Breaking NSE function in new dplyr version

Question

When updating to newest version of the package dplyr, a funcion I have using NSE breaks. I was wondering how the new version changes this and how to fix it. I've tried using the .data$ and .env$ before each variable name but can't seem to get it working.

Here is my custom function:

t_ppond <- function(w, v){
  arguments <- as.list(match.call())

  y <- eval(arguments$w)
  x <- eval(arguments$v)

    d <- data.frame("yy" = y, 
                    "xx" = x)
    tt <- sum(d$yy)

    dff <- d %>%
      mutate("sh" = yy/tt) %>% 
      mutate("rr" = xx*sh)

  sum(dff$rr)
}

And this is what I used it for (calculating a weighted mean from a variable):

data(iris)
iris %>% 
  group_by(Species) %>% 
  summarise("new" = t_ppond(Sepal.Length, Petal.Width))

The above code worked perfectly before the update. Now I get:

Error in summarise_impl(.data, dots) : 
  Evaluation error: object 'Sepal.Length' not found.

Answer 1

Are you sure you need non-standard evaluation for this? You have a custom function that just needs to take in two vectors, so you can write it like:

t_ppond <- function(w, v){
    d <- data.frame("yy" = w, 
                    "xx" = v)
    tt <- sum(d$yy)

    dff <- d %>%
        mutate("sh" = yy/tt) %>% 
        mutate("rr" = xx*sh)

    sum(dff$rr)
}

data(iris)
iris %>% 
    group_by(Species) %>% 
    summarise("new" = t_ppond(Sepal.Length, Petal.Width))

Output:

# A tibble: 3 x 2
     Species       new
      <fctr>     <dbl>
1     setosa 0.2480224
2 versicolor 1.3352089
3  virginica 2.0333030