The road to 2.0 - We need your input!

[cvanelteren]

We’re excited to announce that we’re planning the next major version of UltraPlot: Version 2.0! To make sure we’re building the features and improvements that matter most to you, we’d love to hear your thoughts and ideas.

Your feedback will directly influence the roadmap for UltraPlot 2.0, so this is your chance to have a say in its future!

How You Can Help

1. Fill Out Our Survey
   We’ve created a short Google Form to gather your input. It should take less than 5 minutes to complete:
   👉 Fill out the survey here

2. Share Your Thoughts Below
   If you prefer, feel free to leave a comment on this issue with your ideas, feature requests, or feedback.

Poll will be closed 1-9-2025.

[gepcel]

I've noticed the import is a bit slow. Is there any chance you may speed it up? You can see what I mean by running this code:

C:\Users\user\Downloads>python calc.py
Takes 0.000 seconds to : import os
Takes 0.093 seconds to : import numpy as np
Takes 0.303 seconds to : import pandas as pd
Takes 0.117 seconds to : import geopandas as gpd
Takes 1.156 seconds to : import ultraplot as pplt

[cvanelteren]

@gepcel I think this is due to many dynamical overrides that still exist. For example, the colormap database is recreated on import, the RC is explicitly loaded and overridden to name a few examples. I suspect that moving the explicit loading to lazy will make it a lot more responsive, but would require a lot of work I am afraid. Especially when comparing to matplotlib

Takes 0.000 seconds to : import os
Takes 0.051 seconds to : import numpy as np
Takes 0.213 seconds to : import pandas as pd
Takes 0.057 seconds to : import geopandas as gpd
Takes 0.775 seconds to : import ultraplot as pplt
Takes 0.000 seconds to : import matplotlib.pyplot as plt

[cvanelteren]

@gepcel we have internal benchmarks and my suspicion is confirmed (showing import times):

 python -c "import ultraplot"
pyplot: 0.014441541992709972s
cartopy: 2.792992745526135e-06s
basemap: 0.04044700099620968s
config: 0.18994424300035462s
proj: 0.00033307600824628025s
utils: 8.589995559304953e-06s
colors: 0.03315775000373833s
ticker: 0.008052692006458528s
scale: 0.0005433689948404208s
axes: 0.02696201400249265s
gridspec: 0.0011417740024626255s
figure: 0.0026695639971876517s
constructor: 6.14600139670074e-06s
ui: 0.00023187500482890755s
demos: 0.0006735549977747723s
cmaps: 0.2350437560089631s
cycles: 0.0030879160040058196s
colors: 0.02799595199758187s
fonts: 0.004024214998935349s

[gepcel]

I suspect that moving the explicit loading to lazy will make it a lot more responsive, but would require a lot of work I am afraid. Especially when comparing to matplotlib

Maybe cache it somehow to avoid recreating every time?

Anyway, thanks for the response. I can live with this.

[cvanelteren]

[cvanelteren]

@beckermr ^ pinging just for visibility for the results.

[Yefee]

Curved vector plot?
matplotlib/matplotlib#20038