[
https://issues.apache.org/jira/browse/ARROW-12060?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17456499#comment-17456499
]
Joris Van den Bossche commented on ARROW-12060:
-----------------------------------------------
I would like to get back to this.
> I think it would confuse me if "pc.log10(arr)" returned an array and
> "pc.log10(ds.field('a'))" returned an expression.
A kernel like {{pc.log10}} already returns different things depending on the
input (an array as input gives an array as result, a scalar as input gives a
scalar as result, ..). So that it would return an expression for an expression
as input is also not necessarily that strange IMO.
Now, this also gives some "smelly" interdependence between the compute and
dataset module (eg, if you want to pass FunctionOptions to the
{{Expression.\_call}}, the dataset cython module needs to depend on the cython
_compute module)
> [Python] Enable calling compute functions on Expressions
> --------------------------------------------------------
>
> Key: ARROW-12060
> URL: https://issues.apache.org/jira/browse/ARROW-12060
> Project: Apache Arrow
> Issue Type: Sub-task
> Components: Python
> Reporter: Joris Van den Bossche
> Assignee: Joris Van den Bossche
> Priority: Major
> Fix For: 7.0.0
>
>
> To expose the full power of dataset (projection/filter) expressions, we
> should ensure that all compute kernels can be used in combination with
> expressions.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
