3 releases
0.1.2 | Dec 29, 2020 |
---|---|
0.1.1 | Dec 29, 2020 |
0.1.0 | Dec 28, 2020 |
#2056 in Data structures
57KB
1K
SLoC
The siraph
crate is a node-based digital signal processing crate.
Features
-
nodes: Adds the
nodes
module that contains a couple of basic useful nodes asMap
orConst
. This feature is enabled by default. -
math: Adds the
nodes::math
module that contains useful nodes related to mathematics asAdd
orOscillator
. This feature depends on thenum-traits
crate and thenodes
feature. -
random: Adds the
nodes::random
module that contains useful nodes related to random number generators asSampleAndHold
. This feature depends on therand
crate and thenodes
feature.
Example
// Using the `nodes` feature
use siraph::Graph;
use siraph::nodes::{Hold, Pulse, FromIter, Map};
let mut graph = Graph::new();
// First, we can insert nodes into the graph.
// This node just takes the values given by an iterator
// and send them into its output.
let from_iter = graph.insert(FromIter::new(std::iter::successors(Some(0u32), |&i| Some(i + 1))));
// This node infinitly outputs 4 `false` then 1 `true`.
let pulse = graph.insert(Pulse::new(4));
// This one wait for a pulse and holds the value its has in its input
// until a new pulse.
let hold = graph.insert(Hold::<u32>::new());
// Simply uses the given function to maps its input to its output.
let map = graph.insert(Map::new(|val: u32| val * val));
// Then, we can plug them together.
graph.plug(from_iter, "output", hold, "input").unwrap();
graph.plug(pulse, "output", hold, "resample").unwrap();
graph.plug(hold, "output", map, "input").unwrap();
// Once our graph is done, we can retreive values from it using a sink.
let mut sink = graph.sink(map, "output").unwrap();
// Here is a simple schem of what we have so far
/*
+------------------------+
| from_iter output i32 >----+ +------------------------------+
+------------------------+ +---> input | +----------------------+
| hold output i32 >--> input map output > sink
+---> resample | +----------------------+
+---------------------+ | +------------------------------+
| pulse output bool >-------+
+---------------------+
*/
// Values can be retreived with the `next` function on the sink
// Once again, the `()` is the context provided to the nodes.
assert_eq!(sink.next(), Some(0));
assert_eq!(sink.next(), Some(0));
assert_eq!(sink.next(), Some(0));
assert_eq!(sink.next(), Some(0));
assert_eq!(sink.next(), Some(0));
assert_eq!(sink.next(), Some(25));
assert_eq!(sink.next(), Some(25));
assert_eq!(sink.next(), Some(25));
assert_eq!(sink.next(), Some(25));
assert_eq!(sink.next(), Some(25));
// The sink is an iterator
for (i, val) in sink.take(1000).enumerate() {
let i = ((i/5)*5) as u32 + 10;
assert_eq!(val, i*i)
}
Create your own nodes
You can create your own nodes using the Node
trait.
use siraph::{Node, Register, Input, Output};
// Our node will take an input and smooth it using a basic interpolation function.
#[derive(Default)]
pub struct Smooth {
input: Input<f64>,
output: Output<f64>,
last_value: Option<f64>,
}
impl Node for Smooth {
fn register(&self, r: &mut Register) {
// This function will register the inputs and outputs of this node.
r.input("input", &self.input);
r.output("output", &self.output);
}
fn process(&mut self) {
// It is in this function that all the processing will be done.
const X: f64 = 1.0/3.0;
// In our case, our computation is not very expensive but
// in other cases, things can get complicated.
// We can skip certain part of the processing by
// checking if our outputs are used.
if self.output.is_used() {
if let Some(cur) = self.input.get() {
if let Some(last) = self.last_value {
self.last_value = Some(last * X + (1.0 - X) * cur);
self.output.set(self.last_value);
} else {
self.output.set(cur);
self.last_value = Some(cur);
}
} else {
self.output.set(None);
}
}
}
fn reset(&mut self) {
// In this function, the inputs of the node should not be used even
// if they may return valid values.
self.last_value = None;
}
}
Todo List
- Load VSTs as nodes
Dependencies
~150KB