+ #
+ # Clean up the resulting list of orders such that no column
+ # uselessly appears twice (Postgres might not optimize this out
+ # for us) and no columns uselessly appear after a unique column
+ # (Postgres does not optimize this out for us; as of 9.2, "order
+ # by id, modified_at desc, uuid" is slow but "order by id" is
+ # fast).
+ orders_given_and_default = @orders + model_class.default_orders
+ order_cols_used = {}
+ @orders = []
+ orders_given_and_default.each do |order|
+ otablecol = order.split(' ')[0]