First use experience report

[fkowal]

@dillonkearns I'd like to start with thanking you for the great library.

Today for the first time I tried using elm-graphql.
I admit this a great library and it helped a lot.

I am trying to get my company to start using Elm.
Part of that is educating others and introducing them to Elm.

So I am writing this down to lower the entry barrier.

1. My expectations

I am starting to build a new App, and prepared a query using graphiQL ie. {offers(order_by:{...}) { id name stock {quantity ...}}.

Then I turned to elm-graphql to generate the code. I expected the result would look something like:

module Api.Request exposing (fetchOffers,Data,Offers)

type alias Data = { offers: List Offers, errors: ... }
type alias Offers = {
   id: String
   , name: String 
   , stock: List Stock
   ...
}

fetchOffers baseUrl (Result Error Data -> Msg) = -- generated implementation using SelectionSets and all the details if i decided to write the query by hand

Since the generator didn't ask for a query I knew this would have to be done by hand.
And something that took a few minutes in graphiQL turned into a few hours teaching myself how to write a the same query using the generated code.

Feedback at this point is that:

• code generation was super simple
• writing the queries using "dsl" was challenging, it required quite a bit of Elm knowledge, and I don't see how someone new to the language would manage
• code generation stopped one step away from the same great experience graphiQL provides

2. Amount of generated code
   I am using hasura as the graph server.

After the code was generated It took me a while to explore the generated code to understand how it was structured and how to use it.

At this point, I was overwhelmed because the schema is very rich with plenty of options.
Going back to my use-case where I only needed a small subset of what was actually available.

Feedback:
I understand that this is required in the scenario where the user writes the code by hand
using the generated types.
The cognitive load at this step is high. I can't help it feel that this goes against the idea of graphQL.

3. Order_by
   The model/schema I am using is big 10-20 field.
   As a result Table_order_by type contains 20 OptionalArgument Api.Enum.XX_by.Order_by

The normal use-case is probably order_by: { col1: asc, col2: desc} with 1-2 columns.
I wasn't able to find a default with all columns set to Absent that would allow {default | col1: desc}
And writing this by hand was ... not the best experience.

Overall I feel this library is great.
But I don't feel that I can use it as part of a workshop targeted at people new to Elm.