R Plotly Heatmap Conditional Annotation with ifelse? - r

Question

Follwoing:
R Heatmap: conditionally change label text colours with (ggplot2 or plotly)
I thought the code below would create a heatmap whose annotation is white if the value is greater than 5 (so in this case, when (X,Y)=(3,"C")) but it does not work. Any idea?
Thank you!
df<-tibble(Y=c("A","B","C"),
X=c(1,2,3),
Z=c(1,5,10))
df %>%
plot_ly(x = ~X, y=~Y, z=~Z, type='heatmap') %>%
add_annotations(text = ~Z,
showarrow = FALSE,
font = list(color = ~ifelse(Z>5,'white','black')))

Answer 1

Font colors weren't designed to be dynamic in Plotly. Even though add_annotations inherently understands that you want one 'add_annotation' for each value in Z, it does not assume that's what you mean for the font color. However, you can still have it changed on the fly with a UDF.
This function rebuilds the plot, then parses the annotations to assign colors.
that <- function(plt) {
plt <- plotly_build(plt) # build the plot to collect the data
lapply(1:length(plt$x$layout$annotations), # loop through annotations
function(j) {
this <- plt$x$layout$annotations[[j]] # collect annotation
colr <- ifelse(this$text > 5, "white", "black") # test for color
plt$x$layout$annotations[[j]]$font$color <<- colr # assign color
})
plt # return updated plot
}
The way that you use this is by just piping it to the end of your plot call. I didn't tie in exceptions for cases in which no color was specifically assigned. In other words, assign a color in your original plot call, it doesn't matter what color it is.
df %>%
plot_ly(x = ~X, y = ~Y, z = ~Z, type = 'heatmap') %>%
add_annotations(text = ~Z,
showarrow = FALSE,
font = list(color = "black")) %>% that() # <- I'm new!
I used the color sequence you chose, but I think this is the opposite of what you really wanted.
For example, if I swap the conditions for the colors in the function:
that <- function(plt) {
plt <- plotly_build(plt) # build the plot to collect the data
lapply(1:length(plt$x$layout$annotations), # loop through annotations
function(j) {
this <- plt$x$layout$annotations[[j]] # collect annotation
colr <- ifelse(this$text > 5, "black", "white") # test for color
plt$x$layout$annotations[[j]]$font$color <<- colr # assign color
})
plt # return updated plot
}
df %>%
plot_ly(x = ~X, y = ~Y, z = ~Z, type = 'heatmap') %>%
add_annotations(text = ~Z,
showarrow = FALSE,
font = list(color = "black")) %>% that()
Now you can see the text a lot better:
By the way, you could also call the plot as it is and use that(last_plot()) to rerender it with the new text colors.
Both plotly and ggplot2 have the function last_plot(); make sure you're calling the plotly version if you use this method.

Answer 2

Just an idea, you can also use a heatplot like this if you want to show a thresholded map:
ggplot(df, mapping = aes(x = X, y = Y, fill = Z>5, fill_max = 5)) + geom_tile()