topological-scheduler

Run asynchronous tasks after their dependencies, with a concurrency limit. Worker failures are returned per task instead of rejecting the whole schedule.

npm install topological-scheduler

Quick start

import schedule, { type DependencyGraph, type WorkerFunction } from 'topological-scheduler';

const graph: DependencyGraph<string> = {
  nodes: { fetch: 'fetch data', index: 'build index', report: 'write report' },
  dependencies: {
    fetch: [],
    index: ['fetch'],
    report: ['index'],
  },
};

const worker: WorkerFunction<string, string> = (task, id, done) => {
  console.log(id, task);
  done(null, `completed: ${task}`);
};

const results = await schedule(graph, worker, { concurrency: 2 });

Each result contains id, item, and either result or error. Set failDependents: true to return dependent tasks with skipped: true after a dependency fails. The default concurrency is 1.

The same function accepts a final callback:

schedule(graph, worker, { concurrency: 2 }, (error, results) => {
  if (error) throw error;
  console.log(results);
});

Cycles

The scheduler expects an acyclic graph and does not perform cycle validation. Use topological-sort-group before scheduling when input may contain cycles:

import Graph from 'topological-sort-group';

const checked = Graph.from(graph);
const { cycles } = checked.sort();
if (cycles.length) throw new Error(`Circular dependencies: ${JSON.stringify(cycles)}`);

await schedule(checked.toGraph(), worker, { concurrency: 2 });

Documentation

API documentation