Undefined self attribute in LiquidLegions

[riemanli]

cardinality_estimation_evaluation_framework/src/estimators/liquid_legions.py

Line 171 in 732117e

if len(self.mask[l]) == 1:

The function frequency_sample accesses the undefined attribute self.mask. In addition, I notice that the unit tests doesn't cover the usage of function frequency_sample .

[c-dickens]

I have also just noticed this issue. I think an incorrect copy has been made from cascading_legions.py where this attribute is defined. Notice that in that file, a comment is made that self.mask is used for the key aggregation.

A similar comment is made in liquid_legions.py referencing the same_key_aggregator function. For this reason, I think the dictionary self. unique stores the keys. By this I mean that self.unique operates in accordance with the key array from the LiquidLegions paper. Then replacing self.mask with self.unique should work as a drop in replacement.

However, I also noticed that there is no unit-testing coverage and I don't think the use of frequency capping (f_max from the paper) has been implemented in the current frequency estimation function, although it should be quite straightforward to perform this after the full histogram is returned.