r - How to pass multiple group_by arguments and a dynamic variable argument to a dplyr function

Question

I am trying to pass multiple group_by arguments to a dplyr function as well as a named variable. In understand that I need to use a quosure for dplyr to understand the variables i am passing to it. The following code works fine:

quantileMaker2 <- function(data, groupCol, calcCol) {
  groupCol <- enquo(groupCol)
  calcCol <- enquo(calcCol)

  data %>%
    group_by(!! groupCol) %>%
      summarise('25%' = currency(quantile(!! calcCol, probs = 0.25), digits = 2L),
            '50%' = currency(quantile(!! calcCol, probs = 0.50), digits = 2L),
            '75%' = currency(quantile(!! calcCol, probs = 0.75), digits = 2L),
            avg = currency(mean(!! calcCol), digits = 2L),
            nAgencies = n_distinct('POSIT ID'), 
            nFTEs = sum(FTEs)
  )
}

quantileMaker2(df, employerClass, TCCperFTE)

However when I run the following I have a problem:

quantileMaker3 <- function(data,...,calcCol) {
  groupCol <- quos(...)
  calcCol <- quo(calcCol)

  data %>%
    group_by(!!! groupCol) %>%
    summarise('25%' = currency(quantile(!! calcCol, probs = 0.25), digits = 2L),
          '50%' = currency(quantile(!! calcCol, probs = 0.50), digits = 2L),
          '75%' = currency(quantile(!! calcCol, probs = 0.75), digits = 2L),
          avg = currency(mean(!! calcCol), digits = 2L),
          nAgencies = n_distinct('POSIT ID'), 
          nFTEs = sum(FTEs)
)
}

Which returns the following error:

 Error in summarise_impl(.data, dots) : 
  Evaluation error: anyNA() applied to non-(list or vector) of type 'symbol'. 

Sample data:

Year    employerClass   TCCperFTE   FTEs    POSIT ID
2014    One             5000        20      1
2014    Two             1000        30      2
2015    One             15000       40      1
2015    Two             50000       50      2
2016    One             100000      60      1
2016    Two             500000      70      2

Any help you guys could give would be much appreciated.

Answer 1

You haven't provided sample data, but your function works when modified to use the mtcars data frame.

library(tidyverse)
library(formattable)

quantileMaker3 <- function(data, calcCol, ...) {
  groupCol <- quos(...)
  calcCol <- enquo(calcCol)

  data %>%
    group_by(!!!groupCol) %>%
    summarise('25%' = currency(quantile(!!calcCol, probs = 0.25), digits = 2L),
              '50%' = currency(quantile(!!calcCol, probs = 0.50), digits = 2L),
              '75%' = currency(quantile(!!calcCol, probs = 0.75), digits = 2L),
              avg = currency(mean(!!calcCol), digits = 2L),
              nAgencies = n_distinct(cyl), 
              nFTEs = sum(hp)
    )
}

quantileMaker3(mtcars, mpg, cyl)

# A tibble: 3 x 7
    cyl `25%`             `50%`             `75%`             avg               nAgencies nFTEs
  <dbl> <S3: formattable> <S3: formattable> <S3: formattable> <S3: formattable>     <int> <dbl>
1    4. $22.80            $26.00            $30.40            $26.66                    1  909.
2    6. $18.65            $19.70            $21.00            $19.74                    1  856.
3    8. $14.40            $15.20            $16.25            $15.10                    1 2929.

With multiple grouping arguments:

quantileMaker3(mtcars, mpg, cyl, vs)

# A tibble: 5 x 8
# Groups:   cyl [?]
    cyl    vs `25%`             `50%`             `75%`             avg               nAgencies nFTEs
  <dbl> <dbl> <S3: formattable> <S3: formattable> <S3: formattable> <S3: formattable>     <int> <dbl>
1    4.    0. $26.00            $26.00            $26.00            $26.00                    1   91.
2    4.    1. $22.80            $25.85            $30.40            $26.73                    1  818.
3    6.    0. $20.35            $21.00            $21.00            $20.57                    1  395.
4    6.    1. $18.03            $18.65            $19.75            $19.12                    1  461.
5    8.    0. $14.40            $15.20            $16.25            $15.10                    1 2929.

Incidentally, you can avoid multiple calls to quantile by using nesting. This won't work if any of the output columns are of class formattable (which is what the currency function returns), so I've changed the function to create strings for the currency-format columns.

quantileMaker3 <- function(data, calcCol, ..., quantiles=c(0.25,0.5,0.75)) {

  groupCol <- quos(...)
  calcCol <- enquo(calcCol)

  data %>%
    group_by(!!!groupCol) %>%
    summarise(values = list(paste0("$", sprintf("%1.2f", quantile(!!calcCol, probs=quantiles)))),
              qnames = list(sprintf("%1.0f%%", quantiles*100)),
              nAgencies = n_distinct(cyl), 
              nFTEs = sum(hp),
              avg = paste0("$", sprintf("%1.2f", mean(!!calcCol)))
    ) %>% 
    unnest %>% 
    spread(qnames, values) 
}

quantileMaker3(mtcars, mpg, cyl, vs)

# A tibble: 5 x 8
# Groups:   cyl [3]
    cyl    vs nAgencies nFTEs avg    `25%`  `50%`  `75%` 
  <dbl> <dbl>     <int> <dbl> <chr>  <chr>  <chr>  <chr> 
1    4.    0.         1   91. $26.00 $26.00 $26.00 $26.00
2    4.    1.         1  818. $26.73 $22.80 $25.85 $30.40
3    6.    0.         1  395. $20.57 $20.35 $21.00 $21.00
4    6.    1.         1  461. $19.12 $18.03 $18.65 $19.75
5    8.    0.         1 2929. $15.10 $14.40 $15.20 $16.25