Explorer: Support abstract virtual methods

[Pixep]

Background

The Carbon explorer currently supports virtual and impl prefixed virtual method, but not abstract methods (i.e. pure virtual methods in C++).

Goal

Support abstract method, as described in the class design doc. This includes:

• Checking that the method has no implementation (CallableDeclaration::body())
• Checking that child classes do implement the method

Starting point

A recommended approach would be to modify the switch/case logic for virtual methods declaration in explorer/interpreter/type_checker.cpp (in the TypeChecker::DeclareClassDeclaration function, VirtualOverride::Abstract case) to remove the current error logged when abstract is used, and check if the method has no body first.

This can be tested by creating a running a new lit test under explorer/testdata/class, and run using bazel run //explorer/testdata:class/<my_abstract_method_test.carbon>.run.

Example of tests:

Error case 1

base class A {
  // Error, abstract method cannot have a body (note the curly braces)
  abstract fn Foo[self: Self]() {}
}

Valid case

abstract class A {
  // OK, no implementation
  abstract fn Foo[self: Self]();
}

class B extends A {
  // OK, implemented in child class
  fn Foo[self: Self]() {}
}

Resources

• https://github.com/carbon-language/carbon-lang/blob/trunk/CONTRIBUTING.md
• https://github.com/carbon-language/carbon-lang/blob/trunk/docs/project/contribution_tools.md
• #contributing-help channel on the Discord

[aphsai]

hi, could I take a crack at this?

[Pixep]

hi, could I take a crack at this?

Welcome @aphsai ! We have another person on the Discord who started working on it yesterday (I will ask them to add a little note here to be assigned).
Feel free to join us there, and I will start thinking of another other good first issue as well.

[aphsai]

Thanks for the quick reply, happy to help if they need it.

[n3r0bi0m4n]

@Pixep here I am

[CODER2810]

HELLO I CAN RESOLVE THIS ISSUE KINDLY LET ME WORK IN THIS PROJECT

[ashutosh887]

HI @Pixep I would like to wrap this issue
Please assign it to me

[Pixep]

Thanks for the enthusiasm, let's first see if @n3r0bi0m4n can finish it :)
@n3r0bi0m4n is it something you are still working on?

[Pixep]

Hid a few comments, to make sure we stay on topic. We may open the issue to be re-assigned in the future, but for now someone already is.

[shreya024]

Hid a few comments, to make sure we stay on topic. We may open the issue to be re-assigned in the future, but for now someone already is.

Hi @Pixep I would like to work on this issue if it is open

[Pixep]

Hey everyone, good first issues don't need to be assigned anymore (see https://github.com/carbon-language/carbon-lang/blob/trunk/CONTRIBUTING.md#implement-carbons-design for some explanation), feel free to work on it if you wish!

[AbdullahGhulamNabi]

I'm eager to contribute to the effort to resolve the pure virtual function error problem as a student and to do so in order to improve the community while learning essential software development skills. Can I have a chance to work upon?

[himanshusingh0905]

@Pixep There is no method that can tell if a method has a body or not.

[Pixep]

@Pixep There is no method that can tell if a method has a body or not.

During type checking, you can use fun->body().has_value(). You can see it in the reference work that was done on the same issue for an example here: #2528.

@shreya024 , @AbdullahGhulamNabi , @himanshusingh0905 the issue is open for anyone to work on it, without being assigned :)

[Jayantbhoj]

is it still open can i contribute?

[AbdullahGhulamNabi]

Thanks for mentioning me. I am glad you responded. Can you elaborate the the problem and how can I contribute to it as I only have prior knowledge of C++ and never contributed to open source before. So, it would be really helpful of you could a glimpse of how can I can deep dive into it. Waiting for your reply.

…

On Thu, Sep 14, 2023, 3:10 PM Jayant Bhoj ***@***.***> wrote: is it still open can i contribute? — Reply to this email directly, view it on GitHub <#2512 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AZGRLLCGHOPJGRHGQ4LJJZLX2LJ25ANCNFSM6AAAAAATROLZLU> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[rishitha1238]

Hi Pixep, I would like to work on this if it's open