Converting treemap to ggplot

Question

The good news: I'm able to create a beautiful treemap using the treemap package.

Data:

forTm <- structure(list(
  UnitGroup = c("1N", "BHU", "CSU", "ED", "Med/Surg", "Med/Surg", "Telemetry", 
            "Telemetry", "Telemetry", "Telemetry", "Telemetry"), 
  Unit = c("A", "B", "C", "ED", "D", "E", "F", "G", "H", "I", "J"),
  Count = c(1L, 1L, 1L, 1L, 15L, 10L, 5L, 2L, 3L, 8L, 4L)), 
  class = c("data.frame"), 
  row.names = c(NA, -11L),
  .Names = c("UnitGroup", "Unit", "Count"))

Treemap:

library(treemap)
tm <- treemap(forTm,
          index = c("UnitGroup", "Unit"),
          vSize = "Count",
          vColor = "Count",
          type = "dens",
          palette = "YlGnBu",
          title = "# Patients Sample Title",
          aspRatio = (1.5),
          fontsize.labels = c(16, 12),
          fontsize.title = 16,
          bg.labels = "white",
          border.lwds = c(5,0))

The problem: class(tm) is a list, and I need to plot the treemap on a page with several ggplots. I anticipate needing to add/rearrange the plots for the end user, so I'd like a relatively flexible solution.

Goal: insert treemap into the following dashboard:

#just stand-ins for the plots
samplePlot <- grid.rect(gp = gpar(fill = "grey")) 
treemapHere <- grid.rect(gp = gpar(fill = "blue")) 

grid.arrange(samplePlot, # plot 1
         treemapHere, # plot 2
         samplePlot, # plot 3
         layout_matrix = rbind(c(3, 2), c(1, 1)),
         top = textGrob("Sample Title", 
                        gp = gpar(margin = margin(10, 0, 10, 0))),
         heights = c(5, 5))

But the solution needs to be flexible enough that I can easily add/rearrange the plots, for example as follows:

grid.arrange(samplePlot, # plot 1
         samplePlot, # plot 2
         samplePlot, # plot 3
         samplePlot, # plot 4
         treemapHere, # plot5
         layout_matrix = rbind(c(1, 2, 3), 
                               c(4, 5, 5)),
         top = textGrob("Sample Title", 
                        gp = gpar(margin = margin(10, 0, 10, 0))),
         heights = c(5, 5))

Ideally I'd find a way to recreate the treemap as a ggplot, because I'm very familiar with ggplot's syntax so I'd have an easy time standardizing the sheet's style. But if there's no way to do that I'll accept any solution(save the treemap as a grob?) that allows me to easily rearrange this graph within my page of ggplots.

What I've tried so far: Honestly, not much. Google hasn't been my friend. The only recommendation I've found is to use treemapify, which I can't do for this particular project.

Brantley

The treemap seems to be plotted with grid graphics and it allows you to specify a viewport to draw into (via vp=). What kind of layout with other ggplots are you trying to achieve? How are you arranging the other ggplots? – Adipose

@Adipose I saw that but haven't used viewports before, and the documentation for viewports seemed like I'd be hardcoding the location. I'm arranging my other ggplots as follows, but I'll be adding plots for the end user so was hoping for a more flexible solution. color <- grid.rect(gp = gpar(col = "grey")) grid.arrange(color, # plot 1 color, # treemap here color, # plot 3 layout_matrix = rbind(c(2, 3), c(1, 1)), top = textGrob("Sample Title", gp = gpar(fontsize = 16, font = 2, margin = margin(10, 0, 10, 0))), heights = c(5, 5)) – Brantley

It would be more helpful if your reproducible example made it clear how you want to incorporate these plots. You can take out the output of tm since that really isn't helpful. The treemap function draw the plot as a side effect directly. It's doesn't have a separate plot() or print() method like ggplot objects do. It's hard to say what will work without some testing. – Adipose

@Adipose Done. I added two possible layouts I might end up with, depending on the number of charts the end user decides to keep. – Brantley

There is a package called ggraph (link) which is an extension of ggplot2 that aims at supporting relational data structures such as networks, graphs, and trees etc. It also supports Treemaps. – Inappetence

Answer 1

What if you reconstruct the treemap using ggplot step by step? E.g.

library(tidyverse)
treemapHere <- ggplot(tm$tm, 
   aes(xmin=x0+w, xmax=x0, ymin=y0+h, ymax=y0, fill=ifelse(level==1, NA, color), fill = vSize)) + 
  scale_fill_identity() + 
  geom_rect(aes( size=I((2:1)[level])), color="#303030") + 
  coord_fixed(0.75) + 
  guides(size="none") + 
  cowplot::theme_nothing() + 
  geom_text(aes(x0+w/2, y0+h/2, label=Unit, size=I(scales::rescale(vSize, c(2.5, 3))))) + 
  geom_label(aes(x, y, label=UnitGroup, size=I(scales::rescale(vSize, c(2.5, 7)))), tm$tm %>% group_by(UnitGroup) %>% summarise(x=mean(x0+w/2), y=mean(y0+h/2), vSize=mean(vSize)), inherit.aes=F, fill="white", fontface = "bold")

library(grid)
library(gridExtra)
samplePlot <- grid.rect(gp = gpar(fill = "grey"))   
grid.arrange(samplePlot, # plot 1
         samplePlot, # plot 2
         samplePlot, # plot 3
         samplePlot, # plot 4
         treemapHere, # plot5
         layout_matrix = rbind(c(1, 2, 3), 
                               c(4, 5, 5)),
         top = textGrob("Sample Title", 
                        gp = gpar(margin = margin(10, 0, 10, 0))),
         heights = c(5, 5))

Frumenty