Heads up: duplicate aesthetics will get warning

[teunbrand]

Hi Barett and other maintainers,

We recently did a revdepcheck in ggplot2 and found that a new check we introduced got triggered by GGally. In particular we now throw a warning which violates this test:

ggally/tests/testthat/test-ggsurv.R

Lines 72 to 74 in d21c6d6

	expect_silent(
	ggsurv(sf.kid, CI = TRUE, surv.col = c("black", "red", "blue", "green"))
	)

I believe the warning is thrown because lty is used twice in these layer definitions, once in aes() and once as a fixed parameter:

ggally/R/ggsurv.R

Lines 333 to 343 in d21c6d6

	pl <- pl +
	geom_step(
	aes(y = .data$up, lty = .data$group, col = .data$group),
	lty = stepLty,
	linewidth = size.ci
	) +
	geom_step(
	aes(y = .data$low, lty = .data$group, col = .data$group),
	lty = stepLty,
	linewidth = size.ci
	)

We've no immediate plans to release a ggplot2 that includes the check, but just wanted to simply flag this for the future.

Cheers!

[schloerke]

Thank you!

---

Local testing shows that if we remove aes(lty= .data$group), the plots are the same. We should keep the geom_step arg.

Before making any changes, we should verify that an error is thrown in the test when using the dev version of ggplot2

[smouksassi]

aesthetics specified outside of aes() will always override and win.
in a more complex situation (Shiny) global or inside or outside of aes might be NULL:

library(ggplot2)
ggplot(mtcars,aes(cyl,mpg,col=NULL))+
  geom_point(aes(col=ggplot2::cut_interval(mpg,2)))+
  geom_rug(aes(col=factor(cyl)),col=NULL)
#> Warning in geom_rug(aes(col = factor(cyl)), col = NULL): Ignoring empty
#> aesthetic: `colour`.

^{Created on 2026-02-02 with reprex v2.1.1}