I am not sure if this is the right way to test Step on different backends. I just use this as the start point for collecting better solutions and feedbacks. I tried this solution on two Step: DiscretizeKBins and CreatePolynomialFeatures
The main idea is to see if the operators used in fit_table() and transform_table could be compiled to sql of different backend.
For fit_table(), I save the intermediate fitting expression as a instance variable self.expr and later to check if it is could be transpiled into different dialect.
For transform_table, just to compile the output table.
Hi,
I am not sure if this is the right way to test
Stepon different backends. I just use this as the start point for collecting better solutions and feedbacks. I tried this solution on two Step:DiscretizeKBinsandCreatePolynomialFeaturesThe main idea is to see if the operators used in
fit_table()andtransform_tablecould be compiled to sql of different backend.fit_table(), I save the intermediate fitting expression as a instance variableself.exprand later to check if it is could be transpiled into different dialect.transform_table, just to compile the output table.