[
https://issues.apache.org/jira/browse/ARROW-11279?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17266688#comment-17266688
]
R J commented on ARROW-11279:
-----------------------------
I think this can be solved by incrementally calculating array levels per column
as definition levels appear distinct per column. First time contributing but I
have prepared a pull request
> [Rust][Parquet] ArrowWriter Definition Levels Memory Usage
> ----------------------------------------------------------
>
> Key: ARROW-11279
> URL: https://issues.apache.org/jira/browse/ARROW-11279
> Project: Apache Arrow
> Issue Type: Improvement
> Components: Rust
> Reporter: R J
> Priority: Minor
> Original Estimate: 0.25h
> Remaining Estimate: 0.25h
>
> In the rust implementation of an Arrow RecordBatch writer to parquet
> (3.0.0-SNAPSHOT), the ArrowWriter::write call potentially allocates more
> memory than required.
> For a RecordBatch with m rows and n columns, ArrowWriter::write allocates m*n
> definition levels, leading to m times the required memory usage.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
