Re: [PATCH olang 2/5] checker: scope: populate symbol's type

["Carlos Maniero" <carlos@maniero.me>]

On Sat Sep 21, 2024 at 3:47 PM -03, Johnny Richard wrote:
> I also agree.  I would change a little bit here and make the checker
> calculate the size in bytes and add it to the symbol.

> Let's include the size in bytes for the symbol. I believe this will
> enhance our symbol and make it easier to verify in the next steps

I originally did this way because we may have types that has
platform-dependent sizes such as *pointers*.

But we could address that later than.

> We are traversing the tree twice to populate_scope.  I think the
> populate scope can set the type on scope at this moment.

> If we set extra data to symbol's type we don't need to traverse the ast
> again here.

Makes sense!


> Could you please enlighten the motivation behind setting extra data to
> types on AST nodes?

I don't know if you have anything different in mind, but the alternative
I see is to be to create a type representation for the AST and another
for the scope, which I also think it is fine.

But since I was using the same representation, I opted to first fulfil
the *type_t* on the AST and after to use the same information on the
scope, so it is consistent and you can trust in both informations, the
AST and the scope.

I will change the AST to store the type as *string_view* and keep the
*type_t* on the symbol only. WDYT?

> I don't see other kinds in the future.  What do you think about adding a
> boolean property named "is_primitive"?

Pointers can be a type kind.