Support modal models in Python #1009
Conversation
…l reactors and their modes to reset state variables
| @@ -31,18 +31,18 @@ public static void generateDeclarations( | |||
| if (!allModes.isEmpty()) { | |||
| // Reactor's mode instances and its state. | |||
| body.pr(String.join("\n", | |||
| "reactor_mode_t _lf__modes["+reactor.getModes().size()+"];", | |||
| "reactor_mode_state_t _lf__mode_state;" | |||
There was a problem hiding this comment.
You moved the reactor_mode_state_t _lf__mode_state into the self_base_t struct. If I understand this correctly, now every reactor has a mode state struct (as soon as any modes are present in general) whether it has modes or not. I did not do that to save memory. Is there a benefit in doing it this way that I do not see or does it simply not matter that much?
There was a problem hiding this comment.
Is there a benefit in doing it this way?
We needed this because we wanted to call _LF_ SET_MODE in the .set method of the mode type in the Python target. _LF_ SET_MODE expects a self pointer to a self struct that has a reactor_mode_state_t _lf__mode_state; member. Since self structs are typed, we needed to move it to the base so that it works in the type-agnostic .set method implementation.
The idea is that we want to re-use _LF_ SET_MODE so that any changes in the C runtime to the functionality of this macro is also automatically reflected in the Python runtime.
The implementation of the .set method is here in case you are interested.
|
@housengw Could you please have a look at the conflicts with master? The conflicts seem to primarily be in the imports, but it would be great if you could see if anything needs attention since you are leading the refactoring effort. |
|
will do. |
This is a first step at supporting modal models in Python. We got one test working but ConvertCaseTest.lf gives different results, so this needs more work. This PR depends on branches by the same name (python-modal-models) in reactor-c and reactor-c-py.