MikroORM

Similarities

Both Joist and Mikro use the unit-of-work pattern, a per-request cache that simplifies DX.

Both Joist and Mikro use class entities to hold your domain logic.

Both Joist and Mikro have similar EntityManager APIs (Joist was inspired by Mikro).

Differences

Joist is newer, simpler, & more opinionated.

Mikro is an older project, into it’s V6, and has accumulated “several ways of doing things”, i.e. ~2-3 approaches to config (decorators vs. EntitySchema), two approaches to relations (with & without Refs), opt-in (off by default) dataloader support, optional Repository pattern, etc—where as Joist tries to be more opinionated and have just the single “best” way.

Joist’s relations are always type-safe/load-safe, i.e. always lazy & must be populate-d, which is marked in the type system; Mikro has this feature (inspired by Joist), but has kept its legacy “direct entity” approach, similar to TypeORM, which can lead to very frustrating errors when accessing relations that are not yet loaded.

Mikro can use decorators to define entities (which can get clunky, see their config ocs); Joist generally considers decorators an anti-pattern.

Mikro supports many more databases, including no-SQL like MongoDB; Joist only supports Postgres.

Mikro has a low-level query builder (for SUM, GROUP BY, etc); Joist has users drop-down to Knex (or whatever low-level query builder they prefer). Joist will probably have a low-level query builder someday, but with 90-95% of our queries going through em.find, we’ve not needed to prioritize it yet.

Joist’s EntityManager API has several powerful methods like findOrCreate and findWithNewOrChanged that do both in-database & in-memory lookups (i.e. if you’ve already made a new User with name Bob, findOrCreate will “find it” in-memory and not make “a 2nd bob” instance).

Joist can easily be run with tsx; Mikro’s decorators require a tsc-based tool like ts-node.

Joist’s em.find API supports condition pruning for ergonomic filter/listing endpoints.

Joist has first-class support for GraphQL, including ever-green schema scaffolding.

Joist has out-of-the-box factories for pleasant, idiomatic tests that scale to 10,000s of tests.