Connecting

tsql keeps query building and query execution separate.

You provide:

• schema: output of defineDatabase(...)
• dialect: SQL placeholder and identifier behavior
• connection: your adapter with execute(sql, params)

import { connect, sqliteDialect, type SqlConnection } from "@bms/tsql";
import { schema } from "./schema";

const connection: SqlConnection = {
	execute: async (sql, params) => {
		// Call your runtime driver here.
		return { rows: [] };
	},
};

const db = connect({
	schema,
	dialect: sqliteDialect(),
	connection,
});

SqlConnection

type SqlConnection = {
	execute(
		sql: string,
		params: readonly unknown[],
	): Promise<{
		rows: readonly unknown[];
	}>;
};

execute always receives compiled SQL and bound parameters.

Effect integration

Use @bms/tsql/effect when running queries inside Effect programs. It is a separate export so applications use their own effect package version.

import { Effect } from "effect";
import { connect } from "@bms/tsql/effect";

const db = connect({ schema, dialect: sqliteDialect(), connection });

const users = await Effect.runPromise(
	db.run(db.from("users").select(({ users }) => ({ id: users.id }))),
);

See Effect for Effect-native connections and transaction helpers.

Database object

connect(...) returns Database<TSchema>. Main methods:

• from(...) for select
• insert(...), update(...), delete(...)
• raw and expr template helpers
• aggregate helpers: count, max, min, avg, sum
• transaction methods: begin, and on transactions commit / rollback
• migration runner: migrate