Struct thesis::experiment::Experiment

pub struct Experiment<T, C, E, R, M> { /* fields omitted */ }

An individual experiment. See crate-level documentation for an example on how to use

Implementations

impl<T> Experiment<T, (), (), (), AlwaysControl>

pub fn new(name: &'static str) -> Self

Create a new experiment. The only provided default is accepting the control value in the mismatch handler. All other builder-style functions must be called before run can be called.

impl<T, C, E, R, M> Experiment<T, C, E, R, M>

pub fn control<NC>(self, control_builder: NC) -> Experiment<T, NC, E, R, M> where
    NC: Future<Output = T>, 

Use the future given here as the control, or the existing method for calculating a value

pub fn experimental<NE>(
    self,
    experimental_builder: NE
) -> Experiment<T, C, NE, R, M> where
    NE: Future<Output = T>, 

Use the future given here as the experimental, or the new method for calculating a value

pub fn rollout_strategy<NR>(
    self,
    rollout_strategy: NR
) -> Experiment<T, C, E, NR, M>

Use the given strategy for rolling out the new code

pub fn on_mismatch<NM>(
    self,
    on_mismatch: NM
) -> Experiment<T, C, E, R, FnTrait<NM>> where
    NM: FnOnce(Mismatch<T>) -> T, 

Call this function when running the experiment results in a different value from the control and experimental methods. This can only happen when the rollout strategy returns RolloutDecision::UseExperimentalAndCompare.

pub async fn run(self) -> T where
    T: PartialEq,
    R: RolloutStrategy,
    M: MismatchHandler<T>,
    C: Future<Output = T>,
    E: Future<Output = T>, 

Run the experiment with the parameters provided

impl<T, Err, C, E, R, M> Experiment<Result<T, Err>, C, E, R, M>

pub async fn run_result(self) -> Result<T, Err> where
    T: PartialEq,
    R: RolloutStrategy,
    M: MismatchHandler<Result<T, Err>>,
    C: Future<Output = Result<T, Err>>,
    E: Future<Output = Result<T, Err>>,
    Err: Display, 

Run the experiment with the parameters provided