I can't work out how to do it.
I can premake a list of random numbers. E.g.

// sequence of float rands for when symmetry required
rand_seq_length = 300;  // make longer if run out.
seeds = rands(0, 65535, rand_seq_length, master_seed);

but inside a random function I want to either step through using an index,
or consume the list by popping (for example).

but I can't work out how to alter another variable inside a function as a
side effect.

E.g.
seed_idx = 0;
function randff() = seeds[seed_idx]; // but autoincrement seed_idx each time
this functon is called.
// or
function randff() = [ some list decomation trick that modifies seeds ];

I suspect I'm thinking about the problem incorrectly :(

--
View this message in context: http://forum.openscad.org/children-and-extra-unwanted-evaluations-tp16291p16298.html
Sent from the OpenSCAD mailing list archive at Nabble.com.

excellent question but my simplified example was just that. here's a bit of
code as an example.

module cylinder_protrusions(length, r1, r2, nprotrusions, seed=Gseed,
excess=1.5)  {
if (verbose) {echo("cylinder_protusions");}
intersection() {
// cyl(length, r1 * excess, r2 * excess, true);
cylindrical_thing(length, r1 * excess, r2 * excess, seed);
//
union() {
for (i = [0:nprotrusions]) {
h = ((randn(80,seed) + 20) / 100.0) * length;
z = -((length - h) / 2.0) * (randn(100,seed) / 100.0);
x = min(r1, r2) * randf(0.2,seed);
angle = randf(360.0,seed);
translate([0, 0, z])
rotate([0, 0, angle])
cube(size=[x, max(r1, r2) * 3, h], center=true);
}
}
}
}

as you can see there are two integer random number requests and two floating
point ones.
There are hundreds of these so annotating every single one is going to be
hard.
I'll do it - but I'm looking for a simpler way.

The current seed based mech is almost working and that's in the code above.
It uses two rand functions like this for now:

function randn(n, myseed=false) = floor(rands(0,n,1, (myseed) ? myseed :
rands(0,65535,1)[0])[0]);
function randf(n, myseed=false) = rands(0,n*1000,1, (myseed) ? myseed :
rands(0,65535,1)[0])[0] / 1000.0;

I'd rather remove the seed argument and handle it all in the rand function
itself.
The approach I'm using here is no good if you like a model and do an F6 - it
makes a completely diff model.
So I want to use a master_seed approach and the user can change this to
generate a new sequence of rands for use in the model construction...

--
View this message in context: http://forum.openscad.org/children-and-extra-unwanted-evaluations-tp16291p16301.html
Sent from the OpenSCAD mailing list archive at Nabble.com.