Ordering plotly bars in a grouped-split bar chart - r

Question

I have a data.frame with counts for several groups that are assigned to several loci from two ages:
set.seed(1)
df <- data.frame(group = c(rep("G1",2),rep("G2",2),rep("G3",2)),
locus = c(rep("A",4),rep("B",2)),
age = rep(c("2m","24m"),3),
n = as.integer(runif(6,10,100)),
stringsAsFactors = F)
I want to plot these data in a bar chart using plotly in R, where the x-axis is group, the bars are split by age, and colored by the combination of locus and age.
I set group, locus, and age as factors to give them the order I want them to folow in the figure:
df$group <- factor(df$group, levels = c("G1","G2","G3"))
df$locus <- factor(df$locus, levels = c("A","B"))
df$age <- factor(df$age,levels = c("2m","24m"))
Now I'm creating a data.frame with the specific colors I'd like each of the locus-age combination to have:
library(dplyr)
color.df <- data.frame(locus = c("A","A","B","B"), age = rep(c("2m","24m"),2), color = c("#66C2A5","#488A75","#FC8D62","#B46446"),stringsAsFactors = F) %>%
dplyr::mutate(locus_age=paste0(locus,"_",age))
color.df$locus <- factor(color.df$locus, levels=c("A","B"))
color.df$age <- factor(color.df$age, levels=c("2m","24m"))
color.df$locus_age <- factor(color.df$locus_age,levels=color.df$locus_age)
Then joining df with color.df:
df <- dplyr::left_join(df,color.df)
And finally plotting:
library(plotly)
plot_ly(x = df$group, y = df$n, split = df$age, text = df$n, type = 'bar', color = df$locus_age, colors = color.df$color, showlegend = T,
textposition = "inside", textfont = list(size=12,color='black')) %>%
layout(yaxis=list(title="N"))
Which gives:
My questions are:
Although I defined the df$age order to be c("2m","24m") the "24m" age appears before the "2m" age, as if the split argument in the plot_ly function is ignoring this order. Any idea how to fix this?
Looks like the legend is labelling both age and the locus_age. Any idea how to make it only label by locus_age?

Answer 1

After some experimentation, it seems to me that the split argument is causing the issues.
I have also streamlined the code so that there is a continuous pipe:
library(dplyr)
library(plotly)
data.frame(locus = c("A", "A", "B", "B"),
age = rep(c("2m", "24m"), 2),
color = c("#66C2A5", "#488A75", "#FC8D62", "#B46446"),
stringsAsFactors = FALSE) %>%
mutate(locus_age = paste0(locus, "_", age)) %>%
mutate(locus = factor(locus, levels = c("A", "B"))) %>%
mutate(age = factor(age, levels = c("2m", "24m"))) %>%
mutate(locus_age = forcats::fct_inorder(locus_age)) %>%
left_join(df, .) %>%
plot_ly(x = ~group, y = ~n,
# split = ~age,
text = ~n, type = 'bar',
color = ~locus_age, colors = ~color, showlegend = TRUE,
textposition = "inside", textfont = list(size = 12, color = 'black')) %>%
layout(yaxis = list(title = "N"))
Now, the factor levels seem to be in the expected order and the legend shows only locus_age.
Caveat
Dealing with factors can become tricky sometimes as factor levels are not always created in the expected order.
If we create a character vector in the expected order
x <- outer(c("A", "B"), c("2m", "24m"), paste, sep = "_") %>%
t() %>%
as.vector()
x
[1] "A_2m" "A_24m" "B_2m" "B_24m"
base R's
factor(x)
(without specifying factor levels explicitely)
creates a factor whose levels are sorted in the current locale by default:
[1] A_2m A_24m B_2m B_24m
Levels: A_24m A_2m B_24m B_2m
Alternatively,
forcats::as_factor(x)
creates factor levels in order of appearance (if x is character)
[1] A_2m A_24m B_2m B_24m
Levels: A_2m A_24m B_2m B_24m
or even more explicitely
forcats::fct_inorder(x)