sqlf

A fast SQL query builder for Go.

What sqlf does?

• It helps you efficiently build an SQL statement in run-time.
• You may change the number of affected columns and change the number of arguments in a safe way.
• You may use SQL expressions (like UPDATE counters SET counter = counter + 1) in your SQL statements.
• You may dynamically apply filters by adding where conditions, change result ordering, etc.
• You may safely use ? placeholders in your SQL fragments - sqlf converts them to PostgreSQL-like $1, $2, ... placeholders if needed and does the numbering for you.
• You may .Bind your structure to database columns like you do with other similar libraries.
• sqlf.Stmt has methods to execute a query using any database/sql compatible driver.

What sqlf doesn't?

• sqlf isn't an ORM, you'll still have to use raw SQL.
• There are no database schema migrations or any other database schema maintenance tools.
• There are no compile-time type checks for query arguments, column and table names.
• There is no wrapper for OR clause. It affects performance and in most cases can be avoided by using UNION expressions, WITH clause or window functions. Other option is to split a query into two.
• sqlf doesn't help a developer to pinpoint the cause of issue with SQL statement.

Is It Fast?

It is. See benchmarks: https://github.com/leporo/golang-sql-builder-benchmark

In order to maximize performance and minimize memory footprint, sqlf reuses memory allocated for query building. The heavier load is, the faster sqlf works.

Usage

Build complex statements:

var (
    region       string
    product      string
    productUnits int
    productSales float64
)

sqlf.SetDialect(sqlf.PostgreSQL)

err := sqlf.From("orders").
    With("regional_sales",
        sqlf.From("orders").
            Select("region, SUM(amount) AS total_sales").
            GroupBy("region")).
    With("top_regions",
        sqlf.From("regional_sales").
            Select("region").
            Where("total_sales > (SELECT SUM(total_sales)/10 FROM regional_sales)")).
    // Map query fields to variables
    Select("region").To(&region).
    Select("product").To(&product).
    Select("SUM(quantity)").To(&productUnits).
    Select("SUM(amount) AS product_sales").To(&productSales).
    //
    Where("region IN (SELECT region FROM top_regions)").
    GroupBy("region, product").
    OrderBy("product_sales DESC").
    // Execute the query
    QueryAndClose(ctx, db, func(row *sql.Rows){
        // Callback function is called for every returned row.
        // Row values are scanned automatically to bound variables.
        fmt.Printf("%s\t%s\t%d\t$%.2f\n", region, product, productUnits, productSales)
    })
if err != nil {
    panic(err)
}

Bind a structure:

type Offer struct {
    Id        int64   `db:"id"`
    ProductId int64   `db:"product_id"`
    Price     float64 `db:"price"`
    IsDeleted bool    `db:"is_deleted"`
}

var o Offer

err := sqlf.From("offers").
    Bind(&o).
    Where("id = ?", 42).
    QueryRowAndClose(ctx, db)
if err != nil {
    panic(err)
}

Retrieve data to private fields with more granular control on retrieved fields:

type Offer struct {
    id        int64
    productId int64
    price     float64
    isDeleted bool
}

var o Offer

err := sqlf.From("offers").
    Select("id").To(&o.id).
    Select("product_id").To(&o.productId).
    Select("price").To(&o.price).
    Select("is_deleted").To(&o.isDeleted).
    Where("id = ?", 42).
    QueryRowAndClose(ctx, db)
if err != nil {
    panic(err)
}

Some SQL fragments, like a list of fields to be selected or filtering condition may appear over and over. It can be annoying to repeat them or combine an SQL statement from chunks. Use sqlf.Stmt to construct a basic query and extend it for a case:

func (o *Offer) Select() *sqlf.Stmt {
    return sqlf.From("products").
        .Bind(o)
        // Ignore records, marked as deleted
        Where("is_deleted = false")
}

func (o Offer) Print() {
    fmt.Printf("%d\t%s\t$%.2f\n", o.id, o.name, o.price)
}

var o Offer

// Fetch offer data
err := o.Select().
    Where("id = ?", offerId).
    QueryRowAndClose(ctx, db)
if err != nil {
    panic(err)
}
o.Print()
// ...

// Select and print 5 most recently placed
// offers for a given product
err = o.Select().
    Where("product_id = ?", productId).
    OrderBy("id DESC").
    Limit(5).
    QueryAndClose(ctx, db, func(row *sql.Rows){
        o.Print()
    })
if err != nil {
    panic(err)
}
// ...

SQL Statement Construction and Execution

SELECT

Value Binding

Bind columns to values using To method:

var (
    minAmountRequested = true
    maxAmount float64
    minAmount float64
)

q := sqlf.From("offers").
    Select("MAX(amount)").To(&maxAmount).
    Where("is_deleted = false")

if minAmountRequested {
    q.Select("MIN(amount)").To(&minAmount)
}

err := q.QueryRowAndClose(ctx, db)
if err != nil {
    panic(err)
}
if minAmountRequested {
    fmt.Printf("Cheapest offer: $%.2f\n", minAmount)
}
fmt.Printf("Most expensive offer: $%.2f\n", minAmount)

Joins

There are helper methods to construct a JOIN clause: Join, LeftJoin, RightJoin and FullJoin.

var (
    offerId     int64
    productName string
    price       float64
}

err := sqlf.From("offers o").
    Select("o.id").To(&offerId).
    Select("price").To(&price).
    Where("is_deleted = false").
    // Join
    LeftJoin("products p", "p.id = o.product_id").
    // Bind a column from joined table to variable
    Select("p.name").To(&productName).
    // Print top 10 offers
    OrderBy("price DEST").
    Limit(10).
    QueryAndClose(ctx, db, func(row *sql.Rows){
        fmt.Printf("%d\t%s\t$%.2f\n", offerId, productName, price)
    })
if err != nil {
    panic(err)
}

Use plain SQL for more fancy cases:

var (
    num   int64
    name  string
    value string
)
err := sqlf.From("t1 CROSS JOIN t2 ON t1.num = t2.num AND t2.value IN (?, ?)", "xxx", "yyy").
    Select("t1.num").To(&num).
    Select("t1.name").To(&name).
    Select("t2.value").To(&value).
    QueryAndClose(ctx, db, func(row *sql.Rows){
        fmt.Printf("%d\t%s\ts\n", num, name, value)
    })
if err != nil {
    panic(err)
}

Subqueries

Use SubQuery method to add a sub query to a statement:

	q := sqlf.From("orders o").
		Select("date, region").
		SubQuery("(", ") AS prev_order_date",
			sqlf.From("orders po").
				Select("date").
				Where("region = o.region").
				Where("id < o.id").
				OrderBy("id DESC").
				Clause("LIMIT 1")).
		Where("date > CURRENT_DATE - interval '1 day'").
		OrderBy("id DESC")
	fmt.Println(q.String())
	q.Close()

Note that if a subquery uses no arguments, it's more effective to add it as SQL fragment:

	q := sqlf.From("orders o").
		Select("date, region").
		Where("date > CURRENT_DATE - interval '1 day'").
        Where("exists (SELECT 1 FROM orders po WHERE region = o.region AND id < o.id ORDER BY id DESC LIMIT 1)").
        OrderBy("id DESC")
    // ...
    q.Close()

To select from sub-query pass an empty string to From and immediately call a SubQuery method.

The query constructed by the following example returns top 5 news in each section:

	q := sqlf.Select("").
		From("").
		SubQuery(
			"(", ") counted_news",
			sqlf.From("news").
				Select("id, section, header, score").
				Select("row_number() OVER (PARTITION BY section ORDER BY score DESC) AS rating_in_section").
				OrderBy("section, rating_in_section")).
		Where("rating_in_section <= 5")
    // ...
    q.Close()

Unions

Use Union method to combine results of two queries:

	q := sqlf.From("tasks").
		Select("id, status").
		Where("status = ?", "new").
		Union(true, sqlf.PostgreSQL.From("tasks").
			Select("id, status").
            Where("status = ?", "wip"))
    // ...
	q.Close()

INSERT

sqlf provides a Set method to be used both for UPDATE and INSERT statements:

var userId int64

_, err := sqlf.InsertInto("users").
    Set("email", "new@email.com").
    Set("address", "320 Some Avenue, Somewhereville, GA, US").
    Returning("id").To(&userId).
    Clause("ON CONFLICT (email) DO UPDATE SET address = users.address").
    QueryRowAndClose(ctx, db)

The same statement execution using the database/sql standard library looks like this:

var userId int64

// database/sql
err := db.ExecContext(ctx, "INSERT INTO users (email, address) VALUES ($1, $2) RETURNING id ON CONFLICT (email) DO UPDATE SET address = users.address", "new@email.com", "320 Some Avenue, Somewhereville, GA, US").Scan(&userId)

There are just 2 fields of a new database record to be populated, and yet it takes some time to figure out what columns are being updated and what values are to be assigned to them.

In real-world cases there are tens of fields. On any update both the list of field names and the list of values, passed to ExecContext method, have to to be reviewed and updated. It's a common thing to have values misplaced.

The use of Set method to maintain a field-value map is a way to solve this issue.

Bulk Insert

To insert a multiple rows via a single query, use NewRow method:

_, err := sqlf.InsertInto("users").
    NewRow().
        Set("email", "first@email.com").
        Set("address", "320 Some Avenue, Somewhereville, GA, US").
    NewRow().
        Set("email", "second@email.com").
        Set("address", "320 Some Avenue, Somewhereville, GA, US").
    ExecAndClose(ctx, db)

UPDATE

_, err := sqlf.Update("users").
    Set("email", "new@email.com").
    ExecAndClose(ctx, db)

DELETE

_, err := sqlf.DeleteFrom("products").
    Where("id = ?", 42)
    ExecAndClose(ctx, db)