jorisvandenbossche commented on issue #33843:
URL: https://github.com/apache/arrow/issues/33843#issuecomment-1401679320
@jbrockmendel Using the `divide` kernel with integers does "integer
division" (which is the standard behaviour in C++), so like `//` in python:
```
In [26]: pc.divide([1, 2, 3, 4, 5], 3).to_pylist()
Out[26]: [0, 0, 1, 1, 1]
In [27]: np.array([1, 2, 3, 4, 5]) // 3
Out[27]: array([0, 0, 1, 1, 1])
```
Numpy and pandas actually always return floats for standard division with
integers. If you want this with Arrow, I think the only workaround right now is
to explicitly cast one of both operands to float to trigger standard floating
point division (we should maybe consider enabling float output without having
to do this cast).
For integer division, you can't "behave as float", since there is no Inf or
NaN for integers. It seems that numpy then uses 0 though:
```
In [29]: np.array([0, 1, 2]) // 0
<ipython-input-29-cf533e0a640a>:1: RuntimeWarning: divide by zero
encountered in floor_divide
np.array([0, 1, 2]) // 0
Out[29]: array([0, 0, 0])
```
But I don't know if we want to remove the divide-by-zero error from the
non-checked variant and return 0 instead.
Similar previous issue: https://github.com/apache/arrow/issues/29427
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
