Re: SCM objects on the heap?

[Maciej Stachowiak <mstachow@mit.edu>]

dirk@ida.ing.tu-bs.de writes:
> 
> Thanks for your answer!
> 
> I checked the guile source to figure out how scm_protect_object and
> scm_unprotect_object work: 
> 
> - there is a global list 'scm_protects'
> - whenever a new object is to be protected:
>   scm_protects = scm_cons(obj scm_protects);
> 
> - when an object is to be unprotected:
>   scm_delq_x(obj, scm_protects)
> 
> Besides the fact that this is a time consuming implementation, it has 
> a major problem: regardless of how often an object was protected, a
> single call to unprotect will make it subject to gc.
> 

Yes, I've found this to be a problem as well. That's what I was going
to say in the part of my message that apprently got deleted.

> A simple fix was, not to call scm_delq_x, but a function which only
> deletes the first occurence of the object from scm_protects.
> 
> However, this approach still seems to be quite ineffective: For every
> level of protection a new pair is constructed. Wouldn't it make sense to
> have a SCM subtype for protected objects, which consist of a pair with a
> counter and the 'real' SCM object?
> 

IMO a good solution would be to make the protected list a hash table
from objects to reference counts, with scm_reference_object and
scm_unreference_object which increment or decrement the counts and add
or delete entries as appropriate.

> Also, there should be access to this feature through the gh_ interface.
> 

Currently, the gh_ interface knows nothing about GC because it's meant
to be portable to other Scheme implementations. Of course, since many
other Schemes require C code to mark even objects on the stack,
ignoring GC issues entirely does not make things more portable.

 - Maciej