UniqueCombinations constraint for more than 2 columns

[onacrame]

Environment Details

Please indicate the following details about the environment in which you found the bug:

• SDV version: 0.10
• Python version: 3.9
• Operating System: Windows

Error Description

I was under the impression that adding a list of columns (more than 2) would constrain the model in respect of the combinations of categories represented by those columns but it does not seem to work. I see nulls generated where there should be none

Steps to reproduce

I can’t share my example due to privacy etc. but imagine that a toy example where a list of three categories passed to uniquecombinations would probably generate the same result.

[tjhallum]

This issue of nan values being generated is NOT just limited to cases in which a UniqueCombinations constraint is applied to more than two columns. The generation of nan values likewise occurs when the UniqueCombinations constraint is applied to only two columns. A full reproducible example is provided in the form of a Jupyter Notebook attached to this issue: #521.

[csala]

A full explanation of what caused this issue has been added in the PR that closed it.