Task summary

• Currently, our sync worker pool contains few pitfals related to how we treat and use channels and goroutines
• We need a better implementation that should be generic and robust enough to handle retries logic and maybe more broader use cases.

Few points that needs to be addressed

• worker pool should handle tasks async or should it resolve all batch of tasks and then return?
• setup a max amount of failures in the pool then we can just ignore the worker (forever? or during a timeout)
• worker pool should be able to resolve if there's no workers or all workers failed to resolve the tasks

Proposed interface

by @timwu20

package pools

type Task any
type TaskID string
type TaskResult struct {
	ID         TaskID
	NumRetries uint // if we care
}
type Result any
type BatchStatus struct {
	Failed  map[TaskID]TaskResult
	Success map[TaskID]TaskResult
}
type WorkerPool interface {
	Results() chan Result
	SubmitBatch(tasks ...Task) (batchID string)
	GetBatch(batchID string) BatchStatus
	Shutdown() error // we impl graceful shutdown
}

// Retry, numWorkers, timeouts, etc.
type WorkerPoolOptions any

func NewWorkerPool(opts ...WorkerPoolOptions) WorkerPool {
	panic("unimpl")
}