Community for the Daft project and all things distributed data

Distributed Data Community

I have a question about set_runner_ray(), why do we need `noop_if_initialized = False` option? Since when you initialize the ray runner it will always return the same OnceLock&lt;Runner&gt; to the caller. why can't we just change the default value to True? Or even remove this option?

I believe we have this parameter because the ray runner also takes in parameters such as the Ray cluster address. So if you call `set_runner_ray` twice with different addresses, we will error unless you set `noop_if_initialized` to true.

Hm actually that doesn't make the most sense since we also have parameters on the native runner

It looks like the original intention was something along those lines but I'm not sure how applicable it is today.

<https://github.com/Eventual-Inc/Daft/pull/728>