gh-150942: Optimize stringlib split/splitlines with _PyList_AppendTakeRef - #155922

"""Microbenchmark str/bytes split / rsplit / splitlines (result-list building)."""
import pyperf

# Many small pieces (>12): exercises the append path after the 12-slot prealloc.
import pyperf

# Many small pieces (>12): exercises the append path after the 12-slot prealloc.
STR_MANY = "abc " * 10_000            # .split(" ") -> 10001 pieces
STR_WS = "word  " * 10_000            # .split() whitespace variant
BYTES_MANY = b"abc " * 10_000         # bytes.split
LINES = "line\n" * 10_000             # .splitlines(): SPLIT_APPEND for every line

# Controls
STR_FEW = "a b c d e f g h i j"       # 10 pieces: stays in the SET_ITEM prealloc path
STR_BIG = ("x" * 4096 + " ") * 250    # large pieces: copy-dominated


def bench_method(loops, obj, name, *args):
    method = getattr(obj, name)
    range_it = range(loops)
    t0 = pyperf.perf_counter()
    for _ in range_it:
        method(*args)
    return pyperf.perf_counter() - t0


runner = pyperf.Runner()
runner.bench_time_func("str_split_10k_pieces", bench_method, STR_MANY, "split", " ")
runner.bench_time_func("str_split_whitespace_10k", bench_method, STR_WS, "split")
runner.bench_time_func("str_splitlines_10k", bench_method, LINES, "splitlines")
runner.bench_time_func("bytes_split_10k_pieces", bench_method, BYTES_MANY, "split", b" ")
runner.bench_time_func("str_split_10_pieces_ctl", bench_method, STR_FEW, "split", " ")
runner.bench_time_func("str_split_large_pieces_ctl", bench_method, STR_BIG, "split", " ")