@ronin/compiler

RONIN Compiler

This package compiles RONIN queries to SQL statements.

Setup

You don't need to install this package explicitly, as it is already included in the RONIN client.

However, we would be excited to welcome your feature suggestions or bug fixes for the RONIN compiler. Read on to learn more about how to suggest changes.

Contributing

To start contributing code, first make sure you have Bun installed, which is a JavaScript runtime.

Next, clone the repo and install its dependencies:

bun install

Once that's done, link the package to make it available to all of your local projects:

bun link

Inside your project, you can then run the following command, which is similar to bun add @ronin/compiler or npm install @ronin/compiler, except that it doesn't install @ronin/compiler from npm, but instead uses your local clone of the package:

bun link @ronin/compiler

If your project is not yet compatible with Bun, feel free to replace all of the occurrences of the word bun in the commands above with npm instead.

You will just need to make sure that, once you create a pull request on the current repo, it will not contain a package-lock.json file, which is usually generated by npm. Instead, we're using the bun.lockb file for this purpose (locking sub dependencies to a certain version).

Developing

The programmatic API of the RONIN compiler looks like this:

import { Transaction } from '@ronin/compiler';

const transaction = new Transaction([
  {
    create: { model: { slug: 'account' } }
  },
  {
    get: { accounts: null }
  }
]);

transaction.statements;
// [{
//   statement: 'SELECT * FROM "accounts"',
//   params: [],
//   returning: true,
// }]

transaction.models;
// [{
//   name: 'Account',
//   slug: 'account'
//   ...
// }]

Types

In total, the following types are being exported:

import type {
  Query,

  Model,
  ModelField,
  ModelIndex,
  ModelTrigger,
  ModelPreset,

  Statement,
  Result
} from '@ronin/compiler';

Options

To fine-tune the behavior of the compiler, you can pass the following options:

new Transaction(queries, {
  // A list of models that already existing inside the database.
  models: [
    { slug: 'account' }
  ],

  // Instead of returning an array of parameters for every statement (which allows for
  // preventing SQL injections), all parameters are inlined directly into the SQL strings.
  // This option should only be used if the generated SQL will be manually verified.
  inlineParams: true
});

Transpilation

In order to be compatible with a wide range of projects, the source code of the compiler repo needs to be compiled (transpiled) whenever you make changes to it. To automate this, you can keep this command running in your terminal:

bun run dev

Whenever you make a change to the source code, it will then automatically be transpiled again.

Mental Model

The interface of creating new Transaction instances (thereby creating new transactions) was chosen in order to define the smallest workload unit that the compiler can operate on.

Just like for the database, a transaction for the compiler defines an atomic operation in which a list of queries can be executed serially, and where each query can rely on the changes made by a previous one. In order to facilitate this, a programmatic interface that clarifies the accumulation of in-memory state is required (class instances).

For example, if one query creates a new model, every query after it within the same transaction must be able to interact with the records of that model, or update the model itself, without roundtrips to the database, thereby requiring the accumulation of state while the transaction is being compiled.

Furthermore, since every database transaction causes a lock, the database is inherently not locked between the execution of multiple transactions, which could cause the compilation inputs (e.g. models) of a Transaction to no longer be up-to-date. If the inputs have changed, a new Transaction should therefore be created.

Running Tests

The RONIN compiler has 100% test coverage, which means that every single line of code is tested automatically, to ensure that any change to the source code doesn't cause a regression.

Before you create a pull request on the compiler repo, it is therefore advised to run those tests in order to ensure everything works as expected:

# Run all tests
bun test

# Alternatively, run a single test
bun test -t 'your test name'