Hoist not working for named-list of data frames

tidyverse
#1

Does anyone know why hoist doesn't seem to be working in the following scenario? It works to pluck out a column from a list-column of tibbles, but doesn't work for a named-list-column of data frames...

library(tidyr)
library(dplyr)
library(purrr)
library(rsample)
  
# Works
iris %>% 
  group_nest(Species) %>% 
  hoist(data, Sepal.Length = 'Sepal.Length')
#> # A tibble: 3 x 3
#>   Species    Sepal.Length data             
#>   <fct>      <list>       <list>           
#> 1 setosa     <dbl [50]>   <tibble [50 × 3]>
#> 2 versicolor <dbl [50]>   <tibble [50 × 3]>
#> 3 virginica  <dbl [50]>   <tibble [50 × 3]>

# Doesn't work
iris %>% 
  vfold_cv() %>% 
  mutate(train = map(splits, training)) %>% 
  hoist(train, Sepal.Length = 'Sepal.Length')
#> Error in if (details$repeats > 1) res <- paste(res, "repeated", details$repeats, : argument is of length zero
#2

If you cross post to GitHub (or anywhere, really), can you please include a link here as well? That way, if it gets resolved there, others can follow along.

1 Like
#3

It looks like it's not that the train is named-list-column, but that the larger tibble containining it has a *-list column:

library(tidyverse)
library(rsample)
  
iris %>% 
  vfold_cv() %>% 
  mutate(train = map(splits, training)) 
#> #  10-fold cross-validation 
#> # A tibble: 10 x 3
#>    splits           id     train             
#>  * <named list>     <chr>  <named list>      
#>  1 <split [135/15]> Fold01 <df[,5] [135 × 5]>
#>  2 <split [135/15]> Fold02 <df[,5] [135 × 5]>
#>  3 <split [135/15]> Fold03 <df[,5] [135 × 5]>
#>  4 <split [135/15]> Fold04 <df[,5] [135 × 5]>
#>  5 <split [135/15]> Fold05 <df[,5] [135 × 5]>
#>  6 <split [135/15]> Fold06 <df[,5] [135 × 5]>
#>  7 <split [135/15]> Fold07 <df[,5] [135 × 5]>
#>  8 <split [135/15]> Fold08 <df[,5] [135 × 5]>
#>  9 <split [135/15]> Fold09 <df[,5] [135 × 5]>
#> 10 <split [135/15]> Fold10 <df[,5] [135 × 5]>

iris %>% 
  vfold_cv() %>% 
  mutate(train = map(splits, training)) %>% 
  select(-1)
#> # A tibble: 10 x 2
#>    id     train             
#>    <chr>  <named list>      
#>  1 Fold01 <df[,5] [135 × 5]>
#>  2 Fold02 <df[,5] [135 × 5]>
#>  3 Fold03 <df[,5] [135 × 5]>
#>  4 Fold04 <df[,5] [135 × 5]>
#>  5 Fold05 <df[,5] [135 × 5]>
#>  6 Fold06 <df[,5] [135 × 5]>
#>  7 Fold07 <df[,5] [135 × 5]>
#>  8 Fold08 <df[,5] [135 × 5]>
#>  9 Fold09 <df[,5] [135 × 5]>
#> 10 Fold10 <df[,5] [135 × 5]>

iris %>% 
  vfold_cv() %>% 
  mutate(train = map(splits, training)) %>% 
  select(-1) %>% 
  hoist(train, Sepal.Length = 'Sepal.Length')
#> # A tibble: 10 x 3
#>    id     Sepal.Length train             
#>    <chr>  <named list> <named list>      
#>  1 Fold01 <dbl [135]>  <df[,4] [135 × 4]>
#>  2 Fold02 <dbl [135]>  <df[,4] [135 × 4]>
#>  3 Fold03 <dbl [135]>  <df[,4] [135 × 4]>
#>  4 Fold04 <dbl [135]>  <df[,4] [135 × 4]>
#>  5 Fold05 <dbl [135]>  <df[,4] [135 × 4]>
#>  6 Fold06 <dbl [135]>  <df[,4] [135 × 4]>
#>  7 Fold07 <dbl [135]>  <df[,4] [135 × 4]>
#>  8 Fold08 <dbl [135]>  <df[,4] [135 × 4]>
#>  9 Fold09 <dbl [135]>  <df[,4] [135 × 4]>
#> 10 Fold10 <dbl [135]>  <df[,4] [135 × 4]>

Created on 2020-03-11 by the reprex package (v0.3.0)

#4

Hmm, thanks for finding that quirk. Not sure what to make of your observation. Is there reason to expect this behaviour? What does the asterisk denote anyway?

#5

Upon further inquiry, the asterisk indicates the tibble has row names, and removing the row names solves the issues too:

iris %>% 
  vfold_cv() %>% 
  mutate(train = map(splits, training)) %>% 
  rownames_to_column() %>% 
  hoist(train, Sepal.Length = 'Sepal.Length')
1 Like
#6

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.