feat: THIS keyword
#1454
feat: THIS keyword
#1454
Conversation
compiler/plc_diagnostics/src/diagnostics/diagnostics_registry.rs
Outdated
Show resolved
Hide resolved
| if function_context.linking_context.is_method() | ||
| && self.index.find_pou(type_name).is_some_and(|it| it.is_function_block()) | ||
| { | ||
| let alloca = self.llvm.builder.build_alloca(param_pointer.get_type(), "this"); | ||
| self.llvm.builder.build_store(alloca, param_pointer); | ||
| index.associate_loaded_local_variable(type_name, "__THIS", alloca)?; | ||
| } | ||
| if function_context.linking_context.get_implementation_type() == &ImplementationType::FunctionBlock { | ||
| let alloca = self.llvm.builder.build_alloca(param_pointer.get_type(), "this"); | ||
| self.llvm.builder.build_store(alloca, param_pointer); | ||
| index.associate_loaded_local_variable(type_name, "__THIS", alloca)?; | ||
| } |
There was a problem hiding this comment.
I don't think this is actually neccessary because this is essentially an alias on the function argument %0, no? Specifically I would have assumed define void @FB_Test(%FB_Test* %0) to become define void @FB_Test(%FB_Test* %this) and then keep the codegen unchanged otherwise. However I wasn't able to find a quick solution for that without panicking so maybe theres another reason? Not a blocker though, can perhaps be improved in another PR
There was a problem hiding this comment.
commenting without reading the entire PR, but I would go with ignoring THIS all together in the codegen, the job of this is IMHO done with the resolver, the only exception to that is if we allow THIS^() as in calling the function block in the method but this could then be handled on its own
There was a problem hiding this comment.
They should still resolve to the local variable in the POU, so technically no need for this. it's just a marker to resolve the annotation to local instead of global, so the resolving scope during the annotation phase will be local only or similar, and then the annotation will be that of the local variable, when codegen comes along, it finds the annotation of the local variable on the expression and generates that, again theoretical I did not do a deep dive into the implementation.
There was a problem hiding this comment.
As discussed we leave it as is for now and I created an issue (#1475) so that we don't forget about this
Co-authored-by: Volkan <[email protected]>
Co-authored-by: Volkan <[email protected]>
Co-authored-by: Volkan <[email protected]>
Add
THISkeyword supportImplements validation for the
THISkeyword, which represents a pointer to the current FUNCTION_BLOCK instance.Key points:
THIScan only be used within:FUNCTION_BLOCKimplementationMETHODSbelonging to thatFUNCTION_BLOCK