# Hooks
Hooks are pluggable middleware functions that can be registered before, after or on error(s) of a service method. You can register a single hook function or create a chain of them to create complex work-flows. Most of the time multiple hooks are registered so the examples show the "hook chain" array style registration.
A hook is transport independent, which means it does not matter if it has been called through HTTP(S) (REST), Socket.io, Primus or any other transport Feathers may support in the future. They are also service agnostic, meaning they can be used with any service regardless of whether they have a model or not.
Hooks are commonly used to handle things like validation, logging, populating related entities, sending notifications and more. This pattern keeps your application logic flexible, composable, and much easier to trace through and debug. For more information about the design patterns behind hooks see this blog post (opens new window).
# Quick Example
The following example adds a createdAt
and updatedAt
property before saving the data to the database and logs any errors on the service:
# Hook functions
A hook function can be a normal or async
function that takes the hook context as the parameter and returns the context
object, undefined
or throws an error.
For more information see the hook flow and asynchronous hooks section.
# Hook context
The hook context
is passed to a hook function and contains information about the service method call. It has read only properties that should not be modified and writeable properties that can be changed for subsequent hooks.
Pro Tip: The
context
object is the same throughout a service method call so it is possible to add properties and use them in other hooks at a later time.
# context.app
context.app
is a read only property that contains the Feathers application object. This can be used to retrieve other services (via context.app.service('name')
) or configuration values.
# context.service
context.service
is a read only property and contains the service this hook currently runs on.
# context.path
context.path
is a read only property and contains the service name (or path) without leading or trailing slashes.
# context.method
context.method
is a read only property with the name of the service method (find
, get
, create
, update
, patch
, remove
).
# context.type
context.type
is a read only property with the hook type (one of before
, after
or error
).
# context.params
context.params
is a writeable property that contains the service method parameters (including params.query
). For more information see the service params documentation.
# context.id
context.id
is a writeable property and the id
for a get
, remove
, update
and patch
service method call. For remove
, update
and patch
context.id
can also be null
when modifying multiple entries. In all other cases it will be undefined
.
Note:
context.id
is only available for method typesget
,remove
,update
andpatch
.
# context.data
context.data
is a writeable property containing the data of a create
, update
and patch
service method call.
Note:
context.data
will only be available for method typescreate
,update
andpatch
.
# context.error
context.error
is a writeable property with the error object that was thrown in a failed method call. It is only available in error
hooks.
Note:
context.error
will only be available ifcontext.type
iserror
.
# context.result
context.result
is a writeable property containing the result of the successful service method call. It is only available in after
hooks. context.result
can also be set in
- A
before
hook to skip the actual service method (database) call - An
error
hook to swallow the error and return a result instead
Note:
context.result
will only be available ifcontext.type
isafter
or ifcontext.result
has been set.
# context.dispatch
context.dispatch
is a writeable, optional property and contains a "safe" version of the data that should be sent to any client. If context.dispatch
has not been set context.result
will be sent to the client instead.
Note:
context.dispatch
only affects the data sent through a Feathers Transport like REST or Socket.io. An internal method call will still get the data set incontext.result
.
# context.statusCode
context.statusCode
is a writeable, optional property that allows to override the standard HTTP status code (opens new window) that should be returned.
# context.event
context.event
is a writeable, optional property that allows service events to be skipped by setting it to null
# Hook flow
In general, hooks are executed in the order they are registered with the original service method being called after all before
hooks. This flow can be affected as follows.
# Throwing an error
When an error is thrown (or the promise is rejected), all subsequent hooks - and the service method call if it didn't run already - will be skipped and only the error hooks will run.
The following example throws an error when the text for creating a new message is empty. You can also create very similar hooks to use your Node validation library of choice.
app.service('messages').hooks({
before: {
create: [
context => {
if(context.data.text.trim() === '') {
throw new Error('Message text can not be empty');
}
}
]
}
});
# Setting context.result
When context.result
is set in a before
hook, the original service method call will be skipped. All other hooks will still execute in their normal order. The following example always returns the currently authenticated user instead of the actual user for all get
method calls:
app.service('users').hooks({
before: {
get: [
context => {
// Never call the actual users service
// just use the authenticated user
context.result = context.params.user;
return context;
}
]
}
});
# Asynchronous hooks
When the hook function is async
or a Promise is returned it will wait until all asynchronous operations resolve or reject before continuing to the next hook.
Important: As stated in the hook functions section the promise has to either resolve with the
context
object (usually done with.then(() => context)
at the end of the promise chain) or withundefined
.
# async/await
The use of async/await (opens new window) is highly recommended. This will avoid many common issues when using Promises and asynchronous hook flows. Any hook function can be async
in which case it will wait until all await
operations are completed. Just like a normal hook it should return the context
object or undefined
.
The following example shows an async/await hook that uses another service to retrieve and populate the messages user
when getting a single message:
app.service('messages').hooks({
after: {
get: [
async context => {
const userId = context.result.userId;
// Since context.app.service('users').get returns a promise we can `await` it
const user = await context.app.service('users').get(userId);
// Update the result (the message)
context.result.user = user;
// Returning will resolve the promise with the `context` object
return context;
}
]
}
});
The following example shows an asynchronous hook that uses another service to retrieve and populate the messages user
when getting a single message.
app.service('messages').hooks({
after: {
get: [
async context => {
const userId = context.result.userId;
// Also pass the `params` so we get a secure version of the user
const user = await context.app.service('users').get(userId, context.params);
context.result.user = user;
// Returning will resolve the promise with the `context` object
return context;
}
]
}
});
Note: A common issue when hooks are not running in the expected order is a missing
await
statement.
Important: Most Feathers service calls and newer Node packages already return Promises. They can be returned and chained directly. There usually is no need to instantiate your own
new Promise
instance.
# Converting callbacks
When the asynchronous operation is using a callback instead of returning a promise you have to create and return a new Promise (new Promise((resolve, reject) => {})
) or use util.promisify (opens new window).
The following example reads a JSON file converting fs.readFile (opens new window) with util.promisify
:
const fs = require('fs');
const util = require('util');
const readFile = util.promisify(fs.readFile);
app.service('messages').hooks({
after: {
get: [
async context => {
const data = await readFile('./myfile.json');
context.result.myFile = data.toString();
return context;
}
]
}
});
# Registering hooks
Hook functions are registered on a service through the app.service(<servicename>).hooks(hooks)
method. There are several options for what can be passed as hooks
:
// The standard all at once way (also used by the generator)
// an array of functions per service method name (and for `all` methods)
app.service('servicename').hooks({
before: {
all: [
// Use normal functions
function(context) { console.log('before all hook ran'); }
],
find: [
// Use ES6 arrow functions
context => console.log('before find hook 1 ran'),
context => console.log('before find hook 2 ran')
],
get: [ /* other hook functions here */ ],
create: [],
update: [],
patch: [],
remove: []
},
after: {
all: [],
find: [],
get: [],
create: [],
update: [],
patch: [],
remove: []
},
error: {
all: [],
find: [],
get: [],
create: [],
update: [],
patch: [],
remove: []
}
});
// Register a single hook before, after and on error for all methods
app.service('servicename').hooks({
before(context) {
console.log('before all hook ran');
},
after(context) {
console.log('after all hook ran');
},
error(context) {
console.log('error all hook ran');
}
});
Pro Tip: When using the full object,
all
is a special keyword meaning this hook will run for all methods.all
hooks will be registered before other method specific hooks.
Pro Tip:
app.service(<servicename>).hooks(hooks)
can be called multiple times and the hooks will be registered in that order. Normally all hooks should be registered at once however to see at a glance what the service is going to do.
# Application hooks
To add hooks to every service app.hooks(hooks)
can be used. Application hooks are registered in the same format as service hooks and also work exactly the same. Note when application hooks will be executed however:
before
application hooks will always run before all servicebefore
hooksafter
application hooks will always run after all serviceafter
hookserror
application hooks will always run after all serviceerror
hooks
Here is an example for a very useful application hook that logs every service method error with the service and method name as well as the error stack.
app.hooks({
error(context) {
console.error(`Error in '${context.path}' service method '${context.method}'`, context.error.stack);
}
});
Note: generated applications come with
src/app.hooks.js
containing hooks that run for every service.