Mongoose now requires node.js >= 4.0.0 and MongoDB >= 3.0.0. MongoDB 2.6 and Node.js < 4 where both EOL-ed in 2016.
Query middleware is now compiled when you call mongoose.model()
or db.model()
. If you add query middleware after calling mongoose.model()
, that middleware will not get called.
const schema = new Schema({ name: String });
const MyModel = mongoose.model('Test', schema);
schema.pre('find', () => { console.log('find!'); });
MyModel.find().exec(function() {
// In mongoose 4.x, the above `.find()` will print "find!"
// In mongoose 5.x, "find!" will **not** be printed.
// Call `pre('find')` **before** calling `mongoose.model()` to make the middleware apply.
});
mongoose.connect()
mongoose.connect()
and mongoose.disconnect()
now return a promise if no callback specified, or null
otherwise. It does not return the mongoose singleton.
// Worked in mongoose 4. Does **not** work in mongoose 5, `mongoose.connect()`
// now returns a promise consistently. This is to avoid the horrible things
// we've done to allow mongoose to be a thenable that resolves to itself.
mongoose.connect('mongodb://localhost:27017/test').model('Test', new Schema({}));
// Do this instead
mongoose.connect('mongodb://localhost:27017/test');
mongoose.model('Test', new Schema({}));
useMongoClient
The useMongoClient
option was
removed in Mongoose 5, it is now always true
. As a consequence, Mongoose 5
no longer supports several function signatures for mongoose.connect()
that
worked in Mongoose 4.x if the useMongoClient
option was off. Below are some
examples of mongoose.connect()
calls that do not work in Mongoose 5.x.
mongoose.connect('localhost', 27017);
mongoose.connect('localhost', 'mydb', 27017);
mongoose.connect('mongodb://host1:27017,mongodb://host2:27017');
In Mongoose 5.x, the first parameter to mongoose.connect()
and mongoose.createConnection()
, if specified, must be a MongoDB connection string. The
connection string and options are then passed down to the MongoDB Node.js driver’s MongoClient.connect()
function. Mongoose does not modify the connection string, although mongoose.connect()
and mongoose.createConnection()
support a few additional options in addition to the ones the MongoDB driver supports.
Setters run in reverse order in 4.x:
const schema = new Schema({ name: String });
schema.path('name').
get(() => console.log('This will print 2nd')).
get(() => console.log('This will print first'));
In 5.x, setters run in the order they’re declared.
const schema = new Schema({ name: String });
schema.path('name').
get(() => console.log('This will print first')).
get(() => console.log('This will print 2nd'));
Mongoose 5.1.0 introduced an _id
getter to ObjectIds that lets you get an ObjectId regardless of whether a path
is populated.
const blogPostSchema = new Schema({
title: String,
author: {
type: mongoose.Schema.Types.ObjectId,
ref: 'Author'
}
});
const BlogPost = mongoose.model('BlogPost', blogPostSchema);
await BlogPost.create({ title: 'test', author: author._id });
const blogPost = await BlogPost.findOne();
console.log(blogPost.author); // '5b207f84e8061d1d2711b421'
// New in Mongoose 5.1.0: this will print '5b207f84e8061d1d2711b421' as well
console.log(blogPost.author._id);
await blogPost.populate('author');
console.log(blogPost.author._id); '5b207f84e8061d1d2711b421'
As a consequence, checking whether blogPost.author._id
is no longer viable as a way to check whether author
is populated. Use blogPost.populated('author') != null
or blogPost.author instanceof mongoose.Types.ObjectId
to check whether author
is populated instead.
Note that you can call mongoose.set('objectIdGetter', false)
to change this behavior.
remove()
and deleteX()
deleteOne()
, deleteMany()
, and remove()
now resolve to the result object
rather than the full driver WriteOpResult
object.
// In 4.x, this is how you got the number of documents deleted
MyModel.deleteMany().then(res => console.log(res.result.n));
// In 5.x this is how you get the number of documents deleted
MyModel.deleteMany().then(res => res.n);
The useMongooseAggCursor
option from 4.x is now always on. This is the new syntax for aggregation cursors in mongoose 5:
// When you call `.cursor()`, `.exec()` will now return a mongoose aggregation
// cursor.
const cursor = MyModel.aggregate([{ $match: { name: 'Val' } }]).cursor().exec();
// No need to `await` on the cursor or wait for a promise to resolve
cursor.eachAsync(doc => console.log(doc));
// Can also pass options to `cursor()`
const cursorWithOptions = MyModel.
aggregate([{ $match: { name: 'Val' } }]).
cursor({ batchSize: 10 }).
exec();
Model.geoNear()
has been removed because the MongoDB driver no longer supports it
Due to changes in the MongoDB driver, connection strings must be URI encoded.
If they are not, connections may fail with an illegal character message.
See a full list of affected characters.
If your app is used by a lot of different connection strings, it’s possible that your test cases will pass, but production passwords will fail. Encode all your connection strings to be safe.
If you want to continue to use unencoded connection strings, the easiest fix is to use
the mongodb-uri
module to parse the connection strings, and then produce the properly encoded
versions. You can use a function like this:
const uriFormat = require('mongodb-uri')
function encodeMongoURI (urlString) {
if (urlString) {
let parsed = uriFormat.parse(urlString)
urlString = uriFormat.format(parsed);
}
return urlString;
}
}
// Your un-encoded string.
const mongodbConnectString = "mongodb://...";
mongoose.connect(encodeMongoURI(mongodbConnectString))
The function above is safe to use whether the existing string is already encoded or not.
Domain sockets must be URI encoded. For example:
// Works in mongoose 4. Does **not** work in mongoose 5 because of more
// stringent URI parsing.
const host = '/tmp/mongodb-27017.sock';
mongoose.createConnection(`mongodb://aaron:psw@${host}/fake`);
// Do this instead
const host = encodeURIComponent('/tmp/mongodb-27017.sock');
mongoose.createConnection(`mongodb://aaron:psw@${host}/fake`);
toObject()
OptionsThe options
parameter to toObject()
and toJSON()
merge defaults rather than overwriting them.
// Note the `toObject` option below
const schema = new Schema({ name: String }, { toObject: { virtuals: true } });
schema.virtual('answer').get(() => 42);
const MyModel = db.model('MyModel', schema);
const doc = new MyModel({ name: 'test' });
// In mongoose 4.x this prints "undefined", because `{ minimize: false }`
// overwrites the entire schema-defined options object.
// In mongoose 5.x this prints "42", because `{ minimize: false }` gets
// merged with the schema-defined options.
console.log(doc.toJSON({ minimize: false }).answer);
aggregate()
no longer accepts a spread, you must pass your aggregation pipeline as an array. The below code worked in 4.x:
MyModel.aggregate({ $match: { isDeleted: false } }, { $skip: 10 }).exec(cb);
The above code does not work in 5.x, you must wrap the $match
and $skip
stages in an array.
MyModel.aggregate([{ $match: { isDeleted: false } }, { $skip: 10 }]).exec(cb);
By default, mongoose 4 would coerce any value to a boolean without error.
// Fine in mongoose 4, would save a doc with `boolField = true`
const MyModel = mongoose.model('Test', new Schema({
boolField: Boolean
}));
MyModel.create({ boolField: 'not a boolean' });
Mongoose 5 only casts the following values to true
:
true
'true'
1
'1'
'yes'
And the following values to false
:
false
'false'
0
'0'
'no'
All other values will cause a CastError
Casting for update()
, updateOne()
, updateMany()
, replaceOne()
,
remove()
, deleteOne()
, and deleteMany()
doesn’t happen until exec()
.
This makes it easier for hooks and custom query helpers to modify data, because
mongoose won’t restructure the data you passed in until after your hooks and
query helpers have ran. It also makes it possible to set the overwrite
option
after passing in an update.
// In mongoose 4.x, this becomes `{ $set: { name: 'Baz' } }` despite the `overwrite`
// In mongoose 5.x, this overwrite is respected and the first document with
// `name = 'Bar'` will be replaced with `{ name: 'Baz' }`
User.where({ name: 'Bar' }).update({ name: 'Baz' }).setOptions({ overwrite: true });
Post hooks now get flow control, which means async post save hooks and child document post save hooks execute before your save()
callback.
const ChildModelSchema = new mongoose.Schema({
text: {
type: String
}
});
ChildModelSchema.post('save', function(doc) {
// In mongoose 5.x this will print **before** the `console.log()`
// in the `save()` callback. In mongoose 4.x this was reversed.
console.log('Child post save');
});
const ParentModelSchema = new mongoose.Schema({
children: [ChildModelSchema]
});
const Model = mongoose.model('Parent', ParentModelSchema);
const m = new Model({ children: [{ text: 'test' }] });
m.save(function() {
// In mongoose 5.xm this prints **after** the "Child post save" message.
console.log('Save callback');
});
$pushAll
Operator$pushAll
is no longer supported and no longer used internally for save()
, since it has been deprecated since MongoDB 2.4. Use $push
with $each
instead.
The retainKeyOrder
option was removed, mongoose will now always retain the same key position when cloning objects. If you have queries or indexes that rely on reverse key order, you will have to change them.
Setters now run on queries by default, and the old runSettersOnQuery
option
has been removed.
const schema = new Schema({
email: { type: String, lowercase: true }
});
const Model = mongoose.model('Test', schema);
Model.find({ email: 'FOO@BAR.BAZ' }); // Converted to `find({ email: 'foo@bar.baz' })`
We no longer have a pre-compiled version of mongoose for the browser. If you want to use mongoose schemas in the browser, you need to build your own bundle with browserify/webpack.
The saveErrorIfNotFound
option was removed, mongoose will now always error out from save()
if the underlying document was not found
init
hooks are now fully synchronous and do not receive next()
as a parameter.
Document.prototype.init()
no longer takes a callback as a parameter. It
was always synchronous, just had a callback for legacy reasons.
numAffected
and save()
doc.save()
no longer passes numAffected
as a 3rd param to its callback.
remove()
and debouncingdoc.remove()
no longer debounces
getPromiseConstructor()
getPromiseConstructor()
is gone, just use mongoose.Promise
.
You cannot pass parameters to the next pre middleware in the chain using next()
in mongoose 5.x. In mongoose 4, next('Test')
in pre middleware would call the
next middleware with ‘Test’ as a parameter. Mongoose 5.x has removed support for this.
required
validator for arraysIn mongoose 5 the required
validator only verifies if the value is an array. That is, it will not fail for empty arrays as it would in mongoose 4.
In mongoose 5 the default debug function uses console.info()
to display messages instead of console.error()
.
In Mongoose 4.x, overwriting a filter property that’s a primitive with one that is an object would silently fail. For example, the below code would ignore the where()
and be equivalent to Sport.find({ name: 'baseball' })
Sport.find({ name: 'baseball' }).where({name: {$ne: 'softball'}});
In Mongoose 5.x, the above code will correctly overwrite 'baseball'
with { $ne: 'softball' }
bulkWrite()
resultsMongoose 5.x uses version 3.x of the MongoDB Node.js driver. MongoDB driver 3.x changed the format of
the result of bulkWrite()
calls so there is no longer a top-level nInserted
, nModified
, etc. property. The new result object structure is described here.
const Model = mongoose.model('Test', new Schema({ name: String }));
const res = await Model.bulkWrite([{ insertOne: { document: { name: 'test' } } }]);
console.log(res);
In Mongoose 4.x, the above will print:
BulkWriteResult {
ok: [Getter],
nInserted: [Getter],
nUpserted: [Getter],
nMatched: [Getter],
nModified: [Getter],
nRemoved: [Getter],
getInsertedIds: [Function],
getUpsertedIds: [Function],
getUpsertedIdAt: [Function],
getRawResponse: [Function],
hasWriteErrors: [Function],
getWriteErrorCount: [Function],
getWriteErrorAt: [Function],
getWriteErrors: [Function],
getLastOp: [Function],
getWriteConcernError: [Function],
toJSON: [Function],
toString: [Function],
isOk: [Function],
insertedCount: 1,
matchedCount: 0,
modifiedCount: 0,
deletedCount: 0,
upsertedCount: 0,
upsertedIds: {},
insertedIds: { '0': 5be9a3101638a066702a0d38 },
n: 1 }
In Mongoose 5.x, the script will print:
BulkWriteResult {
result:
{ ok: 1,
writeErrors: [],
writeConcernErrors: [],
insertedIds: [ [Object] ],
nInserted: 1,
nUpserted: 0,
nMatched: 0,
nModified: 0,
nRemoved: 0,
upserted: [],
lastOp: { ts: [Object], t: 1 } },
insertedCount: 1,
matchedCount: 0,
modifiedCount: 0,
deletedCount: 0,
upsertedCount: 0,
upsertedIds: {},
insertedIds: { '0': 5be9a1c87decfc6443dd9f18 },
n: 1 }