Re: Checking for userdata subtypes

[Dirk Laurie <dirk.laurie@...>]

2012/11/10 Kevin Martin <kev82@khn.org.uk>:

> I don't quite follow the situation you describe, but I think you're saying
> that each object created by the library has a different metatable, even
> though it is the same type?

No, only that the metatable is not known at compile time —
neither by name nor by value.

The "usual" (i.e. as in PiL Ch26) way is to create a new metatable
`mymeta` in the C program.  I.e. the C program already knows the
metatable by name and by value at compile time, and it is easy to
check it.

This foreign module looks for `mymeta` in the calling environment.
I.e. the C program only knows the metatable at runtime, when the
userdata object is created.  It's still only one metatable per
userdata type, but the C program does not know the metatable.

It is possible to create a sample object and read off its metatable.
One can memoize it.  Essentially the C API equivalent of

   file_meta = file_meta or getmetatable(io.open("/dev/null"))
   function isfile(val) return getmetatable(val)==file_meta end

Since up to this moment nobody has come up with a standard way
of doing it, maybe that's what I'll do.

Thanks for responding.

Dirk