[
https://issues.apache.org/jira/browse/DRILL-8249?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17565465#comment-17565465
]
ASF GitHub Bot commented on DRILL-8249:
---------------------------------------
jnturton opened a new pull request, #2594:
URL: https://github.com/apache/drill/pull/2594
# [DRILL-8249](https://issues.apache.org/jira/browse/DRILL-8249): Parquet
decoding error reading nation.dict.parquet from test framework.
## Description
Thank you to @ArtTrush for finding and partially debugging this regression.
The Parquet test file attached to the Jira issue cannot be queried with Drill
1.20.1 because it decides from column chunk metadata that a chunk does not use
dictionary encoding and ignores new information found by the Parquet page
reader that indicates that dictionary encoding is used. This fix restores the
consideration of the `usingDictionary` flag which is updated by the page reader.
## Documentation
N/A
## Testing
The Drill Test Framework contains a test query and file nation.dict.parquet
that reveals this bug.
> Parquet decoding error reading nation.dict.parquet from test framework
> ----------------------------------------------------------------------
>
> Key: DRILL-8249
> URL: https://issues.apache.org/jira/browse/DRILL-8249
> Project: Apache Drill
> Issue Type: Bug
> Components: Storage - Parquet
> Affects Versions: 1.20.0, 1.20.1
> Reporter: Artem Trush
> Assignee: Artem Trush
> Priority: Major
> Fix For: 1.20.2
>
> Attachments: nation.dict.parquet
>
>
> After one of the recent fix -
> [DRILL-1282|https://github.com/apache/drill/pull/2338] drill started to throw
> npe. Issue is related to drill 1.20 and above.
> Simple {{SELECT * FROM}} from {{nation.dict.parquet}} file. Attached it.
> Were investigated that the code
> [here|https://github.com/apache/drill/blob/master/exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/columnreaders/VarLenColumnBulkInput.java#L318L341]
> is not working as expected. Before DRILL-1282 these code lines were not used
> for reading this file (or I guess any file).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
