fibo / dflow

is a minimal Dataflow programming engine
MIT License
76 stars 8 forks source link
dataflow dataflow-programming

Dflow

is a minimal Dataflow programming engine

Installation

With npm do

npm install dflow

How it works

A node represents a block of code: it can have inputs and outputs.

An edge connects an input to an output.

A graph represents a program. It can contain nodes and edges. Nodes are executed, sorted by their connections.

Features

NOTA BENE: it is supposed that you implement your own nodes, for example node addition could be implemented using bigint or some floating point library, according to your needs. However an example nodes catalog with basic JavaScript features can be imported from dflow/nodes.

Usage

This is a graph that will compute sin(π / 2) = 1 and print the result.

   ----------------
  | number = π / 2 |
   ----------------
   |
   |
   ---------
  | mathSin |
   ---------
    \
     \
     ------------
    | consoleLog |
     ------------

You can run the following code with any of the following by cloning this repo and launching npm run example:usage.

You should see a number 1 printed on output.

import { Dflow, DflowNode } from "dflow";

const { input, output } = Dflow;

class DflowMathSin extends DflowNode {
  static kind = "mathSin";
  static inputs = [input("number")];
  static outputs = [output("number")];
  run() {
    this.output(0).data = Math.sin(this.input(0).data);
  }
}

class DflowConsoleLog extends DflowNode {
  static kind = "consoleLog";
  static inputs = [input()];
  run() {
    console.log(this.input(0).data);
  }
}

const nodesCatalog = {
  [DflowMathSin.kind]: DflowMathSin,
  [DflowConsoleLog.kind]: DflowConsoleLog,
  // DflowNodeData is a core node
};

function rungraph() {
  const dflow = new Dflow(nodesCatalog);
  const catalog = dflow.nodesCatalog;

  // create nodes
  const numNode = dflow.newNode({
    kind: catalog.data.kind,
    // set numNode output to π / 2
    outputs: [{ data: Math.PI / 2 }],
  });
  const sinNode = dflow.newNode({ kind: catalog.mathSin.kind });
  const consoleLogNode = dflow.newNode({ kind: catalog.consoleLog.kind });

  // connect numNode to sinNode and sinNode to consoleLog
  dflow.connect(numNode).to(sinNode);
  dflow.connect(sinNode).to(consoleLogNode);

  // run graph
  dflow.run();
}

rungraph();

A graph can be executed asynchronously with await dflow.run(): see custom nodes example.

Available examples are listed here.

License

MIT