search too many files fix?

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

diff --git a/apps/search/index.py b/apps/search/index.py
index 993c7d7..b3e9322 100644 (file)
--- a/apps/search/index.py
+++ b/apps/search/index.py
@@ -1,7 +1,7 @@
 # -*- coding: utf-8 -*-
 
 from django.conf import settings
 # -*- coding: utf-8 -*-
 
 from django.conf import settings

-from lucene import SimpleFSDirectory, IndexWriter, CheckIndex, \
+from lucene import SimpleFSDirectory, NIOFSDirectory, IndexWriter, IndexWriterConfig, CheckIndex, \

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


@@ -27,7 +27,7 @@ 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
+from pdcounter.models import Author as PDCounterAuthor, BookStub as PDCounterBook

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


@@ -82,7 +82,7 @@ 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:


@@ -169,11 +169,13 @@ 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):


@@ -212,7 +214,7 @@ class Index(BaseIndex):
         q = NumericRangeQuery.newIntRange("tag_id", 0, Integer.MAX_VALUE, True, True)
         self.index.deleteDocuments(q)
 
         q = NumericRangeQuery.newIntRange("tag_id", 0, Integer.MAX_VALUE, True, True)
         self.index.deleteDocuments(q)
 

-        for tag in catalogue.models.Tag.objects.all():
+        for tag in catalogue.models.Tag.objects.exclude(category='set'):

             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 = 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))


@@ -225,7 +227,16 @@ class Index(BaseIndex):
             doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(int(pdtag.id)))
             doc.add(Field("tag_name", pdtag.name, Field.Store.NO, Field.Index.ANALYZED))
             doc.add(Field("tag_name_pl", pdtag.name, Field.Store.NO, Field.Index.ANALYZED))
             doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(int(pdtag.id)))
             doc.add(Field("tag_name", pdtag.name, Field.Store.NO, Field.Index.ANALYZED))
             doc.add(Field("tag_name_pl", pdtag.name, Field.Store.NO, Field.Index.ANALYZED))

-            doc.add(Field("tag_category", 'pdcounter', Field.Store.NO, Field.Index.NOT_ANALYZED))
+            doc.add(Field("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)
+
+        for pdtag in PDCounterBook.objects.all():
+            doc = Document()
+            doc.add(NumericField("tag_id", Field.Store.YES, True).setIntValue(int(pdtag.id)))
+            doc.add(Field("tag_name", pdtag.title, Field.Store.NO, Field.Index.ANALYZED))
+            doc.add(Field("tag_name_pl", pdtag.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)
 
             doc.add(Field("is_pdcounter", 'true', Field.Store.YES, Field.Index.NOT_ANALYZED))
             self.index.addDocument(doc)
 


@@ -262,7 +273,6 @@ class Index(BaseIndex):
                     book_doc.add(elem)
             else:
                 book_doc.add(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
 


@@ -331,11 +341,13 @@ 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
-        if hasattr(book_info, 'source_name'):
-            match = self.published_date_re.search(source)
+        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:
             if match is not None:

-                fields["published_date"] = Field("published_date", str(match.groups()[0]), Field.Store.YES, Field.Index.NOT_ANALYZED)
+                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
 


@@ -370,11 +382,18 @@ 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):


@@ -446,35 +465,50 @@ class Index(BaseIndex):
 
                 # section content
                 content = []
 
                 # section content
                 content = []

-                footnote = None
+                footnote = []

 
 

-                for start, end in walker(header, ignore_tags=self.ignore_content_tags):
-                    # 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:
-                    #     doc = add_part(snippets, header_index=position, header_type=header.tag,
-                    #                    content=footnote)
+                def all_content(text):
+                    for frag in fragments.values():
+                        frag['content'].append(text)
+                    content.append(text)
+                handle_text = [all_content]

 
 

-                    #     self.index.addDocument(doc)
-
-                    #     footnote = None

 
 

+                for start, text, end in walker(header, ignore_tags=self.ignore_content_tags):
+                    # handle footnotes
+                    if start is not None and start.tag in self.footnote_tags:
+                        footnote = []
+                        def collect_footnote(t):
+                            footnote.append(t)
+                        handle_text.append(collect_footnote)
+                    elif end is not None and footnote is not [] and end.tag in self.footnote_tags:
+                        handle_text.pop()
+                        doc = add_part(snippets, header_index=position, header_type=header.tag,
+                                       content=u''.join(footnote),
+                                       is_footnote=Field("is_footnote", 'true', Field.Store.NO, Field.Index.NOT_ANALYZED))
+                
+                        self.index.addDocument(doc)
+                        #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.


@@ -487,22 +521,20 @@ 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)
 


@@ -531,12 +563,12 @@ 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:
             print("opening index")
             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)
 


@@ -546,13 +578,15 @@ class ReusableIndex(Index):
 
     @staticmethod
     def close_reusable():
 
     @staticmethod
     def close_reusable():

-        if ReusableIndex.index is not None:
+        if ReusableIndex.index:
+            print("closing index")

             ReusableIndex.index.optimize()
             ReusableIndex.index.close()
             ReusableIndex.index = None
 
     def close(self):
             ReusableIndex.index.optimize()
             ReusableIndex.index.close()
             ReusableIndex.index = None
 
     def close(self):

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

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


@@ -616,9 +650,10 @@ class SearchResult(object):
         self.book_id = int(stored.get("book_id"))
 
         pd = stored.get("published_date")
         self.book_id = int(stored.get("book_id"))
 
         pd = stored.get("published_date")

-        if pd is None:
-            pd = 0
-        self.published_date = int(pd)
+        try:
+            self.published_date = int(pd)
+        except ValueError:
+            self.published_date = 0

 
         header_type = stored.get("header_type")
         # we have a content hit in some header of fragment
 
         header_type = stored.get("header_type")
         # we have a content hit in some header of fragment


@@ -708,7 +743,7 @@ 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


@@ -759,8 +794,6 @@ class SearchResult(object):
     def __cmp__(self, other):
         c = cmp(self.score, other.score)
         if c == 0:
     def __cmp__(self, other):
         c = cmp(self.score, other.score)
         if c == 0:

-            if not hasattr(other,'published_date') or not hasattr(self, 'published_date'):
-                import pdb; pdb.set_trace()

             # this is inverted, because earlier date is better
             return cmp(other.published_date, self.published_date)
         else:
             # this is inverted, because earlier date is better
             return cmp(other.published_date, self.published_date)
         else:


@@ -1190,17 +1223,25 @@ class Search(IndexStore):
         if position is None or length is None:
             return None
         # locate content.
         if position is None or length is None:
             return None
         # locate content.

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

         try:
         try:

-            text = snippets.get((int(position),
-                                 int(length)))
-        finally:
-            snippets.close()
+            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, "...")
+            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


@@ -1232,8 +1273,15 @@ class Search(IndexStore):
         for found in tops.scoreDocs:
             doc = self.searcher.doc(found.doc)
             is_pdcounter = doc.get('is_pdcounter')
         for found in tops.scoreDocs:
             doc = self.searcher.doc(found.doc)
             is_pdcounter = doc.get('is_pdcounter')

-            if is_pdcounter:
-                tag = PDCounterAuthor.objects.get(id=doc.get('tag_id'))
+            category = doc.get('tag_category')
+            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
             else:
                 tag = catalogue.models.Tag.objects.get(id=doc.get("tag_id"))
                 # don't add the pdcounter tag if same tag already exists


@@ -1280,7 +1328,7 @@ class Search(IndexStore):
 
         return only_term
 
 
         return only_term
 

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


@@ -1292,14 +1340,14 @@ class Search(IndexStore):
             if prefix:
                 q = self.make_prefix_phrase(toks, field)
             else:
             if prefix:
                 q = self.make_prefix_phrase(toks, field)
             else:

-                q = self.make_term_query(toks, field)
+                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)
 
         return self.search_tags(top, no_book_cat, max_results=max_results, pdcounter=pdcounter)
 
             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, pdcounter=pdcounter)
 

-    def hint_books(self, string, max_results=50, prefix=True):
+    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.


@@ -1310,7 +1358,7 @@ class Search(IndexStore):
         if prefix:
             q = self.make_prefix_phrase(toks, 'title')
         else:
         if prefix:
             q = self.make_prefix_phrase(toks, 'title')
         else:

-            q = self.make_term_query(toks, 'title')
+            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)