Difference between Jena/Joseki and sem_match

Question

Dear All,
There are different ways for us fire SPARQL queries into the Spatial and Graph extension. The two methods we are using at this moment are Jena/Joseki and sem_match. Where Jena/Joseki is an end-point on a WLS.
The same SPARQL queries fired either by sem_match or Jena/Joseki do not result in the same execution plan. We are wondering if it is possible in anyway to let Jena/Joseki use the same rules as sem_match to receive the same execution plan.
If there is not such a method, is there than a plan to bring these two optimizers closer together?
Kind regards,
Max

Answer 1

Hi Max,
Can you please adding s2s and strict_default flags to the beginning of your query?
e.g.
PREFIX ORACLE_SEM_FS_NS: <http://oracle.com/semtech#s2s=f,strict_default=f>
select ?s ?p ?o
where {
  graph ?g {?s ?p ?o }
To answer your question on convergence, Jena Adapter uses SEM_MATCH to answer SPARQL queries.
By default, however, Jena Adapter first constructs a SEM_MATCH and then invokes a server-side logic to translate
the SEM_MATCH in a pure SQL query. This feature is called S2S. S2S has a few benefits including the support
of large queries, read-only database, and a reduced cost in query parsing.
Now, there are a few things/situations that can cause a difference in the execution plan.
- Jena Adapter by default chooses strict_default=t to align with Jena's behavior.
- Given a SPARQL 1.1 query (which uses syntax beyond SPARQL 1.0), Jena Adapter (for Oracle Database 11.2.0.3+ and 12c)  
  breaks up the query into smaller pieces.
We are working on a newer version of Jena Adapter that will translate a complete SPARQL 1.1 query into a single SEM_MATCH for Oracle Database (only for 12c+).
Hope it helps,
Zhe Wu