How to set a quantile-defined colors for heatmaps in R with plotly?

Question

I want to plot a matrix in R with plotly, where the color scale is to be based on quantile values. Here is a self-contained reproducible example:

library(dplyr)
library(plotly)

# Data to plot
d <- data.frame(x = 1:10, y=1:10, z = rnorm(n = 10, mean = 0, sd = 100))

# Quantile values
qtl_val <- c(0, 0.68, 0.95, 0.99, 0.997)

# Colors to use
qtl_colors <- c(
  rgb(0,0,0),
  rgb(246/256,192/256,192/256),
  rgb(244/256,135/256,134/256),
  rgb(219/256,55/256,55/256),
  rgb(164/256,30/256,34/256))

# Function to assign a color to each quantile window
set_color <- function(z,qtl_leg){
  for( i in 1:nrow(qtl_leg)){
    min_val <- qtl_leg$val[i]
    max_val <- ifelse( !is.na(qtl_leg$lead_val[i]), qtl_leg$lead_val[i], Inf)
    if( z >= min_val & z <= max_val ) return(qtl_leg$color[i])
  }
}

qtl_leg <- data.frame(val=round(as.numeric(quantile(d$z, probs=qtl_val)))) %>%
  mutate(lead_val = lead(val)) %>%
  mutate(color = qtl_colors)

d <- d %>%
  rowwise() %>%
  mutate(color = set_color(z,qtl_leg) ) %>%
  as.data.frame()

# Draw the matrix
plot_ly(data=d, x=~x, y=~y, z=~z, color=~color, type='heatmap' )

The plot does not match the colors set in the data.frame, the legend is weird and the tip-tool text as well. In a nutshell, I got something wrong but I don't figure out what.

Answer 1

You could create a custom scale with zauto = FALSE and specify the zmin and zmax like this:

library(plotly)
library(dplyr)
plot_ly(data=d, x=~x, y=~y, z=~z, 
        color=~color, type='heatmap', 
        zauto = FALSE,
        zmin = qtl_val[1],
        zmax = qtl_val[5])

Output: