SQL injection protection module
npm install @nearform/sqlA simple SQL injection protection module that allows you to use ES6 template strings for escaped statements. Works with pg, mysql and mysql2 library.
[![npm version][1]][2] [![build status][3]][4] [![js-standard-style][5]][6]
1. Install
2. Usage
1. Linting
3. Methods
1. glue
2. map
2. (deprecated) append
4. Utilities
1. unsafe
2. quoteIdent
5. How it works?
6. Undefined values and nullable fields
7. Testing, linting, & coverage
8. Benchmark
9. License
``sh`
npm install @nearform/sql
`js
const SQL = require('@nearform/sql')
const username = 'user'
const email = 'user@email.com'
const password = 'Password1'
// generate SQL query
const sql = SQL
INSERT INTO users (username, email, password)
VALUES (${username},${email},${password})
pg.query(sql) // execute query in pg
mysql.query(sql) // execute query in mysql
mysql2.query(sql) // execute query in mysql2
`
We recommend using eslint-plugin-sql to prevent cases in which the SQL tag is forgotten to be added in front of template strings. Eslint will fail if you write SQL queries without sql tag in front of the string.
`sqlSELECT 1
// fails - Message: Use "sql" tag
sqlSELECT 1`
// passes
> ⚠️ Warning
>
> The unsafe option interprets the interpolated values as literals and it should be used carefully to avoid introducing SQL injection vulnerabilities.
`js
const username = 'user1'
const email = 'user1@email.com'
const userId = 1
const updates = []
updates.push(SQLname = ${username})email = ${email}
updates.push(SQL)
const sql = SQLUPDATE users SET ${SQL.glue(updates, ' , ')} WHERE id = ${userId}`
or also
`js
const ids = [1, 2, 3]
const value = 'test'
const sql = SQL
UPDATE users
SET property = ${value}
WHERE id
IN (${SQL.glue(ids.map(id => SQL${id}), ' , ')})`
Glue can also be used statically:
`js(${id})
const ids = [1, 2, 3]
const idsSqls = ids.map(id => SQL)`
SQL.glue(idsSqls, ' , ')
Glue can also be used to generate batch operations:
`js
const users = [
{ id: 1, name: 'something' },
{ id: 2, name: 'something-else' },
{ id: 3, name: 'something-other' }
]
const sql = SQLINSERT INTO users (id, name) VALUES
${SQL.glue(
users.map(user => SQL(${user.id},${user.name}})),
' , '
)}`
Using the default mapperFunction which is just an iteration over the array elements
`js
const ids = [1, 2, 3]
const values = SQL.map(ids)
const sql = SQLINSERT INTO users (id) VALUES (${values})`
Using an array of objects which requires a mapper function
`js
const objArray = [{
id: 1,
name: 'name1'
},
{
id: 2,
name: 'name2'
},
{
id: 3,
name: 'name3'
}]
const mapperFunction = (objItem) => objItem.id
const values = SQL.map(objArray, mapperFunction)
const sql = SQLINSERT INTO users (id) VALUES (${values})`
Append has been deprecated in favour of using template literals:
`jsFROM table
const from = SQLSELECT * ${from}
const sql = SQL`
For now, you can still use append as follows:
`js
const username = 'user1'
const email = 'user1@email.com'
const userId = 1
const sql = SQLUPDATE users SET name = ${username}, email = ${email}, ${dynamicName} = 'dynamicValue'
sql.append(SQL, { unsafe: true })WHERE id = ${userId}
sql.append(SQL)`
Does a literal interpolation of the provided value, interpreting the provided value as-is.
It works similarly to the unsafe option of the append method and requires the same security considerations.
`js
const username = 'john'
const userId = 1
const sql = SQL
UPDATE users
SET username = '${SQL.unsafe(username)}'
WHERE id = ${userId}`
Mimics the native PostgreSQL quote_ident and MySQL quote_identifier functions.
In PostgreSQL, it wraps the provided value in double quotes " and escapes any double quotes existing in the provided value.
In MySQL, it wraps the provided value in backticks ` and escapes any backticks existing in the provided value.
It's convenient to use when schema, table or field names are dynamic and can't be hardcoded in the SQL query string.
`js
const table = 'users'
const username = 'john'
const userId = 1
const sql = SQL
UPDATE ${SQL.quoteIdent(table)}
SET username = ${username}
WHERE id = ${userId}`
The SQL template string tag parses query and returns an objects that's understandable by pg library:
`js
const username = 'user'
const email = 'user@email.com'
const password = 'Password1'
const sql = SQLINSERT INTO users (username, email, password) VALUES (${username}, ${email}, ${password}) // generate SQL query`
sql.text // INSERT INTO users (username, email, password) VALUES ($1 , $2 , $3) - for pg
sql.sql // INSERT INTO users (username, email, password) VALUES (? , ? , ?) - for mysql and mysql2
sql.values // ['user, 'user@email.com', 'Password1']
To help with debugging, you can view an approximate representation of the SQL query with values filled in. It may differ from the actual SQL executed by your database, but serves as a handy reference when debugging. The debug output _should not_ be executed as it is not guaranteed safe. You can may also inspect the SQL object via console.log.
`js
sql.debug // INSERT INTO users (username, email, password) VALUES ('user','user@email.com','Password1')
console.log(sql) // SQL << INSERT INTO users (username, email, password) VALUES ('user','user@email.com','Password1') >>
`
Don't pass undefined values into the sql query string builder. It throws on undefined values as this is a javascript concept and sql does not handle it.
Sometimes you may expect to not have a value to be provided to the string builder, and this is ok as the coresponding field is nullable. In this or similar cases the recommended way to handle this is to coerce it to a null js value.
Example:
`js
const user = { name: 'foo bar' }
const sql = SQLINSERT into users (name, address) VALUES (${user.name},${
user.address || null
})`
sql.debug // INSERT INTO users (name, address) VALUES ('foo bar',null)
The below example functions can be used to generate an INSERT INTO statement from an object, which will convert the object keys to snake case.
`js${value}
function insert(table, insertData, { toSnakeCase } = { toSnakeCase: false }) {
const builder = Object.entries(insertData).reduce(
(acc, [column, value]) => {
if (value !== undefined) {
toSnakeCase
? acc.columns.push(pascalOrCamelToSnake(column))
: acc.columns.push(column)
acc.values.push(SQL)INSERT INTO ${SQL.quoteIdent(table)} (${SQL.unsafe(
}
return acc
},
{ columns: [], values: [] }
)
return SQL
builder.columns.join(', ')
)}) VALUES (${SQL.glue(builder.values, ', ')})
}
const pascalOrCamelToSnake = str =>
str[0].toLowerCase() +
str
.slice(1, str.length)
.replace(/[A-Z]/g, letter => _${letter.toLowerCase()})`
This module can be tested and reported on in a variety of ways...
`sh`
npm run test # runs tap based unit test suite.
npm run test:security # runs sqlmap security tests.
npm run test:typescript # runs type definition tests.
npm run coverage # generates a coverage report in docs dir.
npm run lint # lints via standardJS.
Find more about @nearform/sql` speed here
Copyright NearForm 2021. Licensed under
[Apache 2.0][7]
[1]: https://img.shields.io/npm/v/@nearform/sql.svg?style=flat-square
[2]: https://npmjs.org/package/@nearform/sql
[3]: https://github.com/nearform/sql/workflows/CI/badge.svg
[4]: https://github.com/nearform/sql/actions?query=workflow%3ACI
[5]: https://img.shields.io/badge/code%20style-standard-brightgreen.svg?style=flat-square
[6]: https://github.com/feross/standard
[7]: https://tldrlegal.com/license/apache-license-2.0-(apache-2.0)
