Apply lightness, saturation etc. to functions

[hughjonesd]

Great package.

It would be nice to be able to write e.g.

my_pal <- scales::colour_ramp(c("red", "blue"))
# my_pal is a function that takes numbers and returns colours 
my_pal <- lightness(my_pal, 100)
my_pal(6) # return the lightened colours.

In other words to allow calling lightness, saturation and friends on functions.

One way to do this is as follows:

lightness <- function (shades, values = NULL) UseMethod("lightness")

# original lightness becomes lightness.default

lightness.function <- function(shades, values = NULL) {
  orig_call <- match.call(expand.dots = FALSE)
  orig_call[[1]] <- quote(lightness)
  shades_q <- substitute(shades)
  wrapper <- function (...) {
    inner_call <- match.call(expand.dots = FALSE)
    inner_call[[1]] <- shades_q
    result <- eval(inner_call)
    orig_call$shades <- result
    eval(orig_call)
  }
  formals(wrapper) <- formals(shades)
  
  return(wrapper)
}

I've got a proof of concept if you think this is a cute idea.

An obvious extension is to e.g. ggplot colour scales.

[jonclayden]

Hi David, and thanks for the suggestion. I think the functions returned from scales::colour_ramp expect arguments in the [0,1] range, so my_pal(6) doesn't produce anything useful. But, aside from that, if we have

my_pal <- scales::colour_ramp(c("red", "blue"))
my_pal2 <- lightness.function(my_pal, 100)

(using your lightness.function), then my_pal2(seq(0,1,0.25)) is the same as lightness(my_pal(seq(0,1,0.25)), 100) or my_pal(seq(0,1,0.25)) %>% lightness(100).

Is there a good reason to prefer the two-step higher-order approach? The practical benefit of this isn't obvious to me, so a usage example would be very helpful.

[hughjonesd]

A good use case would be tinkering with plots. For example, if you use scale_colour_viridis(), then you may not have direct access to the colours it uses. Similarly, you may want to change a colourRamp and have it affect a number of plots, where it is spitting out different colours each time.

I often work from the command line, adding bits to my plot commands. It would be nice to just do

ggplot(data, aes(foo, bar)) + lightness(scale_colour_whatever(), 50) 

rather than having to hunt down the specific colours and use scale_colour_manual.

[hughjonesd]

So, I just came across this very example. Here's a plot I'm working on, and I am thinking, great colours, I wish they were a little darker. This comes from scale_color_brewer(type = "qual").

[jonclayden]

Thanks – I see the need. I'll look into getting this running when I have a bit of time.

[jonclayden]

PR #7 is now merged, but I will keep this issue open until a full solution is committed.

[hughjonesd]

Drop me a line when it’s up on CRAN, and I’ll blog about it. The package really fills a gap.

[jonclayden]

OK, the package is now updated so that all the property functions, including opacity, plus complement, addmix and submix, should accept palette functions or scales as their first arguments. As a bonus, gradient() will now return a function if its second argument is missing or NULL, which can then be manipulated further with the other functions.

I think this should handle your requirements, but if you'd be willing to give it a quick test that would be very helpful.

[hughjonesd]

Will do. Sounds awesome! David

…

On Mon, 19 Nov 2018 at 12:01, Jon Clayden ***@***.***> wrote: OK, the package is now updated so that all the property functions, including opacity, plus complement, addmix and submix, should accept palette functions or scales as their first arguments. As a bonus, gradient() will now return a function if its second argument is missing or NULL, which can then be manipulated further with the other functions. I think this should handle your requirements, but if you'd be willing to give it a quick test that would be very helpful. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#5 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AFjJ9-xWw6bzEijxh7AwUl4zKjIkBlnvks5uwp2pgaJpZM4X9dqw> .

[hughjonesd]

This seems to work great. I'll tell you if anything comes up.

[jonclayden]

Great! It's ready for CRAN so I'll get it sent off. If you run into any issues I can always do a point release.

[jonclayden]

The update is now released as version 1.3.0, and I've added an example of darkening existing colours (as above) to the README.