R help - Nov 2020 - RGB -> CYMK, with consistent colors

On Sun, 29 Nov 2020 at 20:55, Derek M Jones <derek at knosof.co.uk> wrote:
[...]
>
> library("colorspace")
>
> two_c=rainbow(2)
> x=runif(20)
> y=runif(20)
> plot(x, y, col=two_c[2])
> pdf(file="cmyk.pdf", colormodel="cmyk")
> plot(x, y, col=two_c[2])
> dev.off()
rainbow(2) gives two RGB-colours: #FF0000 (=pure red) and #00FFFF (=pure
cyan).
So it should be no problem to match two_c[2] perfect in CMYK-colourspace

However after some testing.
I totally agree that CMYK handling in R using pdf(..., colormodel =
"cmyk")
is not correct.

One solution:
Output in tiff or png instead and then convert to CMYK using external tools.

Try this:
png("plot.png")
plot(runif(10), runif(10), col =  ""#00FFFF") # cyan
dev.off()

And then convert plot.png using:
https://www.online-utility.org/image/convert/to/CMYK

Regards
Martin

	[[alternative HTML version deleted]]

Martin,
> However after some testing.
> I totally agree that CMYK handling in R using pdf(..., colormodel =
"cmyk")
> is not correct.
I thought the issue may be OS specific, but I get similar behavior on a Mac.

I have discovered the Cyan tool: http://cyan.fxarena.net/
and will try it out.

As various people have pointed out, RGB -> CMYK conversion is never going
to be perfect.  It just so happens that my book contains many plots that use
red and cyan, and a good conversion to cyan is most of what I am after.

 > One solution:
 > Output in tiff or png instead and then convert to CMYK using external
tools.

There are 600+ figures.  There are png versions here:
http://knosof.co.uk/ESEUR/figures/index.html

Martin,
> However after some testing.
> I totally agree that CMYK handling in R using pdf(..., colormodel =
"cmyk")
> is not correct.
Looking at the source code of the function PostScriptSetCol
in file src/librarygrDevices/src/devPS.c
the conversion to CMYK looks correct.

Looking at the generated pdf (qpdf: https://github.com/qpdf/qpdf is
needed to uncompress the pdf)
the numeric color values look correct, although the pdf contents don't
exactly
match those generated by PostScriptSetCol

It is possible the problem is in the conversion that occurs when
displaying/printing a pdf.