When initializing Unit objects, one would probably add attributes to it later, e.g. unit.is_scout: bool or unit.mining_at_mineral_patch: int to have some sort of persistent data.
This can only be done by extending the Unit class (either by modifying its file or inheriting) and making the Unit object persistent (either by actually making it persistent, or just updating its unit._proto data and invalidating all cache https://docs.python.org/3/library/functools.html#functools.cached_property ).
Instead of invalidating cache it can make sense to convert @cached_property back to @property for example for unit.orders for attributes that may change over multiple frames.
Alternatively the is_scout function can be a property that accesses data from the persistent bot object instead.
For example
@property
def is_scout(self) -> bool:
return self.tag in self._bot_object.scouting_manager.scout_tags
When initializing
Unit
objects, one would probably add attributes to it later, e.g.unit.is_scout: bool
orunit.mining_at_mineral_patch: int
to have some sort of persistent data.This can only be done by extending the Unit class (either by modifying its file or inheriting) and making the Unit object persistent (either by actually making it persistent, or just updating its
unit._proto
data and invalidating all cache https://docs.python.org/3/library/functools.html#functools.cached_property ). Instead of invalidating cache it can make sense to convert@cached_property
back to@property
for example forunit.orders
for attributes that may change over multiple frames.Alternatively the
is_scout
function can be a property that accesses data from the persistent bot object instead. For exampleA solution may be to allow bot authors to set a custom class for
Unit
objects to be initialized with in https://github.com/BurnySc2/python-sc2/blob/76e4a435732d4359e5bd9e15b6283a0498e212ca/sc2/bot_ai_internal.py#L566 as long as they inherit fromUnit
class.