retostauffer/python-colorspace

API Method `hclplot()` from Original R Package Missing?

Closed this issue · 12 comments

avanavana commented

Though the documentation for the original R package says of this python port:

For Python users a beta re-implementation of the full colorspace package in Python 2/Python 3 is also available

I cannot seem to use the hclplot() function, and searching the codebase here turns up nothing for it either.

Am I missing something? Is it here under some other alias, or is it just missing from this port?

zeileis commented

Thanks for the report and apologies for not responding sooner!

Indeed hclplot() is currently not ported yet. The main reason is that some statistical modeling is used to get the interpolation and extrapolation of HCL coordinates if there is not one coordinate that is exactly fixed. This can be done easily in base R without introducing any additional dependencies but not in Python. Hence we put that on our wishlist for the time being.

Reto @retostauffer, maybe we should flag this issue as an open wishlist item here?

retostauffer commented

Thanks @zeileis and sorry @avanavana for the late response. As @zeileis wrote, implementing the hclplot() does require a bit more than just a few lines of code, we've put it on the whishlist for now, hoping to get back to it in the near future.

retostauffer commented

@zeileis getting there, tough not easy to implement as I had to/have to re-implement some R functions from base/stats without adding any more dependencies such as e.g., scipy/statsmodels (not using it so far). Working on it (devel branch; wip). Not the very same results as in R, but close (so far only for sequential_hcl, but) I think we can get it done.

image

retostauffer commented

Added hclplot to the main branch, tough additional testing and adjustments will be required. A few examples can be found here:

zeileis commented

This looks all fantastic, really cool 🚀

In the palette_visualization article hclplot is apparently not imported yet when running hclplot(qualitative_hcl("Dynamic"));. Hence this throws an error at the moment.

  • 👍1
zeileis commented

Regarding the comment:

In R, the colors where C > 0 and L < 1 is TRUE are set to NA, but not the nd object. Thus, the title contains the Hue range from all the colors including those which got killed by fixup = TRUE.
Bug or feature?

This is a good question. I'm pretty sure that this is oversight. I think it would make more sense to compute the hue range without the dropped hues.

retostauffer commented

Agree, before we create conflicts: Shall I fix it quickly?

zeileis commented

Oh, and one more minor and non-important detail that I noticed: If luminance is on the y-axis then by default only axis labels for 0, 20, 40, 60, 80 are shown but not 100. Maybe this is easy to fix. If not, nevermind.

zeileis commented

I'm not working on anything atm, so feel free to go ahead and fix.

retostauffer commented
  • Title: I'll take care of updating R colorspace (but not today).
  • Axis labels: Will check it, matplotlib default, I assume it can be fixed by adjusting the upper limit by a small amount (e.g., 5% as R does).
  • 👍1
retostauffer commented
  • R: Adjusted the R code (svn).
  • Axis labels (Python): Related to the R fixes, choosing y/x-limits on "valid colors only", thus the upper limit was 99 (not 100). Adjusted that by rounding (floor/ceil) to the next 2.5 interval. For diverging palettes the x-axis is now also only showing positive values on the x-axis for the 'left hand side'.
  • Error in article: Missing imports, fixed.
  • 👍1
retostauffer commented

I am closing this issue for now, the hclplot() has been implemented for the different objects and released with 0.4.4 just now.

Please open a new issue in case problems occur or additional features are needed.

  • ❤️1