FAQ

FAQ

Is this a replacement for ReactPHP or Amp?

No. ReactPHP and Amp are full async runtimes: they have I/O event loops that poll sockets, streams, timers and signals, and idle efficiently when there is nothing to do. FiberLoops is just the scheduling core — it interleaves fibers and nothing else. Use it when you want cooperative concurrency over tasks you control, or as a primitive to build something larger on. For real non-blocking I/O, use a full runtime.

Does it do asynchronous I/O?

No. There is no stream/socket/timer integration. A task that does a blocking file_get_contents() or PDO::query() blocks the whole loop for that call. FiberLoops overlaps tasks that yield; it cannot make a blocking call non-blocking.

What is the difference between concurrency and parallelism here?

FiberLoops gives concurrency: multiple tasks make progress by taking turns on a single thread. It does not give parallelism: nothing runs on two CPU cores at once. Only one fiber executes at any instant. See Caveats → Concurrency, not parallelism.

Why did my next() / sleep() throw a LoopException?

You called it outside a fiber — most likely from the main script instead of from inside a task. next() and sleep() suspend the current fiber, which is only legal inside one. Move the call into a task you pass to defer() or await(). See Error Handling.

Can I add tasks while the loop is running?

Yes. defer() is safe to call during run(); the new task is picked up on the next pass. This is how you build dispatchers that spawn workers. See Deferring & Running Tasks → Spawning tasks.

How do I stop the loop early?

There is no built-in stop call — run() ends when the queue drains. Give your long-lived tasks an exit condition (a shared flag, a counter, a deadline) and have them return. See Recipes → Graceful shutdown.

Does sleep() use the CPU?

Yes. sleep() is a busy-wait: it spins, yielding to siblings and re-checking the clock until the deadline. It does not idle the process. If every task is sleeping, the loop runs at 100% CPU. See Caveats → sleep() is a busy-wait.

Can I await() several tasks so they run in parallel?

Not by chaining await() — each call runs one task to completion before the next starts. To overlap tasks and then collect their results, defer() them and have each write into a shared array. See Recipes → Collecting results from concurrent tasks.

Do I need a PHP extension?

No. Fibers are part of core PHP since 8.1. The only requirement is PHP 8.1+.

Why fibers instead of generators?

Generators (yield) are shallow: only the generator function itself can yield, so the suspension point has to be threaded back through every caller. Fibers are stackful: a task can call into helper functions and any of them can yield, with no special plumbing. That makes sleep() and await() — which yield from inside nested calls — natural to write.

Is the execution order deterministic?

Yes. Cooperative scheduling means the interleaving is fixed and reproducible: tasks switch only at the next()/sleep()/await() points you wrote, and the loop advances them round-robin. This makes loop-based code easy to test — see Testing.

Can I run nested loops, or call run() from inside a task?

The intended model is a single run() driving a flat set of tasks (which may spawn more tasks). Calling run() recursively from inside a task is not a supported pattern; to wait for a sub-task and get its value, use await() instead.

Where do I report a bug or ask a question?

• Bugs: open an issue on the repo.
• Questions: the org-wide Discussions.
• Security: see the Security policy — do not open a public issue.