dplyr 0.8.1 grouping functions update

RStudio has just released a minor update to dplyr. They had a rethink of new grouping purrr-style functions used to iterate on grouped tibbles.

The changes include:

• group_map() is now used for iterating on grouped tibbles. It however makes no assumptions about the return type of each operation, combining results in a list - similar to purrr::map().
• The previous behaviour was renamed to group_modify(), always returning and combining grouped tibbles by evaluating each operation with a reconstructed grouping structure - similar to purrr::modify().
  
  

An extract from the wiki states that:

group_modify() is an evolution of do(), if you have used that before.

A typical implementation changes from do(function(.$column)) to group_modify( ~ {function(.$column)}), both yielding the same result. group_map similarly applies a function across each grouped tibble, albeit dropping the grouping variables and outputting lists for each group.

Let’s prepare data for the brief demonstration.

iris_grouped <-
  iris %>% 
  
  # Make the table long and thin by collapsing numeric attributes and associated
  # values describing Sepal and Petal lengh and width into the `Field` and `val`
  # columns respectively.
  gather(Field, val, -Species) %>% 
  
  # group by Species and Field to execute an operation over the val data for
  # each respective group
  group_by(Species, Field)

iris_grouped %>%
  
  # The function returns the first 4 rows of the dataframe, formatting the
  # output to HTML using knitr and kableextra, printing a passed header as shown
  # below
  func_tidy_present(., 
                    header = "Tidied *Gathered & Grouped* Dataframe")

Tidied Gathered & Grouped Dataframe

The demonstration below shows the subtle change, introducing more syntactic cohesion to the coding flow.

iris_grouped %>%
  
  # Apply the function for group using the do method
  do(psych::describe(.$val)) %>%
  
  # Tidyup to present
  func_tidy_present(., header = "Result of the *Do* method")

Result of the Do method

iris_grouped %>% 
  
  # Apply the function for group using the Group_Modify method
  group_modify( ~ {psych::describe(.$val)}) %>% 
  
  # Tidyup to present
  func_tidy_present(., header = "Result of the *Group Modify* method")

Result of the Group Modify method

An implentation of the quantile function returns a list that is subsequently enframed and spread before presentation.

iris_grouped %>%
  
  # Apply the function for each group and enframe results within the
  # group_modify function
  group_modify( ~ {
    quantile(.x$val, probs = c(0.25, 0.5, 0.75)) %>%
      tibble::enframe(name = "prob", value = "quantile")
  }) %>%
  spread(Field, quantile) %>%
  
  # Tidyup to present
  func_tidy_present(.,
                    header = "Result of the *Group Modify* method applying the *quantile* function",
                    return_row_count = 9)

Result of the Group Modify method applying the quantile function