SE filter_ by function taking multiple columns

Question

I would like to filter a data frame to leave only the complete cases based on selected columns. This is easy to do with NSE filter():

library(dplyr)

dd <- data.frame(
  id = 1:4,
  var1 = c(1, 2, NA, 4),
  var2 = c(1, NA, 3, 4),
  var3 = c(1, NA, NA, NA))

dd1 <- dd %>% filter(complete.cases(var1, var2))
dd1
#>   id var1 var2 var3
#> 1  1    1    1    1
#> 2  4    4    4   NA

However I'm running in to a wall in trying to produce an SE version of this operation to which I may pass the quoted names of columns.

library(lazyeval)
filtered_cols <- c("var1", "var2")

dots <- interp(~complete.cases(x), .values = list(x = filtered_cols))

dd2 <- dd %>% filter_(.dots = dots)
#> Error in eval(substitute(expr), envir, enclos): incorrect length (2), expecting: 4

str(dots)
#> Class 'formula'  language ~complete.cases(c("var1", "var2"))
#> ..- attr(*, ".Environment")=<environment: R_GlobalEnv> 

Unfortunately, filtered_cols is getting parsed as a character vector. How can I get interp() to treat filtered_cols as multiple column names to be passed to complete.cases()?

Answer 1

This is what uqs() is for, but you have to use the newer f_interp():

library(lazyeval)
filtered_cols <- c("var1", "var2")
filtered_col_names <- lapply(filtered_cols, as.name)

dots <- f_interp(~complete.cases(uqs(filtered_col_names)))

dd2 <- dd %>% filter_(.dots = dots)