Fastest weather station wins

easy ~15 min

A sailing app needs the current wind speed. Several weather stations report it, but they are flaky: some answer quickly, some slowly, some fail with an exception. Any one good answer is enough.

Implement first_reading(stations, read, max_workers) -> tuple:

Call read(station) for the stations using a thread pool of at most max_workers threads.
Return (station, value) for the first call to finish successfully, measured by clock time, not by position in the list.
Calls that raise are ignored. If every call raises, raise LookupError.
Return as soon as you have an answer. Don't wait for the slow stations still running, and don't start stations that haven't started yet.

# read("north") takes 0.5s -> 12;  read("pier") takes 0.05s -> 9;  read("hill") raises at once
first_reading(["north", "pier", "hill"], read, 3)   # ("pier", 9) after ~0.05s
first_reading(["hill"], read, 1)                    # LookupError
first_reading([], read, 2)                          # LookupError

Constraints: max_workers >= 1, 0 <= len(stations) <= 100.

Show hint

Submit every station, then loop over concurrent.futures.as_completed(futures) and return the first one whose future.exception() is None. The catch: leaving a with ThreadPoolExecutor(...) block waits for every running call. Create the pool yourself and, before returning, call pool.shutdown(wait=False, cancel_futures=True).

Topic: Thread pool / concurrent crawler. ThreadPoolExecutor, asyncio, thread-safe visited set.