[
https://issues.apache.org/jira/browse/SPARK-17091?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16065413#comment-16065413
]
Michael Styles edited comment on SPARK-17091 at 6/27/17 8:26 PM:
-----------------------------------------------------------------
By not pushing the filter to Parquet, are we not preventing Parquet from
skipping blocks during read operations? I have tests that show big improvements
when applying this transformation.
For instance, I have a Parquet file with 162,456,394 rows which is sorted on
column C1.
*IN Predicate*
{noformat}
df.filter[df['C1'].isin([42, 139])).collect()
{noformat}
*OR Predicate*
{noformat}
df.filter((df['C1'] == 42) | (df['C1'] == 139)).collect()
{noformat}
I'm seeing about a 50 -75 % improvement. See attachments.
was (Author: ptkool):
By not pushing the filter to Parquet, are we not preventing Parquet from
skipping blocks during read operations? I have tests that show big improvements
when applying this transformation.
For instance, I have a Parquet file with 162,456,394 rows which is sorted on
column C1.
*IN Predicate*
{noformat}
df.filter[df['C1'].isin([42, 139])).collect()
{noformat}
*OR Predicate*
{noformat}
df.filter((df['C1'] == 42) | (df['C1'] == 139)).collect()
{noformat}
I'm seeing about a 50 -75 % improvement.
> Convert IN predicate to equivalent Parquet filter
> -------------------------------------------------
>
> Key: SPARK-17091
> URL: https://issues.apache.org/jira/browse/SPARK-17091
> Project: Spark
> Issue Type: Bug
> Reporter: Andrew Duffy
> Attachments: IN Predicate.png, OR Predicate.png
>
>
> Past attempts at pushing down the InSet operation for Parquet relied on
> user-defined predicates. It would be simpler to rewrite an IN clause into the
> corresponding OR union of a set of equality conditions.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
