Integrate pyplot interface with reflex-webapps [FEATURE-REQUEST]

[MaximilianNuber]

Dear all,
thank you for this great package. I am interested in using this package for genomics plotting wrappers. I would be interested in notebook-based plots, where xy.pyplot has crucial functionalities xy itself doesn´t have, like imshow, or plot composition where xy.pyplot creates its own xy.pyplot.Figure.subplots layer, and xy itself would depend on reflex, which is made for webapps.

We would be in webapp applications as well. But I haven't found a way to usethe pyplot interface and then put the plots into reflex for the webapp, or notebook-based composition and imshow for xy itself. Alternatively, put an xy plot on an xy.pyplot axis.
I would like to implement one wrapper, which can be used for both notebooks with composition and the specific imshow in xy.pyplot, as well as webapps.
Is this of interest to be implemented? Or did I completely miss something?

Thank you and best regards, Max

[Alek99]

Hey Max, thanks for the kind words! A couple of things you might have missed that make this easier than it looks:

xy itself doesn't depend on Reflex. Reflex is just an optional extra (pip install xy[reflex]), and the adapter lives in a separate reflex_xy package. Core charts render straight into Jupyter / VS Code / Colab / Marimo: https://github.com/reflex-dev/xy/blob/main/docs/integrations/notebooks.md

For imshow, core has xy.heatmap(z, x=..., y=...). Takes a 2-D matrix or an RGB(A) array. If there's a specific imshow knob you lean on (origin, extent, interpolation, aspect) let me know which and I'll check it's honored.

For composition, there are two options right now:

• xy.facets for small multiples with shared domains
• in a Reflex app, one chart per panel and lay them out with regular Reflex components (rx.grid, rx.hstack, etc). That's the intended pattern, and it gives you the same thing a pyplot subplots grid would.

So if you want one wrapper that works in both notebooks and web apps, I'd build it on the core xy API and keep xy.pyplot for porting existing matplotlib code.

On the bridge: you're right that you can't hand a pyplot Figure to reflex_xy today. Under the hood it already builds one core chart per Axes, so exposing that (something like fig.charts() you could pass into reflex_xy.chart) is pretty doable and I'm open to it. The other direction, dropping a core chart onto a pyplot axis, isn't something we'd do since pyplot sits on top of core rather than next to it.

If you can share a small genomics example (array shapes + the layout you're after) I'll point you at the exact calls and see if anything is actually missing.