r - Melt able to remove NA's for some variables but not others? -

i have question regarding melt function reshape2 package in r.

i melt multi-dimensional array, keeping na's of variables, removing na's rest. example illustrate don't want:

library(reshape2) subtotals <- tapply(   mtcars$mpg,   list(mtcars$cyl,mtcars$disp,mtcars$hp,mtcars$drat,mtcars$wt),    sum ) subtotals_melted <- melt(subtotals, na.rm = t) # removes na's vars 

i know there na.rm= option in melt function, removes na's variables. such subtotal-ing operations typically take quite bit of memory due implicit na's, after being melted, wondering if there faster , more memory-efficient way implement such use-case.

thanks in advance!

library(dplyr) library(magrittr)  result =   mtcars %>%   group_by(cyl, disp, hp, drat, wt) %>%   summarize(sum_mpg = sum(mpg, na.rm = true) ) 

this preserve na levels in grouping variables not mpg. in general, though, agree michael bellhouse conditional mutation best done separating 2 separate dataframes.