how to use bracket notation (or an alternative) while programming with dplyr

Question

I'm trying to write a function to calculate toplines (as commonly used in polling data). It needs to include both a "percent" and "valid percent" column.

Here's an example

library(tidyverse)

# prepare some data
d <- gss_cat %>%
  mutate(tvhours2 = tvhours,
         tvhours2 = replace(tvhours2, tvhours > 5 , "6-8"),
         tvhours2 = replace(tvhours2, tvhours > 8 , "9+"),
         tvhours2 = fct_explicit_na(tvhours2),
         # make a weight variable
         fakeweight = rnorm(n(), mean = 1, sd = .25))

The following function works as far as it goes:

make.topline <- function(variable, data, weight){
  variable <- enquo(variable)
  weight <- enquo(weight)

  table <- data %>%
    # calculate denominator
    mutate(total = sum(!!weight)) %>%
    # calculate proportions
    group_by(!!variable) %>%
    summarise(pct = (sum(!!weight)/first(total))*100,
              n = sum(!!weight))

  table
}

make.topline(variable = tvhours2, data = d, weight = fakeweight)

I'm struggling to implement the valid percent column. Here is the syntax I tried.

make.topline2 <- function(variable, data, weight){
  variable <- enquo(variable)
  weight <- enquo(weight)

  table <- data %>%
    # calculate denominator
    mutate(total = sum(!!weight),
           valid.total = sum(!!weight[!!variable != "(Missing)"])) %>%
    # calculate proportions
    group_by(!!variable) %>%
    summarise(pct = (sum(!!weight)/first(total))*100,
              valid.pct = (sum(!!weight)/first(valid.total))*100,
              n = sum(!!weight))

  table
}

make.topline2(variable = tvhours2, data = d, weight = fakeweight)

This yields the following error:

 Error: Base operators are not defined for quosures.
Do you need to unquote the quosure?

  # Bad:
  myquosure != rhs

  # Good:
  !!myquosure != rhs
Call `rlang::last_error()` to see a backtrace 

I know the problem is in this line, but I don't know how to fix it:

mutate(valid.total = sum(!!weight[!!variable != "(Missing)"]))

Answer 1

You can put parentheses around the !!weight. I think of this as making sure we are using the extract brackets only after weight is unquoted (so an order of operations thing).

That line would then look like:

valid.total = sum((!!weight)[!!variable != "(Missing)"])

Alternatively, you could use the new curly-curly operator ({{), which takes the place of the enquo()/!! combination for relatively simple cases like yours. Then your function would look something like

make.topline <- function(variable, data, weight){

    table <- data %>%
        # calculate denominator
        mutate(total = sum({{ weight }}),
               valid.total = sum({{ weight }}[{{ variable }} != "(Missing)"])) %>%
        # calculate proportions
        group_by({{ variable }}) %>%
        summarise(pct = (sum({{ weight }})/first(total))*100,
                  valid.pct = (sum({{ weight }})/first(valid.total))*100,
                  n = sum({{ weight }}))

    table
}

Like the parentheses solution, this runs without error.

make.topline(variable = tvhours2, data = d, weight = fakeweight)

# A tibble: 9 x 4
  tvhours2    pct valid.pct      n
  <fct>     <dbl>     <dbl>  <dbl>
1 0          3.16      5.98   679.
2 1         10.9      20.6   2342.
3 2         14.1      26.6   3022.
4 3          9.10     17.2   1957.
5 4          6.67     12.6   1432.
6 5          3.24      6.13   696.
7 6-8        4.02      7.61   864.
8 9+         1.67      3.16   358.
9 (Missing) 47.2      89.3  10140.