Skip to content

Conversation

@a10y
Copy link

@a10y a10y commented Jul 12, 2016

What changes were proposed in this pull request?

Fix parquet filter pushdown from not reaching all the way down to the file level

Use of previous deprecated constructor defaults to null metadata, which
prevents pushdown from reaching the Parquet level.

How was this patch tested?

Looking at output of collects from SparkShell, before were printing warnings about CorruptStatistics, preventing pushing down filters to individual parquet files. Now able to use the metadata in each file to pushdown.

(If this patch involves UI changes, please attach a screenshot; otherwise, remove this)

@a10y a10y changed the title [PARQUET] Fix for Parquet filter pushdown [PARQUET] Fix for Vectorized Parquet filter pushdown Jul 12, 2016
@a10y a10y changed the title [PARQUET] Fix for Vectorized Parquet filter pushdown [SQL][PARQUET] Fix for Vectorized Parquet filter pushdown Jul 12, 2016
@hvanhovell
Copy link
Contributor

#14160 solves the same thing.

@a10y
Copy link
Author

a10y commented Jul 13, 2016

Yep, looks like the other one was closed by the committer. I saw Sean commented that this might need to be tested against 2.2, is that going to be necessary?

@a10y
Copy link
Author

a10y commented Jul 15, 2016

bump?

@a10y
Copy link
Author

a10y commented Jul 18, 2016

@hvanhovell Anything new on this front?

Use of previous deprecated constructor defaults to null metadata, which
prevents pushdown from reaching the Parquet level.
@a10y a10y force-pushed the bugfix/pushdown branch from f825ad7 to b6f395b Compare August 10, 2016 09:49
@hvanhovell
Copy link
Contributor

@andreweduffy not much. I'll trigger a test.

@liancheng could you take a look at this?

@hvanhovell
Copy link
Contributor

Could you open a JIRA or add the existing JIRA to the PR?

@a10y
Copy link
Author

a10y commented Aug 10, 2016

Actually it appears that since this was opened, later PR #14450 fixes this. Should be safe to close now.

@hvanhovell
Copy link
Contributor

Ok :)... Could you close then?

@a10y
Copy link
Author

a10y commented Aug 10, 2016

Yep, closing now

@a10y a10y closed this Aug 10, 2016
@SparkQA
Copy link

SparkQA commented Aug 10, 2016

Test build #3211 has finished for PR 14159 at commit e64251a.

  • This patch passes all tests.
  • This patch merges cleanly.
  • This patch adds no public classes.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants