Hooks Overview
Hooks allow you to execute your own side effects during specific events of the Document lifecycle. They allow you to do things like mutate data, perform business logic, integrate with third-parties, or anything else, all during precise moments within your application.
With Hooks, you can transform Payload from a traditional CMS into a fully-fledged application framework. There are many use cases for Hooks, including:
- Modify data before it is read or updated
- Encrypt and decrypt sensitive data
- Integrate with a third-party CRM like HubSpot or Salesforce
- Send a copy of uploaded files to Amazon S3 or similar
- Process orders through a payment provider like Stripe
- Send emails when contact forms are submitted
- Track data ownership or changes over time
There are four main types of Hooks in Payload:
Root Hooks
Root Hooks are not associated with any specific Collection, Global, or Field. They are useful for globally-oriented side effects, such as when an error occurs at the application level.
To add Root Hooks, use the hooks
property in your Payload Config:
The following options are available:
Option | Description |
---|---|
| Runs after an error occurs in the Payload application. |
afterError
The afterError
Hook is triggered when an error occurs in the Payload application. This can be useful for logging errors to a third-party service, sending an email to the development team, logging the error to Sentry or DataDog, etc. The output can be used to transform the result object / status code.
The following arguments are provided to the afterError
Hook:
Argument | Description |
---|---|
| The error that occurred. |
| Custom context passed between Hooks. More details. |
| The GraphQL result object, available if the hook is executed within a GraphQL context. |
| The |
| The Collection in which this Hook is running against. This will be |
| The formatted error result object, available if the hook is executed from a REST context. |
Async vs. Synchronous
All Hooks can be written as either synchronous or asynchronous functions. Choosing the right type depends on your use case, but switching between the two is as simple as adding or removing the async
keyword.
Asynchronous
If the Hook should modify data before a Document is updated or created, and it relies on asynchronous actions such as fetching data from a third party, it might make sense to define your Hook as an asynchronous function. This way you can be sure that your Hook completes before the operation's lifecycle continues.
Async hooks are run in series - so if you have two async hooks defined, the second hook will wait for the first to complete before it starts.
Synchronous
If your Hook simply performs a side-effect, such as mutating document data, it might be okay to define it synchronously, so the Payload operation does not have to wait for your hook to complete.
Server-only Execution
Hooks are only triggered on the server and are automatically excluded from the client-side bundle. This means that you can safely use sensitive business logic in your Hooks without worrying about exposing it to the client.
Performance
Hooks are a powerful way to customize the behavior of your APIs, but some hooks are run very often and can add significant overhead to your requests if not optimized.
When building hooks, combine together as many of these strategies as possible to ensure your hooks are as performant as they can be.
Writing efficient hooks
Consider when hooks are run. One common pitfall is putting expensive logic in hooks that run very often.
For example, the read
operation runs on every read request, so avoid putting expensive logic in a beforeRead
or afterRead
hook.
Instead, you might want to use a beforeChange
or afterChange
hook, which only runs when a document is created or updated.
Using Hook Context
Use Hook Context avoid prevent infinite loops or avoid repeating expensive operations across multiple hooks in the same request.
To learn more, see the Hook Context documentation.
Offloading to the jobs queue
If your hooks perform any long-running tasks that don't direct affect request lifecycle, consider offloading them to the jobs queue. This will free up the request to continue processing without waiting for the task to complete.
To learn more, see the Job Queue documentation.