djhoese
commented
5 months ago Kind of a duplicate of https://github.com/pytroll/satpy/issues/2630
I think the question is whether or not this is an option for all datetime types in the metadata or just start_time and end_time. There are other times in the metadata now (observation, nominal, etc), but if we're merging datasets and these differ do we really want to min/max/mean these at all? The alternative is to always min the start_time and max the end_time.
pnuu
Feature Request
Creating a composite that uses a static background image yields an image with a different time than the used satellite data. The current approach sets the
start_timeof the static image toutcnow(), andsatpy.dataset.combine_metadatatakes average of this and the other dataset used in the creation of the composite. Creating images in NRT might not change the time stamp a lot, but doing the same for data 2 days ago back already shifts the result by one full day.Describe the solution you'd like Instead of always taking the average of the times, could we instead have multiple options:
min(datetime_list)- perfect when using timed data with static background, always gives the time of the actual datamean(datetime_list)- when multiple instruments with differing times are combinedmax(datetime_list)- when multiple instruments with differing times are combined, and the more recent one has the more representative timeCould we use the Satpy config system for controlling this? I would argue the default should be
min(), but I'm open to comments.Implementation it self is simple: add new min/max functions and call min/mean/max depending on the setting.
Describe any changes to existing user workflow No compatibility issues if
meanis kept as default. Setting the config item would need one additional line of code in scripts when required.The only place where
combine_metadata(..., average_times=False)is used is in a test method.Additional context I was thinking this could also be a bug, but in the end decided a feature request would be better instead.