could summary() report *which* columns are of a certain type as well?

[dan-reznik]

it would be great if skimr's summary() reported counts and the names of columns under each type category, for example:

skim(iris) %>% summary()

## A skim object    
## 
## Name: iris   
## Number of Rows: 150   
## Number of Columns: 5    
##     
## Column types    
## factor (N): fact_col_1, fact_col_2, ..., fact_col_N   
## dbl (M): dbl_col_1, dbl_col_2, ..., dbl_col_M
## int (...)
## lgl (...)
## chr (...)

this way we could quickly assess type distributions. idea: identify column types with the 3-letter symbols in the purrr::map_* family

[elinw]

That's an interesting possibility, do you want to make a PR for it? I think I might want them in separate rows than the numbers, but I'm not sure. Keeping in mind that there could be 1000s of columns I might want to only display the first 5 (or something) names.

[elinw]

In version 2 there is a single column that contains the variable names (skim_variable) and one with the type (skim_type) so it's easy to get the the of variables for a given type as a vector or to make a named list e.g my_list$numeric etc if you use to_list (which is like skim_to_list in v2). I think we should document this in a vignette. But going forward it might be part of thinking about managing skim with very large data sets.

[elinw]

An example in V2

skim(iris) %>% dplyr::filter(skim_type == "numeric") %>% dplyr::select(skim_variable)

# A tibble: 4 x 1
  skim_variable
  <chr>        
1 Sepal.Length 
2 Sepal.Width  
3 Petal.Length 
4 Petal.Width