Conditional summarize across a row

I am trying to summarize across columns where NA is treated as 0 so long as at least one column has a numerical value, but return NA if both are NA.

I’m trying to call rowSums(x, na.rm = TRUE) but have it return NA when all rows are NA. For example,

# Data example
a <- c(5, NA, NA)
b <- c(6, 6, NA)
df <- data.frame(a, b)

# RowSums
df$c <- rowSums(df[,c("a","b")], na.rm=T)
head(df)

Gives me:

   a  b  c
1  5  6 11
2 NA  6  6
3 NA NA  0

What I really want is:

   a  b  c
1  5  6 11
2 NA  6  6
3 NA NA  NA

How would I go about this?

>Solution :

Check if all values are NA:

df$c <- ifelse(rowSums(is.na(df)) == ncol(df), NA, rowSums(df, na.rm = TRUE))
df
#    a  b  c
# 1  5  6 11
# 2 NA  6  6
# 3 NA NA NA