[
https://issues.apache.org/jira/browse/ARROW-16431?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17531128#comment-17531128
]
Joris Van den Bossche commented on ARROW-16431:
-----------------------------------------------
[~multimeric] thanks for opening the issue. That would indeed be a welcome
improvement of the error message (or at least we could print the two schemas,
if it is not that easy to infer the exact difference between both).
> [C++][Parquet] Improve error message in append_row_groups() when appending
> disjoint metadata
> --------------------------------------------------------------------------------------------
>
> Key: ARROW-16431
> URL: https://issues.apache.org/jira/browse/ARROW-16431
> Project: Apache Arrow
> Issue Type: Improvement
> Components: C++, Parquet
> Reporter: Michael Milton
> Priority: Major
>
> Currently if you try to append together metadata from row groups with
> different schemas (?), you get the following error:
> {code:java}
> File
> "/home/mmilton/.conda/envs/mmilton/envs/driverpipe/lib/python3.9/site-packages/dask/dataframe/io/parquet/arrow.py",
> line 52, in _append_row_groups
> metadata.append_row_groups(md)
> File "pyarrow/_parquet.pyx", line 628, in
> pyarrow._parquet.FileMetaData.append_row_groups
> self._metadata.AppendRowGroups(deref(c_metadata))
> RuntimeError: AppendRowGroups requires equal schemas.
> {code}
> What would be useful here is to actually pass the schema difference in the
> error object in terms of which columns disagree. This information should
> _also_ be in the error message.
> For example if it said:
> {code:java}
> RuntimeError: AppendRowGroups requires equal schemas. Column "foo" was
> previously an int32 but the latest row group is storing it as an int64
> {code}
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
