I have a series of aggregation functions that I want to apply to multiple columns in a data frame (using purrr::map). Some are standard functions, like mean, and others are user-defined functions.
I want the output to be a data frame with the names of those functions as the columns that store their output.
Here's an example that shows the procedure and the desired output:
library(purrr)
foo <- function(x) sd(x) * .1 # arbitrary user-defined function
agg_funcs <- c(mean, median, foo)
names(agg_funcs) <- c("mean", "median", "foo")
fields <- c("mpg", "disp") # fields to aggregate with agg_funcs
fields %>%
set_names(fields) %>%
map_df(function(x) map_df(agg_funcs, function(f) f(mtcars[[x]])), .id = "field")
# A tibble: 2 x 4
field mean median foo
<chr> <dbl> <dbl> <dbl>
1 mpg 20.1 19.2 0.603
2 disp 231. 196. 12.4
But I don't want to name the functions in agg_funcs by hand. Instead, I'd like to do something like
agg_funcs %>% set_names(agg_funcs) # doesn't work
I know that I can get the string representation of a function name like this:
as.character(substitute(mean))
# "mean"
And I can even wrap this operation in a function and pass in the agg funcs I want, one by one:
f <- function(func) as.character(substitute(func))
f(mean) # "mean"
This function f() is the sort of thing I'd want in my nested map operation above, to get the function names as strings.
But this fails when I try to map or sapply over agg_funcs (for reasons I don't totally understand):
sapply(agg_funcs, f) # alt: map(agg_funcs, f)
mean median foo
[1,] "[[" "[[" "[["
[2,] "X" "X" "X"
[3,] "i" "i" "i"
How can I achieve the result I'm looking for (names of functions as columns in mapped output df) without creating names(agg_funcs) by hand?
