Information to list about tools

[jbednar]

The draft tools listing at https://pyviz.org/tools is already very useful, providing lots of information about each project that helps users make sense of them:

What else can/should it show?

• Institutional sponsor? (See Adding the tools dashboard #4). Seems useful for understanding how projects relate to each other, seeing which ones have financial support for multiple developers, which ones are focusing on specific agendas, etc.
• Year of first release? Seems useful for understanding maturity levels and technologies
• Date of latest release? Would need to be updated automatically to be useful
• Underlying library (for high-level libraries)
• Domain (for domain-specific libraries)
• Technology? (Native GUI, JS, JSON, WebGL, OpenGL, SVG, PNG, D3) Would be very helpful, but hard to pin down objectively. Plus many libraries support many of the technologies, which could be hard to track. Maybe would need to make a draft listing and see if it's something everyone would agree on? Would take a bit of work.
• Data size limits? Seems a bit slippery; depends on how much time people think is reasonable to wait. Some libraries have explicit limits, others vary depending on browser and compute capabilities. Hard to make this objective, even though it's measurable in principle.

My personal vote is to remove coverage to make room, then include sponsor, underlying library, and year info, and punt on the rest but consider adding them someday.

• It would also be nice to be able to sort by each column (category, name, stars, contributors, year released, downloads, license), with the first view being categories as shown above, but the rest just being a table (e.g. sorted by year of first release or whatever the user selects).

[jbednar]

Also note that the tools are currently (6/2019) sorted in each category by their rough popularity, using the pypi downloads, conda downloads, stars, and contributors columns. This ordering is currently static, generally having been computed on the day each library was added to the list, but it could presumably be automated based on some formal metric. Across these projects, there are relatively few cases where using one or the other of these metrics would give different orderings; e.g. popular projects tend to have a lot of stars, lots of contributors, lots of PyPi downloads, and lots of Conda downloads. The biggest discrepancies I can see are between PyPi and Conda downloads -- some projects are clearly used more by Conda users, and others more by PyPI users.

Conda download numbers are tricky because of the various channels through which packages can be obtained; I think the current numbers include the main Anaconda channels, conda-forge, and bioconda, not any of the other channels. The holoviz tools are officially distributed using the pyviz channel, not the indexed channels, and other projects may have similar issues. If we could make a single merged "total downloads per month" badge I think I'd favor using that as the sorting order, but hopefully the rough approximate ordering is ok for now.

[jsignell]

In order to sort them by live measures of the total downloads per month, we would need to go to the source instead of using the badge for pypi downloads. This is the source: https://pypistats.org/api/packages/{package_name}/recent

This would be kind of nice to do anyways because then we could coerce the download counts onto the same colormap which would improve legibility of the badges.

[jbednar]

For now I've used the updated matching counts to sort them all by current total downloads per month in #20, which is at least objective, but it won't be updated over time unless we add the above capability.

[jbednar]

I think all of the above is still relevant in 2026, alas.

I am updating this issue to reflect a talk I gave on PyViz at PyTexas 2026 today, and I found these categorizations to be helpful:

• InfoVis vs. SciVis: Abstract data vs 3D real-world data
• Plot Types: Statistical vs Image/Array/Mesh vs Geo vs Networks
• Data size: 10K to 100K points, vs 1 million to 10 billion
• UI/publishing: Static PNG/PDF vs HTML vs Native GUI
• API type: Imperative vs Object-oriented vs Declarative
• Technology: What it’s built on (PDF, JS/HTML, Qt, WebGL, D3/Vega)
• Goal: Plotting vs Tables vs Colormapping vs Dashboarding
• Support: Popularity, Community, Docs, AI support (MCP?)

With that in mind, there are a couple of tasks that someone could take on:

1. Get AI to fill in the various categories of information listed above, for everything in tools.yml. Adding this information by hand the way the existing tools.yml was written would be a tall order for the 175 or so libraries currently covered by PyViz. But maybe we could use Generative AI models, particularly those with tool support like Claude, to do the work for us? I.e. take each of the libraries in tools.yml, then look up any of the information outlined in this issue and create a new tools.yml with the new fields.

2. Once we do that, we'd need a better way to display the data, since the existing tools.yml page is already overwhelmingly wide, ideally with selectable, hideable, sortable columns. Should be able to find a good JS library for that.

Anyone feel like volunteering for either of these two tasks, which can be independent of each other?