Density and correlation values in ggPairs shifted to left in ggplotly()

Question

I am using the package GGally (and method ggpairs) to create scatterplot matrices.

I notice that the static version (object p below) appears to have correctly-centered correlation values and densities. The image below demonstrates this -

However, the interactive version (object ggPS below) appears to have incorrectly-centered correlation values and densities. Namely, they appear to be centered at the left. The image below demonstrates this -

I wonder if there may be a good method to remedy this issue?

library(plotly)
library(GGally)

set.seed(2)
dat <- data.frame(ID = paste0("ID",1:100), A.1=sort(rnorm(100)), A.2=sort(rnorm(100)), A.3=sort(rnorm(100)), B.1=sort(rnorm(100)), B.2=sort(rnorm(100)))

dat$ID <- as.character(dat$ID)
minVal = 0
maxVal = max(dat[,-1])

my_fn <- function(data, mapping, ...){
  x = data[,c(as.character(mapping$x))]
  y = data[,c(as.character(mapping$y))]
  p <- ggplot(data = dat, aes(x=x, y=y)) + coord_cartesian(xlim = c(minVal, maxVal), ylim = c(minVal, maxVal))
  p
}

p <- ggpairs(dat[,-1], lower = list(continuous = my_fn))

ggPS <- ggplotly(p)

As a side note, I tried to use the solution posted at (GGpairs, correlation values are not aligned) by altering the syntax to create object p as follows:

p <- ggpairs(dat[,-1], lower = list(continuous = my_fn), upper = list(continuous = wrap("cor", hjust=5)))

However, this did not appear to make a difference.

Any recommendations would be most appreciated!

Answer 1

Ideally you could access the autoscale option in the modebar using layout specs in plotly. For now the easiest option (probably not optimal) is setting the min/max from the dataframe instead of the min/max per data pair.

my_fn <- function(data, mapping, ...){
  x = data[,c(as.character(mapping$x))]
  y = data[,c(as.character(mapping$y))]
  p <- ggplot(data = dat, aes(x=x, y=y)) + coord_cartesian(xlim = c(min(dat[,-1]),max(dat[,-1])))
  p
}

p <- ggpairs(dat[,-1], lower = list(continuous = my_fn))

ggPS <- ggplotly(p)