[fxttr]

I used to hate SQL when I was a backend engineer. I had difficulties understanding exactly this mental model and tried to avoid writing SQL by using ORMs. At some point I “accidentally” switched to data engineering and was forced to get to grips with SQL. I think that in the meantime I warmed up to a functional programming style which helped me to some extent.

[whatever1]

The thing that trips everyone is that you cannot explain the SQL outputs. In sequential code, you can step through and understand why the logic error produced erroneous output.

With SQL, it is a trial and error. When your query passes your sniff tests, you sign looks good to me, and you ship it to the world. Only to silently break shortly afterwards without any warning.

In enterprise, I am convinced at this point that all of the complex ETL jobs are vending wrong outputs. Just nobody has the tools to diagnose and fix the problems.