Some themes in downstream packages are currently not valid

[paleolimbot]

In the revdep check (#3303), packages that define complete themes fail because the themes don't contain all required elements (several were added by #2934).

• esmisc: https://github.com/EDiLD/esmisc/blob/master/R/theme_edi.R#L13
• fingertipscharts: https://github.com/cran/fingertipscharts/blob/master/R/theme.R#L7
• jcolors: https://github.com/jaredhuling/jcolors/blob/master/R/ggplot2_themes.R#L2
• xpose: https://github.com/UUPharmacometrics/xpose/blob/master/R/themes_gg.R#L29 and https://github.com/UUPharmacometrics/xpose/blob/master/R/themes_gg.R#L46

In addition, some rstan dependencies fail on stan_plot() (breathteststan) and stan_hist() (fergm). rstan defines the default theme at package build time instead of within a function, which might be the cause of these failures.

[clauswilke]

I think these are all cases where all we can do is inform the package maintainers that they need to fix their code.

@paleolimbot Could you reach out to them? And also recommend to them the %+replace% pattern for development of new themes?

ggplot2/R/theme-defaults.r

Lines 245 to 252 in 3b69172

	# Starts with theme_grey and then modify some parts
	theme_grey(
	base_size = base_size,
	base_family = base_family,
	base_line_size = base_line_size,
	base_rect_size = base_rect_size
	) %+replace%
	theme(

[paleolimbot]

On it.

Is there a reason that using validate = TRUE doesn't check for all required elements when complete = TRUE? Right now this is failing at draw time with a cryptic error message, rather than when the theme is created and validated.

[clauswilke]

Many parts of the theme code need revising, including how validation works, but that's for another day. I plan to look into it seriously once this release is out the door.

[paleolimbot]

Sorry, one more theme-related question. Is there a way to use something that isn't an infix operator to replace theme elements? With %+replace% I have to suggest that people do

#' @importFrom ggplot2 %+replace%

which means that some packages that have ggplot2 in Suggests may have to put it in Imports (I think?). I suppose they could also do:

`%+replace%` <- ggplot2::`%+replace%`

[paleolimbot]

All of those packages already import ggplot2 so I rolled with it. For my brain to keep track (or in case anybody is curious):

• esmisc: Forthcoming ggplot2 release and esmisc eduardszoecs/esmisc#2
• fingertipscharts: Forthcoming release of ggplot2 and fingertipscharts ukhsa-collaboration/fingertipscharts#10
• jcolors: Forthcoming release of ggplot2 and jcolors jaredhuling/jcolors#2
• xpose: Forthcoming release of ggplot2 and xpose UUPharmacometrics/xpose#157
• rstan: Forthcoming release of ggplot2 and rstan stan-dev/rstan#631

[clauswilke]

It's unclear to me why xpose breaks. There's something I don't understand yet. They use the recommended pattern and it should work.

[paleolimbot]

I should have checked xpose more carefully...those themes aren't the problem

[clauswilke]

Thanks!

[tungttnguyen]

I didn't see you mentioned ggthemr but it broke after #2934 too

library(ggthemr)
ggthemr('dust')

#> Warning: New theme missing the following elements: axis.ticks.length.x,
#> axis.ticks.length.x.top, axis.ticks.length.x.bottom, axis.ticks.length.y,
#> axis.ticks.length.y.left, axis.ticks.length.y.right

[paleolimbot]

It looks like you've already noted this in Mikata-Project/ggthemr#31 . I'm not sure why our checks didn't catch ggthemr, but feel free to discuss on the ggthemr issue.

[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/