Support accessing associated functions by member access into facets #4872
Conversation
…fic. This looks like it doesn't work out quite right: we want to distinguish between a function in an interface and a function in a particular impl, so giving them different types seems valuable.
Don't attempt to deduce it from the call arguments.
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
Realphabetize.
Moved self binding check to GetSelfBinding.
| sem_ir.insts().GetAs<FunctionType>(inst.interface_function_type_id); | ||
| auto fn_name_id = sem_ir.functions().Get(fn_inst.function_id).name_id; | ||
| out << "<type of " << sem_ir.names().GetFormatted(fn_name_id) << " in "; | ||
| case CARBON_KIND(FunctionTypeWithSelfType inst): { |
There was a problem hiding this comment.
Sorry, I missed this on my first pass. Is this tested somewhere?
It looks like this will format as <type of in self_id>interface_function_type_id -- shouldn't fn_inst be examined last?. Looking through tests (git grep "type of" **/*.carbon), I see <type of F> in check/testdata/operators/builtin/fail_redundant_compound_access.carbon as the only <type of that's tested. Not sure how easy this is to test, but if you have ideas it might be good to add something (also for the surrounding changes, since I think only FunctionType is tested, and nothing qualified).
There was a problem hiding this comment.
Thanks, fixed and tested in toolchain/check/testdata/interface/no_prelude/fail_assoc_fn_invalid_use.carbon.
There was a problem hiding this comment.
Minor and maybe not specific to this PR, but <type of F in T as I> might be clearer as <type of F in (T as I)> to avoid confusion about <type of (F in T) as I>. But I don't know if the as is always there, so maybe not a clear fix... I mean, <type of F in (X)> would be not-so-good, and the current format is probably usually clear enough.
|
LG |
For an expression such as
(Type as Interface).AssocFn(), track theSelftypeTypein the result of the member access so that it's available when checking the function call.This introduces a new kind of type,
ImplFunctionType, that represents the type of a function that is expected within an impl, modeled as the type of the function within the interface plus a value to use asSelf. Calls to values of this type behave like calls to the underlying function except that theSelfparameter is pre-bound to the self type from the facet.In order to support this, fix an issue where the imported list of generic bindings lost their association with their enclosing generic. This adds a little complexity to
import_ref, including a new recursive cycle that I intend to address in a follow-up PR.