managed_runtime

ManagedRuntime[R, ER]: a layer built once and kept, to run many effects with its services, as Effect's ManagedRuntime.

The layer is built on first use (or on entering the runtime as a context manager) and shared by every run until the runtime is disposed. Disposal interrupts and awaits the fibers the runtime started, then releases the layer's resources; later runs die.

Built for an application's lifetime, such as a web server's lifespan::

async with ManagedRuntime.make(AppLive) as runtime:   # builds AppLive
    ...
    response = await runtime.run_async(handle(request))  # per request
# AppLive's resources are released here

Where runs go:

• run_async / run_async_exit: on the running asyncio loop.
• run_sync / run_sync_exit: on a scheduler the runtime owns, driven by the calling thread until the effect is done. Sync runs share it, one thread at a time, so fibers the layer forked go on running during each of them.
• run_fork / run_callback: on the current fiber's run, else the running loop, else the runtime's own scheduler (they then make progress while a sync run or dispose drives it).

dispose (and a sync with) blocks, so it raises inside a fiber or a running asyncio loop, whose fibers it could never see finish: use dispose_effect or dispose_async (async with) there.

Classes#