A (better) node.js lock library.
npm install --save better-lock
# or
yarn add better-lock
Full docs: https://panta82.github.io/better-lock/
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
}
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;
}
}
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');
});
}
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;
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.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.acquireOr([key], failureResult, executor, [callback], [jobOptions])
The same as acquire, except if the lock can't be acquired (the executor never gets called), instead of throwing
an error, the lock will resolve/callback with the provided failureResult value (for example, a null).
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.
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.
Default options are a static member DEFAULT_OPTIONS on the BetterLock class. During runtime, you can change the defaults like this:
import BetterLock from 'better-lock';
BetterLock.DEFAULT_OPTIONS.wait_timeout = 1000;
This library is a good fit if:
The library is not a good fit if:
You need lock reentrancy.
There is no good solution for this in node.js that I know of. And this library doesn't offer any.
You need a shared lock between different nodes
This library is a single process only. If you need to coordinate multiple apps or services, you need a different library.
Add new api canAcquire. A helper for "best effort" lock acquisition.
Add option lock_condition.
Add option queue_ejection_strategy.
Major version bump because the default strategy changed from "newest" (the old assumed value) to "newest" (makes more sense).
To migrate, specify:
{
"queue_ejection_strategy": "newest"
}
in your job options.
Introduce generated API docs. Remove beta tag from npm version.
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:
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.
Handle empty key list
Major version bump.
OVERFLOW_STRATEGIES and related options, which is mostly the reason for the major version bump. The library should otherwise work the same.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.
Can abort jobs waiting in queue.
Code reformat, better pattern for loading options. No feature upgrades.
You can now use a Number as job name
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.
Generated using TypeDoc