`reduce` in purrr with `+` but ignore missing value in R

Question

How can I use reduce function together with lag to obtain the summation but at the same time ignore missing values?

a <- data.frame( b = c(1, 2, NA, 3, 4, 5))
a %>% mutate(addition = reduce(map(0:1, ~lag(b, .x)) , `+`))

The resulting table

b	addition
1	NA
2	3
NA	NA
3	NA
4	7
5	9

My desire result is

b	addition
1	1
2	3
NA	2
3	3
4	7
5	9

Thus, the desire result is to ignore the NA and do the addition for the current value and its lag values. I simplified the example but in the actual case, I need the rolling sum of 12.

I have tried the following, and the results are the same for the column addition in which it returns 25 for all rows. Why does this not work?

a <- data.frame( b = c(1, 2, NA, 3, 4, 5))
a %>% mutate(addition = reduce(map(0:1, ~lag(b, .x)) , ~sum(.x, .y, na.rm = TRUE)))

Answer 1

I am not sure if reduce is the right function for such calculation. Here are couple of options to achieve the same result.

1. Replace NA values with 0 and add using +

library(purrr)
library(dplyr)

a %>% mutate(addition = replace(b, is.na(b),0) + replace(lag(b), is.na(lag(b)),0))

#   b addition
#1  1        1
#2  2        3
#3 NA        2
#4  3        3
#5  4        7
#6  5        9

2. Use map2 and sum the two values ignoring NA values.

a %>% mutate(addition = map2(b, lag(b), ~sum(.x, .y, na.rm = TRUE)))

EDIT

Based on your update if you need to do rolling operations then use rolling functions. For eg - with zoo::rollapplyr

a %>% 
   mutate(addition = zoo::rollapplyr(b, 2, \(x) sum(x, na.rm = TRUE),
                      fill = first(b)))

#   b addition
#1  1        1
#2  2        3
#3 NA        2
#4  3        3
#5  4        7
#6  5        9