Similar to NameError#receiver, this returns the object on which
the modification was attempted. This is useful as it can pinpoint
exactly what is frozen, as in many cases when a FrozenError is
raised, you cannot determine from the context which object is
frozen that you expect not to be.
I'm not sure that using name_err_receiver as the C function for FrozenError#receiver is acceptable, but it doesn't appear to cause
problems.
Users of the current rb_error_frozen C function will have to switch
to using rb_error_frozen_object or the new rb_raise_frozen_error
in order to set the receiver of the FrozenError.
Seems reasonable, but shouldn't FrozenError#initialize use a keyword argument as well as NameError#initialize?
NameError#initialize uses an additional positional argument for the name argument, and I figured that was easier as it is unlikely FrozenError#initialize would need to support additional arguments. However, I don't have strong feelings about it, so attached is a patch that switches to using a keyword argument.
NameError#initialize uses a keyword argument for receiver, so I it makes some sense to be consistent.
OTOH, keyword arguments often imply extra processing, especially if the caller is not Ruby code, so I actually find the positional argument version simpler.
And indeed, I don't think extra arguments will be needed for FrozenError#initialize.
In the patch in rb_raise_frozen_error, exc is used for both the message and the exception object, could you use different variables?
I find it confusing to use the same and re-assign it when there is no need.
In the patch in rb_raise_frozen_error, exc is used for both the message and the exception object, could you use different variables?
I find it confusing to use the same and re-assign it when there is no need.
Sure, I can make that change. Can you and nobu come to an agreement on whether to use the positional argument or the keyword argument so I know which patch to update?
I haven't look C-API well, the additional argument would be ok. rb_raise_frozen_error should be rb_frozen_error_raise like rb_name_err_raise?
As for the test, assert_same would be better for the equality of the receiver.
OK, v3 of the patch is attached, based off the initial patch with the positional argument, with the function renamed to rb_frozen_error_raise and using assert_same in the tests.
Eregon (Benoit Daloze) wrote:
In the patch in rb_raise_frozen_error, exc is used for both the message and the exception object, could you use different variables?
I find it confusing to use the same and re-assign it when there is no need.
A nice extra would be specs in spec/ruby/core/exception/frozen_error_spec.rb (similar to spec/ruby/core/exception/name_error_spec.rb) but this is not a requirement.
Similar to NameError#receiver, this returns the object on which
the modification was attempted. This is useful as it can pinpoint
exactly what is frozen. In many cases when a FrozenError is
raised, you cannot determine from the context which object is
frozen that you attempted to modify.
Users of the current rb_error_frozen C function will have to switch
to using rb_error_frozen_object or the new rb_frozen_error_raise
in order to set the receiver of the FrozenError.
To allow the receiver to be set from Ruby, support an optional
second argument to FrozenError#initialize.
Updated by 
Updated by 
Updated by 
Updated by jeremyevans (Jeremy Evans) 
Updated by