Hook priority

[magik]

It'd be nice to be able to change a hook's priority from the JS itself...

I was writing a script for the API, and I need to get the onPlayerMove hook faster, or look for another way to drop a block under a player before he gets to the next square...

What do you think about a function to de-register/re-register listeners through the API?

[alecgorge]

This wouldn't be possible because every plugin would be loaded at the same priority and it would cause conflicts.

Unless you have a solution for this, I don't think we can implement it.

[magik]

I actually do see an implementation strategy and a possible efficiency gain as well. Right now, the plugin hooks every hook possible to sort of "dummy" functions that go and search to see if there are any bindings for any scripts set.

I think a better way to implement this would have separate Listeners for each script loaded. And these Listenersare only registered to hook when the script itself registers a hook. ( which we know after the eval() )

I see an array of Listeners at the ScriptingApi class level, and for each doFile() called, it would create and push another Listener onto this array. Basically leave the dummy functions there to call trigger, but don't hook them yet. After each eval() in the doFile, check the bindings for what keys have been hooked. And for the local Listener, register the appropriate hooks.

I think a bit change in program flow is needed though. The bindings will need to be separate for each Listener/script, that way the appropriate Listener that was hooked at a certain priority level only calls the bindings hooked at that level. And right now the bindings are part of the MinecraftJSApi class that is shared between all. Maybe the bindings could be moved to the ScriptingApi class so that they could live next to this new array of Listeners.

This way multiple hooks of the same type can be registered, but all to different Listeners. And I think a possible performance gain comes out of this because now the plugin is not being called when it's not needed.

I havn't fully fleshed the idea out, there may be some caveats I may not be thinking of.

[alecgorge]

i think the performance difference would be negligible because in the grand scheme of things the hook calls take practically no time.

the only benefit would be setting priority of listeners.

it also would be a ton of work to make it work because you would need a new class for each different listener which is a bunch of really slow dynamic class creation via the reflection api.

although it is a nice idea, due to the limitations of Java, I don't think this is practical.

[magik]

but this class creation would only ever happen on bootup and on /reloadplugins

you think it would really be that much slower? it's not like it has to hook these dynamically, only marginally dynamically whenever scripts are loaded.... In that one moment I agree this will definitely be slower, but it's only when scripts are loaded. In the grand scheme of things, I would not imagine someone reloading the scripts every few milliseconds to bring the server to a halt

On another note, this limits the API's functionality to medium hook priority. It also makes it so that any external plugin that sets a higher priority could potentially always block the API from receiving bubbling hook calls.

I would love to be able to use anything the hMod API supports.

Maybe this is an idea for a future rev instead, it seems you are busy with python stuff right now anyways.

[alecgorge]

but you can hook inside of closure and after the initial load.

what i am trying to say is that it would be a ton of work for not much benefit.

[magik]

Hrm... how would you hook from inside the closure?

[alecgorge]

When you have a Api.on* call inside a Api.on* call.

Api.onCommand(function (split) {
    ...
    Api.onCommand(function (split) {

    });
    ...
});

[magik]

but that's not the actual hook that's happening, well not the hook I am talking about at least. That's just adding another JS to Java "hook" or binding as you call it in your source. You wouldn't be able to set a plugin to hMod hook priority this way...