Ash Elixir•2y ago

Aggregates with resource relationships

Hello! So maybe there's something I'm not quite understanding from the documentation, but say I have a State and City. A state can have many cities, so:

relationships do
  has_many :cities, MyProject.City
end

relationships do
  has_many :cities, MyProject.City
end

And a city belongs to a state:

relationships do
  belongs_to :state, MyProject.State
end

relationships do
  belongs_to :state, MyProject.State
end

Now for a state, say I want an aggregate for the number of cities in the state. Do I have an aggregates block like the following?

aggregates do
  count :number_of_cities, :cities do
    filterable? true
  end
end

aggregates do
  count :number_of_cities, :cities do
    filterable? true
  end
end

Also, could this be a oneliner as just count :number_of_cities, :cities? No do ... end? Thanks!

15 Replies

frankdugan3•2y ago

Yes, this should do what you want:

aggregates do
  count :number_of_cities, :cities
end

aggregates do
  count :number_of_cities, :cities
end

The do block is an alternative syntax that is sometimes required for passing more complicated options. filterable? is true by default.

AlecStewart1#1125OP•2y ago

Okay, I didn't quite know if when it said

The relationship or relationship path to use for the aggregate

That would also include fields for actual relationships between resources.

frankdugan3•2y ago

In the case of count, you generally don't need to refer to a specific field. But in the case of say a first or sum aggregate, you often want a specific field. For example, sum :total_population, :cities, :population

AlecStewart1#1125OP•2y ago

So for list, if the default field is the primary ID

The field to aggregate. Defaults to the first field in the primary key of the resource

frankdugan3•2y ago

The path can also traverse multiple relationships deep, which is very convenient.

AlecStewart1#1125OP•2y ago

Do I need to just have list :cities_in_state, :cities, :id?

frankdugan3•2y ago

I've actully never used the list aggregate, so I'm not familair w/ the API on that one. 😅 Looking at the docs, it looks like the 3rd argument is the field, yes. list [name], [relationship], [field]

AlecStewart1#1125OP•2y ago

Well I'll try that and see if works. How exactly do aggregates get created and stored? So a state could have like 20000 cities, so if I call for that aggregate somewhere with Ash am I going to have to wait for that aggregate to be initialized the first time? I guess that's more of a question about Postgres than Ash...

frankdugan3•2y ago

Sorry, missed your last question. It will return whichever field (id by default). So if you wanted a different field (like :label), you'd just need to specify it.

AlecStewart1#1125OP•2y ago

Oh okay. Well it says that field is required but it defaults to the primary ID so I was confused as to whether or not it's actually required.

frankdugan3•2y ago

Yeah, that is more of a PG specific question, and I'd assume for something that large, you would want an index to increase efficiency. You can omit it if :id is OK.

AlecStewart1#1125OP•2y ago

Well looks like I found what I might need. https://ash-hq.org/docs/dsl/ash-resource#postgres-custom_indexes

Ash HQ

Ash.Resource

View the documentation for Ash.Resource on Ash HQ.

AlecStewart1#1125OP•2y ago

Also I guess there's always the solution of creating materialized views with SQL in the migrations.

frankdugan3•2y ago

Yeah, I'm using a materialzed view for a report or two and have an Oban job that refreshes it regularly and on demand.

AlecStewart1#1125OP•2y ago

Oh shoot. I forgot about the Oban library. That could fit our use cases pretty well if we don't feel like having cron jobs that run mix.

Gaming

Programming

Aggregates with resource relationships

Did you find this page helpful?