joining dataframe to nested dataframes within purrr::map_*

Question

My aim is to join a dataframe to a dataframes held within a nested list-column, eg:

data(mtcars)
library(tidyr)
library(purrr)

mtcars_nest <- mtcars %>% rownames_to_column() %>% rename(rowname_1 = rowname) %>% select(-mpg) %>% group_by(cyl) %>% nest()
mtcars_mpg <- mtcars %>% rownames_to_column() %>% rename(rowname_2 = rowname) %>% select(rowname_2, mpg)

join_df <- function(df_nest, df_other) {
  df_all <- df_nest %>% inner_join(df_other, by = c("rowname_1" = "rowname_2"))
}

join_df <- mtcars_nest %>%
  mutate(new_mpg = map_df(data, join_df(., mtcars_mpg)))

This returns the following error:

# Error in mutate_impl(.data, dots) : Evaluation error: `by` can't contain join column `rowname_1` which is missing from LHS.

So the dataframe map_* receives from the nested input isn't offering a column name (ie rowname_1 ) to take part in the join. I can't work out why this is the case. I'm passing the data column that contains dataframes from the nested dataframe. I want a dataframe output that can be added to a new column in the input nested dataframe, eg

| rowname_1 | cyl | disp |...|mpg|
|:----------|:----|:-----|:--|:--|

Answer 1

A couple things:

• you should use the tilde to functionize (in purrr ) the function argument to map* ; and
• I think you should be using map instead of map_df , and though I cannot find exactly why map_df doesn't work right, I can get what I think is your desired behavior without it.

Minor point:

• you assign to df_all within join_df() , and the only reason it is working is because that assignment invisibly returns what you assigned to df_all ; I suggest you should be explicit: either follow-up with return(df_all) or just don't assign it, end with inner_join(...) .

Try this:

library(tibble) # rownames_to_column
library(dplyr)
library(tidyr)  # nest
library(purrr)

join_df <- function(df_nest, df_other) {
  df_all <- inner_join(df_nest, df_other, by = c("rowname_1" = "rowname_2"))
  return(df_all)
}

mtcars_nest %>%
  mutate(new_mpg = map(data, ~ join_df(., mtcars_mpg)))
# # A tibble: 3 x 3
#     cyl data               new_mpg           
#   <dbl> <list>             <list>            
# 1    6. <tibble [7 x 10]>  <tibble [7 x 11]> 
# 2    4. <tibble [11 x 10]> <tibble [11 x 11]>
# 3    8. <tibble [14 x 10]> <tibble [14 x 11]>

The new_mpg is effectively the data column with one additional column. Since we know that we have full redundancy, you can always over-write (or remove) data :

mtcars_nest %>%
  mutate(data = map(data, ~ join_df(., mtcars_mpg)))
# # A tibble: 3 x 2
#     cyl data              
#   <dbl> <list>            
# 1    6. <tibble [7 x 11]> 
# 2    4. <tibble [11 x 11]>
# 3    8. <tibble [14 x 11]>

and get your nested and now augmented frames.