R热图：使用（ggplot2或按图）有条件地更改标签文本的颜色

Question

I am trying to produce a heatmap with ggplot2 or plotly in R, where the values associated with a block or tile are used as labels in the respective tile. This was not so difficult, but I have removed the legend and would like to change the colours of the labels conditional on their values to increase their visibility.

Here a reproducible examples to show what I mean.

Data (using data.table and dplyr):

sig <-  rep(c("sig1", "sig2", "sig3"), 100, replace = TRUE, prob = c(0.4, 0.35, 0.25))
date <- c("2019-11-01", "2019-11-02", "2019-11-03")

another <- as.data.table(expand.grid(sig, date))

test_dat_numerics <- another[, number_ok := sample(0:100, 900, replace = TRUE)]
setnames(test_dat_numerics, c("Var1", "Var2"), c("sig", "date"))

test_dat_numerics <- test_dat_numerics[, avg := mean(number_ok), by = .(date, sig)] %>%
  dplyr::select(-number_ok) %>%
  dplyr::rename(number_ok = avg) %>%
  dplyr::mutate(prop = ifelse(number_ok > 50, 1, 0))
  dplyr::distinct()

The heatmap (with ggplot2):

ggp <- ggplot(test_dat_numerics, aes(date, sig, fill = number_ok)) +
  geom_tile() +
  geom_text(aes(label = test_dat_numerics$number_ok)) +
  theme(legend.position="none")

This results in

The darker a block becomes the less visible the text becomes. To prevent this, my intention is to make the text white when a value is below 50 and black otherwise. This is the part where I failed both with ggplot2 and plotly until now and would be grateful for help.

With plotly:

p <- test_dat_numerics %>%
  plot_ly(type = "heatmap",
          x = ~date,
          y = ~sig,
          z = ~number_ok,
          # zmax = 100,
          # zmin = 0,
          showscale = FALSE,
          colorscale = "Blues") %>%
  add_annotations(text = as.character(test_dat_numerics$number_ok),
                  showarrow = FALSE,
                  color = list(if (test_dat_numerics$number_ok[i] > 50) {"black"} else {"white"})) %>%
  layout(title = "Test Heatmap",
         # titlefont = t,
         xaxis = list(title = "Datum"), yaxis = list(title = "Signal")
         )

I found a great plotly example here , but I couldn't manage to get to work for my case. Here the annotation part of my code:

ann <- list()

    for (i in 1:length(unique(test_dat_numerics$sig))) {
      for (j in 1:length(unique(test_dat_numerics$date))) {
        for (k in 1:(length(unique(test_dat_numerics$sig))*length(unique(test_dat_numerics$date)))) {
          ann[[k]] <- list(
          x = i,
          y = j,
          font = list(color = if (test_dat_numerics$number_ok[i] > 50) {"black"} else {"white"}),
          text = as.character(test_dat_numerics$number_ok[[k]]),
          xref = "x", 
          yref = "y", 
          showarrow = FALSE )
        }
      }
    }

p_test_num_heat <- layout(p, annotations = ann)

Here, one of numerous attempts with ggplot2:

ggp <- ggplot(test_dat_numerics, aes(date, sig, fill = number_ok)) +
  geom_tile() +
  geom_text(aes(label = test_dat_numerics$number_ok)) +
  geom_label(aes(colour = factor(test_dat_numerics$prop))) +
  theme(legend.position="none")

(This code produces the plot in the image above if the second to last line is removed.)

I'm pretty stuck on this one... Thanks in advance for any advice!

Answer 1

With ggplot2, you can use colour in the aes of geom_text (+ scale_colour_manual ):

ggplot(test_dat_numerics, aes(date, sig, fill = number_ok)) +
  geom_tile() +
  geom_text(aes(label = number_ok, colour =ifelse(number_ok>50, "black", "white"))) +
  scale_colour_manual(values=c("white"="white", "black"="black")) +
  theme(legend.position="none")