Fix fn selector builder for custom types

[digorithm]

Before, the function selector used for methods with
custom types was being built as "my_func(MyStruct)".

This isn't the right approach, instead, it should be:
"my_func(s(field_1,field_2,...,field_n))".

This fixes the function selector builder to support the
right approach and it also greatly simplifies the whole
process.

[adlerjohn]

Does this functionality not exist in https://github.com/FuelLabs/sway?

[digorithm]

Does this functionality not exist in https://github.com/FuelLabs/sway?

I don't know, I personally haven't seen this when I was exploring the Sway's codebase. But it's a big codebase... so I could've missed it.

[adlerjohn]

cc @sezna

[sezna]

Yes, this is how it works in Sway.

And see Enum behavior right below it.

[sezna]

Should we abstract function selector logic since it is shared? Or is that too much of a hassle?

[digorithm]

Should we abstract function selector logic since it is shared? Or is that too much of a hassle?

I would prefer to avoid premature abstraction (only 2 cases right now) at this stage. Especially that the way this logic is implemented in the compiler is not only different than the way I implemented it here (impl details only, not the logic itself, both are matching the specs) but it is also a bit more complex, as it seems to be holding a lot of context necessary for the compiler itself. It would be quite a bit of work right now to either (1) adapt fuels-rs to use the abstraction in the compiler or (2) abstract the logic so it would be used by both.

[adlerjohn]

utACK

Can you file an issue to track creating some test vectors in the specs that both the compiler and SDKs can use?

[digorithm]

Yup, that's a great idea!