[
https://issues.apache.org/jira/browse/ARROW-16853?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexandre Cagnon updated ARROW-16853:
-------------------------------------
Description:
When applying pyarrow.array() to a list of dict, the dict keys are reordered in
ascending order.
Because of this I cannot preserve the original parquet data schema when doing
transformations from py to arrow.
Is the key reordering the intended behaviour ?
Is there an existing workaround ?
Eg.
x = [\\{'z': '', 'y': {'b': '', 'a': ''}}]
y = pyarrow.array( x )
print( y )
print(y.to_pylist())
was:
When applying pyarrow.array() to a list of dict, the dict keys are reordered in
ascending order.
Because of this I cannot preserve the original parquet data schema when doing
transformations from py to arrow.
Is the key reordering the intended behaviour ?
Is there an existing workaround ?
Eg.
x = [\{'z': '', 'y': {'b': '', 'a': ''}}]
y = pyarrow.array(x)
print(y)
print(y.to_pylist())
> pyarrow.Array affects dict keys order
> -------------------------------------
>
> Key: ARROW-16853
> URL: https://issues.apache.org/jira/browse/ARROW-16853
> Project: Apache Arrow
> Issue Type: Improvement
> Components: Python
> Affects Versions: 8.0.0
> Reporter: Alexandre Cagnon
> Priority: Minor
>
> When applying pyarrow.array() to a list of dict, the dict keys are reordered
> in ascending order.
> Because of this I cannot preserve the original parquet data schema when doing
> transformations from py to arrow.
> Is the key reordering the intended behaviour ?
> Is there an existing workaround ?
>
> Eg.
> x = [\\{'z': '', 'y': {'b': '', 'a': ''}}]
> y = pyarrow.array( x )
> print( y )
> print(y.to_pylist())
>
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
