My jupyter notebook stopped working so I just tried out tech.ml.dataset for the first time on a real-world dataset! My unsolicited opinion: • Filtering rows, adding/removing columns, and joining all behave as I would expect (coming from someone with 10+ years in Python and R). It's a nice combination of pandas and Clojure IMO. ◦ In all of the above, TMD is less verbose than Pandas since we don't have to repeat the name of the dataset if we use the thread macro. In Pandas we often have to repeat the dataset variable 2-3 times per operation ◦ In particular, creating new columns based on if/else statements is so natural in TMD and so un-natural in pandas • I really appreciate that I finally have a one-liner for inspecting CSVs in Clojure. • The most odd function I came across was group-by-column-agg ◦ A bit verbose getting the values of the columns I grouped on and having to repeat the dataset variable for each aggregation field ▪︎ (EDIT: Nevermind, the https://techascent.github.io/tech.ml.dataset/tech.v3.dataset.reductions.html#var-group-by-column-agg was weird but when I tried writing it the more intuitive way, it also worked!) Didn't mess around with dates yet but this has been a great first experience and I think I'll be able to drop pandas much sooner than I expected!