add the ability to have some "undroppable" initial data

[Luca Barone]

In my current custom fake data implementation, I have the ability to seed the database with some initial data, that will remain untouched across drops:

const initialData: DatabaseData = Object.freeze({
  todos: [...],
  users: [...],
})

let data = initialData

/**
 * We need this mechanism as we need to simulate a "database"
 * reset for every test run in order to maintain a clean
 * "database" state for every test.
 *
 * @note it is crucial to not mutate the internal structure of
 * the initialData, in order to achieve this.
 */
export const resetData = () => {
  data = initialData
}

export const getData = () => data

export const setData = (transformer: (data: DatabaseData) => DatabaseData) => {
  data = transformer(data)
}

Currently using @mswjs/data, if I reset the db with drop(db), I lose ALL the data, and I think is not practical to reseed the data on each test 🤔

[Artem Zakharchenko]

Hey, @cloud-walker.

Please note that drop() is not meant for resetting the database state. It's designed to drop all the entities in the database.

If you wish to reset the state, consider combining drop with your seeding function:

import { drop } from '@mswjs/data'

function seed(db) {
  db.user.create({...})
  db.todo.create({...})
}

function resetDb(db) {
  drop(db)
  seed(db)
}

Would this approach work for your use case?

[Artem Zakharchenko]

We've also discussed the reset API here, where the following API was suggested:

import { factory, snapshot } from '@mswjs/data'

const db = factory({...})
// seed your db...
db.user.create(...)

// Take the snapshot of database entities
// in this point of time.
const reset = snapshot(db)

// change the entities...
db.user.update(...)
db.todo.create(...)

// Reset to the exact state when
// the snapshot was taken.
reset()

There's been no investigating whether such API is possible with the current architecture, but if we ever adopt a reset API, it'd look similar to this.

[Luca Barone]

Thats how I solved the issue with my current @mswjs/data experiment. But I think it's a waste of computation to do a reseed afterEach test.

BTW the snapshot proposal could solve the use case!

[Artem Zakharchenko]

Closing this in favor of #86.