An opinionated ORM for TypeScript/node/postgres.

Goals

• Schema-driven code generation (continually-generated classes w/the getter/setter/relation boilerplate)
• Guaranteed N+1 safe (pervasive use of Facebook's dataloader
• All relations are async/await (with an ergonomic, type-safe escape hatch)
• Great performance (all SELECT/INSERT/UPDATE operations are bulk)
• Fast tests (for downstream projects, baseline is 10-20ms/test case)
• Unit of Work (navigate between entities as a consistent graph)

Documentation

See joist-orm.io for documentation.

Building Joist

For contributing to Joist itself, after checkout:

• Run yarn install
• Run yarn build to compile all packages
• Run yarn db to boot up a Docker postgres instance w/the integration test schema.
• Run yarn test to run the tests.
• Prior to committing your changes, run yarn workspaces run format

The build runs two complementary TypeScript checks after tsdown emits the package artifacts:

• yarn typecheck:packages checks each public package's source with its ESM-first NodeNext configuration. It catches source errors that tsdown's isolated transforms do not report and runs after the build so workspace package imports can resolve the emitted declarations.
• yarn typecheck:consumers checks the unit-test project and every joist-tests-* workspace against the emitted package exports. This covers the ESM and CommonJS declaration paths, generated module augmentations, and other issues that only appear when Joist is consumed as a package.

yarn start runs tsdown and TypeScript in watch mode as separate mprocs processes. build:watch updates the ESM and CommonJS artifacts, while typecheck:watch incrementally checks the complete package and consumer project graph. The typecheck watcher emits declarations because TypeScript project-reference builds cannot use --noEmit; build:watch does not clean them while both processes are running, and the next normal build removes them.

License

MIT