Optimized Rect.clipline()

[itzpr3d4t0r]

This PR improves the speed of Rect.clipline() by 45-50% when the line intersects the rectangle, with no noticeable change when there's no intersection. The optimization comes from replacing PyBuildValue with a custom implementation to construct the return tuple.

I've also added a new macro that enables easy dispatching between the double/int custom tuple packing function we have.

I got these results:
Old

Colliding
2.2999593056738377e-06 2.571070200065151e-06
2.3999600671231747e-06 2.5393099116627127e-06
2.2999593056738377e-06 2.440720208687708e-06
2.2999593056738377e-06 2.4487898976076393e-06
2.3999600671231747e-06 2.4846897984389216e-06
Non Colliding
8.999486453831196e-07 1.0697301593609154e-06
8.999486453831196e-07 9.743399394210428e-07
8.999486453831196e-07 9.719599853269757e-07
8.999486453831196e-07 9.75440168986097e-07
8.999486453831196e-07 1.03677986189723e-06

New

Colliding
1.5999539755284786e-06 1.6530799737665803e-06
1.5999539755284786e-06 1.7165300087071954e-06
1.5999539755284786e-06 1.801859913393855e-06
1.5999539755284786e-06 1.6924298892263323e-06
1.5999539755284786e-06 1.8196600314695388e-06
Non Colliding
8.999486453831196e-07 1.036960061173886e-06
8.999486453831196e-07 1.0293898812960834e-06
8.999486453831196e-07 1.0853600455448032e-06
8.999486453831196e-07 1.0411400173325092e-06
8.999486453831196e-07 1.0053500125650316e-06

Using this test program:

from timeit import repeat
from statistics import fmean

from pygame import Rect

r = Rect(100, 100, 100, 100)

G = globals()

print("Colliding")
for _ in range(5):
    data = repeat("r.clipline(0, 0, 105, 150)", number=10, repeat=10000, globals=G)
    print(f"{min(data)} {fmean(data)}")

print("Non Colliding")
for _ in range(5):
    data = repeat("r.clipline(0, 0, 1000, 10)", number=10, repeat=10000, globals=G)
    print(f"{min(data)} {fmean(data)}")

[damusss]

Nice optimization! :)

[ankith26]

LGTM, thanks for the PR 🎉