I'm just getting the hang of graph.cool and apollo coming from reindex and relay, where if a model implemented a Node interface there was more ceremony around the mutations. Seems like that might not be the case here

Can you elaborate with what you mean by ceremony?

Well, how would you design the schema for a survey where you have a Survey model and an Answers model?

and the survey has questions, and each question has a

label

and

hint

, for example...

so a Survey probably also has many questions, right? so that's the first part. One Survey is related to Many Questions

right

SO Questions is a model with hint and label

and has a many to many relation to Survey

so a question can appear in different surveys

sorry, one to many

so only in one?

only in one yeah, each survey questions are unique

but conform to a shape

that's probably it

a drawing of your data model can also help a lot. Depending on your needs, you might choose to directly relate a

survey

to its

answers

for example, then you would have to adjust accordingly

Yeah, I think I was getting confused because I did not realize you could create a Survey and the Questions in one mutation

ah right. it's API sugar that we came up with. we call it nested mutations 🙂

you could split the task up into several mutations anyway

Yeah, thats what I mean by ceremony in my message above. Creating all the Questions, then using the

addtoQuestions

type of mapping.

right

ok so I'm confused how this all has anything to do with

JSON

🙂

I see now that JSON was not what I wanted, I just didn't fully understand the generated mutations that are built in to graph.cool

I think a great thing to do is just play around in the playground, as you already did. GraphiQL makes it fun to explore APIs 🙂

I want to push the documentation further but it's difficult to capture the flexibility of our API in static docs... That's why I'm excited about the integrated GraphiQLs. Have you seen them?

Yeah I have, GraphiQL is the best

👍 it is

yeah so just to clarify all of this, you guys have models that are similar to the List of Type: Question that I was used to with reindex. You basically define all your data shapes as models, even if you do not plan on using the create/delete/update mutations that are generated directly and only really plan on doing your mutation on the Parent nodes.

So I will never create a Question directly, but I will create a Survey, which will create the questions

Interesting to see this differences in APIs. How do you relate a question with an answer if you define them with JSON?