[jira] [Comment Edited] (PARQUET-1951) Allow different strategies to combine key values when merging parquet files

["satish (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/PARQUET-1951?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17249243#comment-17249243 ] 

satish edited comment on PARQUET-1951 at 12/14/20, 7:18 PM:
------------------------------------------------------------

Hi [~gszadovszky] Thanks for the feedback. We don't plan to run this on single machine. We have logic to divide small files in a partition into several groups. We have to 'stitch' the files in a group into a large file. My initial implementation (This is in my library, not in parquet) reads entire data in the group, deserializes and serializes the data to create large file. This is taking significant amount of time compared to parquet-tools merge primarily because of serDe.  Also, we want to stitch older data files for archival etc. The data for these files is not queried often, so we are ok with low performance and we have few workarounds specific to my usecase to workaround these issues.
I'm happy to contribute to add support for column index or bloom filters (I see some PRs adding similar support for other parquet-tools commands).

I also published a PR. [This PR|https://github.com/apache/parquet-mr/pull/847] is not making significant changes. so hope its acceptable.


was (Author: satishkotha):
Hi [~gszadovszky] Thanks for the feedback. We don't plan to run this on single machine. We have logic to divide small files in a partition into several groups. We have to 'stitch' the files in a group into a large file. My initial implementation (This is in my library, not in parquet) reads entire data in the group, deserializes and serializes the data to create large file. This is taking significant amount of time compared to parquet-tools merge primarily because of serDe.  Also, we want to stitch older data files for archival etc. The data for these files is not queried often, so we are ok with low performance and we have few workarounds specific to my usecase to workaround these issues.
I'm happy to contribute to add support for column index or bloom filters (I see we are adding similar support for other parquet-tools commands).

I also published a PR. [This PR|https://github.com/apache/parquet-mr/pull/847] is not making significant changes. so hope its acceptable.

> Allow different strategies to combine key values when merging parquet files
> ---------------------------------------------------------------------------
>
>                 Key: PARQUET-1951
>                 URL: https://issues.apache.org/jira/browse/PARQUET-1951
>             Project: Parquet
>          Issue Type: Improvement
>          Components: parquet-cli
>            Reporter: satish
>            Priority: Minor
>
> I work on Apache Hudi project. We store some additional metadata in parquet files (key range in the file, for example).  So the metadata is different in different parquet files that we want to merge these files. 
> Here is what I'm thinking:
> 1) Merge command takes additional command line option: --strategy <StrategyClassName>. 
> 2) We introduce new strategy class in parquet-hadoop to keep the same behavior as today.  
> We can extend that class and provide our custom implementation.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)