On Thu, 10 Jul 2008, Miller Puckette wrote:
On Tue, Jul 08, 2008 at 03:04:29AM -0400, Mathieu Bouchard wrote:
So, why are singly-linked-lists so crucial to the design of Pd ?
It's not the list management that's the problem,
I think that this in itself might be a problem.
So, why are singly-linked-lists so crucial to the design of Pd ?
Currently this is managed by maintaining a per-list "stub" that the pointers all check through. one could have such a stub for every scalar, but this would add a lot of overhead.
You can get 8589934592 bits for 23$.
Alternatively, all "gpointers" to a given scalar could register themselves somehow, and every time one deleted a scalar one would track down every pointer to it through the registration mechanism.
This concept is already named «weak pointer» nearly everywhere.
I've been unable to convince myself that the machinery could be kept lightweight and efficient.
You still need reference counting as much as last time I said it.
_ _ __ ___ _____ ________ _____________ _____________________ ... | Mathieu Bouchard - tél:+1.514.383.3801, Montréal, Québec