Another thing I remember us talking about for a long time, but never got around to shipping: it would be really good to have some kind of built-in logging mechanism as part of Tumult Analytics, so various methods can print messages at multiple levels (e.g. DEBUG, INFO, WARNING and ERROR).

We should probably adopt a standard pattern for this and don't need anything fancy, though it would be good to have a check around the lines of "you seem to have initialized the SparkSession without disabling Spark's built-in logging feature, so your log is going to be noisy regardless of the log level you chose".