[
https://issues.apache.org/jira/browse/PARQUET-325?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ryan Blue resolved PARQUET-325.
-------------------------------
Resolution: Fixed
Merged #234. Thanks, Tianshuo!
> Do not target row group sizes if padding is set to 0
> ----------------------------------------------------
>
> Key: PARQUET-325
> URL: https://issues.apache.org/jira/browse/PARQUET-325
> Project: Parquet
> Issue Type: Bug
> Components: parquet-mr
> Reporter: Ryan Blue
> Assignee: Ryan Blue
> Fix For: 1.8.0
>
>
> The new padding commit will target row group sizes to the remaining space in
> a block if the space remaining is less than the padding size. This is the
> intended behavior if padding is set, but if padding is 0, then it will target
> a row group to be in any remaining bytes, even if that makes no sense.
> The behavior when padding is 0 should be to always target row groups to the
> default row group size. This will match the existing behavior up to 1.8.0.
> (In the error case, row groups will still span the block boundary, but will
> most likely be created with the minimum number of rows before checking
> whether or not it should be flushed.)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
