dplyr and ggplot in a function: use reorder in aes function

Question

I'm struggling to reorder my data for plotting with ggplot in a function that also uses dplyr:

# example data
library(ggplot2)
library(dplyr)
dat <- data.frame(a = c(rep("l", 10), rep("m", 5), rep("o", 15)),
                  b = sample(100, 30), 
                  c= c(rep("q", 10), rep("r", 5), rep("s", 15)))

Here are my steps outside of a function:

# set a variable
colm <- "a"
# make a table
dat1 <- dat %>% 
  group_by_(colm) %>%  
  tally(sort = TRUE)
# put in order and plot
ggplot(dat2, aes(x = reorder(a, n), y = n)) +
  geom_bar(stat = "identity")

But when I try to make that into a function, I can't seem to use reorder:

f <-  function(the_data, the_column){
       dat %>% group_by_(the_column) %>%  
       tally(sort = TRUE) %>% 
       ggplot(aes_string(x = reorder(the_column, 'n'), y = 'n')) +
       geom_bar(stat = "identity")
}

f(dat, "a")

Warning message:
In mean.default(X[[i]], ...) :
  argument is not numeric or logical: returning NA

The function will work without reorder:

f <-  function(the_data, the_column){
       dat %>% group_by_(the_column) %>%  
       tally(sort = TRUE) %>% 
       ggplot(aes_string(x = the_column, y = 'n')) +
       geom_bar(stat = "identity")
}

f(dat, "a")

And I can get what I want without dplyr, but I'd prefer to use dplyr because it's more efficient in my actual use case:

# without dplyr
ff = function(the_data, the_column) {
  data.frame(table(the_data[the_column])) %>% 
  ggplot(aes(x = reorder(Var1, Freq), y = Freq)) +
  geom_bar(stat = "identity") +
    ylab("n") +
    xlab(the_column)
}

ff(dat, "a")

I see that others have struggled with this (1, 2), but it seems there must be a more efficient dplyr/pipe idiom for this reordering-in-a-function task.

Answer 1

If you are going to use aes_string, then the whole value must be a string, not just partially a string. You can use paste() to help build the expression you want to use for x. For example

f <-  function(the_data, the_column){
       dat %>% group_by_(the_column) %>%  
       tally(sort = TRUE) %>% 
       ggplot(aes_string(x = paste0("reorder(",the_column,", n)"), y = 'n')) +
       geom_bar(stat = "identity")
}

Or you could use expressions rather than strings

f <-  function(the_data, the_column){
       dat %>% group_by_(the_column) %>%  
       tally(sort = TRUE) %>% 
       ggplot(aes_q(x = substitute(reorder(x, n),list(x=as.name(the_column))), y = quote(n))) +
       geom_bar(stat = "identity")
}

but the general idea is that you need to be careful when mixing strings and raw language elements (like names or expressions).