Relationship directives in table defs

[jrf0110]

The purpose of this is to automatically create junction tables if necessary and reason about the relationship between tables. If for instance we have a hasOne relationship, we can safely say get the corresponding records in the the hasOne tables on each query. That is, add a join to whatever queries we're performing. In addition, if there are updates to be made, we can treat a document, say a user with user groups embedded in it, as atomic and when the users array changes, we know how to update related tables.

dirac.register({
  name: 'users'
, hasMany: [ 'groups' ]
, hasOne: [ 'profile' ]
, schema: {
    /* ... */
  }
});

We can automatically create junction tables for hasMany relationships like this:

{
  table: 'books'
, schema: { id, name }
}

{
  table: 'authors'
, schema: { id, name }
}

{
  table: 'books_authors'
, schema: { book_id, author_id }
}

In the case of manyToMany, the junction table schema is the same. We'll just need to detect when a many-to-many relationship exists and only create one version. That is, do not create both books_authors and authors_books when authors specifies hasMany: [ 'books' ] and books specifies `hasMany: [ 'authors' ]``

[jrf0110]

In addition, if a table has as hasOne relationship, we can safely assume that something like this:

dirac.register({
  name: 'users'
, hasMany: [ 'groups' ]
, hasOne: [ 'profile' ]
, schema: {
    /* ... */
  }
});

dirac.insert({
  email: 'bob@bob.com'
, password: 'password'
, profile: {
    firstName: 'Bob'
  , lastName: 'Bobbo'
  }
});

So, we can extract the profile object because it matches the hasOne relationship, and do two inserts in a transaction. One for the user and for the profile.