Technically true, but sql has the ability to induce discipline - every single time I’ve seen folks given the opportunity to do the analysis off a database or wa

Technically true, but sql has the ability to induce discipline - every single time I’ve seen folks given the opportunity to do the analysis off a database or warehouse and use pandas/spark, they almost always end up writing bad code, processing one row or cell at a time. Almost always, they could have done it with sql itself and gotten much better performance but this “extra flexibility” allowed them to be lazier. Sql can also have the same issue with feature creep (snowflakes array columns often invites people to write pathetic sql trying to join on this col for example) but with the right design you can force your users to interact with your data more efficiently.