problems where valid range with a clock will not interpolate across day boundaries

[jamesuber]

this is identified for Louisville.

[jamesuber]

This is related to the valid range having a clock. If the valid range clock is removed, and the resampling clock is added to the subsequent time series (in this case a stats), then the gaps go away and stats seems to interpolate normally across the gaps created by the dropped points.

So not sure this is a bug. But clocks and valid ranges seem to be problematic?

[samhatchett]

this is a difficult edge case involving the set of filters that can drop points. how to expand the source series search space intelligently/efficiently. right now we use a "range-duration" expansion, but probably need to move to an iterative fwd/backwd search algo.

[samhatchett]

settled on using existing code to solve the problem. just swap out the clock, seek to next valid point, restore the clock.