Change ImageCleaner API

[Hckjs]

Change API of ImageCleaner to be __call__(self, tel_id: int, event: ArrayEventContainer). Closes #2015.

@aleberti

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.53%. Comparing base (3952e5a) to head (8a05e27).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2511   +/-   ##
=======================================
  Coverage   92.53%   92.53%           
=======================================
  Files         235      235           
  Lines       20063    20064    +1     
=======================================
+ Hits        18565    18566    +1     
  Misses       1498     1498           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[kosack]

Can you explain why TailCutsDataVolumeReducer was changed as well here? This change seems to remove the ability to change which image cleaner implementation is used inside it, but that is not mentioned at all in the changelog, and seems to be outside of the scope of the PR (aside from the fact that that DataVolumeReducer is confusingly named and should be ImageCleaningDataVolumeReducer or similar, as it is not tail-cuts specific)

[maxnoe]

This change seems to remove the ability to change which image cleaner implementation is used inside it,

Why would something called TailCutsImageCleaner allow to use a different cleaning method than TailCuts?

[kosack]

Why would something called TailCutsImageCleaner allow to use a different cleaning method than TailCuts?

I agree and mentioned that in the request, but that reducer was configurable, presumably for a reason. I suspect more that it was just misnamed, rather than that it should be restricted to one type of cleaning. All it does is iteratively clean and dilate, which should work for any cleaning implementation.

Nevertheless, that should be a different PR if you want to change how it works, and the ACADA people should be involved.

[maxnoe]

Ok...

I think we discussed this already elsewhere: keep the old signature in a method like: clean_low_level (please invent a better name), and have __call__ use that internally.

Then the TailCutsImageReducer (and the TwoStepExtractor) can call the low-level API with minimal changes to them.

[Hckjs]

Thanks for the comments!
Adapting it now to have a __call__(self, tel_id: int, event: ArrayEventContainer) which internally uses the abstract clean_image(self, tel_id, image, arrival_times) method defined in each ImageCleaner-subclass.

The TwoStepExtractor is already using the low level tailcuts_clean method.

I'll create another PR for renaming the TailCutsDataVolumeReducer.

[maxnoe]

Maybe a better change would be to for now just add the event as new optional parameter?

def __call__(self, tel_id, image, peak_time=None, *, event=None):

That is

a) backwards compatible
b) allows to clean other images than what is in event.dl1.tel[tel_id].image / event.dl1.tel[tel_id].peak_time
c) still allows accessing additional information from the event

[maxnoe]

Just to summarize from the discussions that are now in collapsed threads: we decided to not pass the full event, but just the CameraMonitoringContainer, which should have all information necessary for advanced cleaning algorithms.