I personally find the Python plotting landscape a mess and confusing. It always seems to be a lot harder than is necessary to do anything and make it look good.
I find the big benefit of Altair is that the API is so nice and composable, and because vega provides a ton out of the box to build really high quality visualisations.
It's also great being able to add Altair shapes to Folium for geoplotting (as the vega geoplotting is a bit more low-level).
What I really think is missing in the ecosystem is a "vega for tables", so you could be rich, interactive tables with a similar grammar. That would rock.
yes ... I feel like everyone is stuck in a local minima (or maxima, depending on your optimisation direction!) of using matplotlib which is "good enough" but not very good. So there isn't great momentum to improve it, but nobody really likes it. My favorite plotting syntax is that exposed in BeakerX [1], but that's less common than everything else put together.
yep- far too many libraries. Matplotlib, seaborn, bokeh, plotly, altair, ggplot2, vega, now this. And each library covers about 90% of your use cases on its own but you're constantly having to switch to another one for the remaining 10%. Nightmare.
