Reduce data overload by letting frontend specify what fields it wants

[SpaceK33z]

At the moment Binder will return all fields of a model when doing a GET request. In a recent project, we have models with over 50+ fields (I wish I was exaggerating). Often we need to fetch some relations of that model as well, which also have around 30 fields. In the frontend we often only actually use some of these fields.

So I want some way of specifying which fields the API returns, e.g. api/animal/?field=name,kind. By default it would just work like before, no breaking change here.

GraphQL fixes this exact issue, but switching to it now would be a huge refactor. However, we should get some inspiration from this.

I have some suggestions:

Mimicking GraphQL

An example query in GraphQL:

{
  hero {
    name
    friends {
      name
    }
  }
}

GraphQL doesn't use entity url's like api/hero but instead one url, /graphql. That would be too much of a switch so I suggest ignoring the root model (hero) for us is best;

name
friends {
  name
}

Since we need to pass this is in a GET parameter, it would need to be URL encoded. We can also strip all whitespace: .replace(/\s+/g, ' ').
The final query to the backend would be: api/hero/?field=name%20friends%20%7B%20name%20%7D

This syntax also allows filtering (note that the filter key "name_startswith" is not part of the spec, it's just something I made up):

friends(name_startswith: "Henk") {
  name
}

Making up our own syntax

Something like this perhaps:

api/hero/?field=name,friends(name)

Note that ?with=friends wouldn't be necessary here; this is implied. Also every relation should always include id. As a last point this should remove the need for m2m_fields (so it finally fixes #14).