Pandas has officially launched version 3.0.0, and I'm curious about the experiences others have had with the beta or alpha versions. Are there any major breaking changes that we should be aware of? I'm looking to gather as much information as possible about this release!
5 Answers
I’ve been disappointed with some recent design choices in Pandas. They seem to prioritize flexibility, which might be tough for core data science users. For example, the .sum() function concatenates text columns, which isn’t user-friendly. Polars handles this better by ignoring non-numeric columns altogether. While Pandas is trying to be more 'pythonic,' I think they could lose some of their base users.
Good point! But who actually needs to sum text fields anyway?
Honestly, I'm trying to figure out why I should stick with Pandas 3.0 instead of making the switch to Polars. Some of the issues with Pandas have been addressed, but it doesn't seem compelling enough for someone who’s already considering Polars.
There's that, but I've got XML files I need to work with and Polars can't handle them yet, which keeps me tied to Pandas.
Agreed! Plus, if you're using GIS data, GeoPolars isn't stable at the moment, so that's a strike against switching.
Let’s be real—Polars is where it’s at!
If you're looking for details on the changes, check out the release notes: https://pandas.pydata.org/docs/dev/whatsnew/v3.0.0.html
Did you mean this link? https://letmegooglethat.com/?q=pandas+3+release+notes
Wow, real helpful... Not.
One notable change is that every indexing step now creates a copy. This means that 'chained assignment' will no longer function the same way, which will break some existing code. On a positive note, this change removes the 'SettingWithCopyWarning' since developers won't need to make defensive .copy() calls anymore. Overall, while some users may find this disruptive, the change could lead to better code practices!
Yeah, that makes sense! Breaking existing code can be tough, but it should help clean up a lot of confusion surrounding assignments.
Definitely! It might annoy some at first, but I'm hoping it ends up being a win in the long run.
Totally agree! But also, won't most of us just filter out object-type columns when summing? It can look messy, but it's manageable.