The fact that Step depends on DistAlgorithm is pretty inconvenient in some places. Also, any user who isn't concerned with writing something like a generalized wrapper or test framework that is supposed to work for several algorithms shouldn't have to deal with DistAlgorithm at all. For the many algorithms that can only output once, the current Step implementation also makes the API less intuitive, since it always contains a Vec of outputs. Finally, Step could be adapted to the fault kind hierarchy (#285) once that's done.
[x] Make Step generic in the message and output types instead of the algorithm.
[ ] Find a way to make Step statically aware of whether it can potentially have multiple outputs. (E.g. another type parameter, and a method Step::single_output that only exists in the corresponding impl. Or make the output type itself either Option or Vec.)
[x] Make Step generic in the fault kind? (Requires #285 and making Fault and FaultLog generic. See #343)
The fact that
Stepdepends onDistAlgorithmis pretty inconvenient in some places. Also, any user who isn't concerned with writing something like a generalized wrapper or test framework that is supposed to work for several algorithms shouldn't have to deal withDistAlgorithmat all. For the many algorithms that can only output once, the currentStepimplementation also makes the API less intuitive, since it always contains aVecof outputs. Finally,Stepcould be adapted to the fault kind hierarchy (#285) once that's done.Stepgeneric in the message and output types instead of the algorithm.Stepstatically aware of whether it can potentially have multiple outputs. (E.g. another type parameter, and a methodStep::single_outputthat only exists in the correspondingimpl. Or make the output type itself eitherOptionorVec.)Stepgeneric in the fault kind? (Requires #285 and makingFaultandFaultLoggeneric. See #343)