[PD] puredata evolution

[Mathieu Bouchard]

> PD is often said to be meant to be a tool for just prototyping.

Who says that?

> I'm curious: how many of you really re-code their pd patches into 
> something else when you're finished experimenting and you've reached a 
> reasonably "final" version of what you're patching? I don't.

I believe it rarely ever happens and that whoever tells you that is a fan 
of C++ who is snobbing pd in the same way that he'd snob all interpreted 
languages.

> I think pd-vanilla is the data-flow, audio-oriented analoguous of an 
> interpreted programming language.

I've called pd an interpreted programming language for years. I encourage 
people to say a pd program instead of a pd patch.

> Ok, an interpreted language is 1000 times slower than a compiled one,

Usually not, it depends on the primitives of the language, whether they 
are big operations or small operations. Because with default block size 
one pass of [+~] does 64 things in a loop, it's a lot faster than doing 64 
times [+]. Likewise, GridFlow's [# +] often does more than 100000 
equivalents of [# +] and because of that it's rather close to C speeds 
(though it could get a notch closer... and will). GridFlow was a lot about 
making a library that allows to write a fast program in a slow 
interpreter.

Then, nowadays there are several kinds of intermediate languages that are 
used to accelerate interpretation. There's hardly any major scripting 
language interpreter that would reparse commands at every execution (ok, 
Bash does; what else?). Others either use various kinds of bytecode or 
abstract syntax trees.

  _ _ __ ___ _____ ________ _____________ _____________________ ...
| Mathieu Bouchard - tél:+1.514.383.3801, Montréal QC Canada