I think there is a potentially common use case where lessons could be generated from a 3+ column sparse database.
An example is cooking conversions, where the columns would be units: tsp, tbsp, c, pt, qt, gal, ml, l and rows are added until the value “1” is represented in all of the columns. In the rows where tsp and tbsp are 1, I’d expected that the questions created for a column could be filtered by leaving the columns for the larger units blank. The actual behavior is a question or answer that is “” (nothing)
I split the db so the cup - gallon conversion lesson wouldn’t end up including nonsense like 1/12 cup = 1/100 gallon. If I could just leave gallon empty in the tsp row, and have memrise know to ignore that rows that contain a blank value, it would make setting up combinatorial lessons way easier.