[
https://issues.apache.org/jira/browse/SPARK-40011?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17577693#comment-17577693
]
Hyukjin Kwon commented on SPARK-40011:
--------------------------------------
Pandas API on Spark needs that because it reuses several metadata objects from
pandas, and that's explicitly declared at
https://github.com/apache/spark/blob/master/python/setup.py#L271-L275.
> Pandas API on Spark requires Pandas
> -----------------------------------
>
> Key: SPARK-40011
> URL: https://issues.apache.org/jira/browse/SPARK-40011
> Project: Spark
> Issue Type: Bug
> Components: Pandas API on Spark
> Affects Versions: 3.3.0
> Reporter: Daniel Oakley
> Priority: Major
>
> Pandas API on Spark includes code like:
> > import pandas as pd
> > from pandas.api.types import is_hashable, is_list_like # type:
> > ignore[attr-defined]
> This breaks if you don't have pandas installed on your Spark cluster.
> Pandas API was supposed to be an API not pandas integration, why does it
> require pandas to be installed?
> In many places Spark jobs may be run on various Spark clusters with no
> assurance of particular Python packages installed at a root level.
> Can this dependency be removed? Or the required version of Pandas be bundled
> with the Spark distribution? Similar for numpy and other deps.
> If not the docs should clearly state it is not merely a Spark API that mirror
> the Pandas API, but something quite different.
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
