Augusto Radtke created ARROW-2722:
-------------------------------------
Summary: ndarray to arrow conversion fails when downcasted from
pandas to_numeric
Key: ARROW-2722
URL: https://issues.apache.org/jira/browse/ARROW-2722
Project: Apache Arrow
Issue Type: Bug
Components: C++, Python
Affects Versions: 0.9.0
Environment: Windows 10 64-bit
Reporter: Augusto Radtke
The following snippet:
{code:java}
import numpy as np
import pandas as pd
import pyarrow as pa
pa.array(pd.to_numeric(pd.Series(np.array([65536,2,3], dtype=np.uint64)),
downcast='unsigned'),
from_pandas=True, type='uint32')
{code}
fails to convert with message:
{noformat}
ArrowNotImplementedError Traceback (most recent call last)
<ipython-input-2-b259c5cb7044> in <module>()
4
5 pa.array(pd.to_numeric(pd.Series(np.array([65536,2,3], dtype=np.uint64)),
downcast='unsigned'),
----> 6 from_pandas=True, type='uint32')
array.pxi in pyarrow.lib.array()
array.pxi in pyarrow.lib._ndarray_to_array()
error.pxi in pyarrow.lib.check_status()
ArrowNotImplementedError: Unsupported numpy type 6{noformat}
This is a Windows 64-bit machine, running Python 3.6.5, pyarrow 0.9.0, pandas
0.23.1 and numpy 1.14.5.
Seems to be fine for uint16 or uint8 downcasting. Unfortunately I didn't had
the time to dig deeper or try on a Linux machine but it feels like its related
to the LLP64 model.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
