Given the dplyr workflow:

require(dplyr)                                      
mtcars %>% 
    tibble::rownames_to_column(var = "model") %>% 
    filter(grepl(x = model, pattern = "Merc")) %>% 
    group_by(am) %>% 
    summarise(meanMPG = mean(mpg))

I'm interested in conditionally applying filter depending on the value of applyFilter.

Solution

For applyFilter <- 1 the rows are filtered with use of the "Merc" string, without the filter all rows are returned.

applyFilter <- 1


mtcars %>%
  tibble::rownames_to_column(var = "model") %>%
  filter(model %in%
           if (applyFilter) {
             rownames(mtcars)[grepl(x = rownames(mtcars), pattern = "Merc")]
           } else
           {
             rownames(mtcars)
           }) %>%
  group_by(am) %>%
  summarise(meanMPG = mean(mpg))

Problem

The suggested solution is inefficient as the ifelse call is always evaluated; a more desireable approach would only evaluate the filter step for applyFilter <- 1.

Attempt

The inefficient working solution would look like that:

mtcars %>% 
    tibble::rownames_to_column(var = "model") %>% 
    # Only apply filter step if condition is met
    if (applyFilter) { 
        filter(grepl(x = model, pattern = "Merc"))
        }
    %>% 
    # Continue 
    group_by(am) %>% 
    summarise(meanMPG = mean(mpg))

Naturally, the syntax above is incorrect. It's only a illustration how the ideal workflow should look.

Desired answer

• I'm not interested in creating an interim object; the workflow should resemble:

  startingObject
      %>%
      ...
      conditional filter
      ...
      final object
  

• Ideally, I would like to arrive at solution where I can control whether the filter call is being evaluated or not