Will Crichton

Pandas is such a unique learning experience. It's the only tool where my mental model changes substantively every time I use it. Yet, I find Pandas hard to "properly" learn because all the tutorials are boring stock market examples.

There's infinite cases where I can articulate the high-level task, and have no idea how to accomplish it without a for loop & iterrows.

out if curiosity, have you tried taking the tasks you do with for loop and iterrows and tried to translate them into something more "pandas native"?

My knee jerk is that you probably want to break up operations instead of doing them all at once. like, maybe you're using a for loop to accumulate into variables, but you could make new columns and then aggregate them; something like that

The iterrows is a bit of an exaggeration. These days, it's usually -- I can't figure out a simple thing with a multi-index, so I do a complicated thing with a flat table. In the outlier example, I couldn't figure out broadcasting semantics between tables of different sizes.