viirya commented on a change in pull request #23943: [SPARK-27034][SQL] Nested 
schema pruning for ORC
URL: https://github.com/apache/spark/pull/23943#discussion_r263240456
 
 

 ##########
 File path: sql/core/benchmarks/OrcNestedSchemaPruningBenchmark-results.txt
 ##########
 @@ -6,35 +6,35 @@ Java HotSpot(TM) 64-Bit Server VM 1.8.0_202-b08 on Mac OS X 
10.14.3
 Intel(R) Core(TM) i7-8750H CPU @ 2.20GHz
 Selection:                                Best Time(ms)   Avg Time(ms)   
Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 
------------------------------------------------------------------------------------------------------------------------
-Top-level column                                    113            196         
 89          8.8         113.0       1.0X
-Nested column                                      1316           1639         
240          0.8        1315.5       0.1X
+Top-level column                                    116            151         
 36          8.6         116.3       1.0X
+Nested column                                       544            604         
 31          1.8         544.5       0.2X
 
 Review comment:
   I think the read performance is more determined by persistent library than 
Spark side here. As Parquet, at Spark side we provide correctly pruned nested 
schema to ORC library. Pruning nested fields when reading data is done by ORC 
library. I'm not sure if we have much room to optimize at Spark side for this. 
Of course I'm open to any suggestion I'm missing right now.

----------------------------------------------------------------
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]


With regards,
Apache Git Services

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to