Instead of resampling down the stinkbug image down to 64x64 via pillow code (which is a bit distracting for the tutorial, especially if you start having version gating like here), perhaps simpler would be to just crop down to its central region? e.g. img[175:225, 225:275] seems reasonable. (Obviously the surrounding text needs adjusting too.)

To be fully honest, I didn't read another line of the code except for the problematic one...

Doing that, do we actually need to specify the resampling scheme? I mean, it is really a Pillow-function that we are using, so maybe it is not that important? (Default is bicubic.)

(Btw, we also provide an, untested, thumbnail function, although that reads the file, possibly pops it up, and then saves it, as well as returning a Figure.)

I didn't really pay attention to the context either.

I changed it to not specify resampling scheme (as that sort of makes sense as well and my previous approach was wrong).

It seems like the pikepdf install is suddenly failing...

6.0.0 was released yesterday with a 6.0.0.post1 release two hours ago. Maybe it is just to wait for wheels to be published?

Windows and OSX support is currently blocked: pikepdf/pikepdf#390

Pin version (except for Linux)?

Pin version (except for Linux)?

Yes