hmisc::wtd.quantile(..., weights = ...) error when all values are NAs and/or all weights are 0s or NAs

Question

Context: I compute summary statistics by group, I need to use hmisc::wtd.quantile() function because I have frequency of each value.

Problem: sometimes, all values are NAs and/or all weights are 0s or NAs. Then, wtd.quantile() fails if weights = ... is filled. The error message is:

Error in approx(cumsum(wts), x, xout = c(low, high), method = "constant",  : 
  zero non-NA points

Question: do you know how to avoid this behaviour and get the expected NA output (as all values are NA...)?

Here is some code to reproduce the error:

library(Hmisc)

# ex dataset
set.seed(555)
df_NA = data.frame(values = rep(NA, 5),
                   weights = rnorm(5))

# the problem
with(df_NA, wtd.quantile(values, weights = weights, na.rm = TRUE)) # does not work

with(df_NA, wtd.quantile(values, na.rm = TRUE)) # work
# same as
with(df_NA, stats::quantile(values, na.rm = TRUE)) # work

The outputs and the error:

> with(df_NA, wtd.quantile(values, weights = weights, na.rm = TRUE)) # does not work
Error in approx(cumsum(wts), x, xout = c(low, high), method = "constant",  : 
  zero non-NA points
> with(df_NA, wtd.quantile(values, na.rm = TRUE)) # work
  0%  25%  50%  75% 100% 
  NA   NA   NA   NA   NA 
> # same as
> with(df_NA, stats::quantile(values, na.rm = TRUE)) # work
  0%  25%  50%  75% 100% 
  NA   NA   NA   NA   NA