> 3. Its visual austerity is a huge help to me in thinking clearly about
patching and dataflow. It's amazing how often a geometrically elegant
solution turns out to be an elegant solution full stop.
Coincidentally, I was thinking about counterexamples to this today.
Consider a gate:
float
|
v
-----o <- gate!
|
v
(copy of) float
One of the problems with C is that you must take care both in your conditions for
opening the gate, and in avoiding many subtle bugs through spacing, scoping, etc.
that could cause the gate to become ineffectual. Many of those subtle bugs are visual--
I mean, your variable's state is just waiting to be perused by any context inside the
function that's willing to read it. It's amazing Linux works at all when you consider
how few constraints there are on the code one must reason about.
A dataflow language improves this by representing the gate as a visual barrier.
You can't just blithely read "stuff" below the gate[1]. The gate has to pass the data
down the line, and even if you made an additional connection from "float" to something
below, you have to drag the line visually over the gate so that your violation of
common sense and decency is visually obvious.
But now we have a problem, because if we have anything more than "float" that we
want to operate on below the gate, we have no way to read it. There are two possible
solutions to this:
1) store copies of the data below the gate, so that you can read
them if you happen to be allowed through the gate.
2) send some more expressive form of data like an object through the gate, and pack
everything you want to have access to into that object.
#1 is what I see in most patches (and what I end up patching when I just need to get
something done). But think about how that complicates the dataflow. In the gate
example it's especially apparent-- if you want to store some data below the gate you
have to draw a line that does an end run around it to the right inlet of a [float] below
the gate. That violates the very representation of a gate that the visual language was
trying to provide for you.
#2 pack your data into a list and use [route] as the gate. This restores visual clarity
at the expense of anonymous positional arguments. So it works well for routing a list
of, say, two or three values. But not so well for a larger amount of discrete data. An
example of this downside would be a patch that does a good job of showing where
a list flows or branches, but then seeing a big message box at the bottom with something
like [voice $1-$3 $5 $2 $4(. It's a real pain to debug and extend patterns like that.
I'd be interested to hear other approaches to this problem.
-Jonathan
On Wednesday, March 2, 2016 8:20 PM, Matt Barber <brbrofsvl@gmail.com> wrote:
This is a great way to frame it, and it is indeed how I approach composition as well. It sometimes helps to think of a piece as a solution – maybe the only solution – to a set of constraints.
Three limitations that I love in Pd off the top of my head:
1. The relatively small set of core objects really helps with programming ingenuity, and in fact has made me think through some things that have been helpful in other programming contexts. I love when someone throws down a "this can't be done in vanilla" challenge; I've learned lots from thinking, "OK, we'll see about that!"
2. The smallish set of objects also means that Pd is not a black box. It does mean sometimes that you have to know what you're doing, but in general it does not force you to think a specific way about what you're doing. This is not true in my experience with my students who use other programs excluding csound and SC, but including Max – they'll often as a group settle on one massive object or plugin that does 100 things in a really specific way (look at this badass object I found!) and all end up writing more or less the same piece due to the directions the design of the object pushed them in.
3. Its visual austerity is a huge help to me in thinking clearly about patching and dataflow. It's amazing how often a geometrically elegant solution turns out to be an elegant solution full stop.