[PD] Array Enhancements

Mathieu Bouchard matju at artengine.ca
Sat Sep 19 20:26:05 CEST 2009


On Thu, 10 Sep 2009, Jonathan Wilkes wrote:

> You can get all of the above (except onMouseUp behavior) using
> data structures.  Really, there should just be a way to "deputize" a
> ds array as a Pd array usable by [tabread] and the like.
> So if you have a subpatch with
> [struct blah float x array some_array]
> you could put the following in the same subpatch:
> [deputize some_array]
> And now [tabwrite~ some_array]
> will work.

I don't get the point of it. there is already a hidden struct called 
'float' defined by pd in a hidden patch (which doesn't appear in the 
window list) upon startup. Afaik this is a sign that [table] is making 
something that can be used as a DS array. Having it vice-versa shouldn't 
need something like [deputize]. However, such a new object class would be 
nice for introducing a "stride factor", e.g. if you have a struct of 
several elements such as "float x float y float z" and you want 
[tabwrite~] to use only z and not x nor y; this is unless it would be 
better to make it an option in every array-using class, or to make it an 
object-class that does something else: e.g. [virtualarray foo bar y] would 
make a fake "bar" array that is actually a subpart of foo such that 
modifying bar also modifies that subpart of foo (not a copy of it). This 
is like the VIEW feature of the SQL language.

> But I don't know the inner workings of Pd either, so I'm not sure how
> feasible this is.

Your suggestion prompted me to have an idea which in turn sounds like the 
kind of API change that could go hand-in-hand with a replacement to the 
"64-bit fixes" of Pd 0.41, but I'd have to think about it more. It would 
be a generic feature for setting the spacing between relevant elements in 
an array, in bytes. But it may involve too many changes at once for me to 
be comfortable with.

>> I know this is probably flawed for I know not the inner technical 
>> workings of Pd, or for its purposes, there are other more suitable 
>> methods for gui interaction, but is any of this possible? cheers

Well, if you think about the fact that Pd 0.41 fixes the 64-bit problems 
by adding 100% waste (thus the resulting array wastes 50% of its space), 
you can guess that it's because it wasn't easy making an efficient fix, 
therefore it's probably hard to make the [deputize] I suggested; however, 
it's probably easy to make the [deputize] you suggest (?), but I haven't 
tried and I don't know what it would involve, i just know that as long as 
your struct has only one element in it, all versions of Pd so far have 
their internals fairly ready to accept such a feature.

  _ _ __ ___ _____ ________ _____________ _____________________ ...
| Mathieu Bouchard, Montréal, Québec. téléphone: +1.514.383.3801


More information about the Pd-list mailing list