Variable behaviour

[NickAcPT]

Since I want to see this project succeed, I'll try to help with some specifications (since I know the project owner isn't a Skript user himself)

So, let's start with variables, shall we?!

Ok, first: Variables are persisted across (re-)loads.
This means that variables need to be stored somewhere. They will also need to be resolved to Psi elements before they're passed to the executor method.

Also, variables can be null, but, and I'll repeat it, but no nulls should be able to reach the executor methods.

Also, some types can be parsed from strings. These can implement their own parser.

Also, in case support for Skript addons gets added in the future, I'll leave this here:

[Trigary]

Ok, first: Variables are persisted across (re-)loads.

I guess now I know why some Java plugin developers set all their variables to null in JavaPlugin#onDisable.

Also, variables can be null, but, and I'll repeat it, but no nulls should be able to reach the executor methods.

Does that mean that an exception-like something is thrown, or does that mean that the method will just not execute? Would the next method execute?

[NickAcPT]

I guess now I know why some Java plugin developers set all their variables to null in JavaPlugin#onDisable.

You're mixing stuff. Java variables aren't persisted, but Skript ones are.

Does that mean that an exception-like something is thrown

After a bit of looking how Skript does it, I can finally say that I was wrong on my first message.
The null handling needs to be done by the person that writes the effects and expressions.

or does that mean that the method will just not execute?

Yes, it does. For java types that can't be null, Skript assumes that value (and null strings get <none>).
However, you can prevent it on this parser.

Would the next method execute?

Yes, it always executes if no exception gets thrown.
If there's a exception, Skript will scream an error to the console. (example; this error was caused by a NPE with an addon)

[stefvanschie]

The main issue with variables and the current implementation, is that the type of the variable is determined on runtime, but the parser expects a specific type when loading the files.

The implementation would need to be changed, so the psi elements all store psi elements of unknown types, so every element can also store variable elements. Of course the drawback of this is that each execution method will need to check the result and see if it's the correct type.

Of course this comes at several disadvantages. The type returned from any children elements needs to be constantly checked to see if it's the correct type which harms readability. Load times will also be slower because previosuly we only had to check for elements which would return the correct result, whereas we would now have to check them all.

Of course if anyone has a better idea, I'd love to hear it.