plt_add()

[vincentarelbundock]

I’ve been trying to think about ways to make it easier to layer plots. We have add=TRUE, but it is not very convenient, because we need to repeat all the same arguments in every call.

I can’t decide if this is a good or a terrible idea, but here’s a proof of concept for you to comment on:

1. Every time we do plt(), we save the full call somewhere (global option or other).
2. New function plt_add()
   • Takes the last call,
   • Modifies the explicitly-given argument
   • Sets add=TRUE
   • Evaluates the modifed call

This proof of concept PR allows things like:

library(tinyplot)

plt(Sepal.Width ~ Sepal.Length | Species,
    data = iris, 
    type = type_lm())

plt_add(type = "p")

Note: Facets don’t seem to work in this PR.

[grantmcdermott]

This was exactly what I was thinking of in the digression here: #236 (comment)

Great minds etc. ;-)

I really like the idea and think that this is a feature that would get a lot of use. I'll be able to take a look at it tomorrow (or whenever you move it out of draft).

[vincentarelbundock]

@grantmcdermott I added docs and a test. You might want to look at this now.

[grantmcdermott]

Looks great. I requested some changes that should be relatively minor to action (hopefully).

[grantmcdermott]

To be consistent with the rest of the package infrastructure, let's rather save this to the existing .tinyplot_env. This is what we use in R/get_saved_par.R, for an example:

tinyplot/R/get_saved_par.R

Lines 104 to 116 in 02d7806

	#' @export
	get_saved_par = function(when = c("before", "after")) {
	when = match.arg(when)
	par_env_name = paste0(".saved_par_", when)
	return(get(par_env_name, envir = get(".tinyplot_env", envir = parent.env(environment()))))
	}
	# (non-exported) companion function(s) for setting the original pars
	set_saved_par = function(when = c("before", "after"), value) {
	when = match.arg(when)
	par_env_name = paste0(".saved_par_", when)
	assign(par_env_name, value, envir = get(".tinyplot_env", envir = parent.env(environment())))
	}

We'll need to assign NULL values to this call object at startup in R/zzz.R too, e.g.

tinyplot/R/zzz.R

Line 46 in 02d7806

assign(".saved_par_after", NULL, envir = get(".tinyplot_env", envir = parent.env(environment())))

[vincentarelbundock]

Unfortunately, I was unable to get this to work. I kept getting an error in the tinyplot.density tests. The message is obscure, having to do with a do.call and a substitute. I just can figure this out.

My guess is that this will disappear by itself when we refactor the density code.

In the meantime, I left TODOs in both files with the appropriate save code that we can eventually.

Feel free to take a shot at it if you want, but I have to give up for now.

[grantmcdermott]

This is great. Thanks @vincentarelbundock!