Creating tibble or data frame of tibbles or data frames and other class

Question

Is it possible to create a tibble or data.frame, which has columns that are integers and other columns that are tibbles or data.frames?

E.g.:

library(tibble)
set.seed(1)
df.1 <- tibble(name=sample(LETTERS,20,replace = F),score=sample(1:100,20,replace = F))
df.2 <- tibble(name=sample(LETTERS,20,replace = F),score=sample(1:100,20,replace = F))

And then:

df <- tibble(id=1,rank=2,data=df.1)

which gives this error:

Error: Column `data` must be a 1d atomic vector or a list

I guess df.1 has to be a list for this to work?

The length of every column in a data rame (or Ribble) must be the same. Thus you can't add a data frame as a column in another data frame unless its length (number of columns) is equal to the number of rows in the higher-level data frame. — Thomas, Dec 13 '17 at 20:48
You can do `df <- tibble(id=1,rank=2,data=list(df.1))`. But this type of data structure can be a mess to work with if you are not careful. — MrFlick, Dec 13 '17 at 21:08

www · Accepted Answer · 2017-12-13T21:34:12.160

Is this what you are looking for? I think the key is the length of each column should be the same, and we need to use list to create a list column to store df.1 and df.2.

df <- tibble(id = 1:2,
             rank = 2,
             data = list(df.1, df.2))
df
# # A tibble: 2 x 3
#      id  rank              data
#   <int> <dbl>            <list>
# 1     1     2 <tibble [20 x 2]>
# 2     2     2 <tibble [20 x 2]>

head(df$data[[1]])
# # A tibble: 6 x 2
#    name score
#   <chr> <int>
# 1     G    94
# 2     J    22
# 3     N    64
# 4     U    13
# 5     E    26
# 6     S    37

head(df$data[[2]])
# # A tibble: 6 x 2
#    name score
#   <chr> <int>
# 1     V    92
# 2     Q    30
# 3     S    45
# 4     M    33
# 5     L    63
# 6     Y    25

And since the structure of each tibble in the data column are the same. We can use tidyr::unnest to expand the tibble.

library(tidyr)
df_un <- unnest(df)
# # A tibble: 40 x 4
#       id  rank  name score
#    <int> <dbl> <chr> <int>
#  1     1     2     G    94
#  2     1     2     J    22
#  3     1     2     N    64
#  4     1     2     U    13
#  5     1     2     E    26
#  6     1     2     S    37
#  7     1     2     W     2
#  8     1     2     M    36
#  9     1     2     L    81
# 10     1     2     B    31
# # ... with 30 more rows

And we can also nest the tibble, making it back to the original format with a list column.

library(dplyr)
df_n <- df_un %>%
  group_by(id, rank) %>%
  nest() %>%
  ungroup()
df_n
# # A tibble: 2 x 3
#        id  rank              data
#     <int> <dbl>            <list>
#   1     1     2 <tibble [20 x 2]>
#   2     2     2 <tibble [20 x 2]>

# Check if df and df_n are the same
identical(df_n, df)
# [1] TRUE

score -1 · Answer 2 · answered Dec 13 '17 at 21:45

Using tidyr's nest:

set.seed(1)
df.1 <- data.frame(name=sample(LETTERS,20,replace = F),score=sample(1:100,20,replace = F))
df.2 <- data.frame(name=sample(LETTERS,20,replace = F),score=sample(1:100,20,replace = F))

I can create a tibble where df.1 is nested under id and rank:

library(dplyr)
library(tidyr)

data.frame(id=1,rank=2,data=df.1) %>% nest(-id,-rank)

# A tibble: 1 × 3
     id  rank              data
  <dbl> <dbl>            <list>
1     1     2 <tibble [20 × 2]>

For having both df.1 and df.2 in a tibble, I'd simply do:

data.frame(id=c(1,2),rank=c(2,1),data=c(df.1,df.2)) %>% nest(-id,-rank)


# A tibble: 2 × 3
     id  rank              data
  <dbl> <dbl>            <list>
1     1     2 <tibble [10 × 4]>
2     2     1 <tibble [10 × 4]>

Any ideas why both the 20-rows, 2-column data frames become 10-row, 4-column data frames? — www, Dec 13 '17 at 21:58
And any ideas why you did not create list column like MrFlick and I suggested? — www, Dec 13 '17 at 22:06

Creating tibble or data frame of tibbles or data frames and other class

2 Answers2