feat(Ecto.Repo): allow aggregating over *

This is done to provide similar API to `Ecto.Query.API.count/0` as
counting over `*` is faster in some DB than counting over column, even
when column is `NOT NULL` and indexed.

Close elixir-ecto#3175

lib/ecto/repo/queryable.ex

@@ -406,6 +406,26 @@ defmodule Ecto.Repo.Queryable do
     Query.where(queryable, [], ^Enum.to_list(clauses))
   end
 
+  defp query_for_aggregate(queryable, aggregate, :*) do
+    query = %{Queryable.to_query(queryable) | preloads: [], assocs: []}
+    query =
+      case query do
+        %{group_bys: [_ | _]} ->
+          raise Ecto.QueryError, message: "cannot aggregate on query with group_by", query: query
+
+        %{distinct: nil, limit: nil, offset: nil} ->
+          %{query | order_bys: []}
+
+        _ ->
+          query
+          |> Query.subquery()
+          |> Queryable.Ecto.SubQuery.to_query()
+      end
+
+    select = %SelectExpr{expr: {aggregate, [], []}, file: __ENV__.file, line: __ENV__.line}
+    %{query | select: select}
+  end
+
   defp query_for_aggregate(queryable, aggregate, field) do
     query = %{Queryable.to_query(queryable) | preloads: [], assocs: []}
     ast = field(0, field)

test/ecto/repo_test.exs

@@ -198,6 +198,10 @@ defmodule Ecto.RepoTest do
       TestRepo.aggregate(MySchema, :count, :id)
       assert_received {:all, query}
       assert inspect(query) == "#Ecto.Query<from m0 in Ecto.RepoTest.MySchema, select: count(m0.id)>"
+
+      TestRepo.aggregate(MySchema, :count, :*)
+      assert_received {:all, query}
+      assert inspect(query) == "#Ecto.Query<from m0 in Ecto.RepoTest.MySchema, select: count()>"
     end
 
     test "aggregates handle a prefix option" do