[Python] Scalar subclasses should implement Python protocols

[pitrou]

Describe the enhancement requested

To make PyArrow scalars easier to use with non-PyArrow functions, we should define appropriate methods and protocols for specific scalar subclasses.

Examples:

• integer scalars should implement __int__
• floating-point scalars should implement __float__
• binary scalars should implement __bytes__ and perhaps the buffer protocol
• list-like scalars could implement Sequence
• map scalars could implement Mapping
• struct scalars could implement both Sequence and Mapping (?)

Note that some of this exists (unvoluntarily?) but is broken:

>>> a, b = pc.min_max([1,2,3])
>>> a  # expecting 1
'min'
>>> b  # expecting 3
'max'

Component(s)

Python

[pitrou]

cc @AlenkaF

[thisisnic]

Hey folks, @AlenkaF pointed me to this issue as a good place to get started contributing to PyArrow, so I was going to take a look at this. I might be a bit slow as I'll be spending some time getting acquainted with PyArrow in general, so I was going to spend the new few weeks looking at this. If there's any urgency, happy for someone else to take it off my hands though! :)

[pitrou]

There's definitely no urgency here, feel free to take a stab at this!

[pitrou]

Issue resolved by pull request 45818
#45818