Recopie de l'implémentation de ZCatalog.Catalog.Catalog.search, sans les imports.
[Plinn.git] / catalog.py
index fbe323f..382fc57 100644 (file)
@@ -99,7 +99,187 @@ InitializeClass(CatalogTool)
 
 class DelegatedCatalog(Catalog) :
     '''C'est ici qu'on délègue effectivement à Solr '''
 
 class DelegatedCatalog(Catalog) :
     '''C'est ici qu'on délègue effectivement à Solr '''
+    
     def search(self, query, sort_index=None, reverse=0, limit=None, merge=1):
     def search(self, query, sort_index=None, reverse=0, limit=None, merge=1):
-        return Catalog.search(self, query,
-                              sort_index=sort_index,
-                              reverse=reverse, limit=limit, merge=merge)
\ No newline at end of file
+        """Iterate through the indexes, applying the query to each one. If
+        merge is true then return a lazy result set (sorted if appropriate)
+        otherwise return the raw (possibly scored) results for later merging.
+        Limit is used in conjuntion with sorting or scored results to inform
+        the catalog how many results you are really interested in. The catalog
+        can then use optimizations to save time and memory. The number of
+        results is not guaranteed to fall within the limit however, you should
+        still slice or batch the results as usual."""
+
+        rs = None # resultset
+
+        # Indexes fulfill a fairly large contract here. We hand each
+        # index the query mapping we are given (which may be composed
+        # of some combination of web request, kw mappings or plain old dicts)
+        # and the index decides what to do with it. If the index finds work
+        # for itself in the query, it returns the results and a tuple of
+        # the attributes that were used. If the index finds nothing for it
+        # to do then it returns None.
+
+        # Canonicalize the request into a sensible query before passing it on
+        query = self.make_query(query)
+
+        cr = self.getCatalogPlan(query)
+        cr.start()
+
+        plan = cr.plan()
+        if not plan:
+            plan = self._sorted_search_indexes(query)
+
+        indexes = self.indexes.keys()
+        for i in plan:
+            if i not in indexes:
+                # We can have bogus keys or the plan can contain index names
+                # that have been removed in the meantime
+                continue
+
+            index = self.getIndex(i)
+            _apply_index = getattr(index, "_apply_index", None)
+            if _apply_index is None:
+                continue
+
+            cr.start_split(i)
+            limit_result = ILimitedResultIndex.providedBy(index)
+            if limit_result:
+                r = _apply_index(query, rs)
+            else:
+                r = _apply_index(query)
+
+            if r is not None:
+                r, u = r
+                # Short circuit if empty result
+                # BBB: We can remove the "r is not None" check in Zope 2.14
+                # once we don't need to support the "return everything" case
+                # anymore
+                if r is not None and not r:
+                    cr.stop_split(i, result=None, limit=limit_result)
+                    return LazyCat([])
+
+                # provide detailed info about the pure intersection time
+                intersect_id = i + '#intersection'
+                cr.start_split(intersect_id)
+                # weightedIntersection preserves the values from any mappings
+                # we get, as some indexes don't return simple sets
+                if hasattr(rs, 'items') or hasattr(r, 'items'):
+                    _, rs = weightedIntersection(rs, r)
+                else:
+                    rs = intersection(rs, r)
+
+                cr.stop_split(intersect_id)
+
+                # consider the time it takes to intersect the index result with
+                # the total resultset to be part of the index time
+                cr.stop_split(i, result=r, limit=limit_result)
+                if not rs:
+                    break
+            else:
+                cr.stop_split(i, result=None, limit=limit_result)
+
+        # Try to deduce the sort limit from batching arguments
+        b_start = int(query.get('b_start', 0))
+        b_size = query.get('b_size', None)
+        if b_size is not None:
+            b_size = int(b_size)
+
+        if b_size is not None:
+            limit = b_start + b_size
+        elif limit and b_size is None:
+            b_size = limit
+
+        if rs is None:
+            # None of the indexes found anything to do with the query
+            # We take this to mean that the query was empty (an empty filter)
+            # and so we return everything in the catalog
+            warnings.warn('Your query %s produced no query restriction. '
+                          'Currently the entire catalog content is returned. '
+                          'In Zope 2.14 this will result in an empty LazyCat '
+                          'to be returned.' % repr(cr.make_key(query)),
+                          DeprecationWarning, stacklevel=3)
+
+            rlen = len(self)
+            if sort_index is None:
+                sequence, slen = self._limit_sequence(self.data.items(), rlen,
+                    b_start, b_size)
+                result = LazyMap(self.instantiate, sequence, slen,
+                    actual_result_count=rlen)
+            else:
+                cr.start_split('sort_on')
+                result = self.sortResults(
+                    self.data, sort_index, reverse, limit, merge,
+                        actual_result_count=rlen, b_start=b_start,
+                        b_size=b_size)
+                cr.stop_split('sort_on', None)
+        elif rs:
+            # We got some results from the indexes.
+            # Sort and convert to sequences.
+            # XXX: The check for 'values' is really stupid since we call
+            # items() and *not* values()
+            rlen = len(rs)
+            if sort_index is None and hasattr(rs, 'items'):
+                # having a 'items' means we have a data structure with
+                # scores.  Build a new result set, sort it by score, reverse
+                # it, compute the normalized score, and Lazify it.
+
+                if not merge:
+                    # Don't bother to sort here, return a list of
+                    # three tuples to be passed later to mergeResults
+                    # note that data_record_normalized_score_ cannot be
+                    # calculated and will always be 1 in this case
+                    getitem = self.__getitem__
+                    result = [(score, (1, score, rid), getitem)
+                            for rid, score in rs.items()]
+                else:
+                    cr.start_split('sort_on')
+
+                    rs = rs.byValue(0) # sort it by score
+                    max = float(rs[0][0])
+
+                    # Here we define our getter function inline so that
+                    # we can conveniently store the max value as a default arg
+                    # and make the normalized score computation lazy
+                    def getScoredResult(item, max=max, self=self):
+                        """
+                        Returns instances of self._v_brains, or whatever is
+                        passed into self.useBrains.
+                        """
+                        score, key = item
+                        r=self._v_result_class(self.data[key])\
+                              .__of__(aq_parent(self))
+                        r.data_record_id_ = key
+                        r.data_record_score_ = score
+                        r.data_record_normalized_score_ = int(100. * score / max)
+                        return r
+
+                    sequence, slen = self._limit_sequence(rs, rlen, b_start,
+                        b_size)
+                    result = LazyMap(getScoredResult, sequence, slen,
+                        actual_result_count=rlen)
+                    cr.stop_split('sort_on', None)
+
+            elif sort_index is None and not hasattr(rs, 'values'):
+                # no scores
+                if hasattr(rs, 'keys'):
+                    rs = rs.keys()
+                sequence, slen = self._limit_sequence(rs, rlen, b_start,
+                    b_size)
+                result = LazyMap(self.__getitem__, sequence, slen,
+                    actual_result_count=rlen)
+            else:
+                # sort.  If there are scores, then this block is not
+                # reached, therefore 'sort-on' does not happen in the
+                # context of a text index query.  This should probably
+                # sort by relevance first, then the 'sort-on' attribute.
+                cr.start_split('sort_on')
+                result = self.sortResults(rs, sort_index, reverse, limit,
+                    merge, actual_result_count=rlen, b_start=b_start,
+                    b_size=b_size)
+                cr.stop_split('sort_on', None)
+        else:
+            # Empty result set
+            result = LazyCat([])
+        cr.stop()
+        return result