Is there a "Ash way" to group db items?

RicoTrevisan

Say I’ve got a resource

defmodule Log do
...
  attributes do
    uuid_primary_key :id
    attribute :fiber_id, :string
    attribute :process_id, :integer
    attribute :data, :map
    ...
  end
end

I would like to read the db, then group these logs in a certain way

[
  %{
    process_id: 406,
    fiber_id: "1751574781326x565732960809615900",
    logs: [
      %Bubblelog.Bubble.Log{
        id: "3100d75b-069b-411e-885f-ac002217eb9c",
        data: %{...}
       ...
     },
     %Bubblelog.Bubble.Log{...},
   ] 
...

I figured I should create a read action, then put it through a Preparation using Ash.Query.after_action, etc.. But I was running into the problem that this action was returning something that is not a Struct. My imaginary friend Claude suggested a manual_action which didn’t work so we settled on a normal Elixir function do this transformation.

Is there an Ash way of doing this?

6 comments

/ash #ash-postgres #ash-resource-actions

0 362 6

2025-07-04 13:33:27 UTC

Showing Posts 1 to 6

Show Best Posts
Show All (oldest first)
Show All (newest first)

ken-kost

I think what you want is a generic action. That’s like a normal elixir function the ash way.

Post #1

RicoTrevisan

Thanks!

Could I integrate it with a read action into the generic action and use the built-in Ash pagination? In my simple mind it seems easier to keep them separate: first read, then group them. Is there a way to build the read action into the generic action?

Post #2

ken-kost

Don’t see why not. You could even make an interface for the read action and first thing you do in your generic action is call that.

Post #3

olivermt

Make synthetic resource for process and have that as your entry point. You can use a combination of a has many with no attributes linking to a process id attribute on the process resource.

Then fill the process id with an arg in a prepare on the read. This should let you do a Process read and then load the logs.

You can use calcs or aggregates to fill the other columns for process, I didnt study the data model too closely.

You can do a lot of this stuff with datalayerless resources. You could even have two resources point to the same table and just map with a has many

Post #4

RicoTrevisan

Great, thanks! I got a lot of homework for the weekend.

Post #5

FlyingNoodle

Couldn’t you even do something like this?

defmodule Bubblelog.Bubble.Process do
  use Ash.Resource

  attributes do
    integer_primary_key :id
  end
  
  actions do
    defaults [:read, :destroy, create: :*, update: :*]
  end
  
  code_interface do
    define :read
  end

  relationships do
    has_many :logs, Bubblelog.Bubble.Log
  end
end

defmodule Bubblelog.Bubble.Log do
  use Ash.Resource

  attributes do
    uuid_primary_key :id
    attribute :data, :map, description: "I dont know what this is?"
  end

  relationships do
    belongs_to :process, Bubblelog.Bubble.Process
  end
  
  actions do
    defaults [:read, :destroy, create: :*, update: :*]
  end
end

For sure this is missing some keys in the relationships and what not and at least the Log module should probably also have a proper data_layer but I think you could model this with regular relationships?

Post #6

— All posts loaded —