Raw SQL (literals)
We believe that ORMs are inherently leaky abstractions. They are a compromise between the flexibility of SQL and the convenience of an object-oriented programming language. As such, it does not make sense to try to provide a 100% complete abstraction over SQL (which could easily be more difficult to read than the SQL itself).
For this reason, Sequelize treats raw SQL as a first-class citizen. You can use raw SQL almost anywhere in Sequelize1, and thanks to the sql
tag, it's
easy to write SQL that is both safe, readable and reusable.
Writing raw SQL
The sql
tag is a template literal tag that allows you to write raw SQL:
import { sql } from '@sequelize/core';
const id = 5;
await sequelize.query(sql`SELECT * FROM users WHERE id = ${id}`);
As indicated above, raw SQL can be used almost anywhere in Sequelize. For instance, here is one way to use raw SQL to customize the WHERE
clause of a findAll
query:
import { sql } from '@sequelize/core';
const id = 5;
const users = await User.findAll({
where: sql`id = ${id}`,
});
In the above example, we used a variable in our raw SQL. Thanks to the sql
tag, Sequelize will automatically escape that variable to remove any risk of SQL injection.
Sequelize supports two different ways to pass variables in raw SQL: Replacements and Bind Parameters.
Replacements and bind parameters are available in all querying methods, and can be used together in the same query.
Replacements
Replacements are a way to pass variables in your Query. They are an alternative to Bind Parameters.
The difference between replacements and bind parameters is that replacements are escaped and inserted into the query by Sequelize before the query is sent to the database, whereas bind parameters are sent to the database separately from the SQL query text, and 'escaped' by the Database itself.
Replacements can be written in three different ways:
- By using the
sql
tag when the query is in replacement mode - By using numeric identifiers (represented by a
?
) in the query. Thereplacements
option must be an array. The values will be replaced in the order in which they appear in the array and query. - Or by using alphanumeric identifiers (e.g.
:firstName
,:status
, etc…). These identifiers follow common identifier rules (alphanumeric & underscore only, cannot start with a number). Thereplacements
option must be a plain object which includes each parameter (without the:
prefix).
The replacements
option must contain all bound values, or Sequelize will throw an error.
Examples
import { QueryTypes } from '@sequelize/core';
// This query use positional replacements
await sequelize.query('SELECT * FROM projects WHERE status = ?', {
replacements: ['active'],
});
// This query uses named replacements
await sequelize.query('SELECT * FROM projects WHERE status = :status', {
replacements: { status: 'active' },
});
// This query use replacements added by the sql tag
await sequelize.query(sql`SELECT * FROM projects WHERE status = ${'active'}`);
// Replacements are also available in other querying methods
await Project.findAll({
where: {
status: sql`:status`,
},
replacements: { status: 'active' },
});
Bind Parameters
Bind parameters are a way to pass variables in your Query. They are an alternative to Replacements.
The difference between replacements and bind parameters is that replacements are escaped and inserted into the query by Sequelize before the query is sent to the database, whereas bind parameters are sent to the database separately from the SQL query text, and 'escaped' by the Database itself.
A query can have both bind parameters and replacements.
Each database uses a different syntax for bind parameters, but Sequelize provides its own unification layer.
Inconsequentially to which database you use, in Sequelize bind parameters are written following a postgres-like syntax. You can either:
- By using the
sql
tag when the query is in bind parameter mode - Use numeric identifiers (e.g.
$1
,$2
, etc…). Note that these identifiers start at 1, not 0. Thebind
option must be an array which contains a value for each identifier used in the query ($1
is bound to the 1st element in the array (bind[0]
), etc…). - Use alphanumeric identifiers (e.g.
$firstName
,$status
, etc…). These identifiers follow common identifier rules (alphanumeric & underscore only, cannot start with a number). Thebind
option must be a plain object which includes each bind parameter (without the$
prefix).
The bind
option must contain all bound values, or Sequelize will throw an error.
Bind Parameters can only be used for data values. Bind Parameters cannot be used to dynamically change the name of a table, a column, or other non-data values parts of the query,
but you can use sql.attribute
, and sql.identifier
for that.
Your database may have further restrictions with bind parameters.
Examples
import { QueryTypes } from '@sequelize/core';
// This query use positional bind parameters
await sequelize.query('SELECT * FROM projects WHERE status = $1', {
bind: ['active'],
type: QueryTypes.SELECT,
});
// This query uses named bind parameters
await sequelize.query('SELECT * FROM projects WHERE status = $status', {
bind: { status: 'active' },
type: QueryTypes.SELECT,
});
// Bind parameters are also available in other querying methods
await Project.findAll({
where: {
status: sql`$status`,
},
bind: { status: 'active' },
});
Sequelize does not currently support a way to specify the DataType of a bind parameter.
Until such a feature is implemented, you can cast your bind parameters if you need to change their DataType:
import { QueryTypes } from '@sequelize/core';
await sequelize.query('SELECT * FROM projects WHERE id = CAST($1 AS int)', {
bind: [5],
type: QueryTypes.SELECT,
});
Some dialects, such as PostgreSQL and IBM Db2, support a terser cast syntax that you can use if you prefer:
await sequelize.query('SELECT * FROM projects WHERE id = $1::int');
⚠️ Don't put parameters in strings
Never put parameters in strings, including postgres dollar-quoted strings, as this can very easily lead to SQL injection attacks.
You may be tempted to use parameters inside something like DO
blocks,
and it is a common misconception that you can safely use replacements or bind parameters inside dollar-quoted strings, but that is not the case.
For this reason, if you use the ?
, $bind
or :replacements
syntax, Sequelize will not consider these tokens as parameters if they are inside a string or an identifier.
However, when using the sql
tag, Sequelize gives you full control, and you are responsible for ensuring that your query is safe.
Here is an example of a query that is vulnerable to SQL injection:
const id = '$$';
sequelize.query(
sql`
DO $$
DECLARE r record;
BEGIN
SELECT * FROM users WHERE id = ${id};
END
$$;
`,
);
The above query looks like code, has syntax coloring that makes it look like code, but is really a regular dollar-quoted string
that will be interpreted as SQL by the DO
clause (similarly to eval
in JavaScript).
Dollar-quoted strings end as soon as $$
is encountered. If the user passes $$
as the id
parameter, the query will end early and will
at best be invalid SQL, and at worst will allow the user to execute arbitrary SQL.
DO $$
DECLARE r record;
BEGIN
SELECT * FROM users WHERE id = '$$';
-- ^ the contents of the DO clause ends here
END
$$;
Query Variable Mode
While bind parameters written using the $
syntax, and replacements written using the :
and ?
syntaxes, will always be interpreted as
bind parameters and replacements respectively, variables inserted in an sql
-tagged template literal can be interpreted as bind parameters or replacements depending on the Query Variable Mode.
It is not currently possible to configure that mode per query (this feature is planned for a future release). Instead, the mode is pre-determined by the method used to execute the query:
Model.insert
,Model.destroy
andModel.update
are in "bind parameter" mode.- All other methods are in "replacement" mode.
This means that, for instance, variables used in findAll
will be added to the query as replacements:
const fundingStatus = 'funded';
await Project.findAll({
where: and({ status: 'active' }, sql`funding = ${fundingStatus}`),
});
SELECT * FROM projects WHERE status = 'active' AND funding = 'funded'
Whereas variables used in update
will be added to the query as bind parameters:
const fundingStatus = 'funded';
await Project.update(
{ funding: 'pending' },
{
where: and({ status: 'active' }, sql`funding = ${fundingStatus}`),
},
);
UPDATE projects SET funding = $1 WHERE status = $2 AND funding = $3
sql.identifier
The sql.identifier
function can be used to escape the name of an identifier (such as a table or column name) in a query.
import { sql } from '@sequelize/core';
await sequelize.query(sql`SELECT * FROM ${sql.identifier('projects')}`);
-- The identifier quotes are dialect-specific, this is an example for PostgreSQL
SELECT * FROM "projects"
sql.list
When using an array as a variable in a query, Sequelize will by default treat it as an SQL array:
const statuses = ['active', 'pending'];
await sequelize.query(sql`SELECT * FROM projects WHERE status = ANY(${statuses})`);
SELECT * FROM projects WHERE status = ANY(ARRAY['active', 'pending'])
The sql.list
function can be used to tell Sequelize to treat the value as an SQL list instead:
const statuses = ['active', 'pending'];
await sequelize.query(sql`SELECT * FROM projects WHERE status IN ${sql.list(statuses)}`);
SELECT * FROM projects WHERE status IN ('active', 'pending')
When using sql.list
make sure that the array contains at least one value, otherwise ()
will be used as the list, which is invalid SQL.
Read more about this in #15142
sql.where
The sql.where
function can be used to generate an SQL condition from a JavaScript object, using the same syntax as the where
option of the findAll
method.
const where = {
status: 'active',
funding: 'funded',
};
await sequelize.query(sql`SELECT * FROM projects WHERE ${sql.where(where)}`);
SELECT * FROM projects WHERE status = 'active' AND funding = 'funded'
It can also be used to generate an SQL condition where the left operand is something other than an attribute name:
Post.findAll({
where: sql.where(new Date('2012-01-01'), Op.between, [
sql.attribute('createdAt'),
sql.attribute('publishedAt'),
]),
});
-- The left operand is a literal value, and the right operands are column names
-- Something that is not possible to do with the POJO where syntax.
SELECT * FROM "projects" WHERE '2012-01-01' BETWEEN "createdAt" AND "publishedAt"
sql.attribute
The sql.attribute
function can be used to reference the name of a Model attribute. It is similar to the sql.identifier
function,
but the name of the attribute will be mapped to the name of the column, whereas sql.identifier
escapes its value as-is:
class User extends Model {
@Attribute({
type: DataTypes.STRING,
columnName: 'first_name',
})
declare firstName: string;
}
User.findAll({
where: sql.where(
sql.attribute('firstName'),
Op.eq,
'John',
),
});
SELECT * FROM "users" WHERE "first_name" = 'John'
Sequelize is only able to map the attribute name to the column name if it's aware of the Model.
This is typically the case for model methods, but is not the case for sequelize.query
.
On top of this mapping, sql.attribute
also supports the entire range of the attribute syntax. This means that it's possible to:
Use the association reference syntax
When eager loading associated models, you can reference includes using the association reference syntax.
The name of the association must start & end with a $
character, and the name of the attribute must be separated from the association name with a .
character.
User.findAll({
include: [
{
association: User.associations.posts,
where: sql.where(
sql.attribute('$user$.name'),
Op.eq,
'Zoe',
),
},
],
});
Use the Casting Syntax
You can use the ::
syntax to cast the attribute to a different type, just like in POJO attributes
User.findAll({
where: sql.where(
sql.attribute('createdAt::text'),
Op.like,
'2012-%',
),
});
Use the JSON Extraction syntax
You can use the JSON extraction syntax to access JSON properties, just like in POJO attributes
User.findAll({
where: sql.where(
// This will access the property `name` of the JSON column `data`
sql.attribute('data.name'),
Op.eq,
'John',
),
});
sql.cast
The sql.cast
function can be used to cast a value to the type of your choice:
User.findAll({
where: sql.where(
sql.cast(sql.attribute('createdAt'), 'text'),
Op.like,
'2012-%',
),
});
SELECT * FROM "users" WHERE CAST("createdAt" AS text) LIKE '2012-%'
It's also possible to use a Sequelize DataType as the type:
User.findAll({
where: sql.where(
sql.cast(sql.attribute('createdAt'), DataTypes.TEXT),
Op.like,
'2012-%',
),
});
Attributes support a shorthand syntax for casting. See Casting syntax in sql.attribute
and Casting Syntax in POJOs for more information.
sql.uuidV4
& sql.uuidV1
In supported dialects, using sql.uuidV4
and sql.uuidV1
will generate the dialect-specific function to generate a UUID.
In unsupported dialects, using these functions, except as the default value of an attribute, will throw an error.
sequelize.query(sql`INSERT INTO users (id) VALUES (${sql.uuidV4()})`);
-- postgres example
INSERT INTO users (id) VALUES (gen_random_uuid())
These functions are supported by the following dialects:
PostgreSQL | MariaDB | MySQL | MSSQL | SQLite | Snowflake | db2 | ibmi | |
---|---|---|---|---|---|---|---|---|
uuidV1 | uuid_generate_v1 (requires uuid-ossp ) | UUID | UUID | N/A | N/A | N/A | N/A | N/A |
uuidV4 | pg >= v13: gen_random_uuid pg < v13: uuid_generate_v4 (requires uuid-ossp ) | N/A | N/A | NEWID | N/A | N/A | N/A | N/A |
sql.col
This function is available for backwards compatibility, and there are currently no plans to deprecate it,
but it is not recommended to use in new code. Prefer instead to use sql.attribute
, sql.identifier
,
and the sql
template tag.
This function is a third way to reference a column name. It's similar to sql.identifier
, but gives special meaning to the *
characters.
Here are a few examples:
Input | sql.col | sql.identifier |
---|---|---|
* | * | "*" |
users.* | "users".* | "users.*" |
Unlike sql.attribute
, this method does not support any other special syntax, and does not map its input to a column name.
sql.jsonPath
This function can be used to extract a JSON property from a JSON value
sequelize.query(sql`
SELECT ${sql.jsonPath(sql.identifier('data'), ['addresses', 0, 'country'])} AS country
FROM users
`);
-- postgres
SELECT data#>ARRAY['addresses', '0', 'country'] AS country FROM users
-- other dialects
SELECT JSON_EXTRACT(data, '$.addresses[0].country') AS country FROM users
This can be useful to generate a JSON extraction query dynamically in a safe way.
The JSON Path array accepts a mix of strings and numbers. If a string is used, it will be treated as a property name (used to access an object property). If a number is used, it will be treated as an index (used to access an array element).
Make sure to use the correct type for your use case, as using the string '0'
will try to access the property named '0'
instead of the first element of the array.
Read more about this feature in the JSON Extraction chapter.
Attributes support a shorthand syntax for JSON extraction. See Casting syntax in sql.attribute
for more information.
sql.unquote
The sql.unquote
function is used to execute the JSON_UNQUOTE
(or equivalent) function on a JSON value:
sequelize.query(sql`
SELECT ${sql.unquote(sql.jsonPath(sql.identifier('data'), ['addresses', 0, 'country']))} AS country
FROM users
`);
-- postgres (the #>> operator unquotes, unlike the #> operator)
SELECT data#>>ARRAY['addresses', '0', 'country'] AS country FROM users
-- other dialects
SELECT JSON_UNQUOTE(JSON_EXTRACT(data, '$.addresses[0].country')) AS country FROM users
Read more about this feature in the JSON Extraction chapter.
sql.fn
This function exists for backwards compatibility with older versions of Sequelize but is not recommended for new code, as sql
can be used to write
SQL functions in a more natural way.
For instance, the old way of writing a lower
function was:
sql.fn('LOWER', sql.attribute('name'));
Which can now be written as:
sql`LOWER(${sql.attribute('name')})`;
Both result in:
LOWER("name")
sequelize.query
As there are often use cases in which it is just easier to execute raw / already prepared SQL queries, you can use the sequelize.query
method.
By default the function will return two arguments - a results array, and an object containing metadata (such as amount of affected rows, etc). Note that since this is a raw query, the metadata are dialect specific. Some dialects return the metadata "within" the results object (as properties on an array). However, two arguments will always be returned, but for MSSQL and MySQL it will be two references to the same object.
const [results, metadata] = await sequelize.query('UPDATE users SET y = 42 WHERE x = 12');
// Results will be an empty array and metadata will contain the number of affected rows.
When interpolating variables in your query, make absolutely sure that you are tagging your query with the sql
tag. sequelize.query
is
one of the few functions that will interpret plain strings as raw SQL, so forgetting to tag your query with sql
can lead to SQL injection vulnerabilities:
// Dangerous
await sequelize.query(`SELECT * FROM users WHERE first_name = ${firstName}`);
// Safe
await sequelize.query(sql`SELECT * FROM users WHERE first_name = ${firstName}`);
All other functions that accept raw SQL will throw an error if you use a string that has not been tagged with sql
.
This footgun may be removed in a future version of Sequelize.
In cases where you don't need to access the metadata you can pass in a query type to tell sequelize how to format the results. For example, for a simple select query you could do:
import { QueryTypes } from '@sequelize/core';
const users = await sequelize.query('SELECT * FROM `users`', {
type: QueryTypes.SELECT,
});
// We didn't need to destructure the result here - the results were returned directly
Several other query types are available. Peek into the source for details.
A second option is the model. If you pass a model the returned data will be instances of that model.
// Callee is the model definition. This allows you to easily map a query to a predefined model
const projects = await sequelize.query('SELECT * FROM projects', {
model: Projects,
mapToModel: true, // pass true here if you have any mapped fields
});
// Each element of `projects` is now an instance of Project
See more options in the query API reference. Some examples:
import { QueryTypes } from '@sequelize/core';
await sequelize.query('SELECT 1', {
// A function (or false) for logging your queries
// Will get called for every SQL query that gets sent
// to the server.
logging: console.log,
// If plain is true, then sequelize will only return the first
// record of the result set. In case of false it will return all records.
plain: false,
// Set this to true if you don't have a model definition for your query.
raw: false,
// The type of query you are executing. The query type affects how results are formatted before they are passed back.
type: QueryTypes.SELECT,
});
// Note the second argument being null!
// Even if we declared a callee here, the raw: true would
// supersede and return a raw object.
console.log(await sequelize.query('SELECT * FROM projects', { raw: true }));
"Dotted" attributes and the nest
option
If an attribute name of the table contains dots, the resulting objects can become nested objects by setting the nest: true
option. This is achieved with dottie.js under the hood. See below:
-
Without
nest: true
:import { QueryTypes } from '@sequelize/core';
const records = await sequelize.query('select 1 as `foo.bar.baz`', {
type: QueryTypes.SELECT,
});
console.log(JSON.stringify(records[0], null, 2));{
"foo.bar.baz": 1
} -
With
nest: true
:import { QueryTypes } from '@sequelize/core';
const records = await sequelize.query('select 1 as `foo.bar.baz`', {
nest: true,
type: QueryTypes.SELECT,
});
console.log(JSON.stringify(records[0], null, 2));{
"foo": {
"bar": {
"baz": 1
}
}
}
Footnotes
-
If you need to use raw SQL in a place that sequelize does not support, do not hesitate to open a feature request or a pull request. ↩