worker.md

worker

Runs background workers on the Effect clock. The index lists every category.

pollForever

Runs tick at once, then again after each success.

pollForever<A, E, R>(
  tick: Effect.Effect<A, E, R>,
  options: PollOptions<A>
): Effect.Effect<never, E, R>

• Inputs:
  • tick: Effect.Effect<A, E, R>
  • options: PollOptions<A>
• Output: Effect.Effect<never, E, R>
• Errors: E
• Requirements: R
• Side effects: Waits on the Effect clock between ticks; every write is the tick's own.
• Source: apps/backend/src/worker-support.ts:44

How it works

The delay between the end of one tick and the start of the next is options.interval, on Schedule.spaced. When options.skipDelay answers true for the result of a tick, the next tick starts at once, so a worker that made progress drains its queue without waiting. The delay uses the Clock service, so TestClock controls it. The first tick failure stops the loop with that failure, and interruption stops the current tick or delay; the loop never succeeds.

Use

pollForever(tick, {
  interval: Duration.millis(options.pollIntervalMilliseconds),
  skipDelay: Predicate.isTagged("Delivered"),
});

Avoid

A worker loop of setTimeout or setInterval, or a hand-written repeat of the tick and Effect.sleep: timers run on real time, which TestClock cannot drive, and a hand-written loop tends to swallow a failed tick. Pass the tick to pollForever, and let a failure that the tick does not handle end the worker.