polymorphic accessors

[Araq]

An accessor is polymorphic if it has the signature proc access(c: Container; ...): var T where Container does not have the var in the signature, if then later the call is used to mutate we have to check that the first parameter of the call access(first, ...) was mutable to begin with.

In the ABI such a parameter would always become var implicitly tough. In other words, it eventually always becomes a pointer,
so that no two versions of the proc need to be generated.

This also allows us to omit "overloading based on var-ness" entirely. (Though for backwards compat we will still need it.)

[SirOlaf]

How would this interact with ref types, access into a different parameter and things like global variables? I assume this isn’t the entire design you have in mind as the description sounds far too trivial to deal with every scenario.

[Araq]

I don't understand your question, [] exists and becomes "polymorphic" but []= exists too and is like before. Also a [] which does not return var T doesn't allow for mutability at all. (This is important for copy-on-write strings.)

[SirOlaf]

Nevermind, tested and it and points to https://nim-lang.github.io/Nim/var_t_return.html

[planetis-m]

Will this scheme work with copy on write types?

[Araq]

Can't see why not, make the [] not return a var and the []= copy since it's a write.