Remove unnecessary calls to PyNumber_Index, PyLong_Checks

[Starbuck5]

This simplifies code and improves performance. It makes rect[x] ~7% faster and rect[x] = y ~9% faster. Although these functions are really cheap, and therefore I would be surprised if there was any measurable macro impact from this change. You can run them 10 million times in less than a second. I didn't test performance impact in pixelarray.

And in case anyone is curious rect.y is ~8% faster than rect[1] (after this PR). So for those micro-optimizers out there, that's the strategy to use.

PyNumber_AsSsize_t doesn't need the value to be converted to an "index int" first.

If you look at the CPython source, in Objects/abstract.c:

Py_ssize_t
PyNumber_AsSsize_t(PyObject *item, PyObject *err)
{
    Py_ssize_t result;
    PyObject *runerr;
    PyObject *value = _PyNumber_Index(item);
    if (value == NULL)
        return -1;
    ...

The first thing it does is use PyNumber_Index anyway!

Plus it is documented as "Returns o converted to a Py_ssize_t value if o can be interpreted as an integer". Having an __index__ method counts as being interpretable as an integer.

PyIndex_Check(op) || PyLong_Check(op), the PyLong_Check is unnecessary.

Python ints already pass PyIndex_Check, as you can tell because Python ints work in the Rect functions.

[yunline]

LGTM

Testing code:

import pygame
pygame.init()

class IndexLike:
    def __index__(self):
        raise RuntimeError("Nope")

r = pygame.Rect(0,0,1,1)

r[1<<256] # IndexError: Invalid rect Index
r[IndexLike()] # RuntimeError: Nope

[ankith26]

LGTM, thanks! 🎈