prettier-plugin-sql-cst

Prettier plugin SQL-CST

A Prettier plugin for SQL that uses sql-parser-cst and the actual Prettier formatting algorithm.

Try it live!

Like Prettier for JavaScript, this plugin formats SQL expressions differently depending on their length. A short SQL query will be formatted on a single line:

SELECT a, b, c FROM tbl WHERE x > 10;

A longer query, will get each clause printed on a separate line:

SELECT id, client.name, client.priority
FROM client
WHERE client.id IN (12, 18, 121);

An even longer one gets the contents of each clause indented:

SELECT
  client.id,
  client.name AS client_name,
  organization.name AS org_name,
  count(orders.id) AS nr_of_orders
FROM
  client
  LEFT JOIN organization ON client.organization_id = organization.id
  LEFT JOIN orders ON orders.client_id = client.id
WHERE
  client.status = 'active'
  AND client.id IN (28, 214, 457)
  AND orders.status IN ('active', 'pending', 'processing')
GROUP BY client.id
ORDER BY client.name
LIMIT 100;

Formatting philosophy

• Adapt formatting based on expression length.
• Stick to one style and avoid configuration options.
• Format embedded languages (like JSON data and JavaScript programs).
• When unsure, preserve existing syntax.

Currently this plugin preserves most of the syntax elements and concentrates mainly on the layout of whitespace.

See STYLE_GUIDE for overview of the SQL formatting style used.

Getting started

Install it as any other Prettier plugin:

npm install --save-dev prettier prettier-plugin-sql-cst
# or
pnpm add --save-dev prettier prettier-plugin-sql-cst
# or
yarn add --dev prettier prettier-plugin-sql-cst

Then use it on SQL files through Prettier command line tool or Prettier extension for your editor of choice.

For a one-off run you can execute:

prettier --plugin prettier-plugin-sql-cst --parser sqlite myfile.sql

But generally you're better off using a configuration file.

Choosing an SQL dialect

By default the plugin will determine SQL dialect based on file extension:

• .sql or .sqlite - SQLite
• .bigquery - BigQuery

You can override this behavior with a prettier configuration in .prettierrc.json file:

{
  "plugins": ["prettier-plugin-sql-cst"],
  "overrides": [
    {
      "files": ["*.sql"],
      "options": { "parser": "bigquery" }
    }
  ]
}

Or you could also store it inside your package.json:

{
  "prettier": {
    "plugins": ["prettier-plugin-sql-cst"],
    "overrides": [
      {
        "files": ["*.sql"],
        "options": { "parser": "bigquery" }
      }
    ]
  }
}

The plugin provides the following parsers:

• sqlite
• bigquery
• postgresql (experimental! expect crashes)
• mysql (experimental! expect crashes)
• mariadb (experimental! expect crashes)

Configuration

The standard Prettier options printWidth, tabWidth, useTabs apply. There are also several SQL-specific options:

• sqlKeywordCase: "upper" | "lower" | "preserve" (default "upper")

  Enforce case of keywords.

  This effects all keywords except literals (see sqlLiteralCase) and data types (see sqlTypeCase).

  For now preserve is somewhat incompatible with sqlCanonicalSyntax: true (e.g. the added AS keywords will always be in uppercase).

• sqlLiteralCase: "upper" | "lower" | "preserve" (default "upper")

  Enforce case of literals TRUE, FALSE and NULL.

  Also applies to ON & OFF literals in PostgreSQL SET statements.

  Since 0.14.0

• sqlTypeCase: "upper" | "lower" | "preserve" (default "upper")

  Enforce case of type names.

  In PostgreSQL this will only effect a limited set of builtin data types, the rest will be formatted as identifiers. It will effect casing of all the keywords in types like TIMESTAMP WITH TIME ZONE and INTERVAL DAY TO SECOND, but it won't effect the casing of SETOF and TABLE keywords in SETOF type and TABLE (x int, y int). For PostgreSQL the recommended setting of this option is lower.

  In BigQuery this also effects the names parametric types, for example array<struct<x int64>> will get formatted as ARRAY<STRUCT<x INT64>>.

  Since 0.17.0

• sqlIdentifierCase: "upper" | "lower" | "preserve" (default "preserve")

  Enforce case of unquoted identifier names.

  This effects all identifiers except function names (see sqlFunctionCase).

  Beware that in BigQuery unquoted identifiers are case-sensitive, so use of this option with BigQuery is not recommended.

  Experimental: This option is still new and not thoroughly tested. You have been warned.

  Since 0.17.0

• sqlFunctionCase: "upper" | "lower" | "preserve" (default "preserve")

  Enforce case of unquoted function names.

  This option also applies to schema-qualified function names, for example my_schema.foo() will get formatted as my_schema.FOO() when sqlFunctionCase: "upper" is configured.

  Experimental: This option is still new and not thoroughly tested. You have been warned.

  Since 0.17.0

• sqlParamTypes: ("?", "?nr", "$nr", ":name", "$name", "@name" "@``@name``")[] (default [])

  Array of bound parameter types to support.

  By default the parser will reject the following SQL:

  SELECT id, name FROM products WHERE price > ? AND category = ?
  

  Configuring sqlParamTypes: ["?"] will make the parser recognize the ? placeholders.

  Since 0.7.0

• sqlCanonicalSyntax: boolean (default true)

  When enabled, performs some opinionated changes of keywords and operators, like enforcing the use of AS in aliases and replacing <> comparisons with !=. See STYLE_GUIDE for more details.

  Since 0.11.0

• sqlFinalSemicolon: boolean (default true)

  When enabled, enforces a semicolon at the end of last statement. When disabled leaves it up to the author whether to add a final semicolon or not.

  Since 0.13.0

• sqlAcceptUnsupportedGrammar: boolean (default false)

  Normally when the plugin encounters SQL syntax it doesn't support it will throw an error and won't format anything at all. With this option enabled, it will skip over SQL statements it doesn't recognize, leaving them as-is.

  Deprecated: Use conditional comments /* sql-parser-cst-disable */ and /* sql-parser-cst-enable */ instead.

  Since 0.12.0

Usage inside VSCode

To use this plugin inside VSCode, install the Prettier VSCode extension.

Follow Prettier VSCode docs to configure it as the default formatter.

You might also need to configure prettier.documentSelectors to enable Prettier for *.sql files.

To see what Prettier is, or is not doing - open the VSCode Output window and select the Prettier dropdown. On format, the window should show your inferredParser. It should reconfirm that by showing "parser": "sqlite" (or whichever perser you have configured inside your prettier config overrides section) and plugins with the path to this package. If you don't see that part, Prettier not using this package.

FAQ

The SQL dialect I'm using is not supported. Can you add support for it?

Support for new SQL dialects depends on these dialects being supported by sql-parser-cst. If you really want to, you can open a new issue for that in the parser repo. But be aware that implementing parser support for new dialects takes a lot of work. As long as the ongoing implementation of PostgreSQL, MySQL and MariaDB is not finished, it's unlikely that work on any other dialect will start.

How can I format SQL strings inside JavaScript files?

Use prettier-plugin-embed together with prettier-plugin-sql-cst.

Limitations and development status

Currently this plugin supports two SQL dialects:

• SQLite - full support.
• BigQuery - full support.

It also has experimental support for the following dialects:

• PostgreSQL
• MySQL
• MariaDB

The main limitation is that the parser does not support full syntax of these dialects. To overcome this limitation you can use conditional comments to disable parsing of the problematic section of SQL:

SELECT * FROM products;
/* sql-parser-cst-disable */
UNSUPPORTED SQL IN HERE;
/* sql-parser-cst-enable */
UPDATE products SET sold = true WHERE id = 2;

The specifics of the SQL formatting style are still very much subject to change. Though the general principles should be mostly in place by now.