Implement input_sampling_period recursive logic on /algorithms/filter endpoint

Currently, the frontend can make a series of requests to the /algorithms/filter endpoint. If there is no valid data for a specific time period, the frontend retries the request by reducing the input_sampling_period until valid data is returned, using a recursive function. This functionality should exist in the backend to allow all API users to have the same experience.

Frontend code https://code.usgs.gov/ghsc/geomag/geomag-plots/-/blob/master/projects/geomag-lib/src/lib/services/timeseries-manager.service.ts#L135

changed the description

@erigler , @swilbur , @nshavers and I met today to clarify the functionality of sampling_period. sampling_period is being set to the default SamplingPeriod.Minute, which means that filtering on observatory data is only using the EdgeFactory and never using MiniSeedFactory here https://code.usgs.gov/ghsc/geomag/geomag-algorithms/-/blob/master/geomagio/api/ws/data.py#L38.

This behavior is acceptable for now because @erigler is currently combining the EdgeFactory and MiniSeedFactory and this distinction won't be necessary in the future.

Filtering on variometer uses the FDSNFactory which is not affected by this because it ignores sampling_input. Please correct me if I'm wrong anywhere!

Our NOAA colleagues were experimenting with the /ws/algorithms/filter endpoint and noticed that, if no input_sampling_period is specified, it defaults to 1-second, and so things "just work" for most of the variometer stations now, since that is their native sampling period.

Unsurprisingly, this does not work for stations that have been updated to 10Hz sampling periods.

Considering this, is the following set of requirements reasonable:

• if input_sampling_period is not specified (so, only output_sampling_period or even just sampling_period), the backend will recursively attempt successively higher resolution inputs sampling periods?
• if input_sampling_period is specified, then no recursion should take place

That second bullet is a subtle update to what I was originally thinking, but maybe you already came to this conclusion. Or, if you disagree, please explain. Thanks!

This issue was completed with !366 (merged)

closed