Workers

What it is

A small threaded worker pool utility for running Python callables asynchronously via Job objects, tracking status, and optionally collecting completed jobs via a results Queue.

Public API

• JobStatus (Enum)

  • PENDING, RUNNING, COMPLETED, FAILED, RESULT_FETCHED

• class Job

  • Job(queue: Optional[Queue], func: Callable, *args, **kwargs)
    • Creates a unit of work with a UUID id, status tracking, and optional completion-queue push.
  • execute()
    • Runs the callable, sets status, captures result or exception, signals completion, and pushes itself to queue if set.
  • wait(timeout: Optional[float] = None) -> bool
    • Blocks until completion (or timeout). Returns True if completed.
  • is_completed() -> bool
    • Returns True if status is COMPLETED or FAILED.
  • get_result()
    • Marks status as RESULT_FETCHED, re-raises captured exception if any, otherwise returns the result.

• class WorkerPool

  • WorkerPool(num_workers: int)
    • Starts num_workers daemon threads consuming jobs from an internal queue.
  • submit(job: Job)
    • Enqueues a single job for execution.
  • submit_all(jobs: List[Job]) -> Queue[Job]
    • Enqueues multiple jobs; assigns a shared results queue to jobs whose job.queue is None. Returns that queue.
  • shutdown()
    • Signals workers to stop and joins their threads.

Configuration/Dependencies

• Uses standard library: threading, queue, uuid, enum, typing.
• Logs failures and shutdown debug messages via naas_abi_core.logger.

Usage

from queue import Empty
from naas_abi_core.utils.Workers import Job, WorkerPool

def add(a, b):
    return a + b

pool = WorkerPool(num_workers=2)

# Submit multiple jobs and collect them from the returned results queue
jobs = [Job(None, add, i, i) for i in range(3)]
results_q = pool.submit_all(jobs)

completed = []
while len(completed) < len(jobs):
    try:
        job = results_q.get(timeout=2)
        completed.append((job.id, job.get_result()))
    except Empty:
        break

print(completed)

pool.shutdown()

Caveats

• Worker threads are daemon threads; call shutdown() to join them cleanly.
• submit_all() only assigns the returned results queue to jobs where job.queue is None; jobs constructed with an existing queue will push completion to that queue instead.
• Job.get_result() updates status to RESULT_FETCHED even if the job later raises its captured exception.