feat: Add temporary-file-path argument for multi-processing

[iwyoo]

Thank you for the great repo!

I found that using a single predefined temp-file-path may cause error because of race condition. This patch simply resolves the issue by adding option (tmp_fpath).

import os
from multiprocessing import Process, Queue
from py2opsin import py2opsin
import tqdm


def worker(tmp_fpath, q1, q2):
    while True:
        data = q1.get()

        if data is None:
            break

        smiles = py2opsin(data, tmp_fpath=tmp_fpath)
        q2.put(smiles)


if __name__ == "__main__":
    n_processes = 4
    os.makedirs("tmp", exist_ok=True)
    tmp_fpaths = [f"tmp/{i}.txt" for i in range(n_processes)]

    q1 = Queue()
    q2 = Queue()

    processes = [Process(target=worker, args=(tmp_fpaths[i], q1, q2)) for i in range(n_processes)]
    [p.start() for p in processes]

    for i in tqdm.tqdm(range(100)):
        q1.put("2,4,6-trinitrotoluene")

    for _ in tqdm.tqdm(range(100)):
        smiles = q2.get()
        assert smiles == "[N+](=O)([O-])C1=C(C)C(=CC(=C1)[N+](=O)[O-])[N+](=O)[O-]", smiles

    for _ in range(n_processes):
        q1.put(None)

    [p.join() for p in processes]

[JacksonBurns]

Thanks for the PR! I am going to update the doc string and tack on a unit test, then merge.

[JacksonBurns]

A word of caution, too - as you see mentioned in the OPSIN repo (dan2097/opsin#214 (comment)) the underlying OPSIN software will already parallelize itself. Be wary when using multiprocessing in Python that you might end up slowing yourself down if you spawn too many processes and then end up subsequently spawning too many threads.

[JacksonBurns]

Thanks again!