Write function to perform conditional summarize in R using named list

Question

I am trying to write a function that takes a tibble and a list of filter specifications and performs a conditional summarize based on those filter specifications.

# Sample DF with a column to summarize and 2 ID columns.
df <- tibble(
    to_summarize = c(1, 2, 8, 9),
    ID1 = c('A', 'A', 'C', 'A'),
    ID2 = c('X', 'Y', 'Z', 'X')
)

We can conditionally summarize using both IDs (returns 10), or using 1 ID (returns 12).

df %>%
    summarize(
        total1 = sum(to_summarize[ID1 == 'A' & ID2 == 'X']),
        total2 = sum(to_summarize[ID1 == 'A'])
    )

I want to allow the same flexibility within a function. The user should be able to provide a list of filters or an empty list (where the summarize function would execute on the entire column, with no filtering).

I imagine the easiest way being with a named list, where each name is a column to filter on, and each value is the value to filter that column on.

filters <- list(
    ID1 = 'A',
    ID2 = 'X'
)

# Here is my attempt at a function to implement this:
summarise_and_filter <- function(df, filters) {
    df %>%
        summarise(
            total = sum(to_summarize[names(filters) == unname(unlist(filters))]))
}

# It does not work, it just returns zero
df %>%
    summarise_and_filter(
        filters = filters
    )

# I imagine the function might need to call map in some way, or perhaps imap?
map_summarise_and_filter <- function(df, filters) {
    df %>%
        summarise(
            total = sum(
                to_summarize[
                    imap_lgl(
                        filters, 
                        ~.y == .x
                    )]
            )
        )
}

# But this also returns zero
df %>%
    map_summarise_and_filter(
        filters = filters
    )

Answer 1

There are two operations done and one of them can be dynamically calculated

library(dplyr)
df %>%
    mutate(total2 = sum(to_summarize[ID1 == filters[['ID1']]])) %>% 
    filter(across(starts_with("ID"), ~ . == 
                filters[[cur_column()]])) %>%
    summarise(total1 = sum(to_summarize),total2 = first(total2))

-output

# A tibble: 1 x 2
  total1 total2
   <dbl>  <dbl>
1     10     12

---

If we want to do this without filter , then reduce the across output to a single logical vector to subset

library(purrr)
df %>% 
  summarise(total1 = sum(to_summarize[across(starts_with('ID'), 
   ~ . == filters[[cur_column()]]) %>% 
            reduce(`&`)]), 
     total2 = sum(to_summarize[ID1 == filters[['ID1']]]))

-ouptut

# A tibble: 1 x 2
  total1 total2
   <dbl>  <dbl>
1     10     12