Guides with duplicate labels do not get merged properly

[MartinEarle]

I have come across a small issue with manually selecting colour and linetype for my data. I want to manually select values for these aesthetics and relabel them. In this instance, I am showing the "treatment" from a factorial experiment with the label, and an additional factor with the colour. The linetype distinguishes the experiment from the control. Below is my code:

ggplot(df,aes(date,mean))+ 
  geom_line(size=0.5,aes(linetype=sample_name,colour=sample_name))+
  theme_classic()+
  theme(legend.margin=margin(0,0,0,0),legend.box.margin=margin(0,-10,-8,-10),legend.text=element_text(size=11),legend.position="top")+
  scale_x_date(expand=c(0,0),labels=date_format("%m/%d"),breaks = seq.Date(as.Date(date_min),as.Date(date_max),
                                                                              by = "month"),limits = as.Date(c("2018-1-1","2018-9-26")))+
  scale_y_continuous(limits=c(6,9),expand = c(0,0),breaks = seq(6,9.,0.5))+
  scale_colour_manual(labels = c("A", "B", "C", "D", "E", "F"),breaks=c("Inflow","Filter 3","Filter 7","Filter 8", "Filter 9", "Filter 10"),values = c('Black','#004f71','#4c839b','#99b8c6','#C0504D','#d28482'))+
  scale_linetype_manual(labels = c("A", "B", "C", "D", "E", "F"),breaks=c("Inflow","Filter 3","Filter 7","Filter 8", "Filter 9", "Filter 10"),values=c("dashed","solid","solid","solid","solid","solid"))+
  labs(color="", linetype="")+
  xlab("Date")+ylab("Data")

This results in the following plot:

As you can see, the labels for C & D are both duplicated. This does not occur if I rename the second "C" and "D" to "E" and "F", as shown below, or if I hide one of the guides (not shown).

I can provide some data to recreate this, if desired.

Thank you!

[paleolimbot]

Welcome @MartinEarle!

I can't seem to reproduce this...perhaps you could pair down your example and render it using the reprex package? The reprex I was using to try to replicate your error is here:

library(ggplot2)

ggplot(mpg, aes(cty, hwy, col = drv, linetype = drv)) +
  geom_line() +
  scale_colour_manual(
    breaks = c("4", "f", "r"),
    labels = c("4", "f", "r"),
    values = c("red", "blue", "green")
  ) +
  scale_linetype_manual(
    breaks = c("4", "f", "r"),
    labels = c("4", "f", "r"),
    values = c("dashed", "solid", "solid")
  ) +
  labs(color = "", linetype = "")

[MartinEarle]

Hi @paleolimbot!

Sorry, I did not understand how Reprex worked before, but I have now recreated my issue now with the mpg dataset and have copied it below:

library(ggplot2)
#> Warning: package 'ggplot2' was built under R version 3.6.1

mpg$class <- factor(mpg$class, levels = c("pickup", "midsize","compact", "subcompact",  "minivan",  "suv", "2seater"))

ggplot(subset(mpg, class != "2seater"), aes(cty, hwy, col = class, linetype = class)) +
  geom_line() +
  scale_colour_manual(
    breaks = c("pickup", "midsize","compact", "subcompact",  "minivan",  "suv"),
    labels = c("pickup","medium", "small", "small",  "large", "large"),
    values = c("red", "blue", "green", "black", "yellow", "orange")
  ) +
  scale_linetype_manual(
    breaks = c("pickup", "midsize","compact", "subcompact",  "minivan",  "suv"),
    labels = c("pickup","medium", "small", "small",  "large", "large"),
    values = c("dashed", "solid", "solid", "solid", "solid", "solid")
  ) +
  theme(legend.position = "top")+
  labs(color = "", linetype = "")

^{Created on 2019-10-17 by the reprex package (v0.3.0)}

I am using R 3.6.0.

[paleolimbot]

Just simplifying this a tiny bit...the determining factor looks to be that there are two manual scales that are trying to be merged where any(breaks != labels).

library(ggplot2)

breaks <- c("pickup", "midsize","compact", "subcompact",  "minivan",  "suv", "2seater")
labels <- c("pickup","medium", "small", "small",  "large", "large", "2seater")

ggplot(mpg, aes(cty, hwy, col = class, linetype = class)) +
  geom_line() +
  scale_colour_manual(
    breaks = breaks,
    labels = labels,
    values = c("red", "blue", "green", "black", "yellow", "orange", "purple")
  ) +
  scale_linetype_manual(
    breaks = breaks,
    labels = labels,
    values = c("dashed", "solid", "solid", "solid", "solid", "solid", "solid")
  )

^{Created on 2019-10-17 by the reprex package (v0.2.1)}

[paleolimbot]

Upon closer inspection, it looks like it's a general problem with duplicate labels and how guides are merged (using .$labels as an implicit key for merge() doesn't work when there are duplicates):

ggplot2/R/guide-legend.r

Lines 232 to 233 in 115c396

	guide_merge.legend <- function(guide, new_guide) {
	guide$key <- merge(guide$key, new_guide$key, sort = FALSE)

guide1 <- list(
  key = tibble::tibble(
    colour = 1:4, 
    .label = c("low", "low", "high", "high")
  )
)

guide2 <- list(
  key = tibble::tibble(
    linetype = 1:4, 
    .label = c("low", "low", "high", "high")
  )
)

ggplot2:::guide_merge.legend(guide1, guide2)$key
#>   .label colour linetype
#> 1    low      1        1
#> 2    low      1        2
#> 3    low      2        1
#> 4    low      2        2
#> 5   high      3        3
#> 6   high      3        4
#> 7   high      4        3
#> 8   high      4        4

^{Created on 2019-10-17 by the reprex package (v0.2.1)}

We could (1) change the way merging happens (cbind() should work, given that the breaks and labels are hashed to determine if the guides ever get merged), or (2) enforce unique labels.

As a workaround, one could recode the values before they get to ggplot2 (e.g., using fct_recode()).

[clauswilke]

Please don't enforce unique labels. There are valid cases for duplicated labels, e.g. making multiple labels equal to "" to zero them out. Wouldn't the correct approach merge on breaks rather than labels? Breaks have to be unique by definition (I think).

[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/