Utilities

ropt.components.concurrency

Concurrency primitives for workflow coordinators.

run_concurrent runs blocking calls on dedicated threads rather than on the event loop's shared thread pool, which is what lets many optimizations run at once without starving each other. For the helpers meant for scripts and applications, see ropt.utils.

ropt.components.concurrency.run_concurrent

run_concurrent(
    jobs: Sequence[Callable[[], _T]],
    limit: int | None = None,
) -> list[_T]

Run blocking jobs concurrently on dedicated threads and collect results.

Each job runs on its own thread, so the number of jobs that run at once is not capped by any shared thread pool; limit optionally bounds how many run simultaneously. The first job to raise makes its exception propagate as soon as it is observed (fail-fast): jobs that have not started yet are skipped, while any already running are abandoned, since a Python thread cannot be stopped from the outside.

Parameters:

Name	Type	Description	Default
jobs	Sequence[Callable[[], _T]]	The zero-argument callables to run, one result each.	required
limit	int | None	The maximum number to run at once, or None for no limit.	None

Returns:

Type	Description
list[_T]	The job results, in the order of jobs.