Allow reprocessing of a dataset without rereading it in
The user should be able to call klip_dataset multiple times and not have to do any bookkeeping.
Currenlty,values such as dataset.center and dataset.wcs get modified in klip_dataset, meaning that if you want to run klip_dataset multiple times, you have to reset dataset.center and dataset.wcs manually. We need someway to specify input dataset centers/wcs and output dataset centers/wcs (possibly by adding new fields).
Comments (5)
-
-
reporter Yeah, that's what I meant by adding new fields. We would always retain an original copy of everything.
The second issue you address is a different issue as it's more of a memory management problem. If we hold on to the align_and_scale data, that means we'll be potentially holding on to tens of GB of memory after pyklip finishes even if the user doesn't wish to reprocess a dataset. That would be best put in as an optional parameter I would guess.
-
Yeah, a save_aligned flag would suffice. I can probably work on this issue.
-
reporter -
assigned issue to
Sounds good. Note that to not break p1640 data and the interface, we should make new fields for output centers/wcs headers and not modify how the input centers/wcs headers are defined.
-
assigned issue to
-
- changed status to resolved
Added functionality in commit 468d47d
- Log in to comment
How about changing the variable names so that none of the variables set by GPI.GPIData() are changed by klip_dataset? We could also put in some logic so that if klip_dataset is called again, the align_and_scale step can be skipped which would ~halve the time required to process a sequence.