MySQL

Scythe supports MySQL/MariaDB with dialect-specific type handling across all 10 languages. MySQL support operates at the parser and analyzer level – SQL parsing, type inference, and nullability analysis are fully MySQL-aware. The code generation backends work the same regardless of the source database.

Backend support

Every language has at least one MySQL backend. Multi-engine backends (like java-jdbc, php-pdo, rust-sqlx) load engine-specific manifests automatically.

Language	Backend	Library
Rust	rust-sqlx	sqlx
Python	python-aiomysql	aiomysql
TypeScript	typescript-mysql2	mysql2
Go	go-database-sql	database/sql
Java	java-jdbc	JDBC
Kotlin	kotlin-jdbc	JDBC
C#	csharp-mysqlconnector	MySqlConnector
Elixir	elixir-myxql	MyXQL
Ruby	ruby-mysql2	mysql2 gem
PHP	php-pdo	PDO

Differences from PostgreSQL

Feature	PostgreSQL	MySQL
Auto-increment	SERIAL (type alias)	AUTO_INCREMENT (column modifier)
Enums	CREATE TYPE ... AS ENUM	ENUM(...) inline on column
Arrays	TEXT[] native	Not supported
Placeholders	$1, $2	?, ?
JSONB	Native binary JSON	JSON only (no JSONB)
Schemas	schema.table	Database-scoped
RETURNING	Supported	Not supported (use LAST_INSERT_ID())
Range types	Native	Not supported

AUTO_INCREMENT handling

CREATE TABLE users (
    id INT AUTO_INCREMENT PRIMARY KEY,
    name VARCHAR(255) NOT NULL
);

AUTO_INCREMENT columns are treated as NOT NULL, equivalent to PostgreSQL’s SERIAL.

Inline ENUM

MySQL enums are declared inline on the column:

CREATE TABLE users (
    id INT AUTO_INCREMENT PRIMARY KEY,
    status ENUM('active', 'inactive', 'banned') NOT NULL
);

Scythe registers this as a catalog enum named {table}_{column} – here, users_status – and it resolves to enum::users_status, generating a proper enum type rather than a bare string.

Placeholder handling

MySQL uses ? positional placeholders. Scythe translates $N in your SQL to ? for MySQL backends:

-- Written as:
SELECT id, name FROM users WHERE id = $1;

-- Translated to:
SELECT id, name FROM users WHERE id = ?;

Type mapping table

MySQL Type	Neutral Type
INT / INTEGER	int32
BIGINT	int64
SMALLINT	int16
TINYINT	int16
MEDIUMINT	int32
FLOAT (bare, no precision)	float32
DOUBLE	float64
DECIMAL / NUMERIC	decimal
TINYINT UNSIGNED / SMALLINT UNSIGNED	int16
MEDIUMINT UNSIGNED / INT UNSIGNED	int32
BIGINT UNSIGNED	int64
VARCHAR / CHAR / TEXT	string
TINYTEXT / MEDIUMTEXT / LONGTEXT	string
BOOLEAN / BOOL	bool
BIT / BIT(1)	bool
BIT(n) where n > 1	int64
BLOB / TINYBLOB / MEDIUMBLOB / LONGBLOB	bytes
BINARY / VARBINARY	bytes
DATE	date
TIME	time
DATETIME	datetime
TIMESTAMP	datetime
YEAR	int16
JSON	json
ENUM(...)	enum::{table}_{column}
SET(...)	string

UNSIGNED integer columns map to the same-width signed neutral type – there is no dedicated unsigned neutral type, and BIGINT UNSIGNED has no wider type to widen to. Values near the top of the unsigned range (for example, above i16::MAX for SMALLINT UNSIGNED or above i64::MAX for BIGINT UNSIGNED) need application-level handling in the generated code.