deprecate sample_data module#1725
Conversation
This comment was marked as resolved.
This comment was marked as resolved.
maximlt
left a comment
There was a problem hiding this comment.
My main comment is that the original change to the sample_data module wasn't a deprecation. It was direct code removal. Removing a feature has to go through a well-defined process (see HEP2) with discussion, deprecation, and finally removal. I added the sample_data code back. In a future release, we will be able to remove the module and datasets.yaml (that was forgotten here) entirely.
My second comment is that I find some of the replaced plots not always meaningful. I feel like that when we use datasets based on real data (e.g. stocks, penguins), we need to make realistic plots. Otherwise, we risk losing a bit of credibility with our users. I've listed these plots below. Soon (hopefully) these user guides should go away, so I'm ok merging these changes as they are meant to be temporary. Let's keep in mind though for the future that we have to make sure we generate realistic plots.
Plotting with Bokeh
That's not a very pretty plot.

I don't think that it makes so much sense.

Not the most common kind of bar plot.

hexbin plots are better displayed imo with an equal aspect ratio.

Can't really see much on this one.

Plotting Extensions
The first plot isn't great.
Subplots
Lots of empty cells in this subplot, might look a bit off for visitors.

I think there's a missing GIF for the landing page.
|
Just to clarify @Azaya89, the only change that I would like is a new GIF for the landing page. Everything else is good to be merged. |

Fixes #1723
Description
This PR deprecates the
sample_datamodule and recommendssampledatainstead.AI Disclosure
Tools and Models: Copilot Claude Sonnet 4.6
Checklist