[SUGGESTION] unsafe_narrow is not a good name for forced upcasts.

If you have e.g. expression templates, then the upcast to the implementing class needs to be done with unsafe_narrow.

Maybe an alias of unsafe_narrow like unsafe_cast, 'unsafe_pointer_cast` or something like that could be introduced.

Example: https://cpp2.godbolt.org/z/jb1MY3Mcq

Exp : <Impl> type = {  
  cast: (this) -> forward Impl = cpp2::unsafe_narrow<Impl>(this&)*; // unsafe_narrow does not really express what this cast is.
}

A : type = {
  this: Exp<A> = ();
  foo: (this) = std::cout << "foo" << std::endl;
}

main: () -> int = {
  a: A = ();
  e: *Exp<A> = a&;

  e*.cast().foo();

  return 0;
}

I call this static_downcast: https://johelegp.github.io/jegp/static.downcast.
Inspired by https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#Res-casts-named,
I created my own more specific one.

Yes, that is also possible.

By the way,
I noticed in your other issue (and this one)
that you mixed up the terms upcast and downcast.
Then, I tried to find it in the specification for static_cast,
but it wasn't mentioned, so I just left it unsaid.
Now, I did find other examples.
These use "upcast" when converting from derived to base:
https://eel.is/c++draft/class.cdtor#example-3 and https://eel.is/c++draft/class.cdtor#example-1.

Thanks for pointing this out. I knew, that this can also be seen the other way round. I will start using it in terms of the type system and not the functionality.

cpp2::unsafe_narrow is modeled after gsl::narrow_cast.
gsl::narrow_cast seems to be modeled after gsl::narrow
based on its uses in the C++ Core Guidelines.
Although gsl::narrow_cast<T>(x) is specified as being static_cast<T>(x)
[ Note: static_cast<T>(x) is illegal Cpp2. -- end note ]
it's not clear if those should work with non-arithmetic types:
See isocpp/CppCoreGuidelines#1845 and linked issues (and their linked issues/PRs).

Ok, cpp2::unsafe_narrow should be equally handled as the gsl one. But there should be additional casts in cpp2.

Navigation Menu

Repository navigation

[SUGGESTION] unsafe_narrow is not a good name for forced upcasts. #510

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

[SUGGESTION] unsafe_narrow is not a good name for forced upcasts. #510

Description

Activity

JohelEGP commented on Jun 13, 2023

MaxSagebaum commented on Jun 13, 2023

JohelEGP commented on Jun 13, 2023

MaxSagebaum commented on Jun 13, 2023

JohelEGP commented on Jun 19, 2023

MaxSagebaum commented on Jun 19, 2023

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions