purrr mapping more than two variables (for making a pairplot with ggplot)

September 5, 2023

i enjoy using purrr’s map and map2 a lot, but i often find myself in the situation of wanting to map more inputs. I know there is also pmap but i don’t really understand how to use it.

Below is my example for creating a pairplot in ggplot2. It is missing the value of name.x as label for the x-axis.

plots = iris%>%
  pivot_longer(!Species)%>%
  left_join(
    iris%>%
      pivot_longer(!Species),by="Species"
  )%>%
  group_by(name.x,name.y)%>%
  nest()%>%
  mutate(
    plots =
      map2(
        .x=data,.y=name.x,
        ~ggplot(.x)+
          geom_point(aes(value.x,value.y,color=Species))+
          xlab(.y)
  ))

ggarrange(plotlist=plots$plots, common.legend = TRUE, legend="bottom")

I’m looking forward to your suggestions!
It would be ideal if they could be used smoothly in a pipe.

>Solution :

Use pmap like so:

plots <- iris %>%
  pivot_longer(!Species)%>%
  left_join(iris %>% pivot_longer(!Species), by = "Species")%>%
  group_by(name.x, name.y) %>%
  nest() %>%
  mutate(plots = pmap(
    list(data, name.x, name.y),   
    \(d, x, y) ggplot(d)+
      geom_point(aes(value.x, value.y, color = Species))+
      xlab(x) + ylab(y)
  ))

That is:

Collect the variables you want to map over in a list, so list(data, name.x, name.y)
Define a function that takes those arguments.

There’s several ways of defining that function. E.g. you could use the formula syntax like you did for map2, with ..1, ..2 and ..3. But I like the new syntax for anonymous functions (i.e. \(x, y)) quite a bit better, since you can name the arguments.