Cannot have a legend without a pattern in it.

[nfparsons]

When I plot using fills and patterns I get image 1.

What I'm looking for is image 2.

Here's my code:

dat.fac.p %>% filter(visn == 1) %>% ggplot(aes(x = station, y = value, group = visn)) + geom_col_pattern( aes(pattern_density = plot_pattern), fill = 'white', pattern = 'crosshatch', pattern_spacing = 0.02 ) + geom_col( aes(fill = plot_fill), color = 'gray50', alpha = 0.5 ) + scale_pattern_density_manual(name = element_blank(), values = c(a = 0.1, b = 0), labels = c("Positive", "")) + scale_fill_viridis(name = element_blank(), discrete = TRUE, labels = c("Screened", "Not Screened")) + theme_ipsum()

[complexgenome]

can you share sample data?

[M-Colley]

Maybe related:

ggpattern::scale_pattern_fill_manual(values=c(scale_fill_colors4[2],scale_fill_colors4[3]), breaks=c("example1", "example2"), values= c("test1", "test2"))

gives me two legends instead of one where the legends is correctly named (i.e., in this case "test1" etc)

[jkeast]

I'm having the same issue. I copied the pie graph example from the Geometry-based Patterns vignette to illustrate and added a new variable to map the fill aesthetic to. This is what I get:

Are there any work-arounds? Modifying pattern_key_scale_factor isn't helping.

[jkeast]

I found a work-around with override.aes! Here's the code, last line is the important one:

df <- data.frame(
  group = factor(c("Cool", "But", "Use", "Less"), levels = c("Cool", "But", "Use", "Less")),
  value = c(10, 20, 30, 40),
  other = factor(c("a", "b", "c", "d"),levels = c("a", "b", "c", "d"))
)

ggplot(df, aes(x="", y = value, pattern = group, pattern_angle = group))+
  geom_bar_pattern(aes(fill=other), width = 1, stat = "identity", colour = 'black', pattern_aspect_ratio = 1, pattern_density = 0.3, pattern_key_scale_factor = .5) +
  coord_polar("y", start=0) +
  theme_void(20) +
  theme(legend.key.size = unit(1, 'cm')) +
  labs(title = "ggpattern::geom_bar_pattern() + coord_polar()")+
  guides(fill = guide_legend(override.aes = list(pattern = c("none", "none", "none", "none"))))

and the fixed plot:

[trevorld]

@nfparsons, @jkeast seems to provide a solution for legends with "two separate groups" separately controlling color and patterns case. An alternative solution would be to create 3 groups: "Positive" (and screened), "Screened" (and not positive), "Not Screened" and then use the crosshatch pattern for the "Positive" group and the "none" pattern for the other two groups and then color the "Not Screened" group yellow and the other two groups purple.