Make ORMs great again!
record.update(data)
record.isNewRecord
It's now possible to create custom relations
e.g.
// models/User.js
this.has('has_posts_written', {
query: function(store, parentRecords){
const ids = parentRecords.map(r => r.id)
const Post = store.Model('Post')
// query all posts by user_id, group by user_id and count(distinct(id))
return Post.totalCount().group('user_id').where({user_id: ids})
},
convert: function(parent, records){
if(!records) return false
// records => [{user_id: 1, count: 4}, {user_id: 2, count: 1}]
// == the result of the above query!
const result = records.find(r => r.user_id === parent.id)
if(!result) return false
return result.count > 0
}
})
A custom relation could return anything. In the example above it'll return a boolean value.
Works with include()
like any other relation:
User.include('has_posts_written')
The postgres store now supports all geometric data types.
There are also two new store options externalAttributeName
and internalAttributeName
for all stores.
Both options are optional and expect a method in the form of function(fieldName: string): string
With externalAttributeName
you can convert your internal database field names into another format. e.g. from snake_case to camelCase. The conversion is up to you!
internalAttributeName
should handle the opposite way.
Relations got rewritten with optional bulk loading:
const user = await User.limit(3)
const posts = await Promise.all(users.map(user => user.posts))
will only execute 2 queries! To disable bulk loading see bulkFetch.
However, in this case it's better to use const user = await User.include('posts').limit(3)
include was rewritten to load relations in parallel, if possible (SQL only!)
const users = await User.find(1).include('posts')
/*
will execute `SELECT * FROM users WHERE id=1`
and `SELECT * FROM posts WHERE user_id = 1` in parallel
*/
In V1 the above example would have loaded the data in series.
Of course this is only possible if the conditions for the first query will contain all information needed to execute the second query!
Similar to autoloading model attributes, it's now possible to autoload models.
const Store = require('openrecord/store/sqlite3')
const store = new Store({
file: './my-posts-db.sqlite3',
autoLoad: true // enable autoload
})
store.ready(async () => {
const post = await store.Model('Post').find(1) // Post model is automatically available, if a table `posts` is defined in the sqlite3 database before.
console.log(post)
})
Defining a model via ES6 classes is now possible
class User extends Store.BaseModel{
fullName(){
return `${this.first_name} ${this.last_name}`
}
}
Support for GraphQL with automatic relation loading and more.
It's now possible to bundle your store via webpack (Version 3 and 4) There is also a Webpack Plugin to cache your data structure inside your bundle. (Faster startup for serverless apps)
It's now possible to define custom operators and use the whole power of knex
// the new operator is called `regexp`
store.addOperator('regexp', function(field, value, query, condition){
query.where(field, '~', value.toString().replace(/(^\/|\/$)/g, '')) // naiv conversion of js regexp to postgres regexp!
})
// and it will be appended to the `string` type
store.appendOperator('string', 'regexp')
Query via:
const user = await User.where({login_regexp: /open.*/})
The whole core was rewritten to use Promises instead of async.
Docs are now available via https://openrecord.js.org
plugins
and models
store config does not take paths anymore. To get the old behavior back, use the automatic model loading plugin
paranoid
plugin scope to get all records was renamed to withDeleted
instead of with_deleted
join()
does an inner join by default (instead of a left join)save
, delete
, create
,... won't return success
anymore. Instead it will return the record on successdone
callback was removed.create
, save
, destroy
, ... won't take callbacks any more. use e.g. record.save().then(callback)
true
(store or per relation)user.posts
will return a then-able object. To access loaded data directly use user._posts
(Will return null if not loaded)limit(1)
does not return a single record anymore. Use first()
or singleRecord()
insteadlogger
option on store is no longer available. openrecord now uses the debug
module
I've added the ability to run raw joins via .join('JOIN foo ON...')
or via .join(['JOIN foo ON foo.id=? ...', ['bar']])
Thanks to @arthurfranca
Some small bugfixes, dependency upgrades and license change to MIT