[
https://issues.apache.org/jira/browse/PARQUET-128?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14203458#comment-14203458
]
Yash Datta commented on PARQUET-128:
------------------------------------
For a 47 million row parquet table in spark-sql with schema with no
nested/repeating columns ; applying a simple filter :
select * from d_tup_parq_yash where id = 10; (which returns a single record)
or
select * from d_tup_parq_yash where id < 500; (which returns 500 records)
time taken before the patch : 4.4 seconds
time taken after the patch : 2.6 seconds
> Optimize the parquet RecordReader implementation when: A. filterpredicate is
> pushed down , B. filterpredicate is pushed down on a flat schema
> -----------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: PARQUET-128
> URL: https://issues.apache.org/jira/browse/PARQUET-128
> Project: Parquet
> Issue Type: Improvement
> Components: parquet-mr
> Affects Versions: 1.6.0rc2
> Reporter: Yash Datta
> Fix For: parquet-mr_1.6.0
>
>
> The RecordReader implementation currently will read all the columns before
> applying the filter predicate and deciding whether to keep the row or discard
> it.
> We can have a RecordReader which will only assemble the columns on which
> filters are applied (which are usually a few), then apply the filter and
> decide whether to keep the row or not , and then goes on to assemble the
> remaining columns or skip the remaining columns accordingly.
> Also for applications like spark sql , the schema usually applied is a flat
> one with no repeating or nested columns. In such cases, its better to have a
> light-weight, faster RecordReader.
> The performance improvement by this change is seen to be significant , and is
> better in case smaller number of rows are returned by filtering (which is
> usually the case) and there are many number of columns
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
