Fix theme inheritance when not inheriting from blank

[clauswilke]

I think this fixes #3662. I still want to add a few additional unit tests to prevent future such regressions.

I had to update one vdiffr reference image because it wasn't drawn correctly previously. (Line styling wasn't properly inherited across a blank element.)

library(ggplot2)

ggplot(mtcars) + 
  geom_point(aes(disp, mpg)) + 
  facet_wrap(~gear) + 
  theme(strip.text = element_blank(), strip.text.x = element_text())

^{Created on 2019-12-11 by the reprex package (v0.3.0)}

[clauswilke]

The bayesplot examples seem to all run now. Not sure if any of them look wrong.

library(bayesplot)
#> This is bayesplot version 1.7.1
#> - Online documentation and vignettes at mc-stan.org/bayesplot
#> - bayesplot theme set to bayesplot::theme_default()
#>    * Does _not_ affect other ggplot2 plots
#>    * See ?bayesplot_theme_set for details on theme setting

color_scheme_set("blue")
color_scheme_view()

color_scheme_get()
#>      blue
#> 1 #d1e1ec
#> 2 #b3cde0
#> 3 #6497b1
#> 4 #005b96
#> 5 #03396c
#> 6 #011f4b
color_scheme_get(i = c(3, 5)) # 3rd and 5th colors only
#> $mid
#> [1] "#6497b1"
#> 
#> $dark
#> [1] "#03396c"

color_scheme_get("brightblue")
#>   brightblue
#> 1    #cce5ff
#> 2    #99cbff
#> 3    #4ca5ff
#> 4    #198bff
#> 5    #0065cc
#> 6    #004c99
color_scheme_view("brightblue")

# compare multiple schemes
color_scheme_view(c("pink", "gray", "teal"))

color_scheme_view(c("viridis", "viridisA", "viridisB", "viridisC"))

color_scheme_set("pink")
x <- example_mcmc_draws()
mcmc_intervals(x)

color_scheme_set("teal")
color_scheme_view()

mcmc_intervals(x)

color_scheme_set("red")
mcmc_areas(x, regex_pars = "beta")
#> Warning: `expand_scale()` is deprecated; use `expansion()` instead.

color_scheme_set("purple")
color_scheme_view()

y <- example_y_data()
yrep <- example_yrep_draws()
ppc_stat(y, yrep, stat = "mean") + legend_none()
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

############################
### Mixing color schemes ###
############################
color_scheme_set("mix-teal-pink")
ppc_stat(y, yrep, stat = "sd") + legend_none()
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

mcmc_areas(x, regex_pars = "beta")
#> Warning: `expand_scale()` is deprecated; use `expansion()` instead.

##########################
### ColorBrewer scheme ###
##########################
color_scheme_set("brewer-Spectral")
color_scheme_view()

mcmc_trace(x, pars = "sigma")

###########################
### Custom color scheme ###
###########################
orange_scheme <- c("#ffebcc", "#ffcc80",
                   "#ffad33", "#e68a00",
                   "#995c00", "#663d00")
color_scheme_set(orange_scheme)
color_scheme_view()

mcmc_areas(x, regex_pars = "alpha")
#> Warning: `expand_scale()` is deprecated; use `expansion()` instead.

mcmc_dens_overlay(x)

ppc_stat(y, yrep, stat = "var") + legend_none()
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

^{Created on 2019-12-11 by the reprex package (v0.3.0)}

[clauswilke]

@jgabry @tjmahr Could you review the bayesplot examples in the preceding comment for any rendering errors? We did some refactoring of the ggplot2 theme code and I want to make sure there aren't any further regressions.

[thomasp85]

LGTM

[jgabry]

Sorry for not chiming in sooner. LGTM too, and thanks for reaching out!