/Tequel

Primary LanguageTypeScript

Tequel

What is Tequel?

Tequel is an experimental compile-time schema-aware raw SQL query parser for Typescript.

What does this mean?

Tequel brings type safety to writing raw SQL queries.

This is best shown by an example:

// Define a schema: table name, columns names and types
type PersonSchema = {
    columns: {
        name: 'varchar',
        age: 'integer'
    },
    table: 'people',
}
const tql = new Tequel.PG<[PersonSchema]>();

// For legal queries, tql is aware of the returned type: 
const queryResults = await tql.query(`SELECT name FROM people`); // Returned type: { name: string }[]
const names = queryResults.map(result => result.name); 
const ages = queryResults.map(result => result.age); // TS compile error- "age" was not selected

// For illegal queries (selecting a column that doesn't exist), tql is aware of the mistake in the query:
const badQueryResults = await tql.query(`SELECT address FROM people`); 
const addresses = badQueryResults.map(result => result.address); 
// TS Compile error - badQueryResult's type is TequelParseError, not an array of results.
// TequelParserError shows you where the problem is. The exact type of badQueryResults is:
//      TequelParseError<"Query requested column address which does not exist in requested tables">
Definition breakdown:
  • Query parser: Tequel validates your query and returns an error or the type of your query's result set.
  • Raw SQL: Tequel has no query builder or DSL. It accepts raw SQL queries as a string as its input.
  • Schema-aware: Tequel is aware of table schemas that were passed to it. This means that Tequel doesn't just validate that your query has a legal SQL syntax, but it also makes sure that your query matches your DB schema. Tequel will catch things like typos in a column name.
  • Compile-time: All this happens during Typescript compilation. This means you know whether your query will run during compile time. Another use case - You can also change a column name in your DB and immediately knowing which queries need to be refactored.
  • Experimental: This project is not intended for production usage and there is no guarantee that it will be feature-complete. Currently the project is very minimal and implementing more features might expose insurmountable design flaws or issues from the underlying Typescript compilation engine. This library's external API will probably change a lot as it is developed.

Motivation

In the current Node.js ecosystem, there are 3 options to developing the application <> database interface:

  1. Using an ORM and querying with one of its built in query builders - this is type safe but requires learning a new query language which is usually not as flexible or powerful as SQL
  2. Using an ORM and querying with its built in "raw query" functionality - this is not type safe (usually allows type hinting, but mistakes are on you)
  3. Not using an ORM - this is not type safe

You have to choose between compile-time type safety or learning the ORM's query language. This library has 2 goals:

  1. Compile-time schema-aware raw SQL type safety
  2. Minimal learning curve for developers who already know SQL

In its current design this library requires learning how to define a schema. However, the schema definition DSL is minimal and uses the same column names and types as the your DB's DDL.

Future prioritization

This library will prioritize main use cases of the SQL syntax before complete implementation of each feature. This roughly means a breadth-first implementation of the SQL specification. As an example, both "INSERT" and "SELECT" will be partially implemented before all of the specification for an "INSERT" statement.

Caveats

  • This library requires Typescript 4.1 as it heavily relies on template literal types.
  • While I believe that this use case has merit for real world usage, it is not certain that this implementation is feasible. This project might be abandoned.
  • This library is the maintainer's first attempt at writing a production ready parser - critique is deeply appreciated.
  • The SQL specification is very large and will probably not be fully implemented in this library's parser.