Discussion: Skyborn as a reference for more NCL-like curved_quiver rendering

[QianyeSu]

Hi UltraPlot team,
I noticed that UltraPlot already provides curved_quiver, introduced in v1.63.0, and I think this is a very valuable addition.

I would like to suggest considering Skyborn as a reference implementation for possible future improvements to curved_quiver, especially for meteorological vector-field visualization.

In Skyborn, I implemented a Matplotlib-compatible curly_vector function aimed at producing a more NCL-like visual style: curved, flow-following vectors that can look more natural and readable than standard quiver, and in some cases more visually coherent than streamline-like approaches.

I wanted to raise the possibility that, if some aspects of Skyborn’s rendering style are considered visually better than the current curved_quiver output, they might be useful as reference ideas for future refinement.

This may be especially relevant for:

• meteorological wind-field plots
• cyclone and anticyclone structure
• dense lat-lon vector fields
• users who want a style closer to NCL vector rendering

For reference, below are a few example plots produced with this approach.

Thanks.

[cvanelteren]

Thanks for sharing this, and for pointing us to Skyborn as a related implementation. We are open to collaboration and happy to look at external implementations when they help identify a concrete improvement for UltraPlot.

At the moment, though, this is difficult for us to evaluate as a feature request. As written, it mainly asks us to consider Skyborn as a reference for future improvements, but it does not yet show where Skyborn is clearly improving on UltraPlot's current curved_quiver.

We generally prefer to evolve UltraPlot around clearly demonstrated shortcomings or missing capabilities, rather than adopting another package's rendering style by default.

A narrower package focused specifically on meteorological workflows may very well make tradeoffs that feel better in those cases, and we are open to learning from that. But from this issue alone, we cannot yet tell whether the difference is:

• a genuine rendering improvement over UltraPlot
• a matter of parameter tuning or styling choices
• a workflow difference outside the scope of curved_quiver

What would make this much easier for us to evaluate is:

• a side-by-side comparison against UltraPlot's current curved_quiver using the same dataset, projection, sampling, and styling choices
• a short explanation of which specific aspects you think are better and why
• a narrowed proposal for what should change in UltraPlot, for example integration direction, glyph anchoring, handling of dense lat-lon fields, or support for non-uniform/profile grids
• ideally, a minimal reproducible example we can run locally

In other words, we are not dismissing Skyborn as a source of ideas, but we would need a concrete demonstration of what it does better before we could justify treating it as a reference for future UltraPlot work.

If you would like to collaborate on that, feel free to follow up with a direct comparison or a small focused PR.