[jira] [Updated] (ARROW-9780) [Python] Legacy ParquetDataset gives confusing error message when passed a filter for non-directory dataset

["Joris Van den Bossche (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/ARROW-9780?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Joris Van den Bossche updated ARROW-9780:
-----------------------------------------
    Labels: dataset-parquet-legacy  (was: )

> [Python] Legacy ParquetDataset gives confusing error message when passed a filter for non-directory dataset
> -----------------------------------------------------------------------------------------------------------
>
>                 Key: ARROW-9780
>                 URL: https://issues.apache.org/jira/browse/ARROW-9780
>             Project: Apache Arrow
>          Issue Type: Improvement
>          Components: Python
>    Affects Versions: 1.0.0
>            Reporter: Joris Van den Bossche
>            Priority: Major
>              Labels: dataset-parquet-legacy
>
> Not highest priority since it's for the legacy dataset, but this is rather confusing:
> {code}
> In [2]: pq.ParquetDataset("test.parquet", filters=[("x", ">", 4)])   
> ...
> AttributeError: 'NoneType' object has no attribute 'filter_accepts_partition'
> {code}
> Basically the filters argument only makes sense for a directory (because the legacy dataset only filters on partitions), but it shouldn't error like that for non-directories.
> From https://github.com/dask/dask/issues/6512



--
This message was sent by Atlassian Jira
(v8.20.10#820010)