Definitions

[almarklein]

Let's be clear on some terms, so we can be more precise in our discussions. Will update this top post as needed.

Terms

• viz-library: e.g. Datoviz, Pygfx, vispy.visuals, Matplotlib,Plotly, Bokeh, etc.
• gpu-viz-library: Datoviz, Pygfx, and vispy.visuals.
• plotting-api: a plotting API, either as a new library, or a subpackage of a viz-library (like vispy.plot)
• unified-plotting-api: a (new) library providing a plotting API using multiple (gpu-)viz-libraries as backends. It unified multiple viz-libratries.
• backend: in the context of these discussions, this will typically be a viz-library.
• Middleware/intermediate/internal API: an idea of Cyrille, to be continued ...

A unified-plotting-api can follow two approaches:

• Intersection: only use features that all its backends support. This means less features, but all code works on each backend.
• Union: also expose features specific to certain backends. Some code will work on one backend, but not another.

Problem statement

Because of the limited time we had to write the grant proposal, we've rushed this part, and it has never been made very explicit. But it will help discussions if we can nail this down more precisely.

I think we have two problems that we're trying to address to some extend:

1. Vispy, Datoviz and Pygfx all could use a plotting-api (lets combine the effort).
2. We have an ecosystem of awesome visualization libraries, but there is not one library that can address all tasks, which means that scientists must learn multiple APIs to get things done. Examples could be: performant graphics, making publication quality plots, creating (interactive) visualizations for on the web.

I think we've started with (1), but moved somewhat to (2).

Goal

Still not 100% clear.

Approach

Not fully clear yet, but is also part of the grant to figure this out.