Is polars + plotnine currently the best tool on Earth for complex plotting? Or does the crown still go to the tidyverse?
I dropped ggplot like a rock for it's obnoxious ideological limitations. Sometimes I need to draw two series on the same plot and I am not interested in fighting the library for it or reading some redditor quote Tufte at me when he tells me I don't need to, actually.The whole grammar idea is the biggest autist-bait. It really sounds like a good idea to start with and I can see how they ended up going hard on it but it's only actually good for medium-simple applications. It's too complex for very simple plots and it's too much of a pain in the ass to contort it into very complex plots. At some point you're better off just imperatively adding stuff to axes or hell even just drawing it manually in publisher for one offs.
>>107833389That's an interesting opinion. What is your preferred plotting library these days?
>>107832613Polars+plotnine excels in speed for large datasets and Python integration. The tidyverse retains the crown for ecosystem maturity, seamless grammar, and breadth of statistical plotting. The "best" depends entirely on whether your workflow is Python or R-centric.
>>107833389It's not a big deal to pipe through melt and you retain all the other flexibility of ggplot. Procedural plotting always becomes a giant fucking mess.I'd agree that if you're making a few very important charts for publication you should be as manual as possible, but actual investigations involve a lot of adhoc plotting and every python solution I've tried is shit for that. In the time you polished up one beautiful python plot (using matplotlib's ugly api because everything else has limitations) the guy with ggplot made 100 different cross sections and filters to understand every nuance of your dataset. tidyverse still mogsPolars vs data.table is a harder question though, polars is so nice as an api but somehow data.table ends up getting shit done faster for most uses
I expect python libraries to get better more than R libraries because data science is going to be increasingly focused on ML more than any other field and python, mostly by accident than anything, ate R's lunch when it comes to doing ML
>>107834660Didn't post image.
>>107834660If it was that simple it would've happened by now, python has dominated ML for a long time. It has a ton of great libraries available but somehow they are all still worse than ggplot. desu it's possible that they simply made too many instead of working on polishing one good one. The other curse is that the only common ground between the mess of libraries (and thus what works with everything) is matplotlib. which is ass
>>107834733Judging by the thread (unless its just you samefagging), it already happened. Who here is talking good things about tidyverse? Pic unrelated but made with matplotlib and seaborn.
Libreoffice calc
