AI-assisted engineering/Reviewing AI code

Fix the result collector in a worker pool

Review the AI's codeMedium15 minAnthropic and OpenAI style

Your task

An agent's PR moves your job runner to a worker pool. Its version started a goroutine per job and wrote every result into one shared map, so a repeated job id silently overwrote an earlier result. You model what a correct pool should produce, so the PR can be checked against it.

An assistant wrote schedule_pool_results(job_ids, durations, workers). It is in the editor. It has a bug. Find it and fix it.

Input and output:

job_ids, durations: the jobs in submission order. Job i has id job_ids[i] and takes durations[i] time units.
workers: the number of worker slots.
Return one string per job that ran, "<job_id>|<finish_time>".

Rules the tests check:

If the two lists differ in length, use only as many jobs as the shorter list has.
A workers value below 1 counts as 1. A negative duration counts as 0.
Every slot is free at time 0. Take the jobs in submission order. Each job goes to the slot that is free earliest, the lowest slot number on a tie. It starts when that slot is free and finishes duration later, and the slot is busy until then.
A job whose id already appeared earlier in the list is refused: it does not run and takes no slot time.
Sort the output by job id, in ascending order by plain string comparison.

Examples

schedule_pool_results(job_ids: list[str], durations: list[int], workers: int) -> list[str]

Example 1

Inputschedule_pool_results(["a", "b", "c"], [3, 1, 2], 2)

Output["a|3", "b|1", "c|3"]

Example 2

Inputschedule_pool_results(["j2", "j1"], [5, 5], 1)

Output["j1|10", "j2|5"]

Example 3

Inputschedule_pool_results(["a", "a", "b"], [2, 7, 1], 3)

Output["a|2", "b|1"]