X-Git-Url: https://git.mdrn.pl/wolnelektury.git/blobdiff_plain/442e7667f2d3ae96b873dacc1c7f980765ecc57a..159b148e9d37274930d6c0f76c365d333f16fd87:/src/search/index.py

diff --git a/src/search/index.py b/src/search/index.py
index 4c278eab5..f943a4d1d 100644
--- a/src/search/index.py
+++ b/src/search/index.py
@@ -10,16 +10,24 @@ from librarian import dcparser
 from librarian.parser import WLDocument
 from lxml import etree
 import catalogue.models
+import picture.models
 from pdcounter.models import Author as PDCounterAuthor, BookStub as PDCounterBook
 from itertools import chain
-import sunburnt
-import custom
+import scorched
+from . import custom
 import operator
 import logging
 from wolnelektury.utils import makedirs
 
 log = logging.getLogger('search')
 
+if os.path.isfile(settings.SOLR_STOPWORDS):
+    stopwords = set(
+        line.decode('utf-8').strip()
+        for line in open(settings.SOLR_STOPWORDS) if not line.startswith('#'))
+else:
+    stopwords = set()
+
 
 class SolrIndex(object):
     def __init__(self, mode=None):
@@ -121,7 +129,7 @@ class Index(SolrIndex):
         """
         uids = set()
         for q in queries:
-            if isinstance(q, sunburnt.search.LuceneQuery):
+            if isinstance(q, scorched.search.LuceneQuery):
                 q = self.index.query(q)
             q.field_limiter.update(['uid'])
             st = 0
@@ -297,7 +305,6 @@ class Index(SolrIndex):
             book_info = dcparser.parse(open(book.xml_file.path))
 
         fields['slug'] = book.slug
-        fields['tags'] = [t.name for t in book.tags]
         fields['is_book'] = True
 
         # validator, name
@@ -317,9 +324,9 @@ class Index(SolrIndex):
                 elif type_indicator == dcparser.as_person:
                     p = getattr(book_info, field.name)
                     if isinstance(p, dcparser.Person):
-                        persons = unicode(p)
+                        persons = str(p)
                     else:
-                        persons = ', '.join(map(unicode, p))
+                        persons = ', '.join(map(str, p))
                     fields[field.name] = persons
                 elif type_indicator == dcparser.as_date:
                     dt = getattr(book_info, field.name)
@@ -471,7 +478,7 @@ class Index(SolrIndex):
                         fid = start.attrib['id'][1:]
                         handle_text.append(lambda text: None)
                         if start.text is not None:
-                            fragments[fid]['themes'] += map(unicode.strip, map(unicode, (start.text.split(','))))
+                            fragments[fid]['themes'] += map(str.strip, map(str, (start.text.split(','))))
                     elif end is not None and end.tag == 'motyw':
                         handle_text.pop()
 
@@ -508,6 +515,47 @@ class Index(SolrIndex):
         finally:
             snippets.close()
 
+    def remove_picture(self, picture_or_id):
+        """Removes a picture from search index."""
+        if isinstance(picture_or_id, picture.models.Picture):
+            picture_id = picture_or_id.id
+        else:
+            picture_id = picture_or_id
+        self.delete_query(self.index.Q(picture_id=picture_id))
+
+    def index_picture(self, picture, picture_info=None, overwrite=True):
+        """
+        Indexes the picture.
+        Creates a lucene document for extracted metadata
+        and calls self.index_area() to index the contents of the picture.
+        """
+        if overwrite:
+            # we don't remove snippets, since they might be still needed by
+            # threads using not reopened index
+            self.remove_picture(picture)
+
+        picture_doc = {'picture_id': int(picture.id)}
+        meta_fields = self.extract_metadata(picture, picture_info, dc_only=[
+            'authors', 'title', 'epochs', 'kinds', 'genres'])
+
+        picture_doc.update(meta_fields)
+
+        picture_doc['uid'] = "picture%s" % picture_doc['picture_id']
+        self.index.add(picture_doc)
+        del picture_doc['is_book']
+        for area in picture.areas.all():
+            self.index_area(area, picture_fields=picture_doc)
+
+    def index_area(self, area, picture_fields):
+        """
+        Indexes themes and objects on the area.
+        """
+        doc = dict(picture_fields)
+        doc['area_id'] = area.id
+        doc['themes'] = list(area.tags.filter(category__in=('theme', 'thing')).values_list('name', flat=True))
+        doc['uid'] = 'area%s' % area.id
+        self.index.add(doc)
+
 
 class SearchResult(object):
     def __init__(self, doc, how_found=None, query_terms=None):
@@ -551,14 +599,25 @@ class SearchResult(object):
 
             self._hits.append(hit)
 
-    def __unicode__(self):
+    @classmethod
+    def from_book(cls, book, how_found=None, query_terms=None):
+        doc = {
+            'score': book.popularity.count,
+            'book_id': book.id,
+            'published_date': 0,
+        }
+        result = cls(doc, how_found=how_found, query_terms=query_terms)
+        result._book = book
+        return result
+
+    def __str__(self):
         return u"<SR id=%d %d(%d) hits score=%f %d snippets>" % \
             (self.book_id, len(self._hits),
              len(self._processed_hits) if self._processed_hits else -1,
              self._score, len(self.snippets))
 
-    def __str__(self):
-        return unicode(self).encode('utf-8')
+    def __bytes__(self):
+        return str(self).encode('utf-8')
 
     @property
     def score(self):
@@ -566,10 +625,9 @@ class SearchResult(object):
 
     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))
+            raise ValueError("this search result is for book %d; tried to merge with %d" % (self.book_id, other.book_id))
         self._hits += other._hits
-        if other.score > self.score:
-            self._score = other._score
+        self._score += max(other._score, 0)
         return self
 
     def get_book(self):
@@ -647,7 +705,7 @@ class SearchResult(object):
             if self.query_terms is not None:
                 for i in range(0, len(f[self.OTHER]['themes'])):
                     tms = f[self.OTHER]['themes'][i].split(r' +') + f[self.OTHER]['themes_pl'][i].split(' ')
-                    tms = map(unicode.lower, tms)
+                    tms = map(str.lower, tms)
                     for qt in self.query_terms:
                         if qt in tms:
                             themes_hit.add(f[self.OTHER]['themes'][i])
@@ -708,6 +766,110 @@ class SearchResult(object):
             return None
 
 
+class PictureResult(object):
+    def __init__(self, doc, how_found=None, query_terms=None):
+        self.boost = 1.0
+        self.query_terms = query_terms
+        self._picture = None
+        self._hits = []
+        self._processed_hits = None
+
+        if 'score' in doc:
+            self._score = doc['score']
+        else:
+            self._score = 0
+
+        self.picture_id = int(doc["picture_id"])
+
+        if doc.get('area_id'):
+            hit = (self._score, {
+                'how_found': how_found,
+                'area_id': doc['area_id'],
+                'themes': doc.get('themes', []),
+                'themes_pl': doc.get('themes_pl', []),
+            })
+
+            self._hits.append(hit)
+
+    def __str__(self):
+        return u"<PR id=%d score=%f >" % (self.picture_id, self._score)
+
+    def __repr__(self):
+        return str(self)
+
+    @property
+    def score(self):
+        return self._score * self.boost
+
+    def merge(self, other):
+        if self.picture_id != other.picture_id:
+            raise ValueError(
+                "this search result is for picture %d; tried to merge with %d" % (self.picture_id, other.picture_id))
+        self._hits += other._hits
+        self._score += max(other._score, 0)
+        return self
+
+    SCORE = 0
+    OTHER = 1
+
+    @property
+    def hits(self):
+        if self._processed_hits is not None:
+            return self._processed_hits
+
+        hits = []
+        for hit in self._hits:
+            try:
+                area = picture.models.PictureArea.objects.get(id=hit[self.OTHER]['area_id'])
+            except picture.models.PictureArea.DoesNotExist:
+                # stale index
+                continue
+            # Figure out if we were searching for a token matching some word in theme name.
+            themes_hit = set()
+            if self.query_terms is not None:
+                for i in range(0, len(hit[self.OTHER]['themes'])):
+                    tms = hit[self.OTHER]['themes'][i].split(r' +') + hit[self.OTHER]['themes_pl'][i].split(' ')
+                    tms = map(str.lower, tms)
+                    for qt in self.query_terms:
+                        if qt in tms:
+                            themes_hit.add(hit[self.OTHER]['themes'][i])
+                            break
+
+            m = {
+                'score': hit[self.SCORE],
+                'area': area,
+                'themes_hit': themes_hit,
+            }
+            m.update(hit[self.OTHER])
+            hits.append(m)
+
+        hits.sort(key=lambda h: h['score'], reverse=True)
+        hits = hits[:1]
+        self._processed_hits = hits
+        return hits
+
+    def get_picture(self):
+        if self._picture is None:
+            self._picture = picture.models.Picture.objects.get(id=self.picture_id)
+        return self._picture
+
+    picture = property(get_picture)
+
+    @staticmethod
+    def aggregate(*result_lists):
+        books = {}
+        for rl in result_lists:
+            for r in rl:
+                if r.picture_id in books:
+                    books[r.picture_id].merge(r)
+                else:
+                    books[r.picture_id] = r
+        return books.values()
+
+    def __cmp__(self, other):
+        return cmp(self.score, other.score)
+
+
 class Search(SolrIndex):
     """
     Search facilities.
@@ -728,23 +890,51 @@ class Search(SolrIndex):
 
         return q
 
-    def search_words(self, words, fields, book=True):
+    def search_by_author(self, words):
+        from catalogue.models import Book
+        books = Book.objects.filter(parent=None).order_by('-popularity__count')
+        for word in words:
+            books = books.filter(cached_author__iregex='\m%s\M' % word).select_related('popularity__count')
+        return [SearchResult.from_book(book, how_found='search_by_author', query_terms=words) for book in books[:30]]
+
+    def search_words(self, words, fields, required=None, book=True, picture=False):
+        if book and not picture and fields == ['authors']:
+            return self.search_by_author(words)
         filters = []
         for word in words:
-            word_filter = None
-            for field in fields:
-                q = self.index.Q(**{field: word})
-                if word_filter is None:
-                    word_filter = q
-                else:
-                    word_filter |= q
-            filters.append(word_filter)
+            if book or picture or (word not in stopwords):
+                word_filter = None
+                for field in fields:
+                    q = self.index.Q(**{field: word})
+                    if word_filter is None:
+                        word_filter = q
+                    else:
+                        word_filter |= q
+                filters.append(word_filter)
+        if required:
+            required_filter = None
+            for field in required:
+                for word in words:
+                    if book or picture or (word not in stopwords):
+                        q = self.index.Q(**{field: word})
+                        if required_filter is None:
+                            required_filter = q
+                        else:
+                            required_filter |= q
+            filters.append(required_filter)
+        if not filters:
+            return []
+        params = {}
         if book:
-            query = self.index.query(is_book=True)
+            params['is_book'] = True
+        if picture:
+            params['picture_id__gt'] = 0
         else:
-            query = self.index.query()
+            params['book_id__gt'] = 0
+        query = self.index.query(**params)
         query = self.apply_filters(query, filters).field_limit(score=True, all_fields=True)
-        return [SearchResult(found, how_found='search_words', query_terms=words) for found in query.execute()]
+        result_class = PictureResult if picture else SearchResult
+        return [result_class(found, how_found='search_words', query_terms=words) for found in query.execute()]
 
     def get_snippets(self, searchresult, query, field='text', num=1):
         """
@@ -767,13 +957,15 @@ class Search(SolrIndex):
                 text = snippets.get((int(position),
                                      int(length)))
                 snip = self.index.highlight(text=text, field=field, q=query)
+                if not snip and field == 'text':
+                    snip = self.index.highlight(text=text, field='text_nonstem', q=query)
                 if snip not in snips:
                     snips[idx] = snip
                     if snip:
                         num -= 1
                 idx += 1
 
-        except IOError, e:
+        except IOError as e:
             book = catalogue.models.Book.objects.filter(id=book_id)
             if not book:
                 log.error("Book does not exist for book id = %d" % book_id)