Rework how entities_field works #104
Conversation
maciej-szlosarczyk
force-pushed
the
dataloader-ecto
branch
3 times, most recently
from
e42dfd1 to
3aa1cb9
Compare
- Support Async resolution with 0-arity function - Support Dataloader.Ecto - Support function capture resolution
maciej-szlosarczyk
force-pushed
the
dataloader-ecto
branch
from
3aa1cb9 to
893e93b
Compare
There was a problem hiding this comment.
@maciej-szlosarczyk thank you so much for working on this! These are some great changes, and I left a few comments. If you can address those, I'll be happy to run another test suite on my end.
|
@kzlsakal Thanks for taking a look, I added the functionality and required test. Should be all good now, let me know if you need to rebase all changes into a single commit and I'll do just that. |
| def call(%{arguments: %{item_id: item_id, __typename: "ItemWithModuleMiddleware"}} = res, _opts) do | ||
| value = %{item_id: "ModuleMiddleware:#{item_id}", __typename: "ItemWithModuleMiddleware"} | ||
| Map.put(res, :value, value) | ||
| end | ||
|
|
||
| def call(res, _opts) do |
There was a problem hiding this comment.
Absinthe.Middleware.call/2 has the following function spec:
call(Absinthe.Resolution.t(), term) :: Absinthe.Resolution.t()We may want to have only one call/2 definition in this test middleware and ensure we're not calling it with something other than an Absinthe.Resolution struct as the first argument, otherwise it could be a breaking change for many codebases that expect what Absinthe specifies for it. What do you think?
|
|
||
| # setup do | ||
| # end |
There was a problem hiding this comment.
| # setup do | |
| # end |
Hey 👋 Great work on library!
This is a pretty big PR to address the following points:
The main idea here is to use normal middleware pipeline for each resolved representation instead of looking at what those middlewares actually are. Unfortunately it changes some API examples and tests, but it brings them in-line with how Absinthe works in general. I.e async/2 helper requires the first argument to be a 0-arity function, but for some reason the tests used 1-arity.