Currently the extrapolation machinery assumes the same reco toolkit (dlp, pandora, etc.) will be used by Near and Far. This is unlikely ever to be true, and anyways we will want to study what difference it makes.
In practice, extending this to be more general will require several steps:
Currently the extrapolation machinery assumes the same reco toolkit (
dlp,pandora, etc.) will be used by Near and Far. This is unlikely ever to be true, and anyways we will want to study what difference it makes.In practice, extending this to be more general will require several steps:
Extrapclasses (most notably NumuExtrap, but while we're at it also TrivialExtrap for good measure) to support separateRecoTypes for ND and FD, and to pass them along to the relevantLoadersinstance in their constructor (e.g., here)IPredictionGenerator::Generate()interface to accept differentRecoTypes for FD and ND, then propagating that to its derived classes (PredictionNoExtrap and PredictionGenerator3Flavor, at least).RecoTypeonto the extrap instance in thePredictionGenerator's constructor (e.g., here)