Added builton field

[jbednar]

First step to addressing part of #7 , which is to declare which libraries are built on other libraries. Added a "builton" field in the .yml, filling it out for nearly all the libraries. The current values include 20 libraries built on matplotlib, 7 on bokeh, etc.:

• 20 matplotlib
•   7 bokeh
•   6 plotly
•   5 opengl
•   4 vtk
•   2 webgl
•   2 vega
•   2 leaflet

Issues:

1. For now, I counted bokeh, matplotlib, plotly, and vtk as building on themselves, but of course that's up for discussion.
2. Are there any other technologies we should list? My goals here are to (a) help people find the most suitable API for the tool they want to use, (b) understand how the various tools fit together, and (c) get a feeling for the capabilities of each library (e.g. 3D support, browser-based interactivity, etc.). Unfortunately, (c) is a slippery slope -- there are so many things that might be shown (SVG support, Canvas vs WebGL, PNG output, etc.), so I've focused on goals (a) and (b) here. Just listing any tiny fraction of what matplotlib is based on would be daunting!
3. Of course, declaring this information is only the first step; it also needs to be shown in the actual rendered site. I was imagining using a compact logo for each one like , displayed just as for the Sponsor field. But once there are lots of logos they may not be as recognizable and distinguishable as matplotlib/bokeh/plotly are, which is another reason to limit the number of technologies to list. Note that each library may be built on many technologies; so far the maximum is 3 in the current declarations.

[jbednar]

4. Another issue: Jupyter and ipywidgets. Several of the libraries build on ipywidgets, and others support Jupyter without using ipywidgets (bokeh, holoviews, and panel, at least, plus maybe matplotlib). Not sure if it's possible to convey something useful about such support concisely.

[tacaswell]

👍 I really like this idea.

For the next layer down, I would put is "embeddedable" or "widget technology" with things like {Qt, Wx, Tk, Gtk, OSX, react, vue, classic-notebook, jupyter-widgets, html-self-contained, html-with-a-kernel, ...} to let people know "I am already using technology X for a UI, which plotting tools will seamlessly mix in?".

An entry for "output format" with things like {png, ps, pdf, svg, html}.

Using "builton" you can inherit this as well?

[jbednar]

Yes, "UI technology" and "Output formats" would also be great to list. Matplotlib is presumably the worst case in both categories, with at least 5 UI technologies and 6 output formats. It's difficult to cram that many things into a single-line table, but it's probably ok if only the core libraries have that sort of jumbled formatting.

[jsignell]

Sorry for the extremely late response, but I like this idea a lot. I do think that we need to be careful about how much into we try to cram into a table though :) If we add more than a builton column we'll need to get rid of an existing column to make space.

[jsignell]

LGTM

[jbednar]

I'll go ahead and merge this now so that the .yml file won't drift out of sync with master, but for these values to take effect we'll need changes to the build scripts. I don't currently have any volunteers ready to attack that. It doesn't seem too difficult, by just copying the current support for sponsors:

• Create a set of tiny logos (same height as e.g. doc/_static/badges/numfocus.png) for each builton target and put them in doc/_static/badges/
• Copy tools/sponsors.yml to builton.yml and replace the "label" with the builton target, the url with the appropriate target's url, and the logo with the path to that target's badge
• Adapt tools/build.py to have similar support for builton as for sponsors
• Copy and adapt the sponsors support in tools/template.html, renaming as needed for builton
• Test by putting "[build:website_dev]" into a commit message in a PR and examining https://pyviz-dev.github.io/website/ once the build completes
• Once it works, build on the main website by merging the PR and putting "[build:website_release]" into a commit message

[maximlt]

@jbednar this is now deployed (https://pyviz.org/tools.html)

[jbednar]

Wow, that looks great! Thanks for hitting that home!