Motivation

Now, Doris's OLAP_SCAN_NODE can push the conjuncts condition to the storage engine.
But after push it down to storage engine, the conditions still need to be filtered in the OLAP_SCANNER.

It's is a very unreasonable way. Therefore, we need to figure out the filtering conditions that have been pushed down to the storage engine. Delete these pushed down conditions on the OLAP SCANNER. We should not waste CPU resources on the conditions that have been filtered.

It can bring us the following benefits:

I Had Do POC in the data of SSB:

select * from customer where C_NATION = "CHINA" order by C_CUSTKEY limit 10, 15; 

There are 3000W rows of data

It's about 20% performance improvement