SQL Minifier

Why minify a query

Databases ignore formatting, but many places that carry SQL do not. A query embedded in a JSON config, a YAML pipeline, an environment variable or a Java string literal is easier to handle on one line. Logging and APM tools group queries by text, so stripping comments and layout makes the same statement look the same everywhere. ORMs and query builders sometimes emit SQL with odd spacing that a minifier normalises. And for comparing two versions of a query, minifying both removes formatting noise from the diff.

The savings are modest compared with code minifiers, because SQL has no names to shorten: the report query in the first example goes from 163 to 115 bytes. The value is in a predictable, compact form rather than raw size.

What is removed and what is kept

Minification works on the token stream from a dialect-aware tokenizer, so strings, quoted identifiers and comments are recognised exactly as your database would read them.

• Line comments (-- ...) and block comments (/* ... */) are dropped.
• Optimizer hints of the form /*+ INDEX(o idx_orders_created) */ are kept, since removing them can change the execution plan. Oracle’s line hints (--+ FULL(o)) are kept too and still end their line.
• MySQL executable comments, /*!40101 SET NAMES utf8mb4 */ as found in every mysqldump file, are kept because MySQL runs their contents.
• Whitespace is removed wherever two tokens can touch without changing meaning, and a single space is kept only where gluing them would merge or alter tokens: between two words, between two operators (so a - -1 never becomes a--1, which would start a comment), and between a minus and a number.
• Client batch separators stay on their own lines: T-SQL GO with the T-SQL (SQL Server) dialect and the SQL*Plus / after a PL/SQL block with PL/SQL (Oracle). Those tools only recognise them alone on a line.

String literals, including doubled quotes as in 'Ben O''Brien', quoted identifiers and keyword case are left exactly as written. Statements stay in order, separated by their semicolons.

Choose the right Dialect

The Dialect option matters more when minifying than it seems, because it decides which characters start strings, identifiers, parameters and comments. Backticks quote identifiers in MySQL and BigQuery, square brackets do so in T-SQL, $$ ... $$ encloses a string body in PostgreSQL and Snowflake, and # starts a comment in MySQL but not in PostgreSQL. That last difference shows why it matters: SELECT 1 # note loses # note under MySQL, yet keeps it under PostgreSQL, where # is the bitwise XOR operator. Pick the database the query is written for and the tokenizer will treat each character the way that engine does. Names quoted in another dialect’s style, such as [Order Date] in a PostgreSQL query, produce a warning that suggests the matching dialect.

The formatting options (Keyword case, Indent style, Commas and so on) only shape beautified output; the minifier never rewrites keywords.

Validation and privacy

Minify mode still runs the full sql-formatter parse used for beautifying, so an unterminated string, an unclosed bracket or a statement that stops after WHERE halts with the same error and position you would get when formatting. Switch views with Ctrl/Cmd+Shift+M and copy with Ctrl/Cmd+Shift+C. Queries are minified inside your browser, which matters when they contain table names, customer IDs or literal values from production.

Examples

-- daily revenue report
SELECT /*+ INDEX(o idx_orders_created) */
  o.id,
  o.total   -- gross
FROM orders o
WHERE o.created_at >= :since
  AND o.status = 'paid';

SELECT/*+ INDEX(o idx_orders_created) */o.id,o.total FROM orders o WHERE o.created_at>= :since AND o.status='paid';

/*!40101 SET NAMES utf8mb4 */;
-- seed data
INSERT INTO customers (id, name, email)
VALUES
  (1, 'Aisha Tan', '[email protected]'),
  (2, 'Ben O''Brien', '[email protected]');

/*!40101 SET NAMES utf8mb4 */;INSERT INTO customers(id,name,email)VALUES(1,'Aisha Tan','[email protected]'),(2,'Ben O''Brien','[email protected]');

CREATE TABLE dbo.Orders (
  Id INT NOT NULL PRIMARY KEY,
  Total DECIMAL(10, 2) NOT NULL -- gross amount
);
GO
/* seed */
INSERT INTO dbo.Orders (Id, Total) VALUES (1, 129.90);
GO

CREATE TABLE dbo.Orders(Id INT NOT NULL PRIMARY KEY,Total DECIMAL(10,2)NOT NULL);
GO
INSERT INTO dbo.Orders(Id,Total)VALUES(1,129.90);
GO

BEGIN
  -- archive old rows
  DELETE FROM orders_archive WHERE created_at < SYSDATE - 365;
  COMMIT;
END;
/
SELECT COUNT(*) FROM orders_archive;

BEGIN DELETE FROM orders_archive WHERE created_at<SYSDATE- 365;COMMIT;END;
/
SELECT COUNT(*)FROM orders_archive;

Common errors and how to fix them

Related tools

• SQL formatter
• sql validator
• sql formatter/mysql
• sql formatter/postgresql
• sql formatter/plsql
• sql formatter/transactsql
• sql style guide