Your first query

On this page

On MySQL:

  • ? placeholders
  • No RETURNING: CreateTask is execresult, and sql.Result's LastInsertId gives the new id
  • BIGINT AUTO_INCREMENT maps to int64, matching LastInsertId; INT maps to int32
  • DATETIME maps to time.Time, which needs parseTime=true in the DSN
  • migrate_backend_mysql.New is the migration backend, taking a GET_LOCK advisory lock; WithDriverMultiStatements passes on the DSN's multiStatements setting

Seagull reads two things: the migrations that build your schema, and the queries you write against it. From them it generates a Go package with a method for every query, typed from the schema. This first example is a tasks list: one table, and three queries to add a task, complete it and list them all.

The schema

Migrations are plain SQL files, applied in order. Seagull replays them to learn every table and column, including which columns can be NULL.

db/migrations/0001_init.up.sql

CREATE TABLE tasks (
    id         BIGINT       NOT NULL AUTO_INCREMENT PRIMARY KEY,
    title      VARCHAR(200) NOT NULL,
    completed  INT          NOT NULL DEFAULT 0,
    created_at DATETIME     NOT NULL DEFAULT CURRENT_TIMESTAMP
);

The queries

Each query starts with an sgl.query header giving the method's name and its command: what the method returns. ListTasks returns every row, so it is many.

db/queries/tasks.sql

-- sgl.query(ListTasks, many)
SELECT id, title, completed, created_at FROM tasks ORDER BY id;

MySQL has no RETURNING, so adding a task is execresult: the method returns the driver's sql.Result, and LastInsertId gives the id MySQL assigned.

db/queries/tasks.sql

-- sgl.query(CreateTask, execresult)
INSERT INTO tasks (title) VALUES (?);

Completing a task is execrows, which returns how many rows changed, so the program can tell whether the task existed:

db/queries/tasks.sql

-- sgl.query(CompleteTask, execrows)
UPDATE tasks SET completed = 1 WHERE id = ?;

What sgl generates

sgl generate writes the package named in sgl.yaml. Every column of the query becomes a field of a row struct, typed from the schema: a column that can be NULL would be a pointer, and these cannot be, so none are.

internal/store

type ListTasksRow struct {
	ID        int64     `json:"id"`
	Title     string    `json:"title"`
	Completed int32     `json:"completed"`
	CreatedAt time.Time `json:"created_at"`
}

func (queries *Queries) ListTasks(ctx context.Context) ([]ListTasksRow, error) {
	rows, err := queries.reader.QueryContext(ctx, listTasksSQL)
	if err != nil {
		return nil, err
	}
	defer rows.Close()
	var results []ListTasksRow
	if rows.Next() {
		var row ListTasksRow
		scanArgs := []any{&row.ID, &row.Title, &row.Completed, &row.CreatedAt}
		for more := true; more; more = rows.Next() {
			row = ListTasksRow{}
			if err := rows.Scan(scanArgs...); err != nil {
				return nil, err
			}
			results = append(results, row)
		}
	}
	if err := rows.Err(); err != nil {
		return nil, err
	}
	return results, nil
}

Using it

The program opens the database, runs the migrations and calls the generated methods. There is no SQL in it and nothing to scan by hand: the methods take and return Go values.

example.go

package main

import (
	"context"
	"database/sql"
	"fmt"
	"io"

	"seagull.sh/seagull/examples/mysql/001_first_query/internal/store"
)

// run adds two tasks, completes the first and lists them all, through the methods sgl
// generated from db/queries/tasks.sql.
//
// MySQL has no RETURNING, so CreateTask is execresult: it returns the driver's
// sql.Result, and LastInsertId gives the AUTO_INCREMENT id MySQL assigned. The id column
// is BIGINT, so it is an int64 like LastInsertId's result and needs no conversion.
//
// Takes database (*sql.DB) which is a migrated database.
// Takes output (io.Writer) which receives the task list.
//
// Returns error when a query fails.
func run(ctx context.Context, database *sql.DB, output io.Writer) error {
	queries := store.New(database)

	result, err := queries.CreateTask(ctx, "Feed the seagulls")
	if err != nil {
		return fmt.Errorf("creating a task: %w", err)
	}
	feed, err := result.LastInsertId()
	if err != nil {
		return fmt.Errorf("reading the new task's id: %w", err)
	}
	if _, err := queries.CreateTask(ctx, "Count the chips"); err != nil {
		return fmt.Errorf("creating a task: %w", err)
	}

	completed, err := queries.CompleteTask(ctx, feed)
	if err != nil {
		return fmt.Errorf("completing task %d: %w", feed, err)
	}
	fmt.Fprintf(output, "completed %d task\n", completed)

	tasks, err := queries.ListTasks(ctx)
	if err != nil {
		return fmt.Errorf("listing tasks: %w", err)
	}
	for _, task := range tasks {
		fmt.Fprintf(output, "%d\t%-18s completed=%d\n", task.ID, task.Title, task.Completed)
	}
	return nil
}

What it prints

completed 1 task
1	Feed the seagulls  completed=1
2	Count the chips    completed=0

Try it

Change a column's type in the migration, or ask for a column that does not exist, and run sgl generate again: the Go changes with it, or sgl vet says what is wrong, with a code you can look up in the diagnostics.

Across engines

Only these Go types of CreateTask differ between engines; the rest are the same everywhere.

MySQLSQLitePostgreSQLCockroachDBMariaDBDuckDBClickHouse
id column·int64int64int64int64int64·
title column·stringstringstringstringstring·
completed column·int64boolboolint32bool·
created_at column·stringtime.Timetime.Timetime.Timetime.Time·
id parameter······uint64

A · marks a parameter or column that version of the example does not have.

TimescaleDB runs the PostgreSQL version unchanged.

Compare the whole example side by side with SQLite, PostgreSQL, CockroachDB, MariaDB, DuckDB and ClickHouse.

Run it locally

From the directory you cloned the Seagull repository into:

cd seagull/examples/mysql && go run ./001_first_query

Its test runs it too, and checks what it prints against output.txt:

cd seagull/examples/mysql && SGL_EXAMPLE_DSN="$DATABASE_URL" go test ./001_first_query

Needs a running MySQL server: set DATABASE_URL, or pass -dsn, to the address of an empty database.

The files

Every file of the example, as it is in the Seagull repository.

sgl.yaml

# Copyright 2026 PolitePixels Limited
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

# This project stands against fascism, authoritarianism, and all forms of
# oppression. We built this to empower people, not to enable those who would
# strip others of their rights and dignity.

version: 1

databases:
  - name: tasks
    dialect: mysql
    migrations: db/migrations
    queries: db/queries
    go:
      package: store
      out: internal/store

db/migrations/0001_init.up.sql

CREATE TABLE tasks (
    id         BIGINT       NOT NULL AUTO_INCREMENT PRIMARY KEY,
    title      VARCHAR(200) NOT NULL,
    completed  INT          NOT NULL DEFAULT 0,
    created_at DATETIME     NOT NULL DEFAULT CURRENT_TIMESTAMP
);

db/migrations/0001_init.down.sql

DROP TABLE tasks;

db/queries/tasks.sql

-- sgl.query(CreateTask, execresult)
INSERT INTO tasks (title) VALUES (?);

-- sgl.query(ListTasks, many)
SELECT id, title, completed, created_at FROM tasks ORDER BY id;

-- sgl.query(CompleteTask, execrows)
UPDATE tasks SET completed = 1 WHERE id = ?;

example.go

package main

import (
	"context"
	"database/sql"
	"fmt"
	"io"

	"seagull.sh/seagull/examples/mysql/001_first_query/internal/store"
)

// run adds two tasks, completes the first and lists them all, through the methods sgl
// generated from db/queries/tasks.sql.
//
// MySQL has no RETURNING, so CreateTask is execresult: it returns the driver's
// sql.Result, and LastInsertId gives the AUTO_INCREMENT id MySQL assigned. The id column
// is BIGINT, so it is an int64 like LastInsertId's result and needs no conversion.
//
// Takes database (*sql.DB) which is a migrated database.
// Takes output (io.Writer) which receives the task list.
//
// Returns error when a query fails.
func run(ctx context.Context, database *sql.DB, output io.Writer) error {
	queries := store.New(database)

	result, err := queries.CreateTask(ctx, "Feed the seagulls")
	if err != nil {
		return fmt.Errorf("creating a task: %w", err)
	}
	feed, err := result.LastInsertId()
	if err != nil {
		return fmt.Errorf("reading the new task's id: %w", err)
	}
	if _, err := queries.CreateTask(ctx, "Count the chips"); err != nil {
		return fmt.Errorf("creating a task: %w", err)
	}

	completed, err := queries.CompleteTask(ctx, feed)
	if err != nil {
		return fmt.Errorf("completing task %d: %w", feed, err)
	}
	fmt.Fprintf(output, "completed %d task\n", completed)

	tasks, err := queries.ListTasks(ctx)
	if err != nil {
		return fmt.Errorf("listing tasks: %w", err)
	}
	for _, task := range tasks {
		fmt.Fprintf(output, "%d\t%-18s completed=%d\n", task.ID, task.Title, task.Completed)
	}
	return nil
}

main.go

// Command 001_first_query applies its migrations to a MySQL database, then runs the
// example in example.go: two tasks are added, one is completed, and both are listed.
//
// Run it with `go run . -dsn 'user:password@tcp(localhost:3306)/tasks'`, or set
// DATABASE_URL instead of passing -dsn. Give it an empty database of its own, as the
// migrations create its tables and the example adds rows each time it runs.
package main

import (
	"context"
	"database/sql"
	"embed"
	"errors"
	"flag"
	"fmt"
	"log"
	"os"

	"github.com/go-sql-driver/mysql"

	"seagull.sh/seagull/sdk/migrate"
	"seagull.sh/seagull/sdk/migrate/migrate_backend_mysql"
)

// migrations holds the migration files, compiled into the binary.
//
//go:embed db/migrations
var migrations embed.FS

func main() {
	dsn := flag.String("dsn", os.Getenv("DATABASE_URL"), "MySQL data source name; defaults to $DATABASE_URL")
	flag.Parse()
	if err := execute(context.Background(), *dsn); err != nil {
		log.Fatal(err)
	}
}

// execute opens and migrates the database dsn names, and runs the example against it.
//
// Takes dsn (string) which is the database's data source name.
//
// Returns error when the database cannot be prepared or the example fails.
func execute(ctx context.Context, dsn string) error {
	if dsn == "" {
		return errors.New("set -dsn or DATABASE_URL to a MySQL data source name")
	}
	config, err := mysql.ParseDSN(dsn)
	if err != nil {
		return err
	}
	// Seagull maps DATETIME and TIMESTAMP columns to time.Time, which the driver returns
	// only when parseTime is set.
	config.ParseTime = true
	dsn = config.FormatDSN()

	database, err := sql.Open("mysql", dsn)
	if err != nil {
		return err
	}
	defer func() { _ = database.Close() }()

	migrator, err := migrate.NewMigrator(database, migrate_backend_mysql.New(migrate.WithDriverMultiStatements(config.MultiStatements)), migrations, "db/migrations")
	if err != nil {
		return err
	}
	if _, err := migrator.Up(ctx); err != nil {
		return fmt.Errorf("migrating: %w", err)
	}
	return run(ctx, database, os.Stdout)
}

internal/store/models.go

// Code generated by sgl. DO NOT EDIT.

package store

import "time"

type Tasks struct {
	ID        int64     `json:"id"`
	Title     string    `json:"title"`
	Completed int32     `json:"completed"`
	CreatedAt time.Time `json:"created_at"`
}

internal/store/querier.go

// Code generated by sgl. DO NOT EDIT.

// Package store reads DATETIME, DATE and TIMESTAMP columns into time.Time, so the DSN it connects with must set parseTime=true.
package store

import (
	"context"
	"database/sql"
)

type DBTX interface {
	ExecContext(ctx context.Context, query string, args ...any) (sql.Result, error)
	QueryContext(ctx context.Context, query string, args ...any) (*sql.Rows, error)
	QueryRowContext(ctx context.Context, query string, args ...any) *sql.Row
}

type dbtxWrapper interface {
	WrapDBTX(inner any) any
}

type Queries struct {
	reader  DBTX
	writer  DBTX
	wrapper dbtxWrapper
}

func New(db DBTX) *Queries {
	wrapper, _ := db.(dbtxWrapper)
	return &Queries{reader: db, writer: db, wrapper: wrapper}
}

func NewWithReplica(writer DBTX, reader DBTX) *Queries {
	wrapper, _ := writer.(dbtxWrapper)
	return &Queries{reader: reader, writer: writer, wrapper: wrapper}
}

func (queries *Queries) WithTx(transaction *sql.Tx) *Queries {
	if queries == nil {
		return &Queries{reader: transaction, writer: transaction, wrapper: nil}
	}
	db := queries.wrapDBTX(transaction)
	return &Queries{reader: db, writer: db, wrapper: queries.wrapper}
}

func (queries *Queries) wrapDBTX(db DBTX) DBTX {
	if queries.wrapper == nil {
		return db
	}
	wrapped, ok := queries.wrapper.WrapDBTX(db).(DBTX)
	if !ok {
		return db
	}
	return wrapped
}

func (queries *Queries) RunInTx(ctx context.Context, db *sql.DB, fn func(*Queries) error) error {
	transaction, err := db.BeginTx(ctx, nil)
	if err != nil {
		return err
	}
	defer transaction.Rollback()
	if err = fn(queries.WithTx(transaction)); err != nil {
		return err
	}
	return transaction.Commit()
}

internal/store/tasks.sql.go

// Code generated by sgl. DO NOT EDIT.

package store

import (
	"context"
	"database/sql"
	"time"
)

const completeTaskSQL = `UPDATE tasks SET completed = 1 WHERE id = ?;`

func (queries *Queries) CompleteTask(ctx context.Context, id int64) (int64, error) {
	results, err := queries.writer.ExecContext(ctx, completeTaskSQL, id)
	if err != nil {
		return 0, err
	}
	return results.RowsAffected()
}

const createTaskSQL = `INSERT INTO tasks (title) VALUES (?);`

func (queries *Queries) CreateTask(ctx context.Context, title string) (sql.Result, error) {
	return queries.writer.ExecContext(ctx, createTaskSQL, title)
}

const listTasksSQL = `SELECT id, title, completed, created_at FROM tasks ORDER BY id;`

type ListTasksRow struct {
	ID        int64     `json:"id"`
	Title     string    `json:"title"`
	Completed int32     `json:"completed"`
	CreatedAt time.Time `json:"created_at"`
}

func (queries *Queries) ListTasks(ctx context.Context) ([]ListTasksRow, error) {
	rows, err := queries.reader.QueryContext(ctx, listTasksSQL)
	if err != nil {
		return nil, err
	}
	defer rows.Close()
	var results []ListTasksRow
	if rows.Next() {
		var row ListTasksRow
		scanArgs := []any{&row.ID, &row.Title, &row.Completed, &row.CreatedAt}
		for more := true; more; more = rows.Next() {
			row = ListTasksRow{}
			if err := rows.Scan(scanArgs...); err != nil {
				return nil, err
			}
			results = append(results, row)
		}
	}
	if err := rows.Err(); err != nil {
		return nil, err
	}
	return results, nil
}

output.txt

completed 1 task
1	Feed the seagulls  completed=1
2	Count the chips    completed=0

Also on other engines

SQLiteA tasks table and three queries; every INTEGER column becomes int64 and RETURNING hands back the new row.PostgreSQLA tasks table and three queries through pgx's database/sql driver; BIGSERIAL becomes int64, TIMESTAMPTZ becomes time.Time and RETURNING hands back the new row.CockroachDBThe SQLite tasks program on CockroachDB; unique_rowid() fills the INT8 key, STRING and BOOL become string and bool, and the CockroachDB migration backend applies the schema.TimescaleDBA tasks table and three queries through pgx's database/sql driver; BIGSERIAL becomes int64, TIMESTAMPTZ becomes time.Time and RETURNING hands back the new row.MariaDBA tasks table and three queries; MariaDB has INSERT ... RETURNING, so the insert hands back the new row as it would on SQLite or PostgreSQL.DuckDBA tasks table and three queries; a sequence supplies the ids, RETURNING hands back the new row, and BOOLEAN and TIMESTAMP arrive as bool and time.Time.ClickHouseA tasks table on MergeTree and four queries; parameters are {name:Type} placeholders, the caller picks each id, and completing a task is an ALTER TABLE UPDATE mutation.