----- Original Message -----
From: Frank Barknecht fbar@footils.org To: pd-list@iem.at Cc: Sent: Friday, March 30, 2012 3:58 AM Subject: Re: [PD] list-abs and externals [was: Re: [pd] tables as patch storage]
Hi,
On Tue, Mar 27, 2012 at 12:11:10PM -0700, Jonathan Wilkes wrote:
But I find it interesting that you only imagine the guts of list-abs
getting replaced
with externals in order to improve the performance, and not with new core
objects.
Oh, but I can imagine it and it has already happened: list-len.pd was counting the output of [list-drip] before [list length] was introduced, which is now the sole object inside of list-len.pd
But it's not significant enough to hold up as a virtue of your approach. The user benefits from your approach because they can have your consistent interface in all their patches, and if they run into a place where the poor performance of the abstraction becomes a problem, they can replace the guts with an external-- whether it's one already available in someone else's library or one they write themselves.
Another model is to use your library for a consistent patching interface, and when the user runs into a performance issue, he/she can decide whether a solution that replaces part of the abstraction with a compiled object class is a) idiosyncratic to their particular use case and belongs in an external library or b) generally useful and belongs in the set of classes that are common to _every_ pd implementation.*
I can tell you after going through the bulk of externals out there and documenting them that lots of Pd developers are creating generally useful list manipulation classes. An internal class that does [list-drip], for example, is necessary for just about any fast, expressive list processing in Pd vanilla. But it's not there after a very long time, and the evidence on the patch tracker is that it's a waste of time for devs to even add new functionality there, so that's why I say that part of the dev model is broken.
-Jonathan
- there's also c) writing a ridiculously complicated hack to get a big performance boost,
like the current [list-drip], but for obvious reasons this shouldn't be the dominant model.
Ciao
Frank Barknecht Do You RjDj.me? _ ______footils.org__
Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list