add: Dynamic changes to Behaviour Patterns at runtime

[SirPigeonz]

This PR makes both FSM and BTree systems work properly when their nodes are moved, removed, added at runtime.

On top of that many nodes are now fine when they don't have important for them nodes as a child or property references. This allows for preparing "empty" slots for intended behaviours or branches and modeling main behaviours with them, that is "slots", in mind. Upcomming warning system can be used instead for important configuration warnings when behaviours are created by hand in the Editor.

Changes:

• BTRoot, FiniteStateMachine, BTComposite, BTDecorator, FSMStateIntegratedBT and BTIntegratedFSM nodes now can reconfigure themselves when their children are modified at run-time
• allow BTRoot to not have any BTBehaviour as a child
• allow FiniteStateMachine to not have any FSMStates as a children
• allow FiniteStateMachine to not have a initial state
• allow FSMStateIntegratedBT to not have any BTRoot as a child
• allow BTIntegratedFSM to not have any FSM as a child
• BTIntegratedFSM has now default_status property enum for case where it doesn't have FSM child
• new method in BTRoot - swap_entry_point()
• new method in FSMStateIntegratedBT - swap_behaviour_tree()
• new method in BTIntegratedFSM - swap_finite_state_machine()
• make FSMStates and extended nodes update their transitions list when children are added/removed
• FSMStateIntegratedBT disables autostart of its BTRoot
• BTIntegratedFSM disables autostart of its FSM
• made FSMState extensible without need to call super() on _ready()
• made BTRandom extensible without need to call super() on _ready()
• Added few missing documentation comments, to outline how nodes are intended to work and how they should be edited also at run-time.

Changes after review 1

• removed _disable_autostart() in FSM and BTRoot
• made BTRoot and FSMset their autostart to false and hide them in Editor Inspector if their parent is a integration type node
• made BTRoot and FSM a @tool
• added Engine.is_engine_hint guards to ready, callbacks and processes for BTRoot and FSM
• added keep_group optional property to swap_'nodetype'() methods, it allows to preserve original nodes groups from swapped node in the new node

Changes after review 2

If you wish to add, remove, move FSMState nodes at run-time first add new FSMStates stop the FSM with method FiniteStateMachine.exit_active_state_and_stop and re-start it with method method FiniteStateMachine.start providing one of the new states either as start method property or change member FiniteStateMachine.initial_state before running start(). After this procedure you can delete unused states.

• made active property read-only
• modified start() in fsm.gd to accept FSMState property as a start point
• new method exit_active_state_and_stop() to pair with start()
• above two changes make FSM startable and stoppable for example to safely modify FSMStates and resume running of the FSM
• removed some if guards from proccess function and made checks when something changes in the setup
• made BTRoot cleanups and made it to not check for entry point in processing
• changed some configuration warnings. Mostly changed statement that state "nodes must have child nodes", to "nodes SHOULD have child nodes to work" to inform user that they wont work but nothing bad will happen if they don't
• removed warning for BTLeaf that "BTLeaf node must not have any children.". Reason is that there is no issue if it has one, it can prevent user to use some nice composition on top of BTLeaf for no reason :)

[ThePat02]

Keep in mind that some things might change how everything is handled, since #17 requires every script to be a @tool!

[SirPigeonz]

Keep in mind that some things might change how everything is handled, since #17 requires every script to be a @tool!

Yup :) I already added few guards with if Engine.is_editor_hint().

[SirPigeonz]

@ThePat02 Ready for review :)

This time for sure xD

[ThePat02]

Good work! I'll look into it when I finished implementing #17 to avoid any breaking conflicts.

[SirPigeonz]

Good work!

Senkyu! :3

I'll look into it when I finished implementing #17 to avoid any breaking conflicts.

Do as you prefer :) Although, I would personally do opposite, because I changed some behaviours of the addon. Some configurations that were illegal and crash in the past now are legal just do nothing and would need different type of warning and asserts. (I'm just worried that some of your work will be not needed).

That said I tried to test it diligently as I could. Each node was tested alone and with others and main example also runs as used to. I also plan to move to this version (2.0 + this PR) with my prototype so I will probably also catch regressions.

What I want the most is it for you to look into the design and tell me what you think and if I missed something, or should change something :)

[SirPigeonz]

One other thing I noticed: When changing states of a FSM, what happens if you change the active state? Did the current state exit?

Good question! Hmm, I should add active_state guard for processing too.

When you change state with change_state() it runs states exit same with transitions. Or do you mean is it running exit when deleted with Node API? Then no it doesn't run exit. Should it?

We could add method to FiniteStateMachine for deleting states that calls exit if state is current active_state, for situations it's needed.

What do you think?

[SirPigeonz]

@ThePat02 Adressed few points in new commit:

• removed _disable_autostart() in FSM and BTRoot integration nodes
• made BTRoot and FSMset their autostart to false and hide them in Editor Inspector if their parent is a integration type node
• made BTRoot and FSM a @tool (needed for _validate_property(property))
• added Engine.is_engine_hint guards to ready, callbacks and processes for BTRoot and FSM
• added keep_group optional property to swap_'nodetype'() methods, it allows to preserve original nodes groups from swapped node in the new node
• added guard to FSM process if active_state == null

[ThePat02]

One other thing I noticed: When changing states of a FSM, what happens if you change the active state? Did the current state exit?

Good question! Hmm, I should add active_state guard for processing too.

When you change state with change_state() it runs states exit same with transitions. Or do you mean is it running exit when deleted with Node API? Then no it doesn't run exit. Should it?

We could add method to FiniteStateMachine for deleting states that calls exit if state is current active_state, for situations it's needed.

What do you think?

Well what is the workflow with dynamic changing of nodes? There needs to be some kind of failsafe when changin states manually and via code.

[SirPigeonz]

Well what is the workflow with dynamic changing of nodes? There needs to be some kind of failsafe when changin states manually and via code.

Ahhh now I understand, sorry 😁 Ok yeah, there should be some safe API to model states the way user needs for their use case. I will think more about it and write my API ideas.

[ThePat02]

I think it is best to push this feature to a later version of the plugin (possibliy 2.1) as there are some other pitfalls and problems that come with this. I think it is very important to release the new syntax changes, as more and more people are interested in using it. I'd like to avoid problems like people having to rewrite their code with the new syntax too much.

[SirPigeonz]

I think it is best to push this feature to a later version of the plugin (possibliy 2.1) as there are some other pitfalls and problems that come with this. I think it is very important to release the new syntax changes, as more and more people are interested in using it. I'd like to avoid problems like people having to rewrite their code with the new syntax too much.

I'm fine if it will be in 2.1 if you want to release 2.0 early due to changes in API. I use latest code in my game either way.

Furthermore, I just finished a bigger commission for a client, so I will have more free time now to work on this and my game. I will probably finish a new version of this PR today or tomorrow.

[ThePat02]

Sounds great. If you need anything, make sure to add me. I'll unsub for now so I don't get notis for every commit!

[SirPigeonz]

@ThePat02 Ready for review again :)

[ThePat02]

I'll take a look at it after finishing up the other changes. BTW I added a formatting action you can steal for your own projects.