Instead of using parallel pipelines for processing old datasets, this PR refactors the "main" line of processing modules to handle both new and old datasets. This does not affect the processing of the new datasets.
Old dataset processing option is controlled by two boolean variables hard-coded to NWBFileBuilder, convert_timestamps and return_timestamps. The switches are by default turned on (True) for the new datasets, but can be turned off (False) for old datasets that do not include properly adjusted real timestamps. Behavior before this PL was equivalent to both switches set to False.
New default setting for old datasets:
return_timestamps=True, so that some timestamp information is stored with the NWB files (either sample counts or real timestamps).
convert_timestamps=True. Note that conversion relies on adjusted-systime in the continuoustime.dat file, which is actually the same as the unadjusted systime for old datasets. So this still leaves the timepoint alignment issue (#7) unresolved.
Instead of using parallel pipelines for processing old datasets, this PR refactors the "main" line of processing modules to handle both new and old datasets. This does not affect the processing of the new datasets.
Old dataset processing option is controlled by two boolean variables hard-coded to NWBFileBuilder,
convert_timestampsandreturn_timestamps. The switches are by default turned on (True) for the new datasets, but can be turned off (False) for old datasets that do not include properly adjusted real timestamps. Behavior before this PL was equivalent to both switches set to False.New default setting for old datasets:
return_timestamps=True, so that some timestamp information is stored with the NWB files (either sample counts or real timestamps).convert_timestamps=True. Note that conversion relies onadjusted-systimein the continuoustime.dat file, which is actually the same as the unadjustedsystimefor old datasets. So this still leaves the timepoint alignment issue (#7) unresolved.