I tried this as well: Error Handling — ash v3.29.3, but Ash.Error.Changes.Required.new/1 is not defined.
So then I used struct literal syntax, and it resulted in the error appearing in the AshGraph response:
prepare fn query, _context do
error =
%Ash.Error.Changes.Required{
field: :foo,
type: :attribute
}
Ash.Query.add_error(query, error)
end
...
resp_body #=> %{
"data" => nil,
"errors" => [
%{
"code" => "required",
"fields" => ["foo"],
"locations" => [%{"column" => 3, "line" => 2}],
"message" => "is required",
"path" => ["listMemberSelfSubscriptions"],
"short_message" => "is required",
"vars" => []
}
]
}
I think if I play around with one of the other Ash error modules, I’ll find what I need to make this work correctly.






















