Transitive Dependency in spark Dataframe data

Question

I have a spark data frame in the following format.

pid         grouped_ids
------------------------
12          12,13,14,78
6           6,8,12,23
19