Better Lock

A (better) node.js lock library.

Travis tests

Features

  • Typescript-ready
  • Named and keyed locks
  • Queue and execution timeouts
  • Queue size limit
  • Lock on multiple keys
  • Extended stack traces
  • Promise and callback interface
  • JSDoc annotations
  • Good error messages
  • Unit tests, good code coverage
  • No dependencies

Installation

npm install --save better-lock

Usage examples

Minimal example
const lock = new BetterLock();

try {
  const res = await lock.acquire(async () => {
    // Inside the lock. It will stay closed until the promise you return resolves or rejects.
    await doSomeAsyncTask();
    return 'my result';
  });

  // Outside the lock. You will get whatever the promise chain has returned.
  console.log(res); // "my result"
}
catch (err) {
  // Either your or BetterLock's error
}
Advanced example
const lock = new BetterLock({
  name: 'FileLock',                  // To be used in error reporting and logging
  log: winstonLogger.debug,          // Give it your logger with appropeiate level
  wait_timeout: 1000 * 30,           // Max 30 sec wait in queue
  execution_timeout: 1000 * 60 * 5,  // Time out after 5 minutes
  queue_size: 1,                     // At most one pending job
});

async function processFile(filename) {
  try {
    const result = await lock.acquire(filename, async () => {
      const appended = await appendToFile(filename);
      return updateDb(appended);
    });
    return {
      status: true,
      result
    };
  }
  catch (err) {
    if (err instanceof BetterLock.QueueOverflowError) {
      // The job was discarded
      return {
        status: false
      };
    }

    if (err instanceof BetterLock.ExecutionTimeoutError) {
      winstonLogger.warn('Potential swallowed callback! Stack trace to the entry site:', err.stack);
    }
    throw err;
  }
}
Locking on multiple keys
const userLock = new BetterLock({
  name: 'User lock',
  executionTimeout: 1000 * 60 * 60, // Note you can also use camelCase
});

function transferBetweenUsers(fromId, toId, amount) {
  userLock.acquire([fromId, toId], () => {
    return Promise.all([
      User.get(fromId),
      User.get(toId),
    ]).then(([fromUser, toUser]) => {
      fromUser.amount -= amount;
      toUser.amount += amount;
      return Promise.all([
        user1.save(),
        user2.save(),
      ]);
    });
  }).then(() => {
    console.log('Transfer completed');
  });
}
Using callback interface
const BetterLock = require('better-lock');

const lock = new BetterLock();
//...
lock.acquire(done => {
  // Inside the lock
  doMyAsyncStuffHere((err) => {
    // Call done when done
    done(err);
  });
}, (err, result) => {
  // Outside the lock
  if (err) {
    // Either your or BetterLock's error
    console.error(err);
  }
});

You can see a bunch more usage examples in the spec file, here;

API

  • new BetterLock(options)
    Create a new instance of BetterLock. Options should match interface BetterLockOptions. See below for details.

  • BetterLock.acquire([key], executor, [callback], [jobOptions])
    The main method you'll want to call. For each key, given executor will be called only one at a time. If you don't provide callback, it will return a promise that will be resolved with whatever executor returns.

  • key: Arbitrary string under which to lock. It allows you to use the same lock instance for multiple parallel concerns. Eg. this might be a database record id or filename.

  • executor: Function that will be called within the lock. This function should have one of two forms.
    1. Without arguments, in which case it should return a promise. Lock will remain locked until the promise resolves.
    2. With single done argument. In this case, the executor should call done(err, res) once it is done. Arguments passed to done will be passed to the callback of the lock.
  • callback: Optional callback that will be called once executor exits. Results from executor (resolved/rejected value or arguments given to done) will be passed along. This can be used in addition to the returned promise.
  • jobOptions: An object that should match interface BetterLockJobOptions. A subset of main options that will serve as overrides for this particular job (for example, timeout settings).

  • BetterLock.canAcquire([key])
    Returns true if given key can be acquired.

  • BetterLock.abort([key])
    Abort all jobs for a given key (or from the default job queue, if no key is given). Job executors will not be called. Callbacks will be called with JobAbortedError. Currently executing job will not be interrupted.

  • BetterLock.abortAll()
    Abort all jobs for all keys. This is suitable to be called during shutdown of your app.

Options

All available options can be seen here.

BetterLockOptions are provided when you construct a lock instance. A subset of options given in LockJobOptions can be provided when you call lock.acquire, as the last argument.

Example:

lock.acquire(executor, callback, {
  wait_timeout: 1000
});

Most commonly used options are:

  • wait_timeout
    How long can jobs wait in queue before timing out (ms). Null to disable timeout.

  • execution_timeout
    How long can a job be executing before timing out (ms). Null to disable timeout.
    If you do that, though, and you have a swallowed callback, the lock can remain locked permanently.

  • queue_size
    Max queue size for waiting jobs.

Options are presented using snake_case, but you can also provide them using camelCase keys, if that better suits your code style (eg. extend_stack_traces becomes extendStackTraces).

Default options are a static member DEFAULT_OPTIONS on the BetterLock class. That can be seen here. During runtime, you can change the defaults like this:

import BetterLock from 'better-lock';

BetterLock.DEFAULT_OPTIONS.wait_timeout = 1000;

Motivation

I needed something like async-lock, but a bit easier to debug.

This library improves upon async-lock in following ways:

  • Good error messages and log facility
  • Added features (execution timeout, abort jobs...)
  • Extended stack traces (so when you get an error, you have a full stack trace of the original calling code)
  • JSDoc comments (helps if you're using an IDE)

Following features are present in both libraries:

  • Multiple keys per lock instance
  • Acquire multiple keys in a single call
  • Interface, with executor and the callback function
  • Promises
  • Timeout and queue limit

Following features are present in async-lock, but not here:

  • Domain reentrancy (domains are going away)

NOTE: If you want to sync multiple node instances doing the same operation, this library will not help you. You need something that works over network and can use a shared arbiter of who gets the lock (eg. redis).

Change log

0.1.1 (2018/06/04)

You can now use a Number as job name

0.2.0 (2018/09/27)

Code reformat, better pattern for loading options. No feature upgrades.

0.2.1 (2018/09/27)

  • Better and customizable Promise detection.
  • Restored DEFAULT_OPTIONS.

0.3.0 (2018/10/01)

Can abort jobs waiting in queue.

0.3.1 (2018/10/01)

Updated CI to use the current node versions (0.8 & 0.10). Older node versions should continue to work, but are no longer tested. Also, README updates.

1.0.0 (2019/01/28)

Major version bump.

  • Added multi-key locks and refactored a bunch of internals.
  • Removed OVERFLOW_STRATEGIES and related options, which is mostly the reason for the major version bump. The library should otherwise work the same.

1.0.1 (2019/01/28)

Handle empty key list

2.0.0 (2021/05/30)

Major update. The entire library was rewritten in typescript, so you should now get typings in most editors. We also had to switch tests from mocha + chai to jest (easier ts integration).

The API and features have remained largely the same, just a bit of a refresh.

Non-breaking and internal changes:

  • We now export all error names as a type script type. A few other types as well.
  • Errors now have name parameter, which matches these names.

Breaking changes:

  • You can no longer use camel case versions of external-facing objects. Eg. you can no longer pass waitTimeout instead of wait_timeout. In retrospect, this was a pretty flaky API to maintain.

  • Internal LockJob class is no longer exported.

  • Also, errors no longer expose internal LockJob instances (err.job). We now instead provide the most important fields from the job (id and keys).

  • Error names have been renamed to have BetterLock prefix. Eg. WaitTimeoutError -> BetterLockWaitTimeoutError. This will influence err.name and err.message parameters. The idea here is, if you see BetterLock error in the wild, you will know what generated it.

  • We have renamed BetterLock.BetterLockError to BetterLock.BaseError and BetterLock.BetterLockInternalError to BetterLock.InternalError to better match the naming scheme.

  • Since Options are no longer a class but interface, we are no longer exporting them under BetterLock.Options. You can do import {BetterLockOptions} from 'better-lock'; to get the typescript type.

Development

Fork, then git clone. The project is already set up with a WebStorm project, if that's your cup of tee.

To run tests, with coverage:

npm run test

If you want to contribute, create a branch off master, do your work and then make a pull request against master. Unit tests would be appreciated.

License

MIT