Warn when a supplied mapping is going to be overwritten by geom_hline / vline / abline

[RichardJActon]

This pull request addressed issue #2945

geom_hline / vline / abline will now throw a warning if the user supplies both a x|y|intercept/slope value and a mapping such as this:

geom_hline(yintercept = 3, aes(colour = colour))

As the above will cause the user supplied mapping to be ignored/overwritten

Users might resaonably expect the above to work as other geoms do not have this behaviour. The warning messages suggest that the x|y|intercept/slope value(s) be moved inside the aes call thus:

geom_hline(aes(yintercept = 3,colour = colour))

[yutannihilation]

Is it OK to recommend to place values directly in aes()? In this case, it's fine, but I'm a bit afraid this may be a bit confusing because using direct values in aes() is not always good.

Maybe preparing data is the right thing? Or considering the hacky nature of the "annotation mode", is this allowed specially?

[clauswilke]

Maybe preparing data is the right thing?

Actually, geom_hline() will overwrite the dataset if one is provided, so this won't work. I've been wondering whether the code block here:

ggplot2/R/geom-hline.r

Lines 13 to 17 in 3550772

	if (!missing(yintercept)) {
	data <- data.frame(yintercept = yintercept)
	mapping <- aes(yintercept = yintercept)
	show.legend <- FALSE
	}

should only be executed when data is NULL.

[yutannihilation]

I meant not providing *intercept and maping all the variables.

data2 <- data
data2$y <- 3

geom_hline(data = data2,
           aes(yintercept = y, colour = colour))

But, please ignore my comment above. I think I don't get the setantics of variable mappings yet. Sorry for the noise.

[RichardJActon]

Are any further changes needed here?

• I have placed the warning block in the existing check for a missing intercept value and removed the redundant test for the intercept.
• I removed the plot from the test and just checked the geoms directly
• I altered the error message so it does suggest the solution of placing a value directly in a mapping

(I'm not sure what's up with travis - the error says something about exceeding the API rate limit?)

[clauswilke]

I made a few specific comments. We would also need an entry in NEWS.md.

[clauswilke]

This cannot currently be merged because the code in geom_abline(), geom_hline(), and geom_vline() has changed.

@RichardJActon Would you mind updating this? Could you please also add a bullet to NEWS.md? It should briefly describe the change and end with (@yourname, #issuenumber).

[RichardJActon]

Hi @clauswilke,
Was there a concensus for implementing your suggestion to only overwrite the data if none is provided?

At present with just adding the warning message this won't break anyones code but if someone has supplied data that is currently being overwritten silently this change might cause them issues.

I expect I can update this later today but do we want it with or without the change to the data overwrite behaviour?

[clauswilke]

We just opened a separate issue about the overwriting of data: #3090 I suggest you finish this PR as it currently stands.

[clauswilke]

Thanks!

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/