Middleware
Middleware (also called pre and post hooks) are functions which are passed control during execution of asynchronous functions. Middleware is specified on the schema level and is useful for writing plugins.
- Types of Middleware
- Pre
- Errors in Pre Hooks
- Post
- Asynchronous Post Hooks
- Define Middleware Before Compiling Models
- Save/Validate Hooks
- Naming Conflicts
- Notes on findAndUpdate() and Query Middleware
- Error Handling Middleware
- Aggregation Hooks
- Synchronous Hooks
Types of Middleware
Mongoose has 4 types
of middleware: document middleware, model middleware, aggregate middleware, and query middleware.
Document middleware is supported for the following document functions.
In document middleware functions, this
refers to the document.
Query middleware is supported for the following Model and Query functions.
In query middleware functions, this
refers to the query.
- count
- deleteMany
- deleteOne
- find
- findOne
- findOneAndDelete
- findOneAndRemove
- findOneAndUpdate
- remove
- update
- updateOne
- updateMany
Aggregate middleware is for MyModel.aggregate()
. Aggregate middleware
executes when you call exec()
on an aggregate object.
In aggregate middleware, this
refers to the aggregation object.
Model middleware is supported for the following model functions.
In model middleware functions, this
refers to the model.
All middleware types support pre and post hooks. How pre and post hooks work is described in more detail below.
Note: If you specify schema.pre('remove')
, Mongoose will register this
middleware for doc.remove()
by default. If you
want to your middleware to run on Query.remove()
use schema.pre('remove', { query: true, document: false }, fn)
.
Note: Unlike schema.pre('remove')
, Mongoose registers updateOne
and
deleteOne
middleware on Query#updateOne()
and Query#deleteOne()
by default.
This means that both doc.updateOne()
and Model.updateOne()
trigger
updateOne
hooks, but this
refers to a query, not a document. To register
updateOne
or deleteOne
middleware as document middleware, use
schema.pre('updateOne', { document: true, query: false })
.
Note: The create()
function fires save()
hooks.
Pre
Pre middleware functions are executed one after another, when each
middleware calls next
.
const schema = new Schema(..);
schema.pre('save', function(next) {
// do stuff
next();
});
In mongoose 5.x, instead of calling next()
manually, you can use a
function that returns a promise. In particular, you can use async/await
.
schema.pre('save', function() {
return doStuff().
then(() => doMoreStuff());
});
// Or, in Node.js >= 7.6.0:
schema.pre('save', async function() {
await doStuff();
await doMoreStuff();
});
If you use next()
, the next()
call does not stop the rest of the code in your middleware function from executing. Use
the early return
pattern
to prevent the rest of your middleware function from running when you call next()
.
const schema = new Schema(..);
schema.pre('save', function(next) {
if (foo()) {
console.log('calling next!');
// `return next();` will make sure the rest of this function doesn't run
/*return*/ next();
}
// Unless you comment out the `return` above, 'after next' will print
console.log('after next');
});
Use Cases
Middleware are useful for atomizing model logic. Here are some other ideas:
- complex validation
- removing dependent documents (removing a user removes all their blogposts)
- asynchronous defaults
- asynchronous tasks that a certain action triggers
Errors in Pre Hooks
If any pre hook errors out, mongoose will not execute subsequent middleware or the hooked function. Mongoose will instead pass an error to the callback and/or reject the returned promise. There are several ways to report an error in middleware:
schema.pre('save', function(next) {
const err = new Error('something went wrong');
// If you call `next()` with an argument, that argument is assumed to be
// an error.
next(err);
});
schema.pre('save', function() {
// You can also return a promise that rejects
return new Promise((resolve, reject) => {
reject(new Error('something went wrong'));
});
});
schema.pre('save', function() {
// You can also throw a synchronous error
throw new Error('something went wrong');
});
schema.pre('save', async function() {
await Promise.resolve();
// You can also throw an error in an `async` function
throw new Error('something went wrong');
});
// later...
// Changes will not be persisted to MongoDB because a pre hook errored out
myDoc.save(function(err) {
console.log(err.message); // something went wrong
});
Calling next()
multiple times is a no-op. If you call next()
with an
error err1
and then throw an error err2
, mongoose will report err1
.
Post middleware
post middleware are executed after
the hooked method and all of its pre
middleware have completed.
schema.post('init', function(doc) {
console.log('%s has been initialized from the db', doc._id);
});
schema.post('validate', function(doc) {
console.log('%s has been validated (but not saved yet)', doc._id);
});
schema.post('save', function(doc) {
console.log('%s has been saved', doc._id);
});
schema.post('remove', function(doc) {
console.log('%s has been removed', doc._id);
});
Asynchronous Post Hooks
If your post hook function takes at least 2 parameters, mongoose will
assume the second parameter is a next()
function that you will call to
trigger the next middleware in the sequence.
// Takes 2 parameters: this is an asynchronous post hook
schema.post('save', function(doc, next) {
setTimeout(function() {
console.log('post1');
// Kick off the second post hook
next();
}, 10);
});
// Will not execute until the first middleware calls `next()`
schema.post('save', function(doc, next) {
console.log('post2');
next();
});
Define Middleware Before Compiling Models
Calling pre()
or post()
after compiling a model
does not work in Mongoose in general. For example, the below pre('save')
middleware will not fire.
const schema = new mongoose.Schema({ name: String });
// Compile a model from the schema
const User = mongoose.model('User', schema);
// Mongoose will **not** call the middleware function, because
// this middleware was defined after the model was compiled
schema.pre('save', () => console.log('Hello from pre save'));
new User({ name: 'test' }).save();
This means that you must add all middleware and plugins
before calling mongoose.model()
.
The below script will print out "Hello from pre save":
const schema = new mongoose.Schema({ name: String });
// Mongoose will call this middleware function, because this script adds
// the middleware to the schema before compiling the model.
schema.pre('save', () => console.log('Hello from pre save'));
// Compile a model from the schema
const User = mongoose.model('User', schema);
new User({ name: 'test' }).save();
As a consequence, be careful about exporting Mongoose models from the same
file that you define your schema. If you choose to use this pattern, you
must define global plugins
before calling require()
on your model file.
const schema = new mongoose.Schema({ name: String });
// Once you `require()` this file, you can no longer add any middleware
// to this schema.
module.exports = mongoose.model('User', schema);
Save/Validate Hooks
The save()
function triggers validate()
hooks, because mongoose
has a built-in pre('save')
hook that calls validate()
. This means
that all pre('validate')
and post('validate')
hooks get called
before any pre('save')
hooks.
schema.pre('validate', function() {
console.log('this gets printed first');
});
schema.post('validate', function() {
console.log('this gets printed second');
});
schema.pre('save', function() {
console.log('this gets printed third');
});
schema.post('save', function() {
console.log('this gets printed fourth');
});
Naming Conflicts
Mongoose has both query and document hooks for remove()
.
schema.pre('remove', function() { console.log('Removing!'); });
// Prints "Removing!"
doc.remove();
// Does **not** print "Removing!". Query middleware for `remove` is not
// executed by default.
Model.remove();
You can pass options to Schema.pre()
and Schema.post()
to switch whether
Mongoose calls your remove()
hook for Document.remove()
or Model.remove()
. Note here that you need to set both document
and query
properties in the passed object:
// Only document middleware
schema.pre('remove', { document: true, query: false }, function() {
console.log('Removing doc!');
});
// Only query middleware. This will get called when you do `Model.remove()`
// but not `doc.remove()`.
schema.pre('remove', { query: true, document: false }, function() {
console.log('Removing!');
});
Notes on findAndUpdate() and Query Middleware
Pre and post save()
hooks are not executed on update()
,
findOneAndUpdate()
, etc. You can see a more detailed discussion why in
this GitHub issue.
Mongoose 4.0 introduced distinct hooks for these functions.
schema.pre('find', function() {
console.log(this instanceof mongoose.Query); // true
this.start = Date.now();
});
schema.post('find', function(result) {
console.log(this instanceof mongoose.Query); // true
// prints returned documents
console.log('find() returned ' + JSON.stringify(result));
// prints number of milliseconds the query took
console.log('find() took ' + (Date.now() - this.start) + ' millis');
});
Query middleware differs from document middleware in a subtle but
important way: in document middleware, this
refers to the document
being updated. In query middleware, mongoose doesn't necessarily have
a reference to the document being updated, so this
refers to the
query object rather than the document being updated.
For instance, if you wanted to add an updatedAt
timestamp to every
updateOne()
call, you would use the following pre hook.
schema.pre('updateOne', function() {
this.set({ updatedAt: new Date() });
});
You cannot access the document being updated in pre('updateOne')
or
pre('findOneAndUpdate')
query middleware. If you need to access the document
that will be updated, you need to execute an explicit query for the document.
schema.pre('findOneAndUpdate', async function() {
const docToUpdate = await this.model.findOne(this.getQuery());
console.log(docToUpdate); // The document that `findOneAndUpdate()` will modify
});
However, if you define pre('updateOne')
document middleware,
this
will be the document being updated. That's because pre('updateOne')
document middleware hooks into Document#updateOne()
rather than Query#updateOne()
.
schema.pre('updateOne', { document: true, query: false }, function() {
console.log('Updating');
});
const Model = mongoose.model('Test', schema);
const doc = new Model();
await doc.updateOne({ $set: { name: 'test' } }); // Prints "Updating"
// Doesn't print "Updating", because `Query#updateOne()` doesn't fire
// document middleware.
await Model.updateOne({}, { $set: { name: 'test' } });
Error Handling Middleware
New in 4.5.0
Middleware execution normally stops the first time a piece of middleware
calls next()
with an error. However, there is a special kind of post
middleware called "error handling middleware" that executes specifically
when an error occurs. Error handling middleware is useful for reporting
errors and making error messages more readable.
Error handling middleware is defined as middleware that takes one extra parameter: the 'error' that occurred as the first parameter to the function. Error handling middleware can then transform the error however you want.
const schema = new Schema({
name: {
type: String,
// Will trigger a MongoError with code 11000 when
// you save a duplicate
unique: true
}
});
// Handler **must** take 3 parameters: the error that occurred, the document
// in question, and the `next()` function
schema.post('save', function(error, doc, next) {
if (error.name === 'MongoError' && error.code === 11000) {
next(new Error('There was a duplicate key error'));
} else {
next();
}
});
// Will trigger the `post('save')` error handler
Person.create([{ name: 'Axl Rose' }, { name: 'Axl Rose' }]);
Error handling middleware also works with query middleware. You can
also define a post update()
hook that will catch MongoDB duplicate key
errors.
// The same E11000 error can occur when you call `update()`
// This function **must** take 3 parameters. If you use the
// `passRawResult` function, this function **must** take 4
// parameters
schema.post('update', function(error, res, next) {
if (error.name === 'MongoError' && error.code === 11000) {
next(new Error('There was a duplicate key error'));
} else {
next(); // The `update()` call will still error out.
}
});
const people = [{ name: 'Axl Rose' }, { name: 'Slash' }];
Person.create(people, function(error) {
Person.update({ name: 'Slash' }, { $set: { name: 'Axl Rose' } }, function(error) {
// `error.message` will be "There was a duplicate key error"
});
});
Error handling middleware can transform an error, but it can't remove the
error. Even if you call next()
with no error as shown above, the
function call will still error out.
Aggregation Hooks
You can also define hooks for the Model.aggregate()
function.
In aggregation middleware functions, this
refers to the Mongoose Aggregate
object.
For example, suppose you're implementing soft deletes on a Customer
model
by adding an isDeleted
property. To make sure aggregate()
calls only look
at customers that aren't soft deleted, you can use the below middleware to
add a $match
stage to the beginning
of each aggregation pipeline.
customerSchema.pre('aggregate', function() {
// Add a $match state to the beginning of each pipeline.
this.pipeline().unshift({ $match: { isDeleted: { $ne: true } } });
});
The Aggregate#pipeline()
function
lets you access the MongoDB aggregation pipeline that Mongoose will send to
the MongoDB server. It is useful for adding stages to the beginning of the
pipeline from middleware.
Synchronous Hooks
Certain Mongoose hooks are synchronous, which means they do not support
functions that return promises or receive a next()
callback. Currently,
only init
hooks are synchronous, because the init()
function
is synchronous. Below is an example of using pre and post init hooks.
const schema = new Schema({ title: String, loadedAt: Date });
schema.pre('init', pojo => {
assert.equal(pojo.constructor.name, 'Object'); // Plain object before init
});
const now = new Date();
schema.post('init', doc => {
assert.ok(doc instanceof mongoose.Document); // Mongoose doc after init
doc.loadedAt = now;
});
const Test = db.model('Test', schema);
return Test.create({ title: 'Casino Royale' }).
then(doc => Test.findById(doc)).
then(doc => assert.equal(doc.loadedAt.valueOf(), now.valueOf()));
To report an error in an init hook, you must throw a synchronous error. Unlike all other middleware, init middleware does not handle promise rejections.
const schema = new Schema({ title: String });
const swallowedError = new Error('will not show');
// init hooks do **not** handle async errors or any sort of async behavior
schema.pre('init', () => Promise.reject(swallowedError));
schema.post('init', () => { throw Error('will show'); });
const Test = db.model('Test', schema);
return Test.create({ title: 'Casino Royale' }).
then(doc => Test.findById(doc)).
catch(error => assert.equal(error.message, 'will show'));
Next Up
Now that we've covered middleware, let's take a look at Mongoose's approach to faking JOINs with its query population helper.