[ruby-core:107306] [Ruby master Feature#18552] Expose `VALUE rb_singleton_class_get(VALUE)` to extensions

Issue #18552 has been updated by byroot (Jean Boussier).


> Which does seem very clean.

I of course meant "doesn't". Oops.

> In this case, it's exposing yet another operation unavailable to plain Ruby code.

To be fair, I was entertaining the idea to also open a feature request to get a Ruby side API for it, e.g. `Object#has_singletong_class?`.

> We should review how much `rb_singleton_class_get()` is exposing and not exposing

It really doesn't expose much, it's nearly the same code than `rb_singleton_class` except it returns `Qnil` instead of creating the singleton class on the fly.

> whether it's feasible to implement in other runtimes such as TruffleRuby.

Not sure about TruffleRuby, but since `msgpack` has a JRuby extension, I also had to do this same there, and their API already allows to check for the singleton class existence without creating it. 

----------------------------------------
Feature #18552: Expose `VALUE rb_singleton_class_get(VALUE)` to extensions
https://bugs.ruby-lang.org/issues/18552#change-96205

* Author: byroot (Jean Boussier)
* Status: Open
* Priority: Normal
----------------------------------------
Right now the only way to check wether an object has a singleton class is to do something akin to:

```c
!SPECIAL_CONST(obj) && FL_TEST(RBASIC(obj)->klass, FL_SINGLETON);
```

Which doesn't seem very clean.

### Use case

This came up in `msgpack`. The library have a registry of serializers on a per class basis, and wish to support singleton classes too.
So it is calling `rb_singleton_class()` which cause lots of useless singleton classes to be created: https://github.com/msgpack/msgpack-ruby/pull/245


### Proposed patch

https://github.com/ruby/ruby/pull/5499



-- 
https://bugs.ruby-lang.org/

Unsubscribe: <mailto:[e-mail]?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>