I'm nowhere near pycon and I'm no expert in coding but I am spending a lot of time thinking about what a modern python plotting library should look like - if it were designed around users first, prioritising consistency, minimalism, intuition, low cognitive load, conceptual coherence ..
.. matplotlib/pyplot is powerful but very few would say they love the experience of trying to use it.
@pyOpenSci@rzeta0 I once heard about Vega and some data-viz abstractions language across several viz libraries but I couldn't grok any of it. I find matplotlib hard too...
@rzeta0 @pyOpenSci
Yep ggplot is the way, no other plotting system ive ever seen comes close. Vega and Altair are cool but the API is awkward and it couples together some of the wrong things and doesnt allow for the depth of hacking that the underlying grid system does in ggplot2
@rzeta0 @pyOpenSci
I tried it for a little bit when it first popped up, but havent looked since. The R Implementation is very mature and has a lot of surrounding ecosystem with the tidyverse. What I do is basically use python for everything up until the point where I have roughly tabular data and then dump it over into R and plot it there
@pyOpenSci we won’t be at pycon but we are getting a lot of contributor interest in developing type annotations throughout the library, we have had great discussions in our issue tracker and will hopefully start committing code to that effort!
We recognize we're deficient in many areas. Since we're a pure python package, the packaging bits while a bit outdated (still using setup.py, need to migrate to pyproject.toml), but our CI is in good shape. Our documentation needs mountains of content updates, and our test coverage is somewhere between "well at last you tried" and "lol". Our example app (python -m pyqtgraph.examples) does a lot of heavy lifting but it is amazing.
@pyqtgraph@pyOpenSci typing has been so so helpful for us in our stravalib maintenance! i didn't understand how much so until we implemented it. this is awesome!!
@leahawasser@pyOpenSci Most of the type-annotations should be straight forward, we need to introduce some aliases for complex/lengthy type unions, but we also have some class methods that are dynamically composed, and how to annotate those is a head scratcher, at a glance there are no good solutions.
If anyone sees this has some ideas, PLEASE chime in on the linked issue below.
Add comment