Simplify stdlib code by using itertools.batched()

[lgeiger]

#98364 introduced itertools.batched() but since it's pretty new it's still rarely used inside the standard library.

pickle.py includes a custom version multiple times which all could be replaced by itertools.batched() for improved readability.
This probably has a negligible performance impact (at the very least it won't be slower) but should make the code easier to understand.

Let me know if you're interested in a contribution for this or whether you'd rather not have the extra churn for a minor change like this.

Linked PRs

• gh-126317: Simplify stdlib code by using itertools.batched() #126323

[picnixz]

I'm not sure we want to refactor pickle.py, especially not the pure Python implementation. But I'll let this decision fall to @serhiy-storchaka.

[serhiy-storchaka]

There is no great need in such change, but it makes the code slightly smaller, so I have no objections. Please show the results of microbenchmarks for pickling large list, dict and set. Even a small speedup would be an additional argument for this change. A slowdown would be a sign that there is something wrong with itertools.batched().

[dongwooklee96]

bench.py

import pickle
import pyperf

large_list = list(range(10**6))  # 1,000,000 list
large_dict = {str(i): i for i in range(10**6)}  # 1,000,000 dict
large_set = set(range(10**6))  # 1,000,000 set

def pickle_save(data, filename):
    with open(filename, "wb") as f:
        pickle.dump(data, f)

def pickle_load(filename):
    with open(filename, "rb") as f:
        return pickle.load(f)

def run_benchmarks():
    runner = pyperf.Runner()

    # Pickle dump test
    runner.bench_func("pickle_save_large_list", pickle_save, large_list, "large_list.pkl")
    runner.bench_func("pickle_save_large_dict", pickle_save, large_dict, "large_dict.pkl")
    runner.bench_func("pickle_save_large_set", pickle_save, large_set, "large_set.pkl")

    # Pickle load test
    runner.bench_func("pickle_load_large_list", pickle_load, "large_list.pkl")
    runner.bench_func("pickle_load_large_dict", pickle_load, "large_dict.pkl")
    runner.bench_func("pickle_load_large_set", pickle_load, "large_set.pkl")

if __name__ == "__main__":
    run_benchmarks()

result

Benchmark hidden because not significant (6): pickle_save_large_list, pickle_save_large_dict, pickle_save_large_set, pickle_load_large_list, pickle_load_large_dict, pickle_load_large_set

The result of running the benchmark using pyperf.

I'm sorry, my results are a little different from what I've done before, so I'm guessing there's a difference in configure. If you get a different result, I'd appreciate it if you could let me know.

[serhiy-storchaka]

Thank you for your contribution @dongwooklee96.