[Pipet Users] Command Compilation in Piper: A Discourse

[Jean-Marc Valin]

> I understand the difference:
> 
>   * Data type says what kind of data passes through a terminal
>   * Node type says what the node will do to the data
> 
> And, yes, my example was using data types (INTEGER, etc.), but the same
> principle applies: You can make new types ("pseudo-types") by combining
> several node types in a subnet, right?  They are not hard-coded but built as
> a network.  You may not have considered it to be "type defining", but it is,
> in effect, no?

Well, it you combine several nodes in a subnet, you end up with a "new
node" (it's in fact a subnet). You have created a new node type, but you
haven't created a new data type. Right now, there's no way to create new
data types without touching C++ code.

C analogy: you have a couple functions (nodes) and you write a new on
that calls many of the "old" functions. You end up with a new function
(node) but you did not create a new type.

I think I'll write an analogy between Overflow "concepts" and their
Matlab (or C) equivalent.

	Jean-Marc