AngersZhuuuu commented on pull request #28805:
URL: https://github.com/apache/spark/pull/28805#issuecomment-647368853
> > in pr you mentioned, it split join condition by qualifier
>
> I'm a bit confused. That PR is not related to join at all.
Sorry I make I mistake. In current master branch, if we create a DS table
```
create table test_cnf(id string, dt string) using parquet partitioned by
(dt);
```
If we query with a or condition
```
spark-sql> explain select * from test_cnf where (dt='1') or (dt='2' and id =
'3');
== Physical Plan ==
*(1) Filter ((dt#50 = 1) OR ((dt#50 = 2) AND (id#49 = 3)))
+- *(1) ColumnarToRow
+- FileScan parquet default.test_cnf[id#49,dt#50] Batched: true,
DataFilters: [], Format: Parquet, Location:
CatalogFileIndex[file:/Users/angerszhu/Documents/project/AngersZhu/spark/spark-warehouse/test_cnf],
PartitionFilters: [], PushedFilters: [], ReadSchema: struct<id:string>
```
In partition pruning level, it is still not good for this case, and pr
https://github.com/apache/spark/pull/24598 fix similar problem for column
filter level , not partition pruning level.
I did this for partition pruning level, for DataSource and HiveTableScan
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
