stat_function() produces incorrect warning

[clauswilke]

stat_function() generates a warning whenever a mapping is provided, but then it uses the mapping as normal.

These are the relevant lines in the code, and a reprex follows below.

ggplot2/R/stat-function.r

Lines 68 to 71 in 3870c12

	# Warn if supplied mapping and/or data is going to be overwritten
	if (!is.null(mapping)) {
	warning("`mapping` is not used by stat_function()", call. = FALSE)
	}

library(ggplot2)

ggplot(data.frame(x = 0), aes(x = x)) +
  stat_function(aes(color = "linear"), fun = function(x) x) +
  stat_function(aes(color = "quadratic"), fun = function(x) x*x) +
  xlim(-5, 5)
#> Warning: `mapping` is not used by stat_function()

#> Warning: `mapping` is not used by stat_function()

^{Created on 2019-11-04 by the reprex package (v0.3.0)}

I'm not sure whether there are cases where the warning is useful. I would propose to delete it. It's probably too complicated to try to find out whether the mapping is meaningful or not in the given context.

[thomasp85]

Could it be tweaked to say that positional mapping is ignored?

[clauswilke]

I guess we could inspect the mapping and create a warning if positional aesthetics are used.

[clauswilke]

Pinging @paleolimbot as he suggested the warning in the first place. Any thoughts on this from you? In the linked discussion, I see mentioning of the group aesthetic, which also may not work as expected. Should we issue a warning if x, y, or group are used?

[paleolimbot]

Whoops, that was me! Using stat_function() with multiple layers + a manually specified aesthetic should have occurred to me, but didn't...I also didn't know that missing grouping aesthetics from the output of compute_panel() are added back in in compute_panel().

I seem to remember that the issues are caused when stat_function() draws multiple copies of itself because of unknowingly specified groups. Specifying x and y doesn't cause errors but also don't do anything (unless one of them is discrete). Maybe we could (1) warn for mapped x or y in the constructor and/or (2) do a simple override of compute_panel() that just calls Stat$compute_panel() and makes sure that there is exactly one group?

[jan-glx]

A related issue arises with geom_abline (with the difference that the warning is correct in that case -- mapping is ignored):

library(ggplot2)

ggplot(data.frame(x = 0), aes(x = x)) +
  geom_abline(intercept=0, slope = 1, aes(color="true value"))

ggplot2 * 3.3.0.9000 2020-01-22 [1] Github (tidyverse/ggplot2@214f314):

geom_abline(): Ignoring `mapping` because `slope` and/or `intercept` were provided.

ggplot2 * 3.2.1 2019-08-10 [1] CRAN (R 3.6.1):

Warning message:
In geom_abline(intercept = 0, slope = 1, aes(color = "true value")) :
  Using `intercept` and/or `slope` with `mapping` may not have the desired result as mapping is overwritten if either of these is specified

But here a simple workaround is possible by specifying intercept and slope through mapping:

ggplot(data.frame(x = 0), aes(x = x)) +
  geom_abline(aes(intercept=0, slope = 1, color="true value"))

---

Not sure if this belongs here, but I'm leaving this here since that was the only relevant page that showed up googling above error message and using stat_function as described here would pose another workaround (with warning) (and above workaround would help avoid the warning for stat_function in case of a linear function).

---

Some thoughts on this issue itself:
Wouldn't thinks be much easier/consistent if fun of stat_function was an aesthetic & inherit.aes would be FALSE (like in geom_abline)?
Shouldn't the warning disappear at least for stat_function(..., inherit.aes = FALSE)?

[tvatter]

What is the status of this one? I second the latest suggestion: why is the warning not disappearing when inherit.aes = FALSE ?

[clauswilke]

Just came across this on Stackoverflow:
https://stackoverflow.com/q/61505199/4975218
Maybe we can still fix this for 3.3.1.

@paleolimbot If the concern is that some mappings create multiple copies of the function, then I'm not sure the current warning addresses that issue, because most people won't make the connection between the warning and that problem. Instead, it seems to me that what we need is a geom_function() that works like geom_line() but ensures it draws only once per panel and issues a warning otherwise. If you agree I can take a stab at implementing this.

[paleolimbot]

@clauswilke absolutely the right way to go! Let me know if there's something I can help with to get the patch release out the door.

[clauswilke]

@paleolimbot Please take a look here: #3982 and let me know what you think.