Skip to content

Latest commit

 

History

History
1335 lines (1022 loc) · 37.3 KB

select-query-builder.md

File metadata and controls

1335 lines (1022 loc) · 37.3 KB
Raw

Select using Query Builder

What is QueryBuilder

QueryBuilder is one of the most powerful features of TypeORM - it allows you to build SQL queries using elegant and convenient syntax, execute them and get automatically transformed entities.

Simple example of QueryBuilder:

const firstUser = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .where("user.id = :id", { id: 1 })
    .getOne()

It builds the following SQL query:

SELECT
    user.id as userId,
    user.firstName as userFirstName,
    user.lastName as userLastName
FROM users user
WHERE user.id = 1

and returns you an instance of User:

User {
    id: 1,
    firstName: "Timber",
    lastName: "Saw"
}

Important note when using the QueryBuilder

When using the QueryBuilder, you need to provide unique parameters in your WHERE expressions. This will not work:

const result = await dataSource
    .createQueryBuilder('user')
    .leftJoinAndSelect('user.linkedSheep', 'linkedSheep')
    .leftJoinAndSelect('user.linkedCow', 'linkedCow')
    .where('user.linkedSheep = :id', { id: sheepId })
    .andWhere('user.linkedCow = :id', { id: cowId });

... but this will:

const result = await dataSource
    .createQueryBuilder('user')
    .leftJoinAndSelect('user.linkedSheep', 'linkedSheep')
    .leftJoinAndSelect('user.linkedCow', 'linkedCow')
    .where('user.linkedSheep = :sheepId', { sheepId })
    .andWhere('user.linkedCow = :cowId', { cowId });

Note that we uniquely named :sheepId and :cowId instead of using :id twice for different parameters.

How to create and use a QueryBuilder

There are several ways how you can create a Query Builder:

  • Using DataSource:

    const user = await dataSource
    .createQueryBuilder()
    .select("user")
    .from(User, "user")
    .where("user.id = :id", { id: 1 })
    .getOne()

  • Using entity manager:

    const user = await dataSource.manager
    .createQueryBuilder(User, "user")
    .where("user.id = :id", { id: 1 })
    .getOne()

  • Using repository:

    const user = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .where("user.id = :id", { id: 1 })
    .getOne()

There are 5 different QueryBuilder types available:

  • SelectQueryBuilder - used to build and execute SELECT queries. Example:

    const user = await dataSource
    .createQueryBuilder()
    .select("user")
    .from(User, "user")
    .where("user.id = :id", { id: 1 })
    .getOne()

  • InsertQueryBuilder - used to build and execute INSERT queries. Example:

    await dataSource
    .createQueryBuilder()
    .insert()
    .into(User)
    .values([
        { firstName: "Timber", lastName: "Saw" },
        { firstName: "Phantom", lastName: "Lancer" },
    ])
    .execute()

  • UpdateQueryBuilder - used to build and execute UPDATE queries. Example:

    await dataSource
    .createQueryBuilder()
    .update(User)
    .set({ firstName: "Timber", lastName: "Saw" })
    .where("id = :id", { id: 1 })
    .execute()

  • DeleteQueryBuilder - used to build and execute DELETE queries. Example:

    await dataSource
    .createQueryBuilder()
    .delete()
    .from(User)
    .where("id = :id", { id: 1 })
    .execute()

  • RelationQueryBuilder - used to build and execute relation-specific operations [TBD]. Example:

    await dataSource
    .createQueryBuilder()
    .relation(User,"photos")
    .of(id)
    .loadMany();

You can switch between different types of query builder within any of them, once you do, you will get a new instance of query builder (unlike all other methods).

Getting values using QueryBuilder

To get a single result from the database, for example to get a user by id or name, you must use getOne:

const timber = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .where("user.id = :id OR user.name = :name", { id: 1, name: "Timber" })
    .getOne()

getOneOrFail will get a single result from the database, but if no result exists it will throw an EntityNotFoundError:

const timber = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .where("user.id = :id OR user.name = :name", { id: 1, name: "Timber" })
    .getOneOrFail()

To get multiple results from the database, for example, to get all users from the database, use getMany:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .getMany()

There are two types of results you can get using select query builder: entities or raw results. Most of the time, you need to select real entities from your database, for example, users. For this purpose, you use getOne and getMany. But sometimes you need to select some specific data, let's say the sum of all user photos. This data is not an entity, it's called raw data. To get raw data, you use getRawOne and getRawMany. Examples:

const { sum } = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .select("SUM(user.photosCount)", "sum")
    .where("user.id = :id", { id: 1 })
    .getRawOne()
const photosSums = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .select("user.id")
    .addSelect("SUM(user.photosCount)", "sum")
    .groupBy("user.id")
    .getRawMany()

// result will be like this: [{ id: 1, sum: 25 }, { id: 2, sum: 13 }, ...]

Getting a count

You can get the count on the number of rows a query will return by using getCount(). This will return the count as a number rather than an Entity result.

const count = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .where("user.name = :name", { name: "Timber" })
    .getCount()

Which produces the following SQL query:

SELECT count(*) FROM users user WHERE user.name = 'Timber'

What are aliases for?

We used createQueryBuilder("user"). But what is "user"? It's just a regular SQL alias. We use aliases everywhere, except when we work with selected data.

createQueryBuilder("user") is equivalent to:

createQueryBuilder().select("user").from(User, "user")

Which will result in the following SQL query:

SELECT ... FROM users user

In this SQL query, users is the table name, and user is an alias we assign to this table. Later we use this alias to access the table:

createQueryBuilder()
    .select("user")
    .from(User, "user")
    .where("user.name = :name", { name: "Timber" })

Which produces the following SQL query:

SELECT ... FROM users user WHERE user.name = 'Timber'

See, we used the users table by using the user alias we assigned when we created a query builder.

One query builder is not limited to one alias, they can have multiple aliases. Each select can have its own alias, you can select from multiple tables each with its own alias, you can join multiple tables each with its own alias. You can use those aliases to access tables you are selecting (or data you are selecting).

Using parameters to escape data

We used where("user.name = :name", { name: "Timber" }). What does { name: "Timber" } stand for? It's a parameter we used to prevent SQL injection. We could have written: where("user.name = '" + name + "'), however this is not safe, as it opens the code to SQL injections. The safe way is to use this special syntax: where("user.name = :name", { name: "Timber" }), where :name is a parameter name and the value is specified in an object: { name: "Timber" }.

.where("user.name = :name", { name: "Timber" })

is a shortcut for:

.where("user.name = :name")
.setParameter("name", "Timber")

Note: do not use the same parameter name for different values across the query builder. Values will be overridden if you set them multiple times.

You can also supply an array of values, and have them transformed into a list of values in the SQL statement, by using the special expansion syntax:

.where("user.name IN (:...names)", { names: [ "Timber", "Cristal", "Lina" ] })

Which becomes:

WHERE user.name IN ('Timber', 'Cristal', 'Lina')

Adding WHERE expression

Adding a WHERE expression is as easy as:

createQueryBuilder("user").where("user.name = :name", { name: "Timber" })

Which will produce:

SELECT ... FROM users user WHERE user.name = 'Timber'

You can add AND into an existing WHERE expression:

createQueryBuilder("user")
    .where("user.firstName = :firstName", { firstName: "Timber" })
    .andWhere("user.lastName = :lastName", { lastName: "Saw" })

Which will produce the following SQL query:

SELECT ... FROM users user WHERE user.firstName = 'Timber' AND user.lastName = 'Saw'

You can add OR into an existing WHERE expression:

createQueryBuilder("user")
    .where("user.firstName = :firstName", { firstName: "Timber" })
    .orWhere("user.lastName = :lastName", { lastName: "Saw" })

Which will produce the following SQL query:

SELECT ... FROM users user WHERE user.firstName = 'Timber' OR user.lastName = 'Saw'

You can do an IN query with the WHERE expression:

createQueryBuilder("user").where("user.id IN (:...ids)", { ids: [1, 2, 3, 4] })

Which will produce the following SQL query:

SELECT ... FROM users user WHERE user.id IN (1, 2, 3, 4)

You can add a complex WHERE expression into an existing WHERE using Brackets

createQueryBuilder("user")
    .where("user.registered = :registered", { registered: true })
    .andWhere(
        new Brackets((qb) => {
            qb.where("user.firstName = :firstName", {
                firstName: "Timber",
            }).orWhere("user.lastName = :lastName", { lastName: "Saw" })
        }),
    )

Which will produce the following SQL query:

SELECT ... FROM users user WHERE user.registered = true AND (user.firstName = 'Timber' OR user.lastName = 'Saw')

You can add a negated complex WHERE expression into an existing WHERE using NotBrackets

createQueryBuilder("user")
    .where("user.registered = :registered", { registered: true })
    .andWhere(
        new NotBrackets((qb) => {
            qb.where("user.firstName = :firstName", {
                firstName: "Timber",
            }).orWhere("user.lastName = :lastName", { lastName: "Saw" })
        }),
    )

Which will produce the following SQL query:

SELECT ... FROM users user WHERE user.registered = true AND NOT((user.firstName = 'Timber' OR user.lastName = 'Saw'))

You can combine as many AND and OR expressions as you need. If you use .where more than once you'll override all previous WHERE expressions.

Note: be careful with orWhere - if you use complex expressions with both AND and OR expressions, keep in mind that they are stacked without any pretences. Sometimes you'll need to create a where string instead, and avoid using orWhere.

Adding HAVING expression

Adding a HAVING expression is easy as:

createQueryBuilder("user").having("user.name = :name", { name: "Timber" })

Which will produce following SQL query:

SELECT ... FROM users user HAVING user.name = 'Timber'

You can add AND into an exist HAVING expression:

createQueryBuilder("user")
    .having("user.firstName = :firstName", { firstName: "Timber" })
    .andHaving("user.lastName = :lastName", { lastName: "Saw" })

Which will produce the following SQL query:

SELECT ... FROM users user HAVING user.firstName = 'Timber' AND user.lastName = 'Saw'

You can add OR into a exist HAVING expression:

createQueryBuilder("user")
    .having("user.firstName = :firstName", { firstName: "Timber" })
    .orHaving("user.lastName = :lastName", { lastName: "Saw" })

Which will produce the following SQL query:

SELECT ... FROM users user HAVING user.firstName = 'Timber' OR user.lastName = 'Saw'

You can combine as many AND and OR expressions as you need. If you use .having more than once you'll override all previous HAVING expressions.

Adding ORDER BY expression

Adding an ORDER BY expression is easy as:

createQueryBuilder("user").orderBy("user.id")

Which will produce:

SELECT ... FROM users user ORDER BY user.id

You can change the ordering direction from ascending to descending (or versa):

createQueryBuilder("user").orderBy("user.id", "DESC")

createQueryBuilder("user").orderBy("user.id", "ASC")

You can add multiple order-by criteria:

createQueryBuilder("user").orderBy("user.name").addOrderBy("user.id")

You can also use a map of order-by fields:

createQueryBuilder("user").orderBy({
    "user.name": "ASC",
    "user.id": "DESC",
})

If you use .orderBy more than once you'll override all previous ORDER BY expressions.

Adding DISTINCT ON expression (Postgres only)

When using both distinct-on with an order-by expression, the distinct-on expression must match the leftmost order-by. The distinct-on expressions are interpreted using the same rules as order-by. Please note that, using distinct-on without an order-by expression means that the first row of each set is unpredictable.

Adding a DISTINCT ON expression is easy as:

createQueryBuilder("user").distinctOn(["user.id"]).orderBy("user.id")

Which will produce:

SELECT DISTINCT ON (user.id) ... FROM users user ORDER BY user.id

Adding GROUP BY expression

Adding a GROUP BY expression is easy as:

createQueryBuilder("user").groupBy("user.id")

Which will produce the following SQL query:

SELECT ... FROM users user GROUP BY user.id

To add more group-by criteria use addGroupBy:

createQueryBuilder("user").groupBy("user.name").addGroupBy("user.id")

If you use .groupBy more than once you'll override all previous GROUP BY expressions.

Adding LIMIT expression

Adding a LIMIT expression is easy as:

createQueryBuilder("user").limit(10)

Which will produce the following SQL query:

SELECT ... FROM users user LIMIT 10

The resulting SQL query depends on the type of database (SQL, mySQL, Postgres, etc). Note: LIMIT may not work as you may expect if you are using complex queries with joins or subqueries. If you are using pagination, it's recommended to use take instead.

Adding OFFSET expression

Adding an SQL OFFSET expression is easy as:

createQueryBuilder("user").offset(10)

Which will produce the following SQL query:

SELECT ... FROM users user OFFSET 10

The resulting SQL query depends on the type of database (SQL, mySQL, Postgres, etc). Note: OFFSET may not work as you may expect if you are using complex queries with joins or subqueries. If you are using pagination, it's recommended to use skip instead.

Joining relations

Let's say you have the following entities:

import { Entity, PrimaryGeneratedColumn, Column, OneToMany } from "typeorm"
import { Photo } from "./Photo"

@Entity()
export class User {
    @PrimaryGeneratedColumn()
    id: number

    @Column()
    name: string

    @OneToMany((type) => Photo, (photo) => photo.user)
    photos: Photo[]
}
import { Entity, PrimaryGeneratedColumn, Column, ManyToOne } from "typeorm"
import { User } from "./User"

@Entity()
export class Photo {
    @PrimaryGeneratedColumn()
    id: number

    @Column()
    url: string

    @ManyToOne((type) => User, (user) => user.photos)
    user: User
}

Now let's say you want to load user "Timber" with all of his photos:

const user = await createQueryBuilder("user")
    .leftJoinAndSelect("user.photos", "photo")
    .where("user.name = :name", { name: "Timber" })
    .getOne()

You'll get the following result:

{
    id: 1,
    name: "Timber",
    photos: [{
        id: 1,
        url: "me-with-chakram.jpg"
    }, {
        id: 2,
        url: "me-with-trees.jpg"
    }]
}

As you can see leftJoinAndSelect automatically loaded all of Timber's photos. The first argument is the relation you want to load and the second argument is an alias you assign to this relation's table. You can use this alias anywhere in query builder. For example, let's take all Timber's photos which aren't removed.

const user = await createQueryBuilder("user")
    .leftJoinAndSelect("user.photos", "photo")
    .where("user.name = :name", { name: "Timber" })
    .andWhere("photo.isRemoved = :isRemoved", { isRemoved: false })
    .getOne()

This will generate following SQL query:

SELECT user.*, photo.* FROM users user
    LEFT JOIN photos photo ON photo.user = user.id
    WHERE user.name = 'Timber' AND photo.isRemoved = FALSE

You can also add conditions to the join expression instead of using "where":

const user = await createQueryBuilder("user")
    .leftJoinAndSelect("user.photos", "photo", "photo.isRemoved = :isRemoved", {
        isRemoved: false,
    })
    .where("user.name = :name", { name: "Timber" })
    .getOne()

This will generate the following SQL query:

SELECT user.*, photo.* FROM users user
    LEFT JOIN photos photo ON photo.user = user.id AND photo.isRemoved = FALSE
    WHERE user.name = 'Timber'

Inner and left joins

If you want to use INNER JOIN instead of LEFT JOIN just use innerJoinAndSelect instead:

const user = await createQueryBuilder("user")
    .innerJoinAndSelect(
        "user.photos",
        "photo",
        "photo.isRemoved = :isRemoved",
        { isRemoved: false },
    )
    .where("user.name = :name", { name: "Timber" })
    .getOne()

This will generate:

SELECT user.*, photo.* FROM users user
    INNER JOIN photos photo ON photo.user = user.id AND photo.isRemoved = FALSE
    WHERE user.name = 'Timber'

The difference between LEFT JOIN and INNER JOIN is that INNER JOIN won't return a user if it does not have any photos. LEFT JOIN will return you the user even if it doesn't have photos. To learn more about different join types, refer to the SQL documentation.

Join without selection

You can join data without its selection. To do that, use leftJoin or innerJoin:

const user = await createQueryBuilder("user")
    .innerJoin("user.photos", "photo")
    .where("user.name = :name", { name: "Timber" })
    .getOne()

This will generate:

SELECT user.* FROM users user
    INNER JOIN photos photo ON photo.user = user.id
    WHERE user.name = 'Timber'

This will select Timber if he has photos, but won't return his photos.

Joining any entity or table

You can join not only relations, but also other unrelated entities or tables. Examples:

const user = await createQueryBuilder("user")
    .leftJoinAndSelect(Photo, "photo", "photo.userId = user.id")
    .getMany()
const user = await createQueryBuilder("user")
    .leftJoinAndSelect("photos", "photo", "photo.userId = user.id")
    .getMany()

Joining and mapping functionality

Add profilePhoto to User entity, and you can map any data into that property using QueryBuilder:

export class User {
    /// ...
    profilePhoto: Photo
}
const user = await createQueryBuilder("user")
    .leftJoinAndMapOne(
        "user.profilePhoto",
        "user.photos",
        "photo",
        "photo.isForProfile = TRUE",
    )
    .where("user.name = :name", { name: "Timber" })
    .getOne()

This will load Timber's profile photo and set it to user.profilePhoto. If you want to load and map a single entity use leftJoinAndMapOne. If you want to load and map multiple entities use leftJoinAndMapMany.

Getting the generated query

Sometimes you may want to get the SQL query generated by QueryBuilder. To do so, use getSql:

const sql = createQueryBuilder("user")
    .where("user.firstName = :firstName", { firstName: "Timber" })
    .orWhere("user.lastName = :lastName", { lastName: "Saw" })
    .getSql()

For debugging purposes you can use printSql:

const users = await createQueryBuilder("user")
    .where("user.firstName = :firstName", { firstName: "Timber" })
    .orWhere("user.lastName = :lastName", { lastName: "Saw" })
    .printSql()
    .getMany()

This query will return users and print the used sql statement to the console.

Getting raw results

There are two types of results you can get using select query builder: entities and raw results. Most of the time, you need to select real entities from your database, for example, users. For this purpose, you use getOne and getMany. However, sometimes you need to select specific data, like the sum of all user photos. Such data is not an entity, it's called raw data. To get raw data, you use getRawOne and getRawMany. Examples:

const { sum } = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .select("SUM(user.photosCount)", "sum")
    .where("user.id = :id", { id: 1 })
    .getRawOne()
const photosSums = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .select("user.id")
    .addSelect("SUM(user.photosCount)", "sum")
    .groupBy("user.id")
    .getRawMany()

// result will be like this: [{ id: 1, sum: 25 }, { id: 2, sum: 13 }, ...]

Streaming result data

You can use stream which returns you a stream. Streaming returns you raw data, and you must handle entity transformation manually:

const stream = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .where("user.id = :id", { id: 1 })
    .stream()

Using pagination

Most of the time when you develop an application, you need pagination functionality. This is used if you have pagination, page slider, or infinite scroll components in your application.

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .leftJoinAndSelect("user.photos", "photo")
    .take(10)
    .getMany()

This will give you the first 10 users with their photos.

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .leftJoinAndSelect("user.photos", "photo")
    .skip(10)
    .getMany()

This will give you all except the first 10 users with their photos. You can combine those methods:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .leftJoinAndSelect("user.photos", "photo")
    .skip(5)
    .take(10)
    .getMany()

This will skip the first 5 users and take 10 users after them.

take and skip may look like we are using limit and offset, but they aren't. limit and offset may not work as you expect once you have more complicated queries with joins or subqueries. Using take and skip will prevent those issues.

Set locking

QueryBuilder supports both optimistic and pessimistic locking.

Lock modes

Support of lock modes, and SQL statements they translate to, are listed in the table below (blank cell denotes unsupported). When specified lock mode is not supported, a LockNotSupportedOnGivenDriverError error will be thrown.

|                 | pessimistic_read                  | pessimistic_write       | dirty_read    | pessimistic_partial_write (Deprecated, use onLocked instead)   | pessimistic_write_or_fail (Deprecated, use onLocked instead)   | for_no_key_update   | for_key_share |
| --------------- | --------------------------------- | ----------------------- | ------------- | -------------------------------------------------------------- | -------------------------------------------------------------- | ------------------- | ------------- |
| MySQL           | FOR SHARE (8+)/LOCK IN SHARE MODE | FOR UPDATE              | (nothing)     | FOR UPDATE SKIP LOCKED                                         | FOR UPDATE NOWAIT                                              |                     |               |
| Postgres        | FOR SHARE                         | FOR UPDATE              | (nothing)     | FOR UPDATE SKIP LOCKED                                         | FOR UPDATE NOWAIT                                              | FOR NO KEY UPDATE   | FOR KEY SHARE |
| Oracle          | FOR UPDATE                        | FOR UPDATE              | (nothing)     |                                                                |                                                                |                     |               |
| SQL Server      | WITH (HOLDLOCK, ROWLOCK)          | WITH (UPDLOCK, ROWLOCK) | WITH (NOLOCK) |                                                                |                                                                |                     |               |
| AuroraDataApi   | LOCK IN SHARE MODE                | FOR UPDATE              | (nothing)     |                                                                |                                                                |                     |               |
| CockroachDB     |                                   | FOR UPDATE              | (nothing)     |                                                                | FOR UPDATE NOWAIT                                              | FOR NO KEY UPDATE   |               |

To use pessimistic read locking use the following method:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .setLock("pessimistic_read")
    .getMany()

To use pessimistic write locking use the following method:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .setLock("pessimistic_write")
    .getMany()

To use dirty read locking use the following method:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .setLock("dirty_read")
    .getMany()

To use optimistic locking use the following method:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .setLock("optimistic", existUser.version)
    .getMany()

Optimistic locking works in conjunction with both @Version and @UpdatedDate decorators.

setOnLock

Allows you to control what happens when a row is locked. By default, the database will wait for the lock. You can control that behavior by using setOnLocked

To not wait:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .setLock("pessimistic_write")
    .setOnLocked("nowait")
    .getMany()

To skip the row:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .setLock("pessimistic_write")
    .setOnLocked("skip_locked")
    .getMany()

Database support for setOnLocked based on lock mode:

  • Postgres: pessimistic_read, pessimistic_write, for_no_key_update, for_key_share
  • MySQL 8+: pessimistic_read, pessimistic_write
  • MySQL < 8, Maria DB: pessimistic_write
  • Cockroach: pessimistic_write (nowait only)

Use custom index

You can provide a certain index for database server to use in some cases. This feature is only supported in MySQL.

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .useIndex("my_index") // name of index
    .getMany()

Max execution time

We can drop slow query to avoid crashing the server.

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .maxExecutionTime(1000) // milliseconds.
    .getMany()

Partial selection

If you want to select only some entity properties, you can use the following syntax:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .select(["user.id", "user.name"])
    .getMany()

This will only select the id and name of User.

Using subqueries

You can easily create subqueries. Subqueries are supported in FROM, WHERE and JOIN expressions. Example:

const qb = await dataSource.getRepository(Post).createQueryBuilder("post")

const posts = qb
    .where(
        "post.title IN " +
            qb
                .subQuery()
                .select("user.name")
                .from(User, "user")
                .where("user.registered = :registered")
                .getQuery(),
    )
    .setParameter("registered", true)
    .getMany()

A more elegant way to do the same:

const posts = await dataSource
    .getRepository(Post)
    .createQueryBuilder("post")
    .where((qb) => {
        const subQuery = qb
            .subQuery()
            .select("user.name")
            .from(User, "user")
            .where("user.registered = :registered")
            .getQuery()
        return "post.title IN " + subQuery
    })
    .setParameter("registered", true)
    .getMany()

Alternatively, you can create a separate query builder and use its generated SQL:

const userQb = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .select("user.name")
    .where("user.registered = :registered", { registered: true })

const posts = await dataSource
    .getRepository(Post)
    .createQueryBuilder("post")
    .where("post.title IN (" + userQb.getQuery() + ")")
    .setParameters(userQb.getParameters())
    .getMany()

You can create subqueries in FROM like this:

const userQb = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .select("user.name", "name")
    .where("user.registered = :registered", { registered: true })

const posts = await dataSource
    .createQueryBuilder()
    .select("user.name", "name")
    .from("(" + userQb.getQuery() + ")", "user")
    .setParameters(userQb.getParameters())
    .getRawMany()

or using a more elegant syntax:

const posts = await dataSource
    .createQueryBuilder()
    .select("user.name", "name")
    .from((subQuery) => {
        return subQuery
            .select("user.name", "name")
            .from(User, "user")
            .where("user.registered = :registered", { registered: true })
    }, "user")
    .getRawMany()

If you want to add a subselect as a "second from" use addFrom.

You can use subselects in SELECT statements as well:

const posts = await dataSource
    .createQueryBuilder()
    .select("post.id", "id")
    .addSelect((subQuery) => {
        return subQuery.select("user.name", "name").from(User, "user").limit(1)
    }, "name")
    .from(Post, "post")
    .getRawMany()

Hidden Columns

If the model you are querying has a column with a select: false column, you must use the addSelect function in order to retrieve the information from the column.

Let's say you have the following entity:

import { Entity, PrimaryGeneratedColumn, Column } from "typeorm"

@Entity()
export class User {
    @PrimaryGeneratedColumn()
    id: number

    @Column()
    name: string

    @Column({ select: false })
    password: string
}

Using a standard find or query, you will not receive the password property for the model. However, if you do the following:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder()
    .select("user.id", "id")
    .addSelect("user.password")
    .getMany()

You will get the property password in your query.

Querying Deleted rows

If the model you are querying has a column with the attribute @DeleteDateColumn set, the query builder will automatically query rows which are 'soft deleted'.

Let's say you have the following entity:

import { Entity, PrimaryGeneratedColumn, Column, DeleteDateColumn } from "typeorm"

@Entity()
export class User {
    @PrimaryGeneratedColumn()
    id: number

    @Column()
    name: string

    @DeleteDateColumn()
    deletedAt?: Date
}

Using a standard find or query, you will not receive the rows which have a value in that row. However, if you do the following:

const users = await dataSource
    .getRepository(User)
    .createQueryBuilder()
    .select("user.id", "id")
    .withDeleted()
    .getMany()

You will get all the rows, including the ones which are deleted.

Common table expressions

QueryBuilder instances support common table expressions , if minimal supported version of your database supports them. Common table expressions aren't supported for Oracle yet.

const users = await connection.getRepository(User)
    .createQueryBuilder('user')
    .select("user.id", 'id')
    .addCommonTableExpression(`
      SELECT "userId" FROM "post"
    `, 'post_users_ids')
    .where(`user.id IN (SELECT "userId" FROM 'post_users_ids')`)
    .getMany();

Result values of InsertQueryBuilder or UpdateQueryBuilder can be used in Postgres:

const insertQueryBuilder = connection.getRepository(User)
    .createQueryBuilder()
    .insert({
        name: 'John Smith'
    })
    .returning(['id']);

const users = await connection.getRepository(User)
    .createQueryBuilder('user')
    .addCommonTableExpression(insertQueryBuilder, 'insert_results')
    .where(`user.id IN (SELECT "id" FROM 'insert_results')`)
    .getMany();

Time Travel Queries

Time Travel Queries currently supported only in CockroachDB database.

const repository = connection.getRepository(Account)

// create a new account
const account = new Account()
account.name = "John Smith"
account.balance = 100
await repository.save(account)

// imagine we update the account balance 1 hour after creation
account.balance = 200
await repository.save(account)

// outputs { name: "John Smith", balance: "200" }
console.log(account)

// load account state on 1 hour back
account = await repository
    .createQueryBuilder("account")
    .timeTravelQuery(`'-1h'`)
    .getOneOrFail()

// outputs { name: "John Smith", balance: "100" }
console.log(account)

By default timeTravelQuery() uses follower_read_timestamp() function if no arguments passed. For another supported timestamp arguments and additional information please refer to CockroachDB docs.

Debugging

You can get the generated SQL from the query builder by calling getQuery() or getQueryAndParameters().

If you just want the query you can use getQuery()

const sql = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .where("user.id = :id", { id: 1 })
    .getQuery()

Which results in:

SELECT `user`.`id` as `userId`, `user`.`firstName` as `userFirstName`, `user`.`lastName` as `userLastName` FROM `users` `user` WHERE `user`.`id` = ?

Or if you want the query and the parameters you can get an array back using getQueryAndParameters()

const queryAndParams = await dataSource
    .getRepository(User)
    .createQueryBuilder("user")
    .where("user.id = :id", { id: 1 })
    .getQueryAndParameters()

Which results in:

[
 "SELECT `user`.`id` as `userId`, `user`.`firstName` as `userFirstName`, `user`.`lastName` as `userLastName` FROM `users` `user` WHERE `user`.`id` = ?",
 [ 1 ]
]