Use dplyr coalesce in programming

I'd like to use dplyr's programming magic, new to version 0.7.0, to coalesce two columns together. Below, I've listed out a few of my attempts.

df <- data_frame(x = c(1, 2, NA), y = c(2, NA, 3))

# What I want to do:
mutate(df, y = coalesce(x, y))

# Here's the expected output:
#> # A tibble: 3 x 2
#>       x     y
#>   <dbl> <dbl>
#> 1     1     1
#> 2     2     2
#> 3    NA     3

I thought fn1 would work, but it's treating varname as character on the right-hand side.

fn1 <- function(varname) {
  mutate(df, UQ(varname) := coalesce(x, !!varname))
}
fn1("y")
# Error in mutate_impl(.data, dots) : 
#   Evaluation error: Argument 2 must be type double, not character. 

Another attempt with enquo:

fn2 <- function(varname) {
  varname <- enquo(varname)
  mutate(df, varname := coalesce(x, !!varname))
}
fn2("y")  # same error

Maybe I can splice with !!! instead? (Spoiler: I can't.)

fn3 <- function(varname) {
  varnames <- c("x", varname)
  mutate(df, UQ(varname) := coalesce(!!! varnames))
}
fn3("y")
#> # A tibble: 3 x 2
#>       x     y
#>   <dbl> <chr>
#> 1     1     x
#> 2     2     x
#> 3    NA     x

fn4 <- function(varname) {
  varnames <- quo(c("x", varname))
  mutate(df, UQ(varname) := coalesce(!!! varnames))
}
fn4("y")
# Error in mutate_impl(.data, dots) : 
#   Column `y` must be length 3 (the number of rows) or one, not 2