Perform a logical test over a group_by() of rows, but mutate the rows individually?

Question:

How do I perform a logical test over a group of rows, but have the output performed on individual rows in R?

Setup:

I have two items, each broken into 4 quarters of a year. If I shipped just 1 time during the year, I need to do follow-on calculations on that quarter, but if I have no shipments during the year, I need show NA so that I can exclude it from those calculations.

df <- data.frame(item = c(rep("A", 4), rep("B", 4)),
                 value = c(0, NA, 5, 1, 0, NA, NA, NA))

> df
  item value
1    A     0
2    A    NA
3    A     5
4    A     1
5    B     0
6    B    NA
7    B    NA
8    B    NA

Desired Output

The end result should be this:

  item  value new_value
  <chr> <dbl>     <dbl>
1 A         0         0
2 A        NA         0
3 A         5         5
4 A         1         1
5 B         0        NA
6 B        NA        NA
7 B        NA        NA
8 B        NA        NA

What I Have Tried:

I'm generally using the library(tidyverse), but I'm not absolutely wedded to it if there's a base solution. My first thought, however, throws an error:

library(tidyverse)

df %>% 
  group_by(item) %>% 
  mutate(new_value = if_else(sum(value, na.rm = TRUE)>0, coalesce(value, 0), NA_real_))

Error in `mutate()`:
! Problem while computing `new_value = if_else(sum(value, na.rm = TRUE) > 0, coalesce(value, 0), NA_real_)`.
ℹ The error occurred in group 1: item = "A".
Caused by error in `if_else()`:
! `true` must be length 1 (length of `condition`), not 4.

I'm only using the group_by() in order to have the sum(x, na.rm = TRUE) act over the group, but then I want the coalesce() to act on each individual row. I also know that dplyr's if_else() operates differently than the base ifelse(), so I tried that:

df %>% 
  group_by(item) %>% 
  mutate(new_value = ifelse(sum(value, na.rm = TRUE)>0, coalesce(value, 0), NA_real_))

# A tibble: 8 × 3
# Groups:   item [2]
  item  value new_value
  <chr> <dbl>     <dbl>
1 A         0         0
2 A        NA         0
3 A         5         0   # should be '5'
4 A         1         0   # should be '1'
5 B         0        NA
6 B        NA        NA
7 B        NA        NA
8 B        NA        NA

...which is closer to the desired output, but it appears to only use the first row in the group, and I want the action to be on each individual row, as you see in the Desired Output above.

What am I misunderstanding?

You could also use:

df %>%
   group_by(item) %>%
   mutate(new_val = coalesce(value,0) * any(value>0))

# A tibble: 8 × 3
# Groups:   item [2]
  item  value new_val
  <chr> <dbl>   <dbl>
1 A         0       0
2 A        NA       0
3 A         5       5
4 A         1       1
5 B         0      NA
6 B        NA      NA
7 B        NA      NA
8 B        NA      NA

You want to use if/else instead of ifelse because there is a single condition per group to check rather than a condition per row.

library(dplyr)

df |> 
  mutate(value = if (any(value > 0, na.rm = TRUE)) pmax(value, 0, na.rm = TRUE) else NA, .by = item)

  item value
1    A     0
2    A     0
3    A     5
4    A     1
5    B    NA
6    B    NA
7    B    NA
8    B    NA