Writing function using exec() instead of invoke() in R

August 15, 2022

I could use some help with a custom function that I’ve been working on that takes the parallel minima of a vector of variables (all date of diagnosis variables, with dx dates for cases and NA for noncases).

I want to create 2 variables:

{prefix}flag1 sets all nonmissing values to 1 and all missing values to 0
{prefix}flag2 finds the earliest possible date of the vector of variables for each observation/ row.

I’ve been able to generate a function using the invoke() wrapper; however, since the rlang function is being depreciated, I’m trying to execute the same function using exec() or inject() instead.

The actual dataset has 100+ columns, but I am supplying a minimally reproducible example below.

# load libs
library(tidyverse)
library(lubridate)

rdate <- function(x,
                  min = paste0(format(Sys.Date(), '%Y'), '-01-01'),
                  max = paste0(format(Sys.Date(), '%Y'), '-12-31'),
                  sort = TRUE) {
  
  dates <- sample(seq(as.Date(min), as.Date(max), by = "day"), x, replace = TRUE)
  if (sort == TRUE) {
    sort(dates)
  } else {
    dates
  }
}
# set seed for reproducibility
set.seed(42)

# generating ds of random dates
date_ds <- data.frame(v.ar1 = rdate(15), 
                  var_2 = rdate(15), 
                  random3 = rdate(15), 
                  v4= rdate(15), 
                  v__5 = rdate(15), 
                  variable6 = rdate(15), 
                  dates7 = rdate(15)) %>% 
          # remove some observations 
          mutate(var_2 = as_date(ifelse(var_2 < dates7, var_2, NA)), 
                variable6 = as_date(ifelse(variable6 <= v__5, variable6, NA)), 
                v4 = as_date(ifelse(v4 > v.ar1, v4, NA)),
                )
# vector of names 
varnames <- names(date_ds)

# function 
collapse_phenos <- function(varvector, prefix){
  outds <- date_ds %>%
    mutate(!!paste0(prefix,"flag1") := +(rowSums(!is.na(select(., all_of(varvector)))) > 0),
    !!paste0(prefix,"flag2") := invoke(pmin, c(across(all_of(varvector)), na.rm = TRUE))) 
return(outds) 
} 

# return ds
result <- collapse_phenos(varvector = varnames, prefix = "")

Any help would be appreciated! A tidyverse solution would be preferable.

>Solution :

We can modify the function with invoke to exec using the splice operator (!!!) after converting the strings to symbols (syms)


collapse_phenos <- function(varvector, prefix){
  outds <- date_ds %>%
    mutate(!!paste0(prefix,"flag1") := +(rowSums(!is.na(select(., all_of(varvector)))) > 0),
    !!paste0(prefix,"flag2") := exec(pmin, !!! rlang::syms(varvector), na.rm = TRUE)) 
return(outds) 
}

-testing

# return ds
result <- collapse_phenos(varvector = varnames, prefix = "")

-output

> result
        v.ar1      var_2    random3         v4       v__5  variable6     dates7 flag1      flag2
1  2022-01-24       <NA> 2022-01-04       <NA> 2022-01-16       <NA> 2022-01-01     1 2022-01-01
2  2022-02-18 2022-01-20 2022-04-19       <NA> 2022-02-04       <NA> 2022-02-11     1 2022-01-20
3  2022-02-18 2022-01-24 2022-04-24       <NA> 2022-03-17 2022-03-01 2022-02-11     1 2022-01-24
4  2022-03-15       <NA> 2022-05-16       <NA> 2022-03-23       <NA> 2022-03-13     1 2022-03-13
5  2022-03-30       <NA> 2022-05-26 2022-04-13 2022-04-20 2022-04-10 2022-04-18     1 2022-03-30
6  2022-05-02 2022-04-20 2022-07-16       <NA> 2022-04-28       <NA> 2022-04-22     1 2022-04-20
7  2022-05-08       <NA> 2022-08-03 2022-05-10 2022-05-10       <NA> 2022-05-06     1 2022-05-06
8  2022-05-26       <NA> 2022-08-14       <NA> 2022-08-08 2022-07-27 2022-05-21     1 2022-05-21
9  2022-06-02       <NA> 2022-09-02 2022-06-06 2022-08-09       <NA> 2022-07-25     1 2022-06-02
10 2022-06-14       <NA> 2022-10-19 2022-07-05 2022-09-05       <NA> 2022-09-05     1 2022-06-14
11 2022-08-16       <NA> 2022-10-26       <NA> 2022-09-22       <NA> 2022-10-12     1 2022-08-16
12 2022-10-30 2022-10-25 2022-11-10       <NA> 2022-11-21 2022-10-15 2022-10-26     1 2022-10-15
13 2022-11-17       <NA> 2022-11-20       <NA> 2022-11-21 2022-10-23 2022-11-07     1 2022-10-23
14 2022-11-23       <NA> 2022-12-14 2022-11-25 2022-12-23 2022-10-25 2022-12-03     1 2022-10-25
15 2022-12-22       <NA> 2022-12-21 2022-12-24 2022-12-26 2022-12-05 2022-12-07     1 2022-12-05