tanelk commented on pull request #31024:
URL: https://github.com/apache/spark/pull/31024#issuecomment-757633291


   Thats strange, I ran the same test and got this optimized plan for my 
example:
   
   ```
   == Optimized Logical Plan ==
   Filter (((b#3653L > 10000) AND (a#3652L > 10)) OR ((c#3654L < 10) AND 
(b#3653L < 10000))), Statistics(sizeInBytes=312.4 KiB, rowCount=1.00E+4)
   +- Relation[a#3652L,b#3653L,c#3654L] parquet, Statistics(sizeInBytes=625.0 
KiB, rowCount=2.00E+4)
   ````
   
   Something the test does not illustrate is that my proposition can swap the 
sides of the OR.
   
   > ```scala
   > spark.range(20000).selectExpr("id as a", "id as b", "id as 
c").write.saveAsTable("t1")
   > spark.sql("ANALYZE TABLE t1 COMPUTE STATISTICS FOR ALL COLUMNS")
   > spark.sql("set spark.sql.cbo.enabled=true")
   > spark.sql(s"SELECT * FROM t1 WHERE (a > 10 and b > 10000) or (c < 10 and b 
< 10000)").explain("cost")
   > ```
   > 
   > My change result:
   > 
   > ```
   > == Optimized Logical Plan ==
   > Filter (((b#3193L > 10000) AND (a#3192L > 10)) OR ((c#3194L < 10) AND 
(b#3193L < 10000))), Statistics(sizeInBytes=312.4 KiB, rowCount=1.00E+4)
   > +- Relation[a#3192L,b#3193L,c#3194L] parquet, Statistics(sizeInBytes=625.0 
KiB, rowCount=2.00E+4)
   > ```
   > 
   > @tanelk's change result:
   > 
   > ```
   > == Optimized Logical Plan ==
   > Filter (((a#3192L > 10) AND (b#3193L > 10000)) OR ((c#3194L < 10) AND 
(b#3193L < 10000))), Statistics(sizeInBytes=312.4 KiB, rowCount=1.00E+4)
   > +- Relation[a#3192L,b#3193L,c#3194L] parquet, Statistics(sizeInBytes=625.0 
KiB, rowCount=2.00E+4)
   > ```
   > 
   > It seems my change is better.


----------------------------------------------------------------
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:
us...@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org

Reply via email to