Documentation

QuasiQuoter entry point for typed SQL. Disallows SELECT * and SELECT table.* by default to prevent production errors when the schema changes. Use typedSqlStar to opt in to star selects.

Like typedSql but allows SELECT * and SELECT table.* patterns. Use this when you understand that star selects can break at runtime if the schema changes between compilation and deployment.

What the SQL parser can prove about how many rows a query returns.

What kind of result an SQL statement has.

This marker is separate from QueryCardinality because statements without result columns can still have different execution results: DML statements report affected rows, while utility statements such as SET CONSTRAINTS do not report a row count.

Prepared query with a custom row parser. High-level: this is the runtime value produced by the typed SQL quasiquoter.

The result shape returned by sqlQueryTyped.

The result returned by sqlExecTyped.

Run a typed SELECT query.

Also works with INSERT/UPDATE/DELETE ... RETURNING statements that return rows.

The return type is inferred from the query cardinality:

• many rows: [result]
• at most one row: Maybe result
• exactly one row: result

users <- sqlQueryTyped [typedSql| SELECT name FROM users |] -- IO [Text]
total <- sqlQueryTyped [typedSql| SELECT count(*) FROM users |] -- IO Int64

Run a typed query that can return many rows.

This is equivalent to sqlQueryTyped, but fixes the expected cardinality in the function name. It can make type errors easier to read when migrating code from the old list-shaped sqlQueryTyped result.

Run a typed query that can return at most one row.

This is equivalent to sqlQueryTyped, but fixes the expected cardinality in the function name.

Run a typed query that must return exactly one row.

This is equivalent to sqlQueryTyped, but fixes the expected cardinality in the function name.

Run an at-most-one-row query selecting a nullable single column and flatten the two independent failure modes into one Maybe.

Useful for queries like:

email <- sqlQueryTypedMaybeColumn [typedSql|
    SELECT optional_email FROM users WHERE id = ${userId}
|]

Without this helper, the precise sqlQueryTyped result is Maybe (Maybe Text): the outer Maybe is "no row", the inner Maybe is "the selected column was NULL".

Pipeline variant of sqlQueryTyped.

Compose this with pipeline to run independent typed SQL queries in one PostgreSQL pipeline batch.

Pipeline variant of sqlQueryTypedMaybeColumn.

Run a typed statement.

Use sqlQueryTyped instead if your statement has a RETURNING clause. Known utility statements without a row-count result, such as SET CONSTRAINTS, are run with Hasql's no-result decoder and return ().

rowsAffected <- sqlExecTyped [typedSql| DELETE FROM items WHERE id = ${itemId} |]
sqlExecTyped [typedSql| SET CONSTRAINTS ALL DEFERRED |] -- IO ()