We've just been debugging some strange calibration registry problems. Specifically, we have exactly one master bias which has been created but not registered.
in obs_base mapping.py:224 there is a call to
and stepping into the call reveals that:
(the template is of course set in the .paf file).
The butler merrily calls PosixRegistry.LookupData(lookupProperties, dataId) which resolves to the single file that matches — and this means that it uses a regexp to look up the calibDate.
This is not what we wanted; the bias is a calibration type (as stated in the paf file) with a validity range which has been ignored.