joelcarlson / rimagepalette Goto Github PK

This is a big problem!
It seems that ggplot2::discrete_scale() calls the palette (RImagePalette::image_pal()) twice, once for the colors on the plot, and once for the legend. Since there is some randomness in median cut, this often results in a different palette, and thus different colors.

Yikes.

Need to set the seeds properly somewhere in image_pal

I just tried this out and according to the image_palette() documentation, you have to hand the function an image matrix from readJPEG() or similar.

I didn't already have the jpeg package installed and had to install it manually. As far as I can tell, there is no way to create the color palette without reading an image into an object first. If so, should the package make sure jpeg is installed?

If yes, I'm happy to make a small pull request with the change to the DESCRIPTION file.

Other than that, the color palette and examples in the README are great. This will be useful 👍 .

Currently, the implementation of median cut is in R, and uses lists. This is not efficient as every time you modify a list in R it is copied.

Converting this to C++ would greatly speed up the process. Would also allow use of push and pop, which are missing from R lists.

Output palettes are not always particularly vibrant. An option to convert to a more attractive colorspace would be nice.

That is, in the median cut algorithm cut across LAB space, rather than RGB space.

eg.

x <-hex2RGB(lifeAquaticPalette)
as(x,"LAB")

https://cran.r-project.org/web/packages/colorspace/colorspace.pdf