You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
UltraPlot bundles scientific colormaps but I find myself often reaching and failing with regards to discrete colormaps. I think we should explore adding more qualitative colormaps to the UltraPlot family. Examples may be
UltraPlot already has built-in colormaps, so maybe optional bindings would be better than adding data to maintain.
The cmap package supports to_matplotlib().
It would be convenient to be able to set cmap.qualitative: tol:bright-alt in ultraplotrc if cmap is installed.
To elaborate on this. We register all matplotlib colormaps and then create a new database that will be contain a reference to matplotlib's registry. I think this would be best placed within cmap to registry cmaps directly to matplotlib, then it would automagically work also here (without an ordering effect).
I hadn’t thought of registering them as mpl colormaps and then setting them in ultraplotrc.
cmap is designed to depend only on numpy.
It can export to matplotlib, but as far as I know it has no feature to call matplotlib.colormaps.register() itself.
When I registered it manually, the cmap.qualitative: tol:bright setting worked.
That kind of usage works for me.
Activity
pratiman-91 commentedon Oct 5, 2026
@cvanelteren
Docs: https://pratiman-91.github.io/colormaps/
Git Repo: https://github.com/pratiman-91/colormaps
is this helpful?
beskep commentedon Oct 13, 2026
I highly recommend Paul Tol's color schemes, which are colorblind friendly and have a reasonable lightness contrast.
Also, the cmap package has extensive catalogue of colormaps.
cvanelteren commentedon Oct 13, 2026
Yeah I wonder if we should bundle them or just keep references to these resources on the docs.
beskep commentedon Oct 13, 2026
UltraPlot already has built-in colormaps, so maybe optional bindings would be better than adding data to maintain.
The cmap package supports
to_matplotlib().It would be convenient to be able to set
cmap.qualitative: tol:bright-altin ultraplotrc if cmap is installed.cvanelteren commentedon Oct 16, 2026
@beskep isn't that just an order thing? If cmap can register it to mpl we can load them if they are already in their registry
cvanelteren commentedon Oct 16, 2026
To elaborate on this. We register all matplotlib colormaps and then create a new database that will be contain a reference to matplotlib's registry. I think this would be best placed within cmap to registry cmaps directly to matplotlib, then it would automagically work also here (without an ordering effect).
beskep commentedon Oct 16, 2026
I hadn’t thought of registering them as mpl colormaps and then setting them in ultraplotrc.
cmap is designed to depend only on numpy.
It can export to matplotlib, but as far as I know it has no feature to call
matplotlib.colormaps.register()itself.When I registered it manually, the
cmap.qualitative: tol:brightsetting worked.That kind of usage works for me.
cvanelteren commentedon Oct 16, 2026
I think cmaps should do this on import then you can reference them by string directly in mpl or ultraplot
cvanelteren commentedon Oct 22, 2026
So over at
cmapthey added registration features, but auto registrations is not implemented. Not sure if we should proceed to add it here with a dep.