[
https://issues.apache.org/jira/browse/ARROW-16645?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17541866#comment-17541866
]
Joris Van den Bossche commented on ARROW-16645:
-----------------------------------------------
Thanks for the report. It's actually broader than just {{pa.NA}}, as we don't
yet accept any kind of pyarrow scalar in the constructors. For example
{code}
>>> arr = pa.array([1, 2, 3])
>>> scalar= arr[0]
>>> scalar
<pyarrow.Int64Scalar: 1>
>>> pa.array([scalar])
...
ArrowInvalid: Could not convert <pyarrow.Int64Scalar: 1> with type
pyarrow.lib.Int64Scalar: did not recognize Python value type when inferring an
Arrow data type
{code}
And for that general issue, we have ARROW-5295 tracking this.
There is maybe something to say to already special case pa.NA (which can go
alongside the handling of pd.NA/pd.NaT in {{PandasObjectIsNull}}), although I
think I would rather see a general solution for all scalars.
> [Python] Allow pa.array/pa.chunked_array to infer pa.NA when in a non pyarrow
> container
> ---------------------------------------------------------------------------------------
>
> Key: ARROW-16645
> URL: https://issues.apache.org/jira/browse/ARROW-16645
> Project: Apache Arrow
> Issue Type: Improvement
> Components: Python
> Affects Versions: 7.0.0
> Reporter: Matthew Roeschke
> Priority: Major
>
> Example:
>
>
> {code:java}
> In [15]: import pyarrow as pa
> In [16]: pa.array([1, pa.NA])
> ArrowInvalid: Could not convert <pyarrow.NullScalar: None> with type
> pyarrow.lib.NullScalar: did not recognize Python value type when inferring an
> Arrow data type{code}
>
> I would be great if this could be equivalent to
> {code:java}
> In [17]: pa.array([1, pa.NA], mask=[False, True])
> Out[17]:
> <pyarrow.lib.Int64Array object at 0x7fe49c172b80>
> [
> 1,
> null
> ]
> In [18]: pa.__version__
> Out[18]: '7.0.0'{code}
>
>
>
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
