Pandas may be good for adhoc examing some data but I have doubts whether it works well inside a tool like this. It's a bit tricky to code at times, the error messages are usually a bit hard to understand (e.g. a non-empty take on axes?), is it even more performant? I wonder whether flat python + numpy wouldn't just yield simpler code that's easier to debug.
In GitLab by @timdiels on Apr 3, 2019, 17:46
Pandas may be good for adhoc examing some data but I have doubts whether it works well inside a tool like this. It's a bit tricky to code at times, the error messages are usually a bit hard to understand (e.g. a non-empty take on axes?), is it even more performant? I wonder whether flat python + numpy wouldn't just yield simpler code that's easier to debug.