[R] hcl()

[Prof Brian Ripley]

On Fri, 24 Jun 2005, Martin Maechler wrote:

> I have written a nice (IMO) function that lets you explore the
> hcl space quite nicely, and show its calls.
>
> hcl.wheel <-
>    function(chroma = 35, lums = 0:100, hues = 1:360, asp = 1,
>             p.cex = 0.6, do.label = FALSE, rev.lum = FALSE,
>             fixup = TRUE)
> {
>    ## Purpose: show chroma "sections" of hcl() color space; see  ?hcl
>    ## ----------------------------------------------------------------------
>    ## Arguments: chroma: can be vector -> multiple plots are done,
>    ##            lums, hues, fixup : all corresponding to hcl()'s args
>    ##            rev.lum: logical indicating if luminance
>    ## 			should go from outer to inner
>    ## ----------------------------------------------------------------------
>    ## Author: Martin Maechler, Date: 24 Jun 2005
>
>    stopifnot(is.numeric(lums), lums >= 0, lums <= 100,
>              is.numeric(hues), hues >= 0, hues <= 360,
>              is.numeric(chroma), chroma >= 0, (nch <- length(chroma)) >= 1)
>    if(is.unsorted(hues)) hues <- sort(hues)
>    if(nch > 1) {
>        op <- par(mfrow= n2mfrow(nch), mar = c(0,0,0,0))
>        on.exit(par(op))
>    }
>    for(i.c in 1:nch) {
>        plot(-1:1,-1:1, type="n", axes = FALSE, xlab="",ylab="", asp = asp)
>        ## main = sprintf("hcl(h = <angle>, c = %g)", chroma[i.c]),
>        text(0.4, 0.99, paste("chroma =", format(chroma[i.c])),
>             adj = 0, font = 4)
>        l.s <- (if(rev.lum) rev(lums) else lums) / max(lums) # <= 1
>        for(ang in hues) { # could do all this using outer() instead of for()...
>            a. <- ang * pi/180
>            z.a <- exp(1i * a.)
>            cols <- hcl(ang, c = chroma[i.c], l = lums, fixup = fixup)
>            points(l.s * z.a, pch = 16, col = cols, cex = p.cex)
>            ##if(do."text") : draw the 0,45,90,... angle "lines"
>            if(do.label)
>                text(z.a*1.05, labels = ang, col = cols[length(cols)/2],
>                     srt = ang)
>        }
>        if(!fixup) ## show the outline
>            lines(exp(1i * hues * pi/180))
>   }
>   invisible()
> }
>
> ##-- and now a few interesting calls
>
> hcl.wheel() # and watch it redraw when you fiddle with the graphic window
> hcl.wheel(rev.lum= TRUE) # dito
> hcl.wheel(do.lab = TRUE) # dito
>
>
> ## Now watch:
> hcl.wheel(ch = c(25,35,45,55))
>
> hcl.wheel(ch = seq(10, 90, by = 10), p.cex = 0.4)
> hcl.wheel(ch = seq(10, 90, by = 10), p.cex = 0.3, fixup = FALSE)
> hcl.wheel(ch = seq(10, 90, by = 10), p.cex = 0.3, rev.lum = TRUE)
> x11() # new device -- in order to compare with previous :
> hcl.wheel(ch = seq(10, 90, by = 10), p.cex = 0.3, rev.lum = TRUE, fixup=FALSE)
>
> ## the last two, in my eyes show that
> ## 1) fixup = TRUE {the default!} works quite nicely in most cases
> ## 2) Robin's original problem was a sample of a much larger "problem"
> ##    where IMO the 'fixup' algorithm ``breaks down'' and I
> ##    think should be improvable.

It shows that there are slow but massive hue shifts at low luminance.
There are better algorithms than the C code uses but maybe not much better 
for colours way out of gamut.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595