hedy - a functional ORM

Let's face it: Node's ORMs are crap. Mostly at least. But JavaScript is an awesome language. So why? Because we did it wrong. To much class-ish stuff like backbone. And prototypal inheritance, which is bad.

We want to fix this!

Initialisation

const hedy = require('hedy')
const mem = require('hedy/adapter/mem')

const store = hedy(adapter)

The adapter is a function that is called when a query is fired. It gets the query options-object as parameter. This is a database-specific function. In this case we user the build in memory adapter. In the future we want to create adapter for various databases. The goal is that it's really easy to create one for you specific datastore.

See the existing adapters how this works.

Creating Models/Collections/Whatever-you-call-them

..., we just call it query.

const userQuery = store('user') // 'user' is the tablename

The query is a monad-ish structure. You can call methods on it and it returns another query with the method applied. The original query remains untouched. Unter the hood this is done using patchinko.

Fetch array of things

// Array of POJOs containing user data
const users = await userQuery.where(where).load()

The where can be omitted, in this case all items where fetched. The query will only run, if you call load on it.

There are also some utility functions that might be usefull. They can be called on the query without actually fetching it.

// Array of POJOs containing user data
const users = await userQuery
  .where(where)
  .map(fn1)
  .filter(fn2)
  .load()
})

The map and filter will be applied onto the fetched collection in the order as they are applied to the query. They may return a promise.

Because all that is lazy you might do the following:

const usersWithLongNames = userQuery.filter(user => user.name.length > 10)

const femaleUsersWithLongNames = usersWithLongNames.filter(
  user => user.gender === 'female'
)

const kidsWithLongNames = usersWithLongNames.filter(user => user.age < 10)

const namesOfKidsWithLongNames = kidsWithLongNames
  .columns(['name'])
  .map(user => user.name)

// if you now need the names of the kids:
// there you have it.
const names = await namesOfKidsWithLongNames.load()

Sure, in some of those cases the database might do the heavy lifting, but there are cases where code can express much more then a database query. Also code reuse and composition can lead to great improvements here.

Currently there are only a few methods build in: map, reduce, groupBy, indexBy. It's easy to add your own method there. In the upper example lodashs pluck might be a good choice to get the user name:

const store = hedy(adapter, {
  methods: {
    pluck: require('lodash/collection/pluck'),
  },
})

In this case we add the pluck-function of lodash. Now we can do

const usernames = await userQuery.pluck('name').load()
// usernames = ['heiner', 'klaus', 'birgit']

Fetch one thing by pk

const user = await userQuery.get(id)
// user = { id: 1, name: 'Heiner' }

Fetch one first

const user = await userQuery.where(where).first()
// user = { id: 1, name: 'Heiner' }

select columns (with aliasing)

const user = await userQuery
  .columns({ username: 'name' })
  .where(where)
  .get(1)
// user = { id: 1, username: 'Heiner' }

Counting

const count = await userQuery.count()
// count = 5

Create one thing

const user = await userQuery.post(data)
// user = { id: 1, name: 'Heiner' }

Update one thing

const user = await userQuery.put(id, data)
// user = { id: 1, name: 'Heiner' }

You can also patch stuff:

const user = await userQuery.patch(id, data)
// user = { id: 1, name: 'Heiner' }

delete one thing

await userQuery.del(id)

Relations

Relations are a key part of ORMs. In most ORMs relations can only be in the same database. Hedy as a different approach on this. Relations are defined as querys. Let's look at an example:

const data = {
  user: [
    { id: 1, name: 'heiner' },
    { id: 2, name: 'klaus' },
    { id: 3, name: 'manfred' },
  ],
  friend: [{ user1Id: 1, user2Id: 2 }, { user1Id: 2, user2Id: 3 }],
  comment: [
    { id: 1, userId: 2, text: 'gorgeous' },
    { id: 2, userId: 3, text: 'nice' },
    { id: 4, userId: 1, text: 'splended' },
    { id: 5, userId: 2, text: 'awesome' },
  ],
}
const adapter = memAdapter(data)
const store = hedy(adapter)

const userQuery = store('user')
const commentQuery = store('comment')
const friendQuery = store('friend')

Here we have a memory db containing users and their comments.

To fetch the users with the comments we do:

userQuery.withRelated(hedy.hasMany(commentQuery)).then(users => ...)

As you see, we use a helper to declare a to-many-relation and give a query as parameter. The query does not have to request to the same database, so this is perfectly possible.

const memQuery = hedy(memAdapter(data))
const pgQuery = hedy(pgAdapter(config))

const userQuery = memQuery('user')
const commentQuery = pgQuery('comment')

userQuery.withRelated(hedy.hasMany(commentQuery)).then(users => ...)

For a more advanced example see the examples in the examples folder.

Has-many

const users = await userQuery.withRelated(hedy.hasMany(commentQuery)).load()

Has-one

TODO Add example

Belongs-to

const comments = await commentQuery
  .withRelated(hedy.belongsTo(userQuery))
  .load()

Many-to-many

const users = await user
  .withRelated(hedy.hasManyThrough(userQuery, friendQuery))
  .load()

For many-to-many-Relations there are two helper functions to create and delete them:

const friendship = await friends.link(userA, userB)
await friends.unlink(userA, userB)

Current state

This is pretty much WIP. The basics are done. Now it's time to write the adapters. First one will be a memory-adapter. Then we first add a postgres adapter since this is the DB we're using in our project.

Hope you like it.

If you have any ideas, feel free to create a PR/Issue.

Hedy

Install / Use

README