Crate graphity

Graphity offers tools to model signal flow within directed graphs.

There are two main parts playing a role here. First are nodes implemented by the user. These nodes can have arbitrary number of inputs and outputs, and an optional method performing operations over provided data. The second part is then a signal graph structure generated by this library. It allows the user to register previosly defined nodes, wire them up and let signal flow between them the user to register previosly defined nodes, wire them up and let signal flow between them.

All the core concepts are illustrated by the following example. Read through the documentation of individual structs and traits to learn more.

Example

This example will work with a signal graph containing 3 types of nodes:

• Generator –⁠ pushing given value throught its output.
• Sum –⁠ accepting two inputs and returning the sum of them.
• Echo –⁠ accepting one input and printing it on the standard output as a side-effect.

They will be wired up as following:

     [Echo]
 Λ      \     __
 |       \   /  |  |
 |       [Sum]  |  |
 |       /   \__|  V
        /
   [Generator(1)]

Generator sends value 1 to the first input of Sum. Sum sends the calculated result to its own second input, forming a loop. It will also send a copy of the result to Echo, so we can observe the current value.

Defining nodes

First, let’s define all the modules. We’ll start with Sum since it covers most of the important parts:

#[derive(Default)]
pub struct Sum {
    input1: i32,
    input2: i32,
    output: i32,
}

#[derive(PartialEq, Eq, Hash, Clone, Copy, Debug)]
pub enum SumConsumer {
    In1,
    In2,
}

#[derive(PartialEq, Eq, Hash, Clone, Copy, Debug)]
pub struct SumProducer;

impl Node<i32> for Sum {
    type Consumer = SumConsumer;
    type Producer = SumProducer;

    fn write(&mut self, consumer: Self::Consumer, input: i32) {
        match consumer {
            Self::Consumer::In1 => self.input1 = input,
            Self::Consumer::In2 => self.input2 = input,
        }
    }

    fn read(&self, _producer: Self::Producer) -> i32 {
        self.output
    }

    fn tick(&mut self) {
        self.output = self.input1 + self.input2;
    }
}

There is lot to discuss here, let’s split it into parts.

The first part simply defines the structure of the node. In this case, it has one field per each input/output to work with:

#[derive(Default)]
pub struct Sum {
    input1: i32,
    input2: i32,
    output: i32,
}

Next up, there is the definition of node’s consumers. A consumer is simply an input pin of the node to which we can attach an edge. In this case it is an enum that can be one of two invariants, one for each input:

#[derive(PartialEq, Eq, Hash, Clone, Copy, Debug)]
pub enum SumConsumer {
    In1,
    In2,
}

Similarly, node’s producer is an output pin. In the case of this node, there is only one possible output, so we define a primitive struct as the type:

#[derive(PartialEq, Eq, Hash, Clone, Copy, Debug)]
pub struct SumProducer;

Each node provided by the user must implement the Node trait. You can see that this is where we link previously defined consumers and producers. Note that the node is implemented for type i32. That is the payload type that will be used for data flowing between nodes. All consumers and producers within a graph must work with this type:

impl Node<i32> for Sum {
    type Consumer = SumConsumer;
    type Producer = SumProducer;
    // ...
}

The write method will be used to provide data to the node. It passes the assigned Consumer to specify to which consumer is the data meant for. In this case, our Sum node has two inputs, one for each number to be summed:

    fn write(&mut self, consumer: Self::Consumer, input: i32) {
        match consumer {
            Self::Consumer::In1 => self.input1 = input,
            Self::Consumer::In2 => self.input2 = input,
        }
    }

Its counterpart is the read method which will be called from the outside to read data produced by the node. Since our node has only one producer available, we can safely ignore the given producer value:

    fn read(&self, _producer: Self::Producer) -> i32 {
        self.output
    }

Finally, we need to define the tick method which will process all the data set on the input and save the result on the output:

    fn tick(&mut self) {
        self.output = self.input1 + self.input2;
    }

The remaining two nodes for completeness:

pub struct Generator(i32);

#[derive(PartialEq, Eq, Hash, Clone, Copy, Debug)]
pub enum GeneratorConsumer {}

#[derive(PartialEq, Eq, Hash, Clone, Copy, Debug)]
pub struct GeneratorProducer;

impl Node<i32> for Generator {
    type Consumer = GeneratorConsumer;
    type Producer = GeneratorProducer;

    fn read(&self, _producer: Self::Producer) -> i32 {
        self.0
    }
}

#[derive(Default)]
pub struct Echo {
    input: i32,
}

#[derive(PartialEq, Eq, Hash, Clone, Copy, Debug)]
pub struct EchoConsumer;

#[derive(PartialEq, Eq, Hash, Clone, Copy, Debug)]
pub enum EchoProducer {}

impl Node<i32> for Echo {
    type Consumer = EchoConsumer;
    type Producer = EchoProducer;

    fn write(&mut self, _consumer: Self::Consumer, input: i32) {
        self.input = input;
    }

    fn tick(&mut self) {
        println!("Echo: {}", self.input);
    }
}

Defining the graph

The nodes do not bring much value on their own. The goal of this library is to make it easy to model signal flow between these nodes within a graph.

To build such a graph, all you need to do is to call graphity, defining the name of the generated structure, the type of the payload passed between nodes and list of the previously defined nodes:

graphity!(
    Graph<i32>;
    Generator = {Generator, GeneratorConsumer, GeneratorProducer},
    Echo = {Echo, EchoConsumer, EchoProducer},
);

Note that the macro is called within its own module to prevent conflicts with the rest of you code.

Wiring it all up

Finally, let’s instantiate such a graph, add nodes, connect them a and let the signal flow.

First, let’s create an instance of the previously generated graph:

let mut graph = Graph::new();

Then add nodes to it. The returned value is actually an index of the stored node and can be later used to access consumers and producers or remove the node:

let generator = graph.add_node(Generator(1));
let sum = graph.add_node(Sum::default());
let echo = graph.add_node(Echo::default());

As the next step, we can create edges between producers and consumers of available nodes to form the topology described above:

graph.add_edge(
    generator.producer(GeneratorProducer),
    sum.consumer(SumConsumer::In1),
);
graph.add_edge(
    sum.producer(SumProducer),
    sum.consumer(SumConsumer::In2),
);
graph.add_edge(
    sum.producer(SumProducer),
    echo.consumer(EchoConsumer),
);

Once all is wired up, we can trigger tick of the graph. When this happens, the graph is traversed, all individual nodes ticket and their output passed on to the input of their connected nodes. In this example, you can see that the Echo node keeps reporting increasing number due to the cycle feeding back the output of the Sum to its input.

graph.tick();
graph.tick();
graph.tick();

Echo: 1
Echo: 2
Echo: 3

You can find the full executable version of this example under the project sources.

Learn more about individual types in this documentation. Reading about the generated SignalGraph would be a good start.

Re-exports

pub use error::Error;

pub use node::Node;

pub use node::NodeIndex;

pub use node::NodeWrapper;

Modules

error

Public interface of all the possible errors.

node

A set of traits covering registered nodes, consumers, producers and their indices.

signal

Signal components wrap around the nodes provided by the user and internal nodes.

Macros

graphity

A macro generating custom implementation of a signal graph for given node types.