Re: Elisp printer

[Stefan Monnier]

> What I forgot to mention is that this would be coupled with a slot in
> the base class that holds the function to "actually call", and the
> underlying call to make-byte-code would install some bytecode that would
> simply funcall whatever is in this slot, maybe passing in the object as
> a first argument.

But that adds a funcall-indirection.  Given the cost of funcalls in
Elisp, this is a fairly major price to pay.  In that case, you're better
off representing your functions as symbols, so you store the actual
function in the function slot, and the data slots in the symbol-plist or
symbol-value slots.

> I'm not 100% sure this addresses what you're looking for; if not I'd
> appreciate it if you could explain more.

Yes, it would, except I don't like its performance.

> One thing I note is that this doesn't provide the "sugar" of being able
> to refer to slots using their bare name, you'd have to use accessors.

That sugar is just an attractive idea, not indispensable.


        Stefan