Alex Alex - 2 months ago 40
R Question

How to use anonymous functions for mutate_each (and summarise_each)?

As we know, it is possible to call functions in R without assigning them to the environment, e.g.

> (function(x){x/2})(5)
[1] 2.5


I would like to use functions like these, on the fly, in a
mutate_each
(or
summarise_each
) call. For example, with

df <- data.frame(a = runif(10), b = rnorm(10))


I might attempt to do one of the following, but they all return errors:

library(dplyr)

> df %>%
+ mutate_each(funs((function(x){x/2})), a, b)
Error in eval(substitute(expr), envir, enclos) :
Unsupported type CLOSXP for column "a"
>
> df %>%
+ mutate_each(list((function(x){x/2})), a, b)
Error: is.fun_list(calls) is not TRUE
>
>
> df %>%
+ mutate_each(funs((function(x){x/2})(.)), a, b)
Error in vapply(dots[missing_names], function(x) make_name(x$expr), character(1)) :
values must be length 1,
but FUN(X[[1]]) result is length 2
>


However, if I assign the function to the workspace, then all works as expected:

tmp_fn <- function(x){x/2}


and

> df %>%
+ mutate_each(funs(tmp_fn), a, b)
a b
1 0.360048105 -0.452285314
2 0.020175136 0.253063103
3 0.002351454 -0.148997643
4 0.262808493 -0.599555244
5 0.057246370 0.007567076
6 0.400027700 0.264901865
7 0.120505411 -0.346171709
8 0.266166200 0.116066231
9 0.031302148 -0.129739601
10 0.250212897 0.230194217


Is there a way to dynamically define functions in the call to
mutate_each
or
summarise_each
?

Answer

We can wrap the function call with parentheses

df %>%
     mutate_each(funs(((function(x){x/2})(.))))