ca7af9c8cc
* feat: new version of the documentation * feat: add English catalog translation * Update quickstart.md * Update quickstart.zh-CN.md * Update quickstart.zh-CN.md * Update quickstart.zh-CN.md * Update quickstart.zh-CN.md * feat: update quickstart * update doc * update pepository api doc Co-authored-by: ChengLei Shao <chareice@live.com>
1269 lines
38 KiB
Markdown
1269 lines
38 KiB
Markdown
---
|
||
order: 4
|
||
---
|
||
|
||
# 如何选择 ORM
|
||
|
||
本篇文章,需要读者至少熟悉一种较流行的 ORM,对 Model、Migration、QueryBuilder、Repository 也有所了解。在正式介绍 NocoBase 的 Database 设计之前,先来看看大部分 ORM 都有的三个概念:
|
||
|
||
- Model(ModelAttributes)、Entity(EntitySchema):将数据表与模型类或实体类对应起来
|
||
- Migration、Sync API:用于创建、修改、删除数据库表、字段、索引等
|
||
- QueryBuilder、EntityManager、Repository、CRUD API:提供增删改查
|
||
|
||
|
||
|
||
## Model/Entity
|
||
无代码的第一个改造,**Model 动态化**。
|
||
|
||
|
||
简单来说 Model/Entity 的作用就是将数据表、字段、索引、关系映射到类、属性或方法上。我们先来看看 Node.js 里各个 ORM 都是怎么做的。
|
||
### Typeorm
|
||
在 Typeorm 里叫 Entity,通过类属性映射表字段,装饰器风格来配置字段属性
|
||
```typescript
|
||
import { Entity, Column, PrimaryGeneratedColumn } from "typeorm";
|
||
|
||
@Entity()
|
||
export class Photo {
|
||
|
||
@PrimaryGeneratedColumn()
|
||
id: number;
|
||
|
||
@Column({
|
||
length: 100
|
||
})
|
||
name: string;
|
||
|
||
@Column("text")
|
||
description: string;
|
||
|
||
@Column()
|
||
filename: string;
|
||
|
||
@Column("double")
|
||
views: number;
|
||
|
||
@Column()
|
||
isPublished: boolean;
|
||
}
|
||
```
|
||
这种写法,如果 Entity 有修改,需要修改代码,不过 Typeorm 提供了 JSON 风格的 EntitySchema
|
||
```typescript
|
||
import {EntitySchema} from "typeorm";
|
||
|
||
export const CategoryEntity = new EntitySchema({
|
||
name: "category",
|
||
columns: {
|
||
id: {
|
||
type: Number,
|
||
primary: true,
|
||
generated: true
|
||
},
|
||
name: {
|
||
type: String
|
||
}
|
||
}
|
||
});
|
||
|
||
export const PostEntity = new EntitySchema({
|
||
name: "post",
|
||
columns: {
|
||
id: {
|
||
type: Number,
|
||
primary: true,
|
||
generated: true
|
||
},
|
||
title: {
|
||
type: String
|
||
},
|
||
text: {
|
||
type: String
|
||
}
|
||
},
|
||
relations: {
|
||
categories: {
|
||
type: "many-to-many",
|
||
target: "category" // CategoryEntity
|
||
}
|
||
}
|
||
});
|
||
```
|
||
修改 JSON 就容易多了,这种写法非常适用于无代码平台。平台配置 JSON 动态生成对应的 Entity。
|
||
### Prisma
|
||
与 Typeorm 装饰器的风格非常接近,但不同的是 Prisma 另辟蹊径,提供了自成一套的 PSL(Prisma Schema Language):
|
||
```ts
|
||
datasource db {
|
||
url = env("DATABASE_URL")
|
||
provider = "postgresql"
|
||
}
|
||
|
||
generator client {
|
||
provider = "prisma-client-js"
|
||
}
|
||
|
||
model User {
|
||
id Int @id @default(autoincrement())
|
||
createdAt DateTime @default(now())
|
||
email String @unique
|
||
name String?
|
||
role Role @default(USER)
|
||
posts Post[]
|
||
}
|
||
|
||
model Post {
|
||
id Int @id @default(autoincrement())
|
||
createdAt DateTime @default(now())
|
||
updatedAt DateTime @updatedAt
|
||
published Boolean @default(false)
|
||
title String @db.VarChar(255)
|
||
author User? @relation(fields: [authorId], references: [id])
|
||
authorId Int?
|
||
}
|
||
|
||
enum Role {
|
||
USER
|
||
ADMIN
|
||
}
|
||
```
|
||
这种写法,需要修改 PSL 文件,除非自己实现一套 PSL 解析与生成器,不然没办法直接无代码改造。
|
||
### Sequelize
|
||
作为老牌 ORM,下载量和使用量也是惊人,提供了多种配置 Model 的风格:
|
||
|
||
1. 传统的 JSON 风格
|
||
```ts
|
||
sequelize.define('User', {
|
||
// Model attributes are defined here
|
||
firstName: {
|
||
type: DataTypes.STRING,
|
||
allowNull: false
|
||
},
|
||
lastName: {
|
||
type: DataTypes.STRING
|
||
// allowNull defaults to true
|
||
}
|
||
}, {
|
||
// model options
|
||
});
|
||
```
|
||
|
||
2. 改进之后的 Model 风格
|
||
```ts
|
||
const { Sequelize, DataTypes, Model } = require('sequelize');
|
||
const sequelize = new Sequelize('sqlite::memory');
|
||
|
||
class User extends Model {}
|
||
|
||
User.init({
|
||
// Model attributes are defined here
|
||
firstName: {
|
||
type: DataTypes.STRING,
|
||
allowNull: false
|
||
},
|
||
lastName: {
|
||
type: DataTypes.STRING
|
||
// allowNull defaults to true
|
||
}
|
||
}, {
|
||
// Other model options go here
|
||
sequelize, // We need to pass the connection instance
|
||
modelName: 'User' // We need to choose the model name
|
||
});
|
||
|
||
// the defined model is the class itself
|
||
console.log(User === sequelize.models.User); // true
|
||
```
|
||
|
||
3. 装饰器风格
|
||
```ts
|
||
import { Table, Column, Model, HasMany } from 'sequelize-typescript'
|
||
|
||
@Table
|
||
class Person extends Model {
|
||
@Column
|
||
name: string
|
||
|
||
@Column
|
||
birthday: Date
|
||
|
||
@HasMany(() => Hobby)
|
||
hobbies: Hobby[]
|
||
}
|
||
```
|
||
Sequelize 可以使用 JSON 风格配置 Model,非常适用于无代码平台。平台配置 JSON 动态生成对应的 Model。
|
||
### Objection.js
|
||
基于 Knex,Model 如下:
|
||
```typescript
|
||
const { Model } = require('objection');
|
||
|
||
class Person extends Model {
|
||
// Table name is the only required property.
|
||
static get tableName() {
|
||
return 'persons';
|
||
}
|
||
|
||
// Each model must have a column (or a set of columns) that uniquely
|
||
// identifies the rows. The column(s) can be specified using the `idColumn`
|
||
// property. `idColumn` returns `id` by default and doesn't need to be
|
||
// specified unless the model's primary key is something else.
|
||
static get idColumn() {
|
||
return 'id';
|
||
}
|
||
|
||
// Methods can be defined for model classes just as you would for
|
||
// any JavaScript class. If you want to include the result of these
|
||
// methods in the output json, see `virtualAttributes`.
|
||
fullName() {
|
||
return this.firstName + ' ' + this.lastName;
|
||
}
|
||
|
||
// Optional JSON schema. This is not the database schema!
|
||
// No tables or columns are generated based on this. This is only
|
||
// used for input validation. Whenever a model instance is created
|
||
// either explicitly or implicitly it is checked against this schema.
|
||
// See http://json-schema.org/ for more info.
|
||
static get jsonSchema() {
|
||
return {
|
||
type: 'object',
|
||
required: ['firstName', 'lastName'],
|
||
|
||
properties: {
|
||
id: { type: 'integer' },
|
||
parentId: { type: ['integer', 'null'] },
|
||
firstName: { type: 'string', minLength: 1, maxLength: 255 },
|
||
lastName: { type: 'string', minLength: 1, maxLength: 255 },
|
||
age: { type: 'number' },
|
||
|
||
// Properties defined as objects or arrays are
|
||
// automatically converted to JSON strings when
|
||
// writing to database and back to objects and arrays
|
||
// when reading from database. To override this
|
||
// behaviour, you can override the
|
||
// Model.jsonAttributes property.
|
||
address: {
|
||
type: 'object',
|
||
properties: {
|
||
street: { type: 'string' },
|
||
city: { type: 'string' },
|
||
zipCode: { type: 'string' }
|
||
}
|
||
}
|
||
}
|
||
};
|
||
}
|
||
|
||
// This object defines the relations to other models.
|
||
static get relationMappings() {
|
||
// Importing models here is a one way to avoid require loops.
|
||
const Animal = require('./Animal');
|
||
const Movie = require('./Movie');
|
||
|
||
return {
|
||
pets: {
|
||
relation: Model.HasManyRelation,
|
||
// The related model. This can be either a Model
|
||
// subclass constructor or an absolute file path
|
||
// to a module that exports one. We use a model
|
||
// subclass constructor `Animal` here.
|
||
modelClass: Animal,
|
||
join: {
|
||
from: 'persons.id',
|
||
to: 'animals.ownerId'
|
||
}
|
||
},
|
||
|
||
movies: {
|
||
relation: Model.ManyToManyRelation,
|
||
modelClass: Movie,
|
||
join: {
|
||
from: 'persons.id',
|
||
// ManyToMany relation needs the `through` object
|
||
// to describe the join table.
|
||
through: {
|
||
// If you have a model class for the join table
|
||
// you need to specify it like this:
|
||
// modelClass: PersonMovie,
|
||
from: 'persons_movies.personId',
|
||
to: 'persons_movies.movieId'
|
||
},
|
||
to: 'movies.id'
|
||
}
|
||
},
|
||
|
||
children: {
|
||
relation: Model.HasManyRelation,
|
||
modelClass: Person,
|
||
join: {
|
||
from: 'persons.id',
|
||
to: 'persons.parentId'
|
||
}
|
||
},
|
||
|
||
parent: {
|
||
relation: Model.BelongsToOneRelation,
|
||
modelClass: Person,
|
||
join: {
|
||
from: 'persons.parentId',
|
||
to: 'persons.id'
|
||
}
|
||
}
|
||
};
|
||
}
|
||
}
|
||
```
|
||
直观感受,并没有 Typeorm、Prisma、Sequelize 精炼。不过 Objection.js 提供了 mixin,也可以将 Model 进一步抽象,代码就精炼多了,也可以支持装饰器风格,比如这样:
|
||
```ts
|
||
const { compose, Model } = require('objection');
|
||
|
||
const mixins = compose(
|
||
SomeMixin,
|
||
SomeOtherMixin,
|
||
EvenMoreMixins,
|
||
LolSoManyMixins,
|
||
ImAMixinWithOptions({ foo: 'bar' })
|
||
);
|
||
|
||
class Person extends mixins(Model) {}
|
||
|
||
@SomeMixin
|
||
@MixinWithOptions({ foo: 'bar' })
|
||
class Person extends Model {}
|
||
```
|
||
从构思上来说,可改造空间大,也可以巧妙的将各种配置 JSON 化,从而达到动态生成的 Model 目的。
|
||
### Bookshelf.js
|
||
基于 Knex,也是类 JSON 的配置风格,一些非常流行的开源项目 Ghost、Strapi 就用的它。
|
||
```typescript
|
||
const knex = require('knex')({
|
||
client: 'mysql',
|
||
connection: process.env.MYSQL_DATABASE_CONNECTION
|
||
})
|
||
const bookshelf = require('bookshelf')(knex)
|
||
|
||
const User = bookshelf.model('User', {
|
||
tableName: 'users',
|
||
posts() {
|
||
return this.hasMany(Posts)
|
||
}
|
||
})
|
||
|
||
const Post = bookshelf.model('Post', {
|
||
tableName: 'posts',
|
||
tags() {
|
||
return this.belongsToMany(Tag)
|
||
}
|
||
})
|
||
|
||
const Tag = bookshelf.model('Tag', {
|
||
tableName: 'tags'
|
||
})
|
||
|
||
```
|
||
同样支持动态化改造。Bookshelf 的可改造空间巨大,不过看似作者已经不再维护了。
|
||
### 总结
|
||
到底哪个好呢?**仅从动态化 Model/Entity 角度来说:**
|
||
|
||
- Sequelize 细节做的最好
|
||
- Typeorm 非常活跃,才迭代了 v0.2.38,但 stars 就已经超 25k+ 以上所有 ORM 关注度最多,装饰器的风格也被大家所喜爱。但是 EntitySchema 的细节做的还不够,需要进一步优化和改造,存在许多未知
|
||
- Objection.js 的构思非常不错,尤其是 mixin,灵活性和可改造性非常强
|
||
- Bookshelf.js 也不错,如果是早几年,Bookshelf 可能会是我的第一选择
|
||
- 至于 Prisma,特立独行的 PSL,深受大家喜爱,但是 PSL 并不支持动态化 Model
|
||
|
||
|
||
|
||
以上 ORM 都是以关系型数据库为主,不过 Typeorm 和 Prisma 也支持 MongoDB(细节有差异),只支持 MongoDB 的 ORM 不在讨论范围内。
|
||
|
||
|
||
## Migration
|
||
|
||
|
||
有了 Model/Entity 之后,需要创建对应的数据库表、字段和索引。上文提及的大多数 Model/Entity 都可以详细的描述字段的属性和关系(Model 的 DSL),理论上就可以直接生成表和关系约束了,而并不需要单独再配置 migration 文件了。比如:
|
||
### Sequelize
|
||
提供了 `sequelize.sync()` 和 `Model.sync()` 方法,可以快速的根据 Model Attributes 生成数据表、字段和索引。sync 提供了丰富的参数,支持删掉重建、只新增不删除、只同步某些表等等处理。
|
||
```typescript
|
||
const User = sequelize.define('User', {
|
||
// Model attributes are defined here
|
||
firstName: {
|
||
type: DataTypes.STRING,
|
||
allowNull: false
|
||
},
|
||
lastName: {
|
||
type: DataTypes.STRING
|
||
// allowNull defaults to true
|
||
}
|
||
}, {
|
||
// Other model options go here
|
||
});
|
||
|
||
await sequelize.sync();
|
||
// Or
|
||
await User.sync();
|
||
```
|
||
除了 sync,Sequelize 也提供了 migration 工具,具体写法如:
|
||
```typescript
|
||
module.exports = {
|
||
up: (queryInterface, Sequelize) => {
|
||
return queryInterface.createTable('Person', {
|
||
name: Sequelize.DataTypes.STRING,
|
||
isBetaMember: {
|
||
type: Sequelize.DataTypes.BOOLEAN,
|
||
defaultValue: false,
|
||
allowNull: false
|
||
}
|
||
});
|
||
},
|
||
down: (queryInterface, Sequelize) => {
|
||
return queryInterface.dropTable('Person');
|
||
}
|
||
};
|
||
```
|
||
在生产环境或者希望精确控制时,都可以通过 Sequelize 提供的 queryInterface 来处理。
|
||
### Typeorm
|
||
没有给力的 sequelize.sync() 方法,但是提供了 synchronize: true 配置参数,效果类似,会自动创建表,因为会重新建表,并不适用于生产环境。生产环境建议使用更为安全的 migration 方式。
|
||
```typescript
|
||
createConnection({
|
||
type: "mysql",
|
||
host: "localhost",
|
||
port: 3306,
|
||
username: "root",
|
||
password: "admin",
|
||
database: "test",
|
||
entities: [
|
||
Photo
|
||
],
|
||
synchronize: true,
|
||
logging: false
|
||
});
|
||
```
|
||
Migration 如下,具体的 QueryRunner 细节大家可以看官网,提供了一套标准的 Migration API
|
||
```typescript
|
||
import {MigrationInterface, QueryRunner, Table, TableIndex, TableColumn, TableForeignKey } from "typeorm";
|
||
|
||
export class QuestionRefactoringTIMESTAMP implements MigrationInterface {
|
||
|
||
async up(queryRunner: QueryRunner): Promise<void> {
|
||
await queryRunner.createTable(new Table({
|
||
name: "question",
|
||
columns: [
|
||
{
|
||
name: "id",
|
||
type: "int",
|
||
isPrimary: true
|
||
},
|
||
{
|
||
name: "name",
|
||
type: "varchar",
|
||
}
|
||
]
|
||
}), true)
|
||
|
||
await queryRunner.createIndex("question", new TableIndex({
|
||
name: "IDX_QUESTION_NAME",
|
||
columnNames: ["name"]
|
||
}));
|
||
|
||
await queryRunner.createTable(new Table({
|
||
name: "answer",
|
||
columns: [
|
||
{
|
||
name: "id",
|
||
type: "int",
|
||
isPrimary: true
|
||
},
|
||
{
|
||
name: "name",
|
||
type: "varchar",
|
||
},
|
||
{
|
||
name: 'created_at',
|
||
type: 'timestamp',
|
||
default: 'now()'
|
||
}
|
||
]
|
||
}), true);
|
||
|
||
await queryRunner.addColumn("answer", new TableColumn({
|
||
name: "questionId",
|
||
type: "int"
|
||
}));
|
||
|
||
await queryRunner.createForeignKey("answer", new TableForeignKey({
|
||
columnNames: ["questionId"],
|
||
referencedColumnNames: ["id"],
|
||
referencedTableName: "question",
|
||
onDelete: "CASCADE"
|
||
}));
|
||
}
|
||
|
||
async down(queryRunner: QueryRunner): Promise<void> {
|
||
const table = await queryRunner.getTable("answer");
|
||
const foreignKey = table.foreignKeys.find(fk => fk.columnNames.indexOf("questionId") !== -1);
|
||
await queryRunner.dropForeignKey("answer", foreignKey);
|
||
await queryRunner.dropColumn("answer", "questionId");
|
||
await queryRunner.dropTable("answer");
|
||
await queryRunner.dropIndex("question", "IDX_QUESTION_NAME");
|
||
await queryRunner.dropTable("question");
|
||
}
|
||
|
||
}
|
||
```
|
||
### Knex
|
||
Objection.js 和 Bookshelf 都基于 Knex,所以 Migration 都使用的 Knex。
|
||
```typescript
|
||
knex.schema.createTable('users', function (table) {
|
||
table.increments();
|
||
table.string('name');
|
||
table.timestamps();
|
||
})
|
||
// Outputs:
|
||
// create table `users` (`id` int unsigned not null auto_increment primary key, `name` varchar(255), `created_at` datetime, `updated_at` datetime)
|
||
```
|
||
Knex 和 Laravel 的 QueryBuilder 非常接近,Schema Builder 也如此。单纯从 Migration API 设计来说,个人更喜欢这种语法风格,干净、简洁。
|
||
### Prisma
|
||
Migration 的思路与上述的做法不同,Prisma 提供了完整的配置 Model 的 PSL 语法,为开发环境提供了 migrate dev 支持,每次修改 PSL 文件之后,可以执行 migrate dev 生成对应变更的 sql 文件,生产环境再执行 migrate 命令来同步修改,详情查看官方文档 [Prisma Migrate](https://www.prisma.io/docs/concepts/components/prisma-migrate) 章节
|
||
```bash
|
||
migrations/
|
||
└─ 20210305110829_first_migration/
|
||
└─ migration.sql
|
||
└─ 20210305120829_add_fields/
|
||
└─ migration.sql
|
||
└─ 20210308102042_type-change/
|
||
└─ migration.sql
|
||
```
|
||
### 总结
|
||
|
||
- 流程上,Prisma 的做法最省事儿,只配置一份 PSL,每次修改之后,通过 migrate dev 命令生成 sql,无需修改。
|
||
- 其次是 Sequelize,提供了 sequelize.sync 方法,配置的 Model 都可以通过 sync 方法同步给数据库。但是这种方式有些暴力和冗余,如果能稍加改进就更好了。
|
||
- 至于传统的 Migration 做法,配置 Model 已经写了一份 DSL 了,配置 Migration 再写另外一份 DSL,非常不友好,无代码改造也非常困难。
|
||
|
||
|
||
|
||
**怎么无代码改造来解决 Migration 问题呢?**
|
||
Sequelize 的方案非常不错,虽然有些暴力和冗余,但是可以再稍加改进,尤其是需要达到生产环境的精准控制。
|
||
|
||
|
||
**sequelize.sync 是否会有安全问题呢?**
|
||
因为 sync 支持 force: true 参数,会强制删除重建,在生产环境要关掉。
|
||
修改了 Model 的 DSL,在 sync 里怎么判断是创建、修改或删除呢?篇幅有限这里就不细说了。
|
||
|
||
|
||
## QueryBuilder、EntityManager、Repository、CRUD API
|
||
|
||
|
||
有了 Model/Entity,也创建数据表和字段了,那接下来就能操控数据库了。Model/Entity 常见有两种模式 Active Record 和 Data Mapper,Typeorm 支持的最完整,所以我们先来看看 Typeorm 吧。
|
||
### Typeorm
|
||
|
||
1. Active Record 模式
|
||
```typescript
|
||
import {BaseEntity, Entity, PrimaryGeneratedColumn, Column} from "typeorm";
|
||
|
||
@Entity()
|
||
export class User extends BaseEntity {
|
||
|
||
@PrimaryGeneratedColumn()
|
||
id: number;
|
||
|
||
@Column()
|
||
firstName: string;
|
||
|
||
@Column()
|
||
lastName: string;
|
||
|
||
@Column()
|
||
isActive: boolean;
|
||
|
||
static findByName(firstName: string, lastName: string) {
|
||
return this.createQueryBuilder("user")
|
||
.where("user.firstName = :firstName", { firstName })
|
||
.andWhere("user.lastName = :lastName", { lastName })
|
||
.getMany();
|
||
}
|
||
}
|
||
|
||
// example how to save AR entity
|
||
const user = new User();
|
||
user.firstName = "Timber";
|
||
user.lastName = "Saw";
|
||
user.isActive = true;
|
||
await user.save();
|
||
|
||
// example how to remove AR entity
|
||
await user.remove();
|
||
|
||
// example how to load AR entities
|
||
const users = await User.find({ skip: 2, take: 5 });
|
||
const newUsers = await User.find({ isActive: true });
|
||
const timber = await User.findOne({ firstName: "Timber", lastName: "Saw" });
|
||
const timber = await User.findByName("Timber", "Saw");
|
||
```
|
||
|
||
2. Data Mapper 模式
|
||
```typescript
|
||
import {Entity, EntityRepository, Repository, PrimaryGeneratedColumn, Column} from "typeorm";
|
||
|
||
@Entity()
|
||
export class User {
|
||
|
||
@PrimaryGeneratedColumn()
|
||
id: number;
|
||
|
||
@Column()
|
||
firstName: string;
|
||
|
||
@Column()
|
||
lastName: string;
|
||
|
||
@Column()
|
||
isActive: boolean;
|
||
|
||
}
|
||
|
||
// custom repository
|
||
@EntityRepository()
|
||
export class UserRepository extends Repository<User> {
|
||
|
||
findByName(firstName: string, lastName: string) {
|
||
return this.createQueryBuilder("user")
|
||
.where("user.firstName = :firstName", { firstName })
|
||
.andWhere("user.lastName = :lastName", { lastName })
|
||
.getMany();
|
||
}
|
||
|
||
}
|
||
|
||
const userRepository = connection.getRepository(User);
|
||
|
||
// example how to save DM entity
|
||
const user = new User();
|
||
user.firstName = "Timber";
|
||
user.lastName = "Saw";
|
||
user.isActive = true;
|
||
await userRepository.save(user);
|
||
|
||
// example how to remove DM entity
|
||
await userRepository.remove(user);
|
||
|
||
// example how to load DM entities
|
||
const users = await userRepository.find({ skip: 2, take: 5 });
|
||
const newUsers = await userRepository.find({ isActive: true });
|
||
const timber = await userRepository.findOne({ firstName: "Timber", lastName: "Saw" });
|
||
|
||
// custom repository
|
||
const userRepository = connection.getCustomRepository(UserRepository);
|
||
const timber = await userRepository.findByName("Timber", "Saw");
|
||
```
|
||
从无代码、低代码改造角度来说,Data Mapper 模式更适合,通过 EntitySchema 生成 Entity,再交给 EntityManager 或 Repository 来处理数据的增删改查,也可以根据需要自定义 Repository。这就是低代码的扩展能力,通用的 Repository 处理常规需求,特殊需求自定义扩展。除了常用的 CRUD API,Typeorm 也提供了强大的 QueryBuilder 来自定义其他 API,如上文例子的 findByName。
|
||
### Sequelize
|
||
Sequelize 的 Model 是 Active Record 模式,提供了常用的 CRUD API,如:
|
||
```ts
|
||
const jane = await User.create({ name: "Jane" });
|
||
console.log(jane.name); // "Jane"
|
||
jane.name = "Ada";
|
||
// the name is still "Jane" in the database
|
||
await jane.save();
|
||
// Now the name was updated to "Ada" in the database!
|
||
|
||
Model.findAll({
|
||
attributes: ['foo', 'bar']
|
||
});
|
||
// SELECT foo, bar FROM ...
|
||
|
||
Post.findAll({
|
||
where: {
|
||
authorId: 2
|
||
}
|
||
});
|
||
// SELECT * FROM post WHERE authorId = 2
|
||
|
||
const { Op } = require("sequelize");
|
||
Post.findAll({
|
||
where: {
|
||
[Op.and]: [{ a: 5 }, { b: 6 }], // (a = 5) AND (b = 6)
|
||
[Op.or]: [{ a: 5 }, { b: 6 }], // (a = 5) OR (b = 6)
|
||
someAttribute: {
|
||
// Basics
|
||
[Op.eq]: 3, // = 3
|
||
[Op.ne]: 20, // != 20
|
||
[Op.is]: null, // IS NULL
|
||
[Op.not]: true, // IS NOT TRUE
|
||
[Op.or]: [5, 6], // (someAttribute = 5) OR (someAttribute = 6)
|
||
|
||
// Using dialect specific column identifiers (PG in the following example):
|
||
[Op.col]: 'user.organization_id', // = "user"."organization_id"
|
||
|
||
// Number comparisons
|
||
[Op.gt]: 6, // > 6
|
||
[Op.gte]: 6, // >= 6
|
||
[Op.lt]: 10, // < 10
|
||
[Op.lte]: 10, // <= 10
|
||
[Op.between]: [6, 10], // BETWEEN 6 AND 10
|
||
[Op.notBetween]: [11, 15], // NOT BETWEEN 11 AND 15
|
||
|
||
// Other operators
|
||
|
||
[Op.all]: sequelize.literal('SELECT 1'), // > ALL (SELECT 1)
|
||
|
||
[Op.in]: [1, 2], // IN [1, 2]
|
||
[Op.notIn]: [1, 2], // NOT IN [1, 2]
|
||
|
||
[Op.like]: '%hat', // LIKE '%hat'
|
||
[Op.notLike]: '%hat', // NOT LIKE '%hat'
|
||
[Op.startsWith]: 'hat', // LIKE 'hat%'
|
||
[Op.endsWith]: 'hat', // LIKE '%hat'
|
||
[Op.substring]: 'hat', // LIKE '%hat%'
|
||
[Op.iLike]: '%hat', // ILIKE '%hat' (case insensitive) (PG only)
|
||
[Op.notILike]: '%hat', // NOT ILIKE '%hat' (PG only)
|
||
[Op.regexp]: '^[h|a|t]', // REGEXP/~ '^[h|a|t]' (MySQL/PG only)
|
||
[Op.notRegexp]: '^[h|a|t]', // NOT REGEXP/!~ '^[h|a|t]' (MySQL/PG only)
|
||
[Op.iRegexp]: '^[h|a|t]', // ~* '^[h|a|t]' (PG only)
|
||
[Op.notIRegexp]: '^[h|a|t]', // !~* '^[h|a|t]' (PG only)
|
||
|
||
[Op.any]: [2, 3], // ANY ARRAY[2, 3]::INTEGER (PG only)
|
||
|
||
// In Postgres, Op.like/Op.iLike/Op.notLike can be combined to Op.any:
|
||
[Op.like]: { [Op.any]: ['cat', 'hat'] } // LIKE ANY ARRAY['cat', 'hat']
|
||
|
||
// There are more postgres-only range operators, see below
|
||
}
|
||
}
|
||
});
|
||
```
|
||
常规的 CRUD API 支持的还不错,但是自定义查询或者基于 QueryBuilder 实现更复杂查询的支持就弱爆了,比如:
|
||
```ts
|
||
Post.findAll({
|
||
where: {
|
||
[Op.or]: [
|
||
sequelize.where(sequelize.fn('char_length', sequelize.col('content')), 7),
|
||
{
|
||
content: {
|
||
[Op.like]: 'Hello%'
|
||
}
|
||
},
|
||
{
|
||
[Op.and]: [
|
||
{ status: 'draft' },
|
||
sequelize.where(sequelize.fn('char_length', sequelize.col('content')), {
|
||
[Op.gt]: 10
|
||
})
|
||
]
|
||
}
|
||
]
|
||
}
|
||
});
|
||
|
||
Model.findAll({
|
||
attributes: [
|
||
'id', 'foo', 'bar', 'baz', 'qux', 'hats', // We had to list all attributes...
|
||
[sequelize.fn('COUNT', sequelize.col('hats')), 'n_hats'] // To add the aggregation...
|
||
]
|
||
});
|
||
|
||
// This is shorter, and less error prone because it still works if you add / remove attributes from your model later
|
||
Model.findAll({
|
||
attributes: {
|
||
include: [
|
||
[sequelize.fn('COUNT', sequelize.col('hats')), 'n_hats']
|
||
]
|
||
}
|
||
});
|
||
```
|
||
Sequelize 在配置 Model 和 Migration 的表现上都比较不错,但是在 QueryBuilder 的支持上弱爆了,几乎不解决各数据库的兼容性问题,提供的 queryInterface 也非常难用,一点也不 SQL-Friendly。
|
||
### Prisma
|
||
提供了 Prisma Client 用于支持数据的 CRUD,用法与 Sequelize 相似。没有提供 QueryBuilder 如果常规 CRUD API 支持的不好,可能很难改造,这部分了解的不多,细节存在非常多未知。感兴趣的看官网 [Prisma Client](https://www.prisma.io/docs/concepts/components/prisma-client) 章节。
|
||
```ts
|
||
const user = await prisma.user.create({
|
||
data: {
|
||
email: 'elsa@prisma.io',
|
||
name: 'Elsa Prisma',
|
||
},
|
||
})
|
||
|
||
const getPosts = await prisma.post.findMany({
|
||
where: {
|
||
title: {
|
||
contains: 'cookies',
|
||
},
|
||
},
|
||
include: {
|
||
author: true, // Return all fields
|
||
},
|
||
})
|
||
```
|
||
这种风格的 CRUD API 非常不利于调试,一旦出现问题将很难改造。
|
||
### Objection.js
|
||
得益于 Knex,Objection 在 QueryBuilder 的表现上非常不错 SQL-Friendly,可造性非常强。
|
||
```ts
|
||
const jennifer = await Person.query().insert({
|
||
firstName: 'Jennifer',
|
||
lastName: 'Lawrence'
|
||
});
|
||
// insert into "persons" ("firstName", "lastName") values ('Jennifer', 'Lawrence')
|
||
|
||
console.log(jennifer instanceof Person); // --> true
|
||
console.log(jennifer.firstName); // --> 'Jennifer'
|
||
console.log(jennifer.fullName()); // --> 'Jennifer Lawrence'
|
||
|
||
const numUpdated = await Person.query()
|
||
.patch({ lastName: 'Dinosaur' })
|
||
.where('age', '>', 60);
|
||
// update "persons" set "lastName" = 'Dinosaur' where "age" > 60
|
||
|
||
console.log(numUpdated, 'people were updated');
|
||
|
||
const middleAgedJennifers = await Person.query()
|
||
.select('age', 'firstName', 'lastName')
|
||
.where('age', '>', 40)
|
||
.where('age', '<', 60)
|
||
.where('firstName', 'Jennifer')
|
||
.orderBy('lastName');
|
||
|
||
console.log('The last name of the first middle aged Jennifer is');
|
||
console.log(middleAgedJennifers[0].lastName);
|
||
|
||
// select "age", "firstName", "lastName"
|
||
// from "persons"
|
||
// where "age" > 40
|
||
// and "age" < 60
|
||
// and "firstName" = 'Jennifer'
|
||
// order by "lastName" asc
|
||
```
|
||
### Bookshelf
|
||
一样基于 Knex,细节就不一一罗列了
|
||
|
||
|
||
### 总结
|
||
除了常规的增删改查,还有一个非常重要的能力就是关系数据的处理能力,关系数据的 eager loading 也是各 ORM 永恒的话题,这里就不细说了
|
||
|
||
|
||
- 从完整性来说,Typeorm 的最完整
|
||
- 如果说有什么理由让我选择 Objection.js,我会说它的 QueryBuilder 非常给力,写起来非常舒服,可造性也非常强
|
||
- Sequelize 的常规 CRUD API 并没有太大问题,但是在关系数据的处理上问题太多了。而且黑箱设计的 CRUD API,非常难以调试和改造
|
||
- Prisma 的 CRUD API 与 Sequelize 相似,这让我非常担心它是否也会有 Sequelize 的各种糟心问题
|
||
- 至于 Bookshelf,已经不维护了就不深入讨论了,Objection.js 是类似最好的替代品
|
||
|
||
|
||
|
||
## 如何选择?
|
||
|
||
- 综合实力 Typeorm 最强,各方面表现的都不差,但细节还差那么点,需要更深入使用才能知道细节表现力,存在非常多未知,但因为社区活跃,成长空间巨大
|
||
- 从无代码改造角度来说,Sequelize 工作量是最少的,尤其是给力的 sync 方法。但我非常不喜欢它的 QueryBuilder 设计。Sequelize 的社区活跃度也很高,但是核心团队看起来出现了些问题 [issue #12956](https://github.com/sequelize/sequelize/issues/12956)
|
||
- 我非常喜欢 Objection.js 在 QueryBuilder 上的表现力,可造性非常强,但是要完整的无代码支持,工作量也非常多
|
||
|
||
|
||
|
||
**在这样的大环境下,无代码的 Database 要如何选择 ORM 和改造呢?**
|
||
综合考虑,Sequelize 最适合作为蓝本,基于 Sequelize 优先实现第一版,即使 Sequelize 存在问题,后续也可以替换为 Typeorm 或 Objection.js 等。
|
||
|
||
|
||
## 改造开始
|
||
### Collection
|
||
NocoBase 首先基于 Sequelize 的 ModelOptions、 ModelAttributes、ModelAssociations 提炼了一套更适合无代码配置的 JSON 风格的配置协议,称之为 Collection Schema Language,简称 CSL。示例如下:
|
||
```typescript
|
||
// 用户
|
||
db.collection({
|
||
name: 'users',
|
||
fields: {
|
||
username: { type: 'string', unique: true },
|
||
password: { type: 'password', unique: true },
|
||
posts: { type: 'hasMany' },
|
||
},
|
||
});
|
||
|
||
// 文章
|
||
db.collection({
|
||
name: 'posts',
|
||
fields: {
|
||
title: 'string',
|
||
content: 'text',
|
||
tags: 'belongsToMany',
|
||
comments: 'hasMany',
|
||
author: { type: 'belongsTo', target: 'users' },
|
||
},
|
||
});
|
||
|
||
// 标签
|
||
db.collection({
|
||
name: 'tags',
|
||
fields: [
|
||
{ type: 'string', name: 'name' },
|
||
{ type: 'belongsToMany', name: 'posts' },
|
||
],
|
||
});
|
||
|
||
// 评论
|
||
db.collection({
|
||
name: 'comments',
|
||
fields: [
|
||
{ type: 'text', name: 'content' },
|
||
{ type: 'belongsTo', name: 'user' },
|
||
],
|
||
});
|
||
```
|
||
与现在流行的装饰器配置的结构很像,只不过它是用 JSON 编写的,便于动态生成 Model。与 Typeorm 的 EntitySchema 做法也非常接近。
|
||
|
||
|
||
**那为什么不直接使用 EntitySchema 呢?**
|
||
一方面对 EntitySchema 细节表现力如何未知,另一方面需要考虑后续的自定义字段需求,尤其是 UISchema 的扩展能力,自成一体的 CSL 更适合。而且结构上与 Typeorm 和 Prisma 的装饰器风格非常接近,学习成本并不高。
|
||
### Model & Repository
|
||
配置好了 CSL 之后,会根据它自动初始化 ORM 的 Model(Active Record)和 Repository(Data Mapper):
|
||
```typescript
|
||
const User = app.collection({
|
||
name: 'users',
|
||
fields: {
|
||
username: { type: 'string', unique: true },
|
||
password: { type: 'password', unique: true },
|
||
posts: { type: 'hasMany' },
|
||
},
|
||
});
|
||
|
||
User.model;
|
||
User.repository;
|
||
```
|
||
Model 的具体形态取决于适配的 ORM,保留原滋原味,细节就不多说了。Repository 自成一体,提供更适合 NocoBase 的 CRUD API。Model 和 Repository 都可以自定义,如:
|
||
```ts
|
||
class UserModel extends Model {
|
||
|
||
}
|
||
|
||
class UserRepository extends Repository {
|
||
|
||
}
|
||
|
||
const User = app.collection({
|
||
name: 'users',
|
||
model: UserModel,
|
||
repository: UserRepository,
|
||
fields: {
|
||
username: { type: 'string', unique: true },
|
||
password: { type: 'password', unique: true },
|
||
posts: { type: 'hasMany' },
|
||
},
|
||
});
|
||
```
|
||
支持 Model 也支持 Repository,与 Typeorm 的做法非常接近,但是 Repository 并不是 Typeorm 的 Repository,我们来看看 Repository API 吧。
|
||
### Repository API
|
||
NocoBase 的 Repository API 有:
|
||
|
||
- repository.findMany()
|
||
- repository.findOne()
|
||
- repository.create()
|
||
- repository.update()
|
||
- repository.destroy()
|
||
- repository.relation().of()
|
||
- findMany()
|
||
- findOne()
|
||
- create()
|
||
- update()
|
||
- destroy()
|
||
- set()
|
||
- add()
|
||
- remove()
|
||
- toggle()
|
||
#### findMany
|
||
```ts
|
||
repository.findMany({
|
||
// 过滤
|
||
filter: {
|
||
$and: [{ a: 5 }, { b: 6 }], // (a = 5) AND (b = 6)
|
||
$or: [{ a: 5 }, { b: 6 }], // (a = 5) OR (b = 6)
|
||
someAttribute: {
|
||
// Basics
|
||
$eq: 3, // = 3
|
||
$ne: 20, // != 20
|
||
$is: null, // IS NULL
|
||
$not: true, // IS NOT TRUE
|
||
$gt: 6, // > 6
|
||
$gte: 6,
|
||
},
|
||
'someAttribute.$eq': 3,
|
||
'nested.someAttribute': {
|
||
//
|
||
},
|
||
nested: {
|
||
someAttribute: {},
|
||
},
|
||
},
|
||
// 字段白名单
|
||
fields: [],
|
||
// 附加字段,主要用于附加关系字段
|
||
appends: [],
|
||
// 字段黑名单
|
||
expect: [],
|
||
sort: [],
|
||
page: 1,
|
||
pageSize: 2,
|
||
});
|
||
```
|
||
#### findOne
|
||
```typescript
|
||
repository.findOne({
|
||
// 过滤
|
||
filter: {},
|
||
// 为更快速的 pk 过滤提供
|
||
filterByPk: 1, // 通常情况等同于 filter: {id: 1}
|
||
// 字段白名单
|
||
fields: [],
|
||
// 附加字段,主要用于附加关系字段
|
||
appends: [],
|
||
// 字段黑名单
|
||
expect: [],
|
||
});
|
||
```
|
||
#### create
|
||
```ts
|
||
repository.create({
|
||
// 待存数据
|
||
values: {
|
||
a: 'a',
|
||
// 快速建立关联
|
||
o2o: 1, // 建立一对一关联
|
||
m2o: 1, // 建立多对一关联
|
||
o2m: [1,2] // 建立一对多关联
|
||
m2m: [1,2] // 建立多对多关联
|
||
// 新建关联数据并建立关联
|
||
o2o: {
|
||
key1: 'val1',
|
||
},
|
||
o2m: [{key1: 'val1'}, {key2: 'val2'}],
|
||
// 子表格数据
|
||
subTable: [
|
||
// 如果数据存在,更新处理
|
||
{id: 1, key1: 'val1111'},
|
||
// 如果数据不存在,直接创建并关联
|
||
{key2: 'val2'},
|
||
],
|
||
},
|
||
// 字段白名单
|
||
whitelist: [],
|
||
// 字段黑名单
|
||
blacklist: [],
|
||
// 关系数据默认会新建并建立关联处理,如果是已存在的数据只关联,但不更新关系数据
|
||
// 如果需要更新关联数据,可以通过 updateAssociations 指定
|
||
updateAssociations: ['subTable'],
|
||
});
|
||
```
|
||
#### update
|
||
```ts
|
||
repository.update({
|
||
// 待更新数据
|
||
values: {},
|
||
// 过滤,哪些数据要更新
|
||
filter: {},
|
||
// 为更快速的 pk 过滤提供
|
||
filterByPk: 1, // 通常情况等同于 filter: {id: 1}
|
||
// 字段白名单
|
||
whitelist: [],
|
||
// 字段黑名单
|
||
blacklist: [],
|
||
// 指定需要更新数据的关联字段
|
||
updateAssociations: [],
|
||
});
|
||
```
|
||
#### destroy
|
||
```ts
|
||
// 特定 primary key 值
|
||
repository.destroy(1);
|
||
|
||
// 批量 primary key 值
|
||
repository.destroy([1, 2, 3]);
|
||
|
||
// 复杂的 filter
|
||
repository.destroy({
|
||
filter: {},
|
||
});
|
||
```
|
||
#### relation
|
||
关系数据的 CRUD,用法与常规 Repository 的一致
|
||
```ts
|
||
// user_id = 1 的 post 的 repository
|
||
const userPostsRepository = repository.relation('posts').of(1);
|
||
|
||
userPostsRepository.findMany({
|
||
|
||
});
|
||
|
||
userPostsRepository.findOne({
|
||
|
||
});
|
||
|
||
userPostsRepository.create({
|
||
values: {},
|
||
});
|
||
|
||
userPostsRepository.update({
|
||
values: {},
|
||
});
|
||
|
||
userPostsRepository.destroy({
|
||
|
||
});
|
||
```
|
||
关联操作,只处理关系约束的建立与解除
|
||
```ts
|
||
// user_id = 1 的 post 的 relatedQuery
|
||
const userPostsRepository = repository.relation('posts').of(1);
|
||
|
||
// 建立关联
|
||
userPostsRepository.set(1);
|
||
|
||
// 批量,仅用于 HasMany 和 BelongsToMany
|
||
userPostsRepository.set([1,2,3]);
|
||
|
||
// BelongsToMany 的中间表
|
||
userPostsRepository.set([
|
||
[1, {/* 中间表数据 */}],
|
||
[2, {/* 中间表数据 */}],
|
||
[3, {/* 中间表数据 */}],
|
||
]);
|
||
|
||
// 仅用于 HasMany 和 BelongsToMany
|
||
userPostsRepository.add(1);
|
||
|
||
// BelongsToMany 的中间表
|
||
userPostsRepository.add(1, {/* 中间表数据 */});
|
||
|
||
// 删除关联
|
||
userPostsRepository.remove(1);
|
||
|
||
// 建立或解除
|
||
userPostsRepository.toggle(1);
|
||
userPostsRepository.toggle([1, 2, 3]);
|
||
```
|
||
### Collection Sync
|
||
在 Migration 章节,介绍了各种 ORM 的 Migration 做法。得出结论 Sequelize.sync() 的方案较优,但不够精细,而且有些暴力,不过也没有关系,Collection 也打算这样做,再进一步的改进 sync 的细节,流程上就变得非常友好了。
|
||
|
||
|
||
只执行某个 collection 的 sync。虽然有 hasMany 的 posts,但因为关系表不存在并不会创建
|
||
```ts
|
||
const User = db.collection({
|
||
name: 'users',
|
||
fields: {
|
||
username: { type: 'string', unique: true },
|
||
password: { type: 'password', unique: true },
|
||
posts: { type: 'hasMany' },
|
||
},
|
||
});
|
||
|
||
await User.sync();
|
||
```
|
||
我们也可以通过 db.sync 批量的将多个 collections 同步给数据库,通常不需要关注建表顺序、关系主外键和关系约束的先后顺序等等,collection 内部通通自动帮你处理好。比如下面例子:
|
||
|
||
- 不需要特意声明外键 user_id
|
||
- 不需要考虑关系外键要怎么建立,在哪里建立,也不需要考虑顺序问题
|
||
- 自动创建多对多中间表以及相关外键及约束
|
||
```ts
|
||
// 文章
|
||
db.collection({
|
||
name: 'posts',
|
||
fields: {
|
||
title: 'string',
|
||
content: 'text',
|
||
tags: 'belongsToMany',
|
||
comments: 'hasMany',
|
||
author: { type: 'belongsTo', target: 'users' },
|
||
},
|
||
});
|
||
|
||
// 用户
|
||
db.collection({
|
||
name: 'users',
|
||
fields: {
|
||
username: { type: 'string', unique: true },
|
||
password: { type: 'password', unique: true },
|
||
posts: { type: 'hasMany' },
|
||
},
|
||
});
|
||
|
||
// 标签
|
||
db.collection({
|
||
name: 'tags',
|
||
fields: [
|
||
{ type: 'string', name: 'name' },
|
||
{ type: 'belongsToMany', name: 'posts' },
|
||
],
|
||
});
|
||
|
||
// 评论
|
||
db.collection({
|
||
name: 'comments',
|
||
fields: [
|
||
{ type: 'text', name: 'content' },
|
||
{ type: 'belongsTo', name: 'user' },
|
||
],
|
||
});
|
||
|
||
await db.sync();
|
||
```
|
||
### 动态化
|
||
为了更好的支持动态配置 collection,提供了以下 API:
|
||
|
||
- db.collection 创建
|
||
- db.getCollection 获取
|
||
- collection.mergeOptions 和 collection 配置,不删除
|
||
- collection.hasField
|
||
- collection.getField
|
||
- collection.addField
|
||
- collection.setFields
|
||
- collection.mergeField 合并参数,如果 field 不存在则添加
|
||
- collection.removeField 移除
|
||
- collection.sync
|
||
|
||
备注:name 是非常重要的标识,但如果涉及 name 的修改如何处理比较合适?
|
||
|
||
|
||
有了 Collection API 就可以实现动态 Collection 了。解决动态数据的持久化问题,可以将数据储存在数据表里。为此,我们可以创建 collections 和 fields 两张表。
|
||
```ts
|
||
const Collection = db.collection({
|
||
name: 'collections',
|
||
fields: [
|
||
{ name: 'name', type: 'string', unique: true },
|
||
{ name: 'fields', type: 'hasMany', foreignKey: 'collectionName' },
|
||
],
|
||
});
|
||
|
||
const Field = db.collection({
|
||
name: 'fields',
|
||
fields: [
|
||
{ name: 'name', type: 'string' },
|
||
],
|
||
});
|
||
|
||
db.on('collections.afterCreate', async (model) => {
|
||
const collection = db.collection(model.toJSON()); // 实际可能不是用 model.toJSON() 方法
|
||
await collection.sync();
|
||
});
|
||
|
||
db.on('collections.afterUpdate', async (model) => {
|
||
const collection = db.getCollection(model.get('name'));
|
||
// 更新配置
|
||
collection.mergeOptions(model.get());
|
||
await collection.sync();
|
||
});
|
||
|
||
db.on('fields.afterCreate', async (model) => {
|
||
const collection = db.getCollection(model.get('collectionName'));
|
||
// 新增字段
|
||
collection.addField(model.toJSON()); // 实际可能不是用 model.toJSON() 方法
|
||
await collection.sync();
|
||
});
|
||
|
||
db.on('fields.afterUpdate', async (model) => {
|
||
const collection = db.getCollection(model.get('collectionName'));
|
||
// 更新字段配置
|
||
collection.mergeField(model.toJSON()); // 实际可能不是用 model.toJSON() 方法
|
||
await collection.sync();
|
||
});
|
||
```
|
||
备注:db.on 之后会另起一篇和 app.on 一起介绍
|
||
|
||
|
||
这样就可以用 Collection.repository.create() 来动态创建表了,比如:
|
||
```ts
|
||
await Collection.repository.create({
|
||
values: {
|
||
name: 'test',
|
||
fields: [
|
||
{ name: 'name', type: 'string' },
|
||
],
|
||
},
|
||
});
|
||
```
|
||
以上就是 plugin-collections 的核心实现逻辑了。
|