Proposal for Extensible Primitives (was: FFI)

Andreas Raab andreas.raab at gmx.de
Tue Aug 15 17:14:15 UTC 2006


Lukas Renggli wrote:
> No changes are required per se, but the suggested modifications would
> make the system simpler, smaller and more clean in my opinion.

In this case my proposal would be to solve the problem we have, not the 
one we don't have ;-) I don't mind some discussion about ways in which 
to change the FFI but they should be weighed relative to the fact that 
the FFI is an already deployed subsystem and changing its syntax means 
that every single line that was ever written in it will break instantly 
(or worse, it won't even load).

To come back to the problem we have -the dependency problem- I would 
again suggest the solution that I mentioned before: A class variable 
that, if installed, handles the requests such that the FFI can hook 
itself up to the parser by providing the appropriate entry point.

Cheers,
   - Andreas

> Primitives would have a consistent syntax and would work searchable
> such as senders and implementors of a method. Moreover no code would
> be left over in the compiler package, namely the function
> Parser>>#externalFunctionDeclaration could be part of the FFI package.
> 
> Of course I am wasting my time with minor details ;-)
> 
> Cheers,
> Lukas
> 




More information about the Squeak-dev mailing list