Support attributes in symbolic expressions #1369
Conversation
…n the corresponding field type if necessary.
… name. No more explicit replacement of dots in the allocation methods.
| @@ -405,6 +405,9 @@ def _infer_dtype(t: Union[ast.Name, ast.Attribute]): | |||
|
|
|||
| def _Attribute(t, symbols, inferred_symbols): | |||
| inferred_type = _dispatch(t.value, symbols, inferred_symbols) | |||
| if (isinstance(inferred_type, dtypes.pointer) and isinstance(inferred_type.base_type, dtypes.struct) and | |||
There was a problem hiding this comment.
@alexnick83 what about plain structs (i.e., not pointers)
There was a problem hiding this comment.
Since we have decided that Structures will always resolve to a pointer to a struct, I think it is better to not handle that case (now), since it might be something unrelated to nested data and cause further confusion.
There was a problem hiding this comment.
Note: This assumption was valid for the time being, but I am now arguing to change that assumption. For cases where the type of an expression is being evaluated, where the t.value type of an Attribute expression is not merely a structure, but an array of structures (indexed through a Slice), the returned inferred_type is no longer a pointer but directly a plain struct.
Example: my_array_of_structs[i].my_attribute
Commit e2561ed implements a handle for this additional case.
I do not know how to make tests for that.