Decorators and some other features for sequelize (v6).
- Installation
- Model Definition
- Usage
- Model association
- Indexes
- Repository mode
- Model validation
- Scopes
- Hooks
- Why
() => Model
? - Recommendations and limitations
- this assumes usage of
sequelize@6
- sequelize-typescript requires sequelize
- additional typings as documented here and reflect-metadata
npm install --save-dev @types/node @types/validator npm install sequelize reflect-metadata sequelize-typescript
Your tsconfig.json
needs the following flags:
"target": "es6", // or a more recent ecmascript version "experimentalDecorators": true, "emitDecoratorMetadata": true
SequelizeConfig
renamed toSequelizeOptions
modelPaths
property renamed tomodels
The @Scopes
and @DefaultScope
decorators now take lambda's as options
@DefaultScope(() => ({...})) @Scopes(() => ({...}))
instead of deprecated way:
@DefaultScope({...}) @Scopes({...}))
import { Table, Column, Model, HasMany } from 'sequelize-typescript'; @Table class Person extends Model { @Column name: string; @Column birthday: Date; @HasMany(() => Hobby) hobbies: Hobby[]; }
import { Table, Model } from 'sequelize-typescript'; @Table class Person extends Model {}
import { Optional } from 'sequelize'; import { Table, Model } from 'sequelize-typescript'; interface PersonAttributes { id: number; name: string; } interface PersonCreationAttributes extends Optional<PersonAttributes, 'id'> {} @Table class Person extends Model<PersonAttributes, PersonCreationAttributes> {}
The model needs to extend the Model
class and has to be annotated with the @Table
decorator. All properties that should appear as a column in the database require the @Column
annotation.
See more advanced example here.
The @Table
annotation can be used without passing any parameters. To specify some more define options, use an object literal (all define options from sequelize are valid):
@Table({ timestamps: true, ... }) class Person extends Model {}
Decorator | Description |
---|---|
@Table | sets options.tableName=<CLASS_NAME> and options.modelName=<CLASS_NAME> automatically |
@Table(options: DefineOptions) | sets define options (also sets options.tableName=<CLASS_NAME> and options.modelName=<CLASS_NAME> if not already defined by define options) |
A primary key (id
) will be inherited from base class Model
. This primary key is by default an INTEGER
and has autoIncrement=true
(This behaviour is a native sequelize thing). The id can easily be overridden by marking another attribute as primary key. So either set @Column({primaryKey: true})
or use @PrimaryKey
together with @Column
.
Annotations to define custom and type safe createdAt
, updatedAt
and deletedAt
attributes:
@CreatedAt creationDate: Date; @UpdatedAt updatedOn: Date; @DeletedAt deletionDate: Date;
Decorator | Description |
---|---|
@CreatedAt | sets timestamps=true and createdAt='creationDate' |
@UpdatedAt | sets timestamps=true and updatedAt='updatedOn' |
@DeletedAt | sets timestamps=true , paranoid=true and deletedAt='deletionDate' |
The @Column
annotation can be used without passing any parameters. But therefore it is necessary that the js type can be inferred automatically (see Type inference for details).
@Column name: string;
If the type cannot or should not be inferred, use:
import {DataType} from 'sequelize-typescript'; @Column(DataType.TEXT) name: string;
Or for a more detailed column description, use an object literal (all attribute options from sequelize are valid):
@Column({ type: DataType.FLOAT, comment: 'Some value', ... }) value: number;
Decorator | Description |
---|---|
@Column | tries to infer dataType from js type |
@Column(dataType: DataType) | sets dataType explicitly |
@Column(options: AttributeOptions) | sets attribute options |
If you're in love with decorators: sequelize-typescript provides some more of them. The following decorators can be used together with the @Column annotation to make some attribute options easier available:
Decorator | Description | Options |
---|---|---|
@AllowNull(allowNull?: boolean) | sets attribute.allowNull (default is true ) | |
@AutoIncrement | sets attribute.autoIncrement=true | |
@Unique(options? UniqueOptions) | sets attribute.unique=true | UniqueOptions |
@Default(value: any) | sets attribute.defaultValue to specified value | |
@PrimaryKey | sets attribute.primaryKey=true | |
@Comment(value: string) | sets attribute.comment to specified string | |
Validate annotations | see Model validation |
The following types can be automatically inferred from javascript type. Others have to be defined explicitly.
Design type | Sequelize data type |
---|---|
string | STRING |
boolean | BOOLEAN |
number | INTEGER |
bigint | BIGINT |
Date | DATE |
Buffer | BLOB |
Get/set accessors do work as well
@Table class Person extends Model { @Column get name(): string { return 'My name is ' + this.getDataValue('name'); } set name(value: string) { this.setDataValue('name', value); } }
Except for minor variations sequelize-typescript will work like pure sequelize. (See sequelize docs)
To make the defined models available, you have to configure a Sequelize
instance from sequelize-typescript
(!).
import { Sequelize } from 'sequelize-typescript'; const sequelize = new Sequelize({ database: 'some_db', dialect: 'sqlite', username: 'root', password: '', storage: ':memory:', models: [__dirname + '/models'], // or [Player, Team], });
Before you can use your models you have to tell sequelize where they can be found. So either set models
in the sequelize config or add the required models later on by calling sequelize.addModels([Person])
or sequelize.addModels([__dirname + '/models'])
:
sequelize.addModels([Person]); sequelize.addModels(['path/to/models']);
import {Sequelize} from 'sequelize-typescript'; const sequelize = new Sequelize({ ... models: [__dirname + '/**/*.model.ts'] }); // or sequelize.addModels([__dirname + '/**/*.model.ts']);
A model is matched to a file by its filename. E.g.
// File User.ts matches the following exported model. export class User extends Model {}
This is done by comparison of the filename against all exported members. The matching can be customized by specifying the modelMatch
function in the configuration object.
For example, if your models are named user.model.ts
, and your class is called User
, you can match these two by using the following function:
import {Sequelize} from 'sequelize-typescript'; const sequelize = new Sequelize({ models: [__dirname + '/models/**/*.model.ts'] modelMatch: (filename, member) => { return filename.substring(0, filename.indexOf('.model')) === member.toLowerCase(); }, });
For each file that matches the *.model.ts
pattern, the modelMatch
function will be called with its exported members. E.g. for the following file
//user.model.ts import {Table, Column, Model} from 'sequelize-typescript'; export const UserN = 'Not a model'; export const NUser = 'Not a model'; @Table export class User extends Model { @Column nickname: string; }
The modelMatch
function will be called three times with the following arguments.
user.model UserN -> false user.model NUser -> false user.model User -> true (User will be added as model)
Another way to match model to file is to make your model the default export.
export default class User extends Model {}
⚠️ When using paths to add models, keep in mind that they will be loaded during runtime. This means that the path may differ from development time to execution time. For instance, using.ts
extension within paths will only work together with ts-node.
Instantiation and inserts can be achieved in the good old sequelize way
const person = Person.build({ name: 'bob', age: 99 }); person.save(); Person.create({ name: 'bob', age: 99 });
but sequelize-typescript also makes it possible to create instances with new
:
const person = new Person({ name: 'bob', age: 99 }); person.save();
Finding and updating entries does also work like using native sequelize. So see sequelize docs for more details.
Person.findOne().then((person) => { person.age = 100; return person.save(); }); Person.update( { name: 'bobby', }, { where: { id: 1 } } ).then(() => {});
Relations can be described directly in the model by the @HasMany
, @HasOne
, @BelongsTo
, @BelongsToMany
and @ForeignKey
annotations.
@Table class Player extends Model { @Column name: string; @Column num: number; @ForeignKey(() => Team) @Column teamId: number; @BelongsTo(() => Team) team: Team; } @Table class Team extends Model { @Column name: string; @HasMany(() => Player) players: Player[]; }
That's all, sequelize-typescript does everything else for you. So when retrieving a team by find
Team.findOne({ include: [Player] }).then((team) => { team.players.forEach((player) => console.log(`Player ${player.name}`)); });
the players will also be resolved (when passing include: Player
to the find options)
@Table class Book extends Model { @BelongsToMany(() => Author, () => BookAuthor) authors: Author[]; } @Table class Author extends Model { @BelongsToMany(() => Book, () => BookAuthor) books: Book[]; } @Table class BookAuthor extends Model { @ForeignKey(() => Book) @Column bookId: number; @ForeignKey(() => Author) @Column authorId: number; }
To access the through-table instance (instanceOf BookAuthor
in the upper example) type safely, the type need to be set up manually. For Author
model it can be achieved like so:
@BelongsToMany(() => Book, () => BookAuthor) books: Array<Book & {BookAuthor: BookAuthor}>;
For one-to-one use @HasOne(...)
(foreign key for the relation exists on the other model) and @BelongsTo(...)
(foreign key for the relation exists on this model)
Decorator | Description |
---|---|
@ForeignKey(relatedModelGetter: () => typeof Model) | marks property as foreignKey for related class |
@BelongsTo(relatedModelGetter: () => typeof Model) | sets SourceModel.belongsTo(RelatedModel, ...) while as is key of annotated property and foreignKey is resolved from source class |
@BelongsTo(relatedModelGetter: () => typeof Model, foreignKey: string) | sets SourceModel.belongsTo(RelatedModel, ...) while as is key of annotated property and foreignKey is explicitly specified value |
@BelongsTo(relatedModelGetter: () => typeof Model, options: AssociationOptionsBelongsTo) | sets SourceModel.belongsTo(RelatedModel, ...) while as is key of annotated property and options are additional association options |
@HasMany(relatedModelGetter: () => typeof Model) | sets SourceModel.hasMany(RelatedModel, ...) while as is key of annotated property and foreignKey is resolved from target related class |
@HasMany(relatedModelGetter: () => typeof Model, foreignKey: string) | sets SourceModel.hasMany(RelatedModel, ...) while as is key of annotated property and foreignKey is explicitly specified value |
@HasMany(relatedModelGetter: () => typeof Model, options: AssociationOptionsHasMany) | sets SourceModel.hasMany(RelatedModel, ...) while as is key of annotated property and options are additional association options |
@HasOne(relatedModelGetter: () => typeof Model) | sets SourceModel.hasOne(RelatedModel, ...) while as is key of annotated property and foreignKey is resolved from target related class |
@HasOne(relatedModelGetter: () => typeof Model, foreignKey: string) | sets SourceModel.hasOne(RelatedModel, ...) while as is key of annotated property and foreignKey is explicitly specified value |
@HasOne(relatedModelGetter: () => typeof Model, options: AssociationOptionsHasOne) | sets SourceModel.hasOne(RelatedModel, ...) while as is key of annotated property and options are additional association options |
@BelongsToMany(relatedModelGetter: () => typeof Model, through: (() => typeof Model)) | sets SourceModel.belongsToMany(RelatedModel, {through: ThroughModel, ...}) while as is key of annotated property and foreignKey /otherKey is resolved from through class |
@BelongsToMany(relatedModelGetter: () => typeof Model, through: (() => typeof Model), foreignKey: string) | sets SourceModel.belongsToMany(RelatedModel, {through: ThroughModel, ...}) while as is key of annotated property, foreignKey is explicitly specified value and otherKey is resolved from through class |
@BelongsToMany(relatedModelGetter: () => typeof Model, through: (() => typeof Model), foreignKey: string, otherKey: string) | sets SourceModel.belongsToMany(RelatedModel, {through: ThroughModel, ...}) while as is key of annotated property and foreignKey /otherKey are explicitly specified values |
@BelongsToMany(relatedModelGetter: () => typeof Model, through: string, foreignKey: string, otherKey: string) | sets SourceModel.belongsToMany(RelatedModel, {through: throughString, ...}) while as is key of annotated property and foreignKey /otherKey are explicitly specified values |
@BelongsToMany(relatedModelGetter: () => typeof Model, options: AssociationOptionsBelongsToMany) | sets SourceModel.belongsToMany(RelatedModel, {through: throughString, ...}) while as is key of annotated property and options are additional association values, including foreignKey and otherKey . |
Note that when using AssociationOptions, certain properties will be overwritten when the association is built, based on reflection metadata or explicit attribute parameters. For example, as
will always be the annotated property's name, and through
will be the explicitly stated value.
sequelize-typescript resolves the foreign keys by identifying the corresponding class references. So if you define a model with multiple relations like
@Table class Book extends Model { @ForeignKey(() => Person) @Column authorId: number; @BelongsTo(() => Person) author: Person; @ForeignKey(() => Person) @Column proofreaderId: number; @BelongsTo(() => Person) proofreader: Person; } @Table class Person extends Model { @HasMany(() => Book) writtenBooks: Book[]; @HasMany(() => Book) proofedBooks: Book[]; }
sequelize-typescript cannot know which foreign key to use for which relation. So you have to add the foreign keys explicitly:
// in class "Books": @BelongsTo(() => Person, 'authorId') author: Person; @BelongsTo(() => Person, 'proofreaderId') proofreader: Person; // in class "Person": @HasMany(() => Book, 'authorId') writtenBooks: Book[]; @HasMany(() => Book, 'proofreaderId') proofedBooks: Book[];
With the creation of a relation, sequelize generates some method on the corresponding models. So when you create a 1:n relation between ModelA
and ModelB
, an instance of ModelA
will have the functions getModelBs
, setModelBs
, addModelB
, removeModelB
, hasModelB
. These functions still exist with sequelize-typescript. But TypeScript wont recognize them and will complain if you try to access getModelB
, setModelB
or addModelB
. To make TypeScript happy, the Model.prototype
of sequelize-typescript has $set
, $get
, $add
functions.
@Table class ModelA extends Model { @HasMany(() => ModelB) bs: ModelB[]; } @Table class ModelB extends Model { @BelongsTo(() => ModelA) a: ModelA; }
To use them, pass the property key of the respective relation as the first parameter:
const modelA = new ModelA(); modelA .$set('bs', [ /* instance */ ]) .then(/* ... */); modelA.$add('b' /* instance */).then(/* ... */); modelA.$get('bs').then(/* ... */); modelA.$count('bs').then(/* ... */); modelA.$has('bs').then(/* ... */); modelA.$remove('bs' /* instance */).then(/* ... */); modelA.$create('bs' /* value */).then(/* ... */);
The @Index
annotation can be used without passing any parameters.
@Table class Person extends Model { @Index // Define an index with default name @Column name: string; @Index // Define another index @Column birthday: Date; }
To specify index and index field options, use an object literal (see indexes define option):
@Table class Person extends Model { @Index('my-index') // Define a multi-field index on name and birthday @Column name: string; @Index('my-index') // Add birthday as the second field to my-index @Column birthday: Date; @Index({ // index options name: 'job-index', parser: 'my-parser', type: 'UNIQUE', unique: true, where: { isEmployee: true }, concurrently: true, using: 'BTREE', operator: 'text_pattern_ops', prefix: 'test-', // index field options length: 10, order: 'ASC', collate: 'NOCASE', }) @Column jobTitle: string; @Column isEmployee: boolean; }
Decorator | Description |
---|---|
@Index | adds new index on decorated field to options.indexes |
@Index(name: string) | adds new index or adds the field to an existing index with specified name |
@Table(options: IndexDecoratorOptions) | sets both index and index field options |
The createIndexDecorator()
function can be used to create a decorator for an index with options specified with an object literal supplied as the argument. Fields are added to the index by decorating properties.
const SomeIndex = createIndexDecorator(); const JobIndex = createIndexDecorator({ // index options name: 'job-index', parser: 'my-parser', type: 'UNIQUE', unique: true, where: { isEmployee: true }, concurrently: true, using: 'BTREE', operator: 'text_pattern_ops', prefix: 'test-', }); @Table class Person extends Model { @SomeIndex // Add name to SomeIndex @Column name: string; @SomeIndex // Add birthday to SomeIndex @Column birthday: Date; @JobIndex({ // index field options length: 10, order: 'ASC', collate: 'NOCASE', }) @Column jobTitle: string; @Column isEmployee: boolean; }
With sequelize-typescript@1
comes a repository mode. See docs for details.
The repository mode makes it possible to separate static operations like find
, create
, ... from model definitions. It also empowers models so that they can be used with multiple sequelize instances.
Enable repository mode by setting repositoryMode
flag:
const sequelize = new Sequelize({ repositoryMode: true, ..., });
Retrieve repository to create instances or perform search operations:
const userRepository = sequelize.getRepository(User); const luke = await userRepository.create({ name: 'Luke Skywalker' }); const luke = await userRepository.findOne({ where: { name: 'luke' } });
For now one need to use the repositories within the include options in order to retrieve or create related data:
const userRepository = sequelize.getRepository(User); const addressRepository = sequelize.getRepository(Address); userRepository.find({ include: [addressRepository] }); userRepository.create({ name: 'Bear' }, { include: [addressRepository] });
⚠️ This will change in the future: One will be able to refer the model classes instead of the repositories.
Nested scopes and includes in general won't work when using @Scope
annotation together with repository mode like:
@Scopes(() => ({ // includes withAddress: { include: [() => Address], }, // nested scopes withAddressIncludingLatLng: { include: [() => Address.scope('withLatLng')], }, })) @Table class User extends Model {}
⚠️ This will change in the future: Simple includes will be implemented.
Validation options can be set through the @Column
annotation, but if you prefer to use separate decorators for validation instead, you can do so by simply adding the validate options as decorators: So that validate.isEmail=true
becomes @IsEmail
, validate.equals='value'
becomes @Equals('value')
and so on. Please notice that a validator that expects a boolean is translated to an annotation without a parameter.
See sequelize docs for all validators.
The following validators cannot simply be translated from sequelize validator to an annotation:
Validator | Annotation |
---|---|
validate.len=[number, number] | @Length({max?: number, min?: number}) |
validate[customName: string] | For custom validators also use the @Is(...) annotation: Either @Is('custom', (value) => { /* ... */}) or with named function @Is(function custom(value) { /* ... */}) |
const HEX_REGEX = /^#([A-Fa-f0-9]{6}|[A-Fa-f0-9]{3})$/; @Table export class Shoe extends Model { @IsUUID(4) @PrimaryKey @Column id: string; @Equals('lala') @Column readonly key: string; @Contains('Special') @Column special: string; @Length({ min: 3, max: 15 }) @Column brand: string; @IsUrl @Column brandUrl: string; @Is('HexColor', (value) => { if (!HEX_REGEX.test(value)) { throw new Error(`"${value}" is not a hex color value.`); } }) @Column primaryColor: string; @Is(function hexColor(value: string): void { if (!HEX_REGEX.test(value)) { throw new Error(`"${value}" is not a hex color value.`); } }) @Column secondaryColor: string; @Is(HEX_REGEX) @Column tertiaryColor: string; @IsDate @IsBefore('2017-02-27') @Column producedAt: Date; }
Scopes can be defined with annotations as well. The scope options are identical to native sequelize (See sequelize docs for more details)
@DefaultScope(() => ({ attributes: ['id', 'primaryColor', 'secondaryColor', 'producedAt'], })) @Scopes(() => ({ full: { include: [Manufacturer], }, yellow: { where: { primaryColor: 'yellow' }, }, })) @Table export class ShoeWithScopes extends Model { @Column readonly secretKey: string; @Column primaryColor: string; @Column secondaryColor: string; @Column producedAt: Date; @ForeignKey(() => Manufacturer) @Column manufacturerId: number; @BelongsTo(() => Manufacturer) manufacturer: Manufacturer; }
Hooks can be attached to your models. All Model-level hooks are supported. See the related unit tests for a summary.
Each hook must be a static
method. Multiple hooks can be attached to a single method, and you can define multiple methods for a given hook.
The name of the method cannot be the same as the name of the hook (for example, a @BeforeCreate
hook method cannot be named beforeCreate
). That’s because Sequelize has pre-defined methods with those names.
@Table export class Person extends Model { @Column name: string; @BeforeUpdate @BeforeCreate static makeUpperCase(instance: Person) { // this will be called when an instance is created or updated instance.name = instance.name.toLocaleUpperCase(); } @BeforeCreate static addUnicorn(instance: Person) { // this will also be called when an instance is created instance.name += ' 🦄'; } }
@ForeignKey(Model)
is much easier to read, so why is @ForeignKey(() => Model)
so important? When it comes to circular-dependencies (which are in general solved by node for you) Model
can be undefined
when it gets passed to @ForeignKey. With the usage of a function, which returns the actual model, we prevent this issue.
Unless you are using the repository mode, you won't be able to add one and the same model to multiple Sequelize instances with differently configured connections. So that one model will only work for one connection.
This is not only good practice regarding design, but also matters for the order of execution. Since Typescript creates a __metadata("design:type", SomeModel)
call due to emitDecoratorMetadata
compile option, in some cases SomeModel
is probably not defined(not undefined!) and would throw a ReferenceError
. When putting SomeModel
in a separate file, it would look like __metadata("design:type", SomeModel_1.SomeModel)
, which does not throw an error.
If you need to minify your code, you need to set tableName
and modelName
in the DefineOptions
for @Table
annotation. sequelize-typescript uses the class name as default name for tableName
and modelName
. When the code is minified the class name will no longer be the originally defined one (So that class User
will become class b
for example).
To contribute you can:
- Open issues and participate in discussion of other issues.
- Fork the project to open up PR's.
- Update the types of Sequelize.
- Anything else constructively helpful.
In order to open a pull request please:
- Create a new branch.
- Run tests locally (
npm install && npm run build && npm run cover
) and ensure your commits don't break the tests. - Document your work well with commit messages, a good PR description, comments in code when necessary, etc.
In order to update the types for sequelize please go to the Definitely Typed repo, it would also be a good idea to open a PR into sequelize so that Sequelize can maintain its own types, but that might be harder than getting updated types into microsoft's repo. The Typescript team is slowly trying to encourage npm package maintainers to maintain their own typings, but Microsoft still has dedicated and good people maintaining the DT repo, accepting PR's and keeping quality high.
Keep in mind sequelize-typescript
does not provide typings for sequelize
- these are seperate things. A lot of the types in sequelize-typescript
augment, refer to, or extend what sequelize already has.