets_position", Field.Store.YES, True).setIntValue(snip_pos[0]))
             doc.add(NumericField("snippets_length", Field.Store.YES, True).setIntValue(snip_pos[1]))

+            if snippets.revision:
+                doc.add(NumericField("snippets_revision", Field.Store.YES, True).setIntValue(snippets.revision))

 
             if 'fragment_anchor' in fields:
                 doc.add(Field("fragment_anchor", fields['fragment_anchor'],
 
             if 'fragment_anchor' in fields:
                 doc.add(Field("fragment_anchor", fields['fragment_anchor'],


@@ -426,9 +541,8 @@ class Index(BaseIndex):
 
         fragments = {}
         snippets = Snippets(book.id).open('w')
 
         fragments = {}
         snippets = Snippets(book.id).open('w')

-        position = 0

         try:
         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
 
                 if header.tag in self.skip_header_tags:
                     continue


@@ -437,35 +551,50 @@ class Index(BaseIndex):
 
                 # section content
                 content = []
 
                 # section content
                 content = []

-                footnote = None
+                footnote = []

 
 

-                for start, end in walker(header, ignore_tags=self.ignore_content_tags):
+                def all_content(text):
+                    for frag in fragments.values():
+                        frag['content'].append(text)
+                    content.append(text)
+                handle_text = [all_content]
+
+
+                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:
                     # 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,
                         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)
                         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}
 
                     # 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:]
                     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(',')))
                         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
 
                     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.


@@ -478,25 +607,22 @@ class Index(BaseIndex):
                                        fragment_anchor=fid,
                                        content=fix_format(frag['content']),
                                        themes=frag['themes'])
                                        fragment_anchor=fid,
                                        content=fix_format(frag['content']),
                                        themes=frag['themes'])

-
+                        #print '@ FRAG %s' % frag['content']

                         self.index.addDocument(doc)
 
                         # Collect 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))
 
                         # 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)
 
                 self.index.addDocument(doc)

-                position += 1

 
         finally:
             snippets.close()
 
         finally:
             snippets.close()


@@ -507,7 +633,7 @@ def log_exception_wrapper(f):
         try:
             f(*a)
         except Exception, e:
         try:
             f(*a)
         except Exception, e:

-            print("Error in indexing thread: %s" % e)
+            log.error("Error in indexing thread: %s" % e)

             traceback.print_exc()
             raise e
     return _wrap
             traceback.print_exc()
             raise e
     return _wrap


@@ -523,12 +649,11 @@ class ReusableIndex(Index):
     """
     index = None
 
     """
     index = None
 

-    def open(self, analyzer=None, threads=4):
-        if ReusableIndex.index is not None:
+    def open(self, analyzer=None, **kw):
+        if ReusableIndex.index:

             self.index = ReusableIndex.index
         else:
             self.index = ReusableIndex.index
         else:

-            print("opening index")
-            Index.open(self, analyzer)
+            Index.open(self, analyzer, **kw)

             ReusableIndex.index = self.index
             atexit.register(ReusableIndex.close_reusable)
 
             ReusableIndex.index = self.index
             atexit.register(ReusableIndex.close_reusable)
 


@@ -538,13 +663,16 @@ class ReusableIndex(Index):
 
     @staticmethod
     def close_reusable():
 
     @staticmethod
     def close_reusable():

-        if ReusableIndex.index is not None:
+        if ReusableIndex.index:

             ReusableIndex.index.optimize()
             ReusableIndex.index.close()
             ReusableIndex.index = None
 
             ReusableIndex.index.optimize()
             ReusableIndex.index.close()
             ReusableIndex.index = None
 

+            index_changed.send_robust(None)
+

     def close(self):
     def close(self):

-        pass
+        if ReusableIndex.index:
+            ReusableIndex.index.commit()

 
 
 class JoinSearch(object):
 
 
 class JoinSearch(object):


@@ -607,27 +735,26 @@ class SearchResult(object):
         stored = search.searcher.doc(scoreDocs.doc)
         self.book_id = int(stored.get("book_id"))
 
         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
+        pd = stored.get("published_date")
+        try:
+            self.published_date = int(pd)
+        except ValueError:
+            self.published_date = 0

 
 

-        fragment = stored.get("fragment_anchor")
+        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

 
 

-        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)
+            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 []})
+            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.search = search
         self.searched = searched
 
         self.search = search
         self.searched = searched


@@ -646,6 +773,8 @@ class SearchResult(object):
         return self
 
     def get_book(self):
         return self
 
     def get_book(self):

+        if hasattr(self, '_book'):
+            return self._book

         return catalogue.models.Book.objects.get(id=self.book_id)
 
     book = property(get_book)
         return catalogue.models.Book.objects.get(id=self.book_id)
 
     book = property(get_book)


@@ -664,7 +793,10 @@ class SearchResult(object):
 
         # to sections and fragments
         frags = filter(lambda r: r[FRAGMENT] is not None, self._hits)
 
         # 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 r: r[FRAGMENT] is None, self._hits)

+
+        # sections not covered by fragments

         sect = filter(lambda s: 0 == len(filter(
             lambda f: s[POSITION][POSITION_INDEX] >= f[POSITION][POSITION_INDEX]
             and s[POSITION][POSITION_INDEX] < f[POSITION][POSITION_INDEX] + f[POSITION][POSITION_SPAN],
         sect = filter(lambda s: 0 == len(filter(
             lambda f: s[POSITION][POSITION_INDEX] >= f[POSITION][POSITION_INDEX]
             and s[POSITION][POSITION_INDEX] < f[POSITION][POSITION_INDEX] + f[POSITION][POSITION_SPAN],


@@ -672,15 +804,20 @@ class SearchResult(object):
 
         hits = []
 
 
         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()
+        def remove_duplicates(lst, keyfn, compare):
+            els = {}
+            for e in lst:
+                eif = keyfn(e)
+                if eif in els:
+                    if compare(els[eif], e) >= 1:
+                        continue
+                els[eif] = e
+            return els.values()
+
+        # remove fragments with duplicated fid's and duplicated snippets
+        frags = remove_duplicates(frags, lambda f: f[FRAGMENT], lambda a, b: cmp(a[SCORE], b[SCORE]))
+        frags = remove_duplicates(frags, lambda f: f[OTHER]['snippets'] and f[OTHER]['snippets'][0] or f[FRAGMENT],
+                                  lambda a, b: cmp(a[SCORE], b[SCORE]))

 
         # remove duplicate sections
         sections = {}
 
         # remove duplicate sections
         sections = {}


@@ -702,7 +839,7 @@ class SearchResult(object):
 
         for f in frags:
             try:
 
         for f in frags:
             try:

-                frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT])
+                frag = catalogue.models.Fragment.objects.get(anchor=f[FRAGMENT], book__id=self.book_id)

             except catalogue.models.Fragment.DoesNotExist:
                 # stale index
                 continue
             except catalogue.models.Fragment.DoesNotExist:
                 # stale index
                 continue


@@ -745,7 +882,6 @@ class SearchResult(object):
             for r in rl:
                 if r.book_id in books:
                     books[r.book_id].merge(r)
             for r in rl:
                 if r.book_id in books:
                     books[r.book_id].merge(r)

-                    #print(u"already have one with score %f, and this one has score %f" % (books[book.id][0], found.score))

                 else:
                     books[r.book_id] = r
         return books.values()
                 else:
                     books[r.book_id] = r
         return books.values()


@@ -862,12 +998,31 @@ class Search(IndexStore):
         IndexStore.__init__(self)
         self.analyzer = WLAnalyzer()  # PolishAnalyzer(Version.LUCENE_34)
         # self.analyzer = WLAnalyzer()
         IndexStore.__init__(self)
         self.analyzer = WLAnalyzer()  # PolishAnalyzer(Version.LUCENE_34)
         # self.analyzer = WLAnalyzer()

-        self.searcher = IndexSearcher(self.store, True)
+        reader = IndexReader.open(self.store, True)
+        self.searcher = IndexSearcher(reader)

         self.parser = QueryParser(Version.LUCENE_34, default_field,
                                   self.analyzer)
 
         self.parent_filter = TermsFilter()
         self.parent_filter.addTerm(Term("is_book", "true"))
         self.parser = QueryParser(Version.LUCENE_34, default_field,
                                   self.analyzer)
 
         self.parent_filter = TermsFilter()
         self.parent_filter.addTerm(Term("is_book", "true"))

+        index_changed.connect(self.reopen)
+
+    def close(self):
+        reader = self.searcher.getIndexReader()
+        self.searcher.close()
+        reader.close()
+        super(Search, self).close()
+        index_changed.disconnect(self.reopen)
+
+    def reopen(self, **unused):
+        reader = self.searcher.getIndexReader()
+        rdr = reader.reopen()
+        if not rdr.equals(reader):
+            log.debug('Reopening index')
+            oldsearch = self.searcher
+            self.searcher = IndexSearcher(rdr)
+            oldsearch.close()
+            reader.close()

 
     def query(self, query):
         """Parse query in default Lucene Syntax. (for humans)
 
     def query(self, query):
         """Parse query in default Lucene Syntax. (for humans)


@@ -911,7 +1066,8 @@ class Search(IndexStore):
 
         return toks
 
 
         return toks
 

-    def fuzziness(self, fuzzy):
+    @staticmethod
+    def fuzziness(fuzzy):

         """Helper method to sanitize fuzziness"""
         if not fuzzy:
             return None
         """Helper method to sanitize fuzziness"""
         if not fuzzy:
             return None


@@ -932,7 +1088,6 @@ class Search(IndexStore):
                 fuzzterms = []
 
                 while True:
                 fuzzterms = []
 
                 while True:

-                    #                    print("fuzz %s" % unicode(fuzzterm.term()).encode('utf-8'))

                     ft = fuzzterm.term()
                     if ft:
                         fuzzterms.append(ft)
                     ft = fuzzterm.term()
                     if ft:
                         fuzzterms.append(ft)


@@ -949,7 +1104,8 @@ class Search(IndexStore):
                 phrase.add(term)
         return phrase
 
                 phrase.add(term)
         return phrase
 

-    def make_term_query(self, tokens, field='content', modal=BooleanClause.Occur.SHOULD, fuzzy=False):
+    @staticmethod
+    def make_term_query(tokens, field='content', modal=BooleanClause.Occur.SHOULD, fuzzy=False):

         """
         Returns term queries joined by boolean query.
         modal - applies to boolean query
         """
         Returns term queries joined by boolean query.
         modal - applies to boolean query


@@ -1103,7 +1259,6 @@ class Search(IndexStore):
         topDocs = self.searcher.search(q, only_in, max_results)
         for found in topDocs.scoreDocs:
             books.append(SearchResult(self, found, how_found='search_everywhere_themesXcontent', searched=searched))
         topDocs = self.searcher.search(q, only_in, max_results)
         for found in topDocs.scoreDocs:
             books.append(SearchResult(self, found, how_found='search_everywhere_themesXcontent', searched=searched))

-            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()


@@ -1122,7 +1277,6 @@ class Search(IndexStore):
         topDocs = self.searcher.search(q, only_in, max_results)
         for found in topDocs.scoreDocs:
             books.append(SearchResult(self, found, how_found='search_everywhere', searched=searched))
         topDocs = self.searcher.search(q, only_in, max_results)
         for found in topDocs.scoreDocs:
             books.append(SearchResult(self, found, how_found='search_everywhere', searched=searched))

-            print "* %s scatter search: %s" % (searched, books[-1]._hits)

 
         return books
 
 
         return books
 


@@ -1181,18 +1335,36 @@ class Search(IndexStore):
         length = stored.get('snippets_length')
         if position is None or length is None:
             return None
         length = stored.get('snippets_length')
         if position is None or length is None:
             return None

+        revision = stored.get('snippets_revision')
+        if revision: revision = int(revision)
+

         # locate content.
         # locate content.

-        snippets = Snippets(stored.get('book_id')).open()
+        book_id = int(stored.get('book_id'))
+        snippets = Snippets(book_id, revision=revision)
+

         try:
         try:

-            text = snippets.get((int(position),
-                                 int(length)))
-        finally:
-            snippets.close()
+            snippets.open()
+        except IOError, e:
+            log.error("Cannot open snippet file for book id = %d [rev=%d], %s" % (book_id, revision, e))
+            return []

 
 

-        tokenStream = TokenSources.getAnyTokenStream(self.searcher.getIndexReader(), scoreDoc.doc, field, self.analyzer)
-        #  highlighter.getBestTextFragments(tokenStream, text, False, 10)
-        snip = highlighter.getBestFragments(tokenStream, text, 3, "...")
+        try:
+            try:
+                text = snippets.get((int(position),
+                                     int(length)))
+            finally:
+                snippets.close()

 
 

+            tokenStream = TokenSources.getAnyTokenStream(self.searcher.getIndexReader(), scoreDoc.doc, field, self.analyzer)
+            #  highlighter.getBestTextFragments(tokenStream, text, False, 10)
+            snip = highlighter.getBestFragments(tokenStream, text, 3, "...")
+
+        except Exception, e:
+            e2 = e
+            if hasattr(e, 'getJavaException'):
+                e2 = unicode(e.getJavaException())
+            raise Exception("Problem fetching snippets for book %d, @%d len=%d" % (book_id, int(position), int(length)),
+                e2)

         return snip
 
     @staticmethod
         return snip
 
     @staticmethod


@@ -1212,33 +1384,56 @@ class Search(IndexStore):
         if terms:
             return JArray('object')(terms, Term)
 
         if terms:
             return JArray('object')(terms, Term)
 

-    def search_tags(self, query, filter=None, max_results=40):
+    def search_tags(self, query, filt=None, max_results=40, pdcounter=False):

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

-        tops = self.searcher.search(query, filter, max_results)
+        if not pdcounter:
+            filters = self.chain_filters([filt, self.term_filter(Term('is_pdcounter', 'true'), inverse=True)])
+        tops = self.searcher.search(query, filt, max_results)

 
         tags = []
         for found in tops.scoreDocs:
             doc = self.searcher.doc(found.doc)
 
         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')
+            category = doc.get('tag_category')
+            try:
+                if is_pdcounter == 'true':
+                    if category == 'pd_author':
+                        tag = PDCounterAuthor.objects.get(id=doc.get('tag_id'))
+                    elif category == 'pd_book':
+                        tag = PDCounterBook.objects.get(id=doc.get('tag_id'))
+                        tag.category = 'pd_book'  # make it look more lik a tag.
+                    else:
+                        print "Warning. cannot get pdcounter tag_id=%d from db; cat=%s" % (int(doc.get('tag_id')), category)
+                else:
+                    tag = catalogue.models.Tag.objects.get(id=doc.get("tag_id"))
+                    # don't add the pdcounter tag if same tag already exists
+
+                tags.append(tag)
+
+            except catalogue.models.Tag.DoesNotExist: pass
+            except PDCounterAuthor.DoesNotExist: pass
+            except PDCounterBook.DoesNotExist: pass
+
+        log.debug('search_tags: %s' % tags)

 
         return tags
 
 
         return tags
 

-    def search_books(self, query, filter=None, max_results=10):
+    def search_books(self, query, filt=None, max_results=10):

         """
         Searches for Book objects using query
         """
         bks = []
         """
         Searches for Book objects using query
         """
         bks = []

-        tops = self.searcher.search(query, filter, max_results)
+        tops = self.searcher.search(query, filt, max_results)

         for found in tops.scoreDocs:
             doc = self.searcher.doc(found.doc)
         for found in tops.scoreDocs:
             doc = self.searcher.doc(found.doc)

-            bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id")))
+            try:
+                bks.append(catalogue.models.Book.objects.get(id=doc.get("book_id")))
+            except catalogue.models.Book.DoesNotExist: pass

         return bks
 
         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])
         q = MultiPhraseQuery()
         for i in range(len(toks)):
             t = Term(field, toks[i])


@@ -1264,7 +1459,7 @@ class Search(IndexStore):
 
         return only_term
 
 
         return only_term
 

-    def hint_tags(self, string, max_results=50):
+    def hint_tags(self, string, max_results=50, pdcounter=True, prefix=True, fuzzy=False):

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


@@ -1273,14 +1468,17 @@ class Search(IndexStore):
         top = BooleanQuery()
 
         for field in ['tag_name', 'tag_name_pl']:
         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, fuzzy=fuzzy)

             top.add(BooleanClause(q, BooleanClause.Occur.SHOULD))
 
         no_book_cat = self.term_filter(Term("tag_category", "book"), inverse=True)
 
             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, fuzzy=False):

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


@@ -1288,7 +1486,10 @@ class Search(IndexStore):
         """
         toks = self.get_tokens(string, field='SIMPLE')
 
         """
         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', fuzzy=fuzzy)

 
         return self.search_books(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)