Tim Hoffmann

I'm fine with the global solution for matplotlib (which motivated the PR). The disadvantage from sphinx-gallery's perspective is that every library has to activate tat setting themselves (and know about...

> Whats a reasonable time across the widget examples to aim for? For release builds, the time does not matter. Only size may need to be considered. Our current CI...

> If we really think that 99.9% of the users who pass plot(['1', '2', '3']) mean for that to be a float array, such that we are willing to issue...

I think we all agree that - if there was only the unintended strings case, an immediate warning (I.e. warnings.warn) is more helpful than an opt-in (logging). - opt-in warnings...

@mwaskom Do I understand you correctly? You intentionally send string-tryped numbers. But you still want us to warnings.warn on them? Do you plan to filter the warning?

A warning at runtime is certainly helpful. I’m tempted to deprecate labels-only completely. The documentation will get harder to understand with correct description of container order. And even without the...

> I think `legend.remap/rename_labels` per [#27725 (comment)](https://github.com/matplotlib/matplotlib/issues/27725#issuecomment-1920452030) is relatively simple and not likely to lead to confusion. May be ok. We need to make clear though that legend labels are...

The printing via ghostscript did not have the same user experience and options as a native print dialog. It was more like a workaround. Printing is not one of the...

Thanks for the explanations. In case it helps, I provide the exact case that triggered me opening this issue. In the example in https://github.com/matplotlib/matplotlib/issues/27005 the build runs without error. The...

Thanks for the comments. Makes sense. `__dtype__` may even be something to be formalized in the array API standard. Do you have any opinion on whether one should first bring...