Skip to content

Instantly share code, notes, and snippets.

@avdi
Created June 12, 2012 19:08
Show Gist options
  • Star 1 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save avdi/2919483 to your computer and use it in GitHub Desktop.
Save avdi/2919483 to your computer and use it in GitHub Desktop.
Class reloading breaks type-checking
class Foo
end
Foo.object_id # => 34581480
f = Foo.new # => #<Foo:0x000000041f53c0>
f.class.object_id # => 34581480
# This just prevents the "const redefined" warning
Object.send(:remove_const, :Foo)
# "Reload" class Foo
class Foo
end
Foo.object_id # => 34580180
f.class.object_id # => 34581480
f.is_a?(Foo) # => false
f.class == Foo # => false
f.class.name == "Foo" # => true
@mrcasals
Copy link

Isn't this an expected behaviour? The second Foo has a different object id (lines 4, 17), but f keeps being an instance of the first Foo definition. So, obviously, lines 20 and 21 fail, but line 21 still passes.

Looks right to me.

@avdi
Copy link
Author

avdi commented Jun 12, 2012

Yes, it is. But surprising to newbs. This was born out of an OoR mailing list discussion.

@mrcasals
Copy link

It's an interesting point, I hadn't thought about this until this gist. Thanks for sharing anyway :)

Copy link

ghost commented Jun 13, 2012

It should not be surprising to newbs because the code specifically instructed ruby to remove the constant Foo.

Any other behaviour would be just as surprising, because you would then see that the constant would still exist, despite the code specifically telling ruby to remove that constant before.

If you would want that behaviour different, by treating class Foo the same, then you would have to disallow remove_const, because remove_const could no longer change ANY namespace (classes and modules are constants just as any other upcased first character)

@geeksam
Copy link

geeksam commented Jun 14, 2012

@shevegen, Avdi is saying that actual newbs are actually surprised when they first encounter this behavior. Telling someone that they shouldn't be surprised by something that has surprised them is, in effect, a "well, actually".

The fundamental source of the surprise, as you point out, is remove_const. (Or, as I like to think of it, Ruby providing a concrete example of [the first half of] the adage "constants aren't, variables don't.") In many other languages, there's no equivalent -- once a reference is in the symbol table, it's there to stay.

@avdi
Copy link
Author

avdi commented Jun 14, 2012 via email

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment