Improve is_isogram performance by removing unnecessary sorting - #15318

from itertools import product
from statistics import median
from timeit import repeat


def is_isogram_old(string: str) -> bool:
    if not all(char.isalpha() for char in string):
        raise ValueError("String must only contain alphabetic characters.")

    letters = sorted(string.lower())
    return len(letters) == len(set(letters))


def is_isogram_new(string: str) -> bool:
    if not all(char.isalpha() for char in string):
        raise ValueError("String must only contain alphabetic characters.")

    letters = string.lower()
    return len(letters) == len(set(letters))


def get_outcome(function, text: str) -> tuple:
    try:
        return ("return", function(text))
    except Exception as error:
        return ("raise", type(error), str(error))


def verify_results() -> None:
    expected_cases = [
        ("", True),
        ("a", True),
        ("A", True),
        ("Aa", False),
        ("isogram", True),
        ("lamp", True),
        ("Uncopyrightable", True),
        ("Dermatoglyphics", True),
        ("ambidextrously", True),
        ("allowance", False),
        ("Alphabet", False),
        ("abcdefghijklmnopqrstuvwxyz", True),
        ("ABCDEFGHIJKLMNOPQRSTUVWXYZ", True),
        ("abcdefghijklmnopqrstuvwxyza", False),
        ("é", True),
        ("Éé", False),
        ("αβγδεζηθικλμνξοπρστυφχψω", True),
        ("hello world", ValueError),
        ("six-year-old", ValueError),
        ("copy1", ValueError),
        ("test!", ValueError),
        ("a_b", ValueError),
        ("a" * 10_000, False),
    ]

    for text, expected in expected_cases:
        for function in (is_isogram_old, is_isogram_new):
            outcome = get_outcome(function, text)
            if isinstance(expected, type) and issubclass(expected, Exception):
                assert outcome[0] == "raise" and outcome[1] is expected
            else:
                assert outcome == ("return", expected)

    generated_count = 0
    for size in range(6):
        for characters in product("aAbB1 ", repeat=size):
            text = "".join(characters)
            assert get_outcome(is_isogram_old, text) == get_outcome(
                is_isogram_new, text
            )
            generated_count += 1

    print(
        f"Passed {len(expected_cases)} expected-result cases and "
        f"{generated_count:,} exhaustive comparison cases."
    )


def benchmark() -> None:
    repeat_count = 5
    calls_per_repeat = 5

    for repetitions in (10_000, 50_000, 100_000):
        text = "abcdefghijklmnopqrstuvwxyz" * repetitions
        print(f"\nBenchmarking {len(text):,} characters...")

        old_times = repeat(
            lambda: is_isogram_old(text),
            repeat=repeat_count,
            number=calls_per_repeat,
        )
        new_times = repeat(
            lambda: is_isogram_new(text),
            repeat=repeat_count,
            number=calls_per_repeat,
        )

        old_times_ms = [time / calls_per_repeat * 1_000 for time in old_times]
        new_times_ms = [time / calls_per_repeat * 1_000 for time in new_times]
        old_median = median(old_times_ms)
        new_median = median(new_times_ms)

        print(
            "Old version: "
            f"min={min(old_times_ms):.3f}, median={old_median:.3f}, "
            f"max={max(old_times_ms):.3f} ms/call"
        )
        print(
            "New version: "
            f"min={min(new_times_ms):.3f}, median={new_median:.3f}, "
            f"max={max(new_times_ms):.3f} ms/call"
        )
        print(f"Median speedup: {old_median / new_median:.2f}x")


if __name__ == "__main__":
    verify_results()
    benchmark()