Is it possible to bypass/ignore an argument in a custom function

Question

Just for clarification, not intended to reopen, as I will start a follow-up question:

This not a question about using an ifelse in dplyr.
This is a question about passing by an argument in a custom function. This is important for certain (not mentioned) reasons!

Original question:

I have this example with the mtcars dataset:

Where basically the dataframe is filtered and then plotted.
I wonder if it is possible to make the plot of the unfiltered data by this function ignoring the filtering variable z:

# create a custom function
my_function <- function(df, x, y, z) {
  df %>% 
    filter(am == z) %>% 
  ggplot(aes(x = {{x}}, y={{y}}))+
    geom_col()
}

apply function with filter am==0 <- WORKS

my_function(mtcars, cyl, mpg, 0)

apply function with filter am==1 <- WORKS

my_function(mtcars, cyl, mpg, 1)

would like to ignore the filter argument to get plot of all data <- WORKS NOT

my_function(mtcars, cyl, mpg)

Error in `filter()`:
  ! Problem while computing `..1 = am == z`.
Caused by error:
  ! argument "z" is missing, with no default

You could check if the argument was supplied and conditinally perform the filter: `my_function <- function(df, x, y, z) {hasz <- !missing(z); df %>% {if (hasz) filter(., am==z) else .} %>% ggplot(aes(x = {{x}}, y={{y}})) +geom_col()}` — MrFlick, Jun 17 '22 at 20:05
@MrFlick. It seems that you are quite experienced here. This question is closed. But I don't understand why. This not a question about using an ifelse in dplyr. This is a question about passing by an argument in a custom function. As you can see that I have already a little experience and I am following a certain quest. In this case my ultimate goal is to solve the last line of my function: `anim_save(paste0("images/survival_", z, ".gif"))`. And if z = NULL it starts to save all other ever filtered images. But the first was to understand how to bypass an argument. Could you clarify. Many thnx — TarJae, Jun 17 '22 at 20:10
You framed the question as you were interesting in conditionally filtering data only under certain circumstances. I was unable to guess what your true intention was. But there are plenty of existing questions about testing if a parameter has passed. I can add a duplicate for that if you like as well. — MrFlick, Jun 17 '22 at 20:12
@MrFlick. Now after going through all again. I think you are right. Thanks for clarifying! — TarJae, Jun 17 '22 at 20:37

olivroy · Accepted Answer · 2022-06-17T20:01:32.123

3

It is about working with the default.

I'd suggest having a default z = NULL, then your function would look like this.\

# create a custom function
my_function <- function(df, x, y, z = NULL) {
  
  if (!is.null(z)) {
   df <-  df %>% 
      filter(am == z)
  }
  
df %>% 
    ggplot(aes(x = {{ x }}, y = {{ y }})) +
    geom_col()
}

my_function(mtcars, cyl, mpg, 0)

my_function(mtcars, cyl, mpg, 1) 
my_function(mtcars, cyl, mpg)

edited Jun 17 '22 at 20:01

answered Jun 17 '22 at 19:59

olivroy

548
3
13

1

You can also use `if (missing(z))` (and don't need a default setting), but I prefer your solution. – Ben Bolker Jun 17 '22 at 20:00
Yes, it looks more obvious that `z` is different variable type than x and y. – olivroy Jun 17 '22 at 20:05
@olivroy. Thanks this solved also my main issue, i.e how to save without the filter argument! – TarJae Jun 17 '22 at 20:37
@olivroy, also easier if the function is being called indirectly (it's easy to force a parameter to be `NULL`, very hard to force it to be "missing") – Ben Bolker Jun 17 '22 at 20:51

Is it possible to bypass/ignore an argument in a custom function

1 Answers1