Attack action design

[734162621]

In version 0.1, we set the action <Attack_Unit(tag)> as {'name': 'Attack_Unit', 'arg': ['tag'], 'func': [(12, F.Attack_screen, ('queued', 'screen_tag'))]}, which directly call pysc2 function Attack_screen.

However, the pysc2 function Attack_screen('queued', 'screen') acts like a "right-click" on an enemy unit. Unless the target unit is dead, your unit will not use next queued actions, leading to difficulties in implementing micro-operations for Attack-Move-Attack-Move.

We initiate this discussion to see if everyone has suggestions for the Attack action design.

[734162621]

How about add an action called <Attack_Unit_then_Move(tag, screen)> as {'name': 'Attack_Unit_then_Move', 'arg': ['tag', 'screen'], 'func': [(12, F.Attack_screen, ('now', 'screen_tag')), (0, F.no_op, ()), (331, F.Move_screen, ('now', 'screen'))]}?

This action may be useful for scenarios with high requirements for micro operations like SMAC tasks.