CutPlugin class for cuts in straxen

[JoranAngevaare]

The cuts currently in straxen (https://github.com/XENONnT/straxen/blob/master/straxen/plugins/x1t_cuts.py ) have a quite lot of overlap with one another. Most of the time a simple Boolean array is all that is returned. However, we each time have to add the time fields et cetera which may somewhat obscure the functionality.
To simplify the cuts somewhat I've written a simple Plugin Class. In straxen it could be used as for example:

class CutS1s(CutPlugin):
    depends_on = ('events', 'corrected_areas')
    provides = 'cs1_cut'
    cut_description = "Remove cS1 above 100 PE"
    cut_name ='cut_high_cs1'
    __version__ = '0.0.0'
    
    def cut_by(self, events):
        return events['cs1'] > 100

class CutS2s(CutPlugin):
    depends_on = ('events', 'corrected_areas')
    cut_description = "Remove cS2s below 100 PE"
    
    def cut_by(self, events):
        return events['cs2'] < 100

class CutRecord0(CutPlugin):
    depends_on = ('records',)
    provides = 'is_not_record0_cut'
    cut_description = "is record 0"
    cut_name ='is_not_record0_cut'
    __version__ = '0.0.0'
    
    def cut_by(self, records):
        return records['record_i'] != 0

[WenzDaniel]

Hej Joran,

nice Idea. I have not checked your code yet, but one quick question. In your current version cuts would be the same data_kind as events would not they be? Maybe it is better to have a new "cut"-data_kind instead?

[JoranAngevaare]

Hej Joran,

nice Idea. I have not checked your code yet, but one quick question. In your current version cuts would be the same data_kind as events would not they be? Maybe it is better to have a new "cut"-data_kind instead?

Hi Daniel,
Thanks for the suggestion! The datakind is derived from the dependency specified in the cut_by(self <datakind>), so in principle you could cut on peak level. I think it would make analysts lives harder if we create a new datatype that is not events as that is where most analysts will actually apply the cuts. That can get tricky if you want to load multiple datakinds as get_df (the function most analysts will use) will complain that this is not possible.

[JoranAngevaare]

As @WenzDaniel just correctly pointed out the endtime needs to be computed when we are working with lower level plugins. Although it's not really the envisaged application it's easy enough to implement.

See now for example:

class CutRecord0(CutPlugin):
    depends_on = ('records',)
    provides = 'is_not_record0_cut'
    cut_description = "is record 0"
    cut_name ='is_not_record0_cut'
    __version__ = '0.0.0'
    
    def cut_by(self, records):
        return records['record_i'] != 0

[JoranAngevaare]

In 752f521 I've added a test mainly since I think I should learn how how to write tests. Also it's completely fine to add more tests.

If someone foresees a good metric to add to this test, please let me know.

[JoranAngevaare]

@ ACs this is ready for merge

[WenzDaniel]

Maybe just add one additional test checking the endtime since we had these issues in peaks and you are computing some new endtimes here. Does not hurt much to have this little check. All the rest looks fine.

[JoranAngevaare]

Thanks, this is done in 3673c28 !