Your first query

On DuckDB:

• $N placeholders
• No auto-incrementing key: the id column defaults to nextval('task_ids') from a CREATE SEQUENCE
• BIGINT maps to int64 and INTEGER to int32
• BOOLEAN maps to bool and TIMESTAMP to time.Time
• DuckDB runs in the program's process through cgo, with an in-memory database by default
• Migrations run with the migrate_backend_duckdb.New() backend, which takes no lock

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 SEQUENCE task_ids;

CREATE TABLE tasks (
    id         BIGINT    PRIMARY KEY DEFAULT nextval('task_ids'),
    title      VARCHAR   NOT NULL,
    completed  BOOLEAN   NOT NULL DEFAULT false,
    created_at TIMESTAMP 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;

Adding a task uses RETURNING, so the new row comes straight back and the command is one:

db/queries/tasks.sql

-- sgl.query(CreateTask, one)
INSERT INTO tasks (title) VALUES ($1) RETURNING id, title, completed, created_at;

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 = true WHERE id = $1;

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 bool      `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/duckdb/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.
//
// DuckDB has no auto-incrementing INTEGER PRIMARY KEY, so the migration draws each id
// from a sequence with DEFAULT nextval('task_ids'), and RETURNING hands it back.
//
// 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)

	feed, err := queries.CreateTask(ctx, "Feed the seagulls")
	if err != nil {
		return fmt.Errorf("creating a task: %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.ID)
	if err != nil {
		return fmt.Errorf("completing task %d: %w", feed.ID, 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=%t\n", task.ID, task.Title, task.Completed)
	}
	return nil
}

What it prints

completed 1 task
1	Feed the seagulls  completed=true
2	Count the chips    completed=false

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.

	DuckDB	SQLite	PostgreSQL	CockroachDB	MySQL	MariaDB	ClickHouse
id column	int64	int64	int64	int64	·	int64	·
title column	string	string	string	string	·	string	·
completed column	bool	int64	bool	bool	·	int32	·
created_at column	time.Time	string	time.Time	time.Time	·	time.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, MySQL, MariaDB and ClickHouse.

Run it locally

From the directory you cloned the Seagull repository into:

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

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

cd seagull/examples/duckdb && go test ./001_first_query

Needs cgo and a C compiler (the DuckDB driver).

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: duckdb
    migrations: db/migrations
    queries: db/queries
    go:
      package: store
      out: internal/store

db/migrations/0001_init.up.sql

CREATE SEQUENCE task_ids;

CREATE TABLE tasks (
    id         BIGINT    PRIMARY KEY DEFAULT nextval('task_ids'),
    title      VARCHAR   NOT NULL,
    completed  BOOLEAN   NOT NULL DEFAULT false,
    created_at TIMESTAMP NOT NULL DEFAULT current_timestamp
);

db/migrations/0001_init.down.sql

DROP TABLE tasks;
DROP SEQUENCE task_ids;

db/queries/tasks.sql

-- sgl.query(CreateTask, one)
INSERT INTO tasks (title) VALUES ($1) RETURNING id, title, completed, created_at;

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

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

example.go

package main

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

	"seagull.sh/seagull/examples/duckdb/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.
//
// DuckDB has no auto-incrementing INTEGER PRIMARY KEY, so the migration draws each id
// from a sequence with DEFAULT nextval('task_ids'), and RETURNING hands it back.
//
// 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)

	feed, err := queries.CreateTask(ctx, "Feed the seagulls")
	if err != nil {
		return fmt.Errorf("creating a task: %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.ID)
	if err != nil {
		return fmt.Errorf("completing task %d: %w", feed.ID, 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=%t\n", task.ID, task.Title, task.Completed)
	}
	return nil
}

main.go

// Command 001_first_query applies its migrations to an in-memory DuckDB database, then
// runs the example in example.go: two tasks are added, one is completed, and both are
// listed.
//
// Run it with `go run .`, or `go run . -db example.duckdb` to keep the database in a
// file.
package main

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

	_ "github.com/duckdb/duckdb-go/v2"

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

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

func main() {
	path := flag.String("db", "", "DuckDB database file; an in-memory database when empty")
	flag.Parse()
	if err := execute(context.Background(), *path); err != nil {
		log.Fatal(err)
	}
}

// execute opens and migrates the DuckDB database at path, or an in-memory one when path
// is empty, and runs the example against it.
//
// Takes path (string) which is the database file, or empty.
//
// Returns error when the database cannot be prepared or the example fails.
func execute(ctx context.Context, path string) error {
	database, err := sql.Open("duckdb", path)
	if err != nil {
		return err
	}
	defer func() { _ = database.Close() }()

	migrator, err := migrate.NewMigrator(database, migrate_backend_duckdb.New(), 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 bool      `json:"completed"`
	CreatedAt time.Time `json:"created_at"`
}

internal/store/querier.go

// Code generated by sgl. DO NOT EDIT.

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"
	"time"
)

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

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 ($1) RETURNING id, title, completed, created_at;`

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

func (queries *Queries) CreateTask(ctx context.Context, title string) (CreateTaskRow, error) {
	var row CreateTaskRow
	err := queries.writer.QueryRowContext(ctx, createTaskSQL, title).Scan(&row.ID, &row.Title, &row.Completed, &row.CreatedAt)
	if err != nil {
		return CreateTaskRow{}, err
	}
	return row, nil
}

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 bool      `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=true
2	Count the chips    completed=false