Optimized Rect/FRect pgRect_FromObject

[itzpr3d4t0r]

I have ported #1843 since the last addition of FRect. I also updated my test program to produce more compact results. This PR should speed up all functions that accept a Rect-like as param. Actual Rect/FRect object as parameters are not affected.
This means that functions that take sequences like (x, y, w, h) or ((x, y), (w, h)) should benefit from this.
Results:

FRect
OLD

FRect (ms)	4a	1a	2a	1a 2s	r attr	r meth
instantiation	142.775	148.58	151.93	165.547	258.922	321.953
update	108.586	119.54	129.14	150.608	241.638	320.861
union	128.721	139.383	144.534	164.283	264.903	330.44
union_ip	110.288	122.154	131.417	156.187	236.39	319.557
clip	124.958	135.148	144.719	169.749	266.152	331.327
colliderect	68.5006	90.9471	82.5043	120.149	203.799	284.795
collidelist	1503.06	1630.91
collidelistall	1539.78	1669.13

NEW

FRect (ms)	4a	1a	2a	1a 2s	r attr	r meth
instantiation	112.192	124.763	154.137	161.03	207.193	259.056
update	106.381	114.321	115.644	130.511	170.518	238.515
union	121.665	133.924	142.434	151.088	201.634	256.291
union_ip	108.603	123.94	119.543	137.935	177.864	237.649
clip	120.511	130.499	136.267	151.63	201.553	256.948
colliderect	64.6672	83.9012	89.2537	100.712	157.369	201.935
collidelist	1284.65	1557.18
collidelistall	1299.62	1572.09

Rect
OLD

Rect (ms)	4a	1a	2a	1a 2s	r attr	r meth
instantiation	145.682	155.88	151.38	164.339	268.44	323.51
update	134.111	135.066	138.892	150.072	249.852	320.009
union	154.835	154.755	157.354	171.846	272.256	331.964
union_ip	128.519	140.884	142.469	157.663	258.865	314.057
clip	148.828	156.624	160.743	179.83	285.872	353.953
colliderect	65.0292	90.2959	97.4342	114.931	210.152	278.185
collidelist	1007.44	1316.01
collidelistall	1035.64	1382.66

NEW

Rect (ms)	4a	1a	2a	1a 2s	r attr	r meth
instantiation	125.46	127.815	144.663	160.908	206.661	255.443
update	90.2074	93.5331	123.818	133.744	177.901	240.214
union	117.424	123.701	153.282	157.854	201.07	256.451
union_ip	107.671	104.778	127.645	139.339	183.731	246.307
clip	123.323	129.958	150.536	160.204	197.421	259.764
colliderect	62.8959	75.0202	86.1815	104.539	148.968	207.309
collidelist	668.871	1179.3
collidelistall	712.745	1223.26

The code I used:

import timeit
from statistics import fmean

from pygame.rect import Rect, FRect
from tabulate import tabulate


def create_table(title: str, *headers, data: list):
    table_headers = [title, *headers]
    table = []
    for row in data:
        table.append(row)

    print(tabulate(table, headers=table_headers, tablefmt="github", numalign="left",
                   stralign="left", ))


NUMBER = 1_000_000

r1 = Rect(799, 799, 10, 10)
r2 = Rect(0, 0, 5, 5)
rects = [
    (0, 0, 100, 100),
    (50, 50, 100, 100),
    (100, 100, 100, 100),
    (150, 150, 100, 100),
    (200, 200, 100, 100),
    (250, 250, 100, 100),
    (300, 300, 100, 100),
    (350, 350, 100, 100),
    (400, 400, 100, 100),
    (450, 450, 100, 100),
    (500, 500, 100, 100),
    (550, 550, 100, 100),
    (600, 600, 100, 100),
    (650, 650, 100, 100),
    (700, 700, 100, 100),
]
rects2 = [((a, b), (c, d)) for a, b, c, d in rects]


class RectAttr:
    def __init__(self, x, y, w, h):
        self.rect = Rect(x, y, w, h)


r3 = RectAttr(799, 799, 10, 10)


class RectMethod:
    def __init__(self, x, y, w, h):
        self.p = Rect(x, y, w, h)

    def rect(self):
        return self.p


r4 = RectMethod(799, 799, 10, 10)

GLOB = globals()


def test(name: str, funcs: list[str], data_table) -> None:
    t = [name]
    for func in funcs:
        val = fmean(timeit.repeat(func, globals=GLOB, number=NUMBER))
        t.append(val * 1000)
    data_table.append(t)
    print(f"{name} done")


data = []

test("instantiation",
     [
         "Rect(10.0, 10.0, 2.0, 2.0)",
         "Rect((10.0, 10.0, 2.0, 2.0))",
         "Rect((10.0, 10.0), (2.0, 2.0))",
         "Rect(((10.0, 10.0), (2.0, 2.0)))",
         "Rect(r3)",
         "Rect(r4)",
     ],
     data
     )
test("update",
     [
         "r1.update(10.0, 10.0, 2.0, 2.0)",
         "r1.update((10.0, 10.0, 2.0, 2.0))",
         "r1.update((10.0, 10.0), (2.0, 2.0))",
         "r1.update(((10.0, 10.0), (2.0, 2.0)))",
         "r1.update(r3)",
         "r1.update(r4)",
     ],
     data
     )
test("union",
     [
         "r1.union(10.0, 10.0, 2.0, 2.0)",
         "r1.union((10.0, 10.0, 2.0, 2.0))",
         "r1.union((10.0, 10.0), (2.0, 2.0))",
         "r1.union(((10.0, 10.0), (2.0, 2.0)))",
         "r1.union(r3)",
         "r1.union(r4)",
     ],
     data
     )
test("union_ip",
     [
         "r2.union_ip(10.0, 10.0, 2.0, 2.0)",
         "r2.union_ip((10.0, 10.0, 2.0, 2.0))",
         "r2.union_ip((10.0, 10.0), (2.0, 2.0))",
         "r2.union_ip(((10.0, 10.0), (2.0, 2.0)))",
         "r2.union_ip(r3)",
         "r2.union_ip(r4)",
     ],
     data
     )
test("clip",
     [
         "r1.clip(10.0, 10.0, 2.0, 2.0)",
         "r1.clip((10.0, 10.0, 2.0, 2.0))",
         "r1.clip((10.0, 10.0), (2.0, 2.0))",
         "r1.clip(((10.0, 10.0), (2.0, 2.0)))",
         "r1.clip(r3)",
         "r1.clip(r4)",
     ],
     data
     )
test("colliderect",
     [
         "r1.colliderect(10.0, 10.0, 2.0, 2.0)",
         "r1.colliderect((10.0, 10.0, 2.0, 2.0))",
         "r1.colliderect((10.0, 10.0), (2.0, 2.0))",
         "r1.colliderect(((10.0, 10.0), (2.0, 2.0)))",
         "r1.colliderect(r3)",
         "r1.colliderect(r4)",
     ],
     data
     )
test("collidelist",
     [
         "r1.collidelist(rects)",
         "r1.collidelist(rects2)",
     ],
     data
     )
test("collidelistall",
     [
         "r1.collidelistall(rects)",
         "r1.collidelistall(rects2)",
     ],
     data
     )

create_table(
    "Rect (ms)", "4a", "1a", "2a", "1a 2s", "r attr", "r meth",
    data=data
)

[MyreMylar]

LGTM 👍

I get the same speedup testing locally with this testing. noticed one line that looks to need deleting, but otherwise looks alright to me.

[ankith26]

LGTM, thanks for the PR 🍰