How feasible would it be to retrofit the stop-on-idle changes you previously made to eslint_d, here in this project?

[DinoChiesa]

@mantoni thanks for this project.

In July 2024, you introduced this stop-on-idle code into eslint_d, as part of a significant "new implementation".

As you may remember, core_d is used within prettierd, with a similar philosophy to eslint_d. But you didn't retrofit the stop-on-idle stuff into core_d. and I think eslint_d now no longer depends on core_d. So the operational improvements you made to eslint_d at that time, do not also accrue to projects depending on core_d.

How difficult would it be to retrofit the stop-on-idle behavior into the core_d project? And could you do it? I can make an effort ... the setTimeout and clearTimeout is simple enough. But what am I not aware of? You are much more familiar with the eslint_d and core_d code. There is ground that you covered there that I will be unaware of, so I'd rather not pursue a journey of discovery on my own if you already know it.

Is the implementation of IDLE ... not useful without also implementing the PPID thing - the one that allows the "host" to explicitly specify the PPID of the existing process?

[mantoni]

That was a complete rewrite of eslint_d and you're right, it doesn't depend on core_d anymore. The major problem I was facing is that the daemon was consuming too much memory, especially when used across multiple projects with many plugins. The new architecture spans a daemon per project and auto-stops it after the IDLE timeout or when the spawning process dies (the PPID). These are two concepts that don't depend on each other.

It took quite some time and I just didn't get to also extracting a "general purpose" core. I'm not sure if it's even worth it. I certainly won't have time to do this in the foreseeable future.

Do you actually see a need to change the architecture for your use-case?

[DinoChiesa]

oh - well ya, the prettierd that depends on the original core_d just never exits. You probably don't remember but you weighed in on the ticket 2 and a half years ago saying "yes, someone asked about this for eslint_d too, I don't have time now, but I'd be happy to review."

I'll see if I can just implement the Idle timeout idea.