5 releases
0.2.3 | Apr 12, 2023 |
---|---|
0.2.2 | Sep 30, 2021 |
0.2.1 | Sep 30, 2021 |
0.2.0 | Sep 28, 2021 |
0.1.0 | Sep 27, 2021 |
#13 in #finite-state-machine
34 downloads per month
19KB
275 lines
Dirty FSM
dirty-fsm
is a "Quick and Dirty" implementation of a finite state machine. Most of the concepts come from code I wrote as part of io.github.frc5024.lib5k.libkontrol
.
Example
In the following example, I model a state machine that represents a claw and a button. When the button is pressed, the claw will toggle between open and closed.
We start by setting a feature flag and loading the library.
// This feature is required to use the new `#[default]` macro on enum variants
#![feature(derive_default_enum)]
use dirty_fsm::*;
use thiserror::Error;
Next, we define the states of the machine.
/// The possible states of the claw
#[derive(Debug, Default, PartialEq, Eq, Clone, Hash)]
enum ClawState {
/// The claw is closed
#[default]
ClawClosed,
/// The claw is open
ClawOpen
}
Along with the states, we need to define some kind of error type (although if not needed at all, we can just use ()
).
/// Defines errors that can occur while running actions
#[derive(Debug, Error)]
enum ClawError {
/// An example error
#[error("Example error")]
ExampleError,
}
Next, we define the code to actually run during our first state (ClawClosed
). This is a regular Rust struct, that implements the Action
trait.
Action
contains a few simple functions that are called at various points throughout the action's life:
on_register
: Called once when the action is registered with a state machine (viaStateMachine::add_action
)on_first_run
: Called once right before the firstexecute
call after this state has been started or switched to. This should be treated like an initializer function. Usually used to save information about the environment before performing an operation inexecute
.execute
: Called multiple times during the action's life. This is where the action's code should go. It should be treated as the body of awhile true
loop, since it will be run over and over until it returns anActionFlag
that indicates the action is done.on_finish
: Called once right after the lastexecute
call once this action is finished.
/// Action that actually handles the claw being closed
#[derive(Debug)]
struct ClawClosedAction;
impl Action<ClawState, ClawError, bool> for ClawClosedAction {
fn on_register(&mut self) -> Result<(), ClawError> {
println!("ClawClosedAction has been registered with the state machine");
Ok(())
}
fn on_first_run(&mut self, context: &bool) -> Result<(), ClawError> {
println!("Button has been pressed, claw is closing");
Ok(())
}
fn execute(
&mut self,
delta: &chrono::Duration,
context: &bool,
) -> Result<crate::action::ActionFlag<ClawState>, ClawError> {
println!("Claw code is running now");
// If the button is pressed, switch to the next claw state
if context {
Ok(ActionFlag::SwitchState(ClawState::ClawOpen))
} else {
Ok(ActionFlag::Continue)
}
}
fn on_finish(&mut self, interrupted: bool) -> Result<(), ClawError> {
println!("ClawClosedAction is done executing");
Ok(())
}
}
Since we have two states, this needs to be done again for the other state.
/// Action that actually handles the claw being opened
#[derive(Debug)]
struct ClawOpenedAction;
impl Action<ClawState, ClawError, bool> for ClawOpenedAction {
fn on_register(&mut self) -> Result<(), ()> {
println!("ClawOpenedAction has been registered with the state machine");
Ok(())
}
fn on_first_run(&mut self, context: &bool) -> Result<(), ClawError> {
println!("Button has been pressed, claw is opening");
Ok(())
}
fn execute(
&mut self,
delta: &chrono::Duration,
context: &bool,
) -> Result<crate::action::ActionFlag<ClawState>, ClawError> {
println!("Claw code is running now");
// If the button is pressed, throw an error as an example
if *context {
Err(ClawError::ExampleError)
} else {
Ok(ActionFlag::Continue)
}
}
fn on_finish(&mut self, interrupted: bool) -> Result<(), ClawError> {
println!("ClawOpenedAction is done executing");
Ok(())
}
}
Finally, the code to start and run the state machine:
fn main() {
// Create the state machine
let mut claw_machine = StateMachine::new();
claw_machine.add_action(ClawState::ClawClosed, ClawClosedAction {}).unwrap();
claw_machine.add_action(ClawState::ClawOpen, ClawOpenedAction {}).unwrap();
// State. This example assumes some outside "force" is changing this value
let mut button_pressed = false;
// Run the state machine
loop {
claw_machine.run(&mut button_pressed).unwrap();
}
}
Dependencies
~1–1.3MB
~20K SLoC