HAL embedding

[maxsz]

Hi, great work!

HAL supports embedding resources (https://github.com/mikekelly/hal_specification/blob/master/hal_specification.md) using the special _embedded keyword. This has the benefit of saving some network requests. Are you planning to support these?

[vicmosin]

Hi, thanks!
What exactly do you mean? All the response is deserialized either to JSON or Dictionary. From there you can retrieve it. Or your idea is to have some shortcut to _embedded collection?

[maxsz]

My first thought was, that a .follow(...).get { } combination could omit the actual network request, when the _embedded collection already contains the data from the link.

[vicmosin]

Ok, now I get it. But I wouldn't do. My believe is that traverson's duty is to do very simple thing - follow the links provided by server. Adding logics such as checking whether the embedded collection contains needed resource somewhere between traversing the links will create more issues. How will we identify such resource? What if embedded resource differs from actual one? How client will know whether the resource was retrieved from embedded or from real call? All this will complicate the traversing process as well as produce possible side effects.
If we are talking about performance, nobody said you can't use caching policies, Etag for example, so this unnecessary last request will be cached anyway.

[vicmosin]

But I'd like to hear some ideas of how it might be implemented... Even in pseudo code))
Thanks for your response

[systemfreund]

If i might chime in:
We actually are using _embedded resources, see here
https://developers.smoope.com/docs/rest-api#_example_response_4

However, our API responses will never contain a relation X in both sections (_links, _embedded) at the same time. Therefore there's no need for the traverson client to be smart about it.

Example, which you won't happen in our API:

{
   "_links": {
       "user" : {
           "href" : "https://api.smoope.com/users/a8c5438f-41ed-4bf8-8fc8-b8e187d2280b"
       },
   },

   "_embedded": {
       "user": {
           "_links": {
               "self" : {
                   "href" : "https://api.smoope.com/users/a8c5438f-41ed-4bf8-8fc8-b8e187d2280b"
               },
           },

           "name": "Gregor Gysi",
           [...]
       }
   }
}

But, I agree of course that the traverson client should consider the _embedded resources in the .follow() call. Example:

{
   "_embedded": {
       "user": {
           "_links": {
               "self" : {
                   "href" : "https://api.smoope.com/users/a8c5438f-41ed-4bf8-8fc8-b8e187d2280b"
               },
           },

           "name": "Gregor Gysi",
           [...]
       }
   }
}

Here resource.follow("user") should indeed directly return the user resource which is contained in the parent resource without doing any network calls.