search: word1 word2 word3 to use AND-ing instead of OR-ing

[tiborsimko]

Current behaviour

Search for cms event. You get 5040 hits including PHENIX record etc which are not what users would expect.

This is because searching for cms event is equivalent to searching for cms OR event, followed by appied ranking.

If you look for cms AND event, you get 1038 hits which are closer to what users expect. (I.e. no PHENIX record, etc).

Expected behaviour

When a user searches for word1 word2 word3, one would expect to do word1 AND word2 AND word3 query instead of current word1 OR word2 OR word3. The former is much more common in general search engines.

Let's change the behaviour to "silent AND" instead of "silent OR" for multi-word queries.

[psaiz]

Our inspire colleagues do already the AND. I like that behaviour much better.

[psaiz]

This comes from the default parser. A change like:

--- a/invenio_records_rest/query.py
+++ b/invenio_records_rest/query.py
@@ -25,7 +25,7 @@ def default_search_factory(self, search, query_parser=None):
     def _default_parser(qstr=None):
         """Default parser that uses the Q() from search engine dsl."""
         if qstr:
-            return dsl.Q("query_string", query=qstr)
+            return dsl.Q("query_string", query=qstr, default_operator='AND')

would do the trick. I'll check with our invenio colleagues if this is worth a new parameter in records_rest, or if we shouldjust pass our parser.

[psaiz]

This one depends on two other:

• Ensure that the pipelines check the code style with black #3459
• search: possibility to specify a different query_parser inveniosoftware/invenio-records-rest#329