数据框行按组嵌套列表元素

Question

I have a data frame like this

  id key value
1  x   a     1
2  x   b     2
3  y   a     3
4  y   b     4

read.table(text = "id   key value
x   a   1
x   b   2
y   a   3
y   b   4", header = TRUE, sep = "\t")

And I would like to get a list for each id with sub lists for each key

So with my example the expected output would be :

$x
$x$a
$x$a$value
[1] 1

$x$b
$x$b$value
[1] 2

$y
$y$a
$y$a$value
[1] 3

$y$b
$y$b$value
[1] 4

list(
  x = list(
    a = list(value = 1), 
    b = list(value = 2)
  ), 
  y = list(
    a = list(value = 3), 
    b = list(value = 4)
  )
)

I can achieve it with nested lapply and split but I think there should be a more straightforward way to do it.

Any help would be appreciated.

Answer 1

Overview

Two methods - one using base and the other using plyr - to split your data frame by a group, apply a function over each group, and return the results in a list.

Use base::split.data.frame() followed by an lapply() to extract the value element for each unique id - key pair.

# split data frame
# based on 'id' and 'key' pairs
df.split <-
    split.data.frame(
        x = df
        , f = list( df$id, df$key )
    )
# keep only the value
# element within each list
df.split <-
    lapply(
        X = df.split
        , FUN = function( i )
            i[["value"]]
    )

# view results
df.split
# $x.a
# [1] 1
# 
# $y.a
# [1] 3
# 
# $x.b
# [1] 2
# 
# $y.b
# [1] 4

# end of script #

Use plyr::dlply() to do the same thing, without the need for lapply() .

# load necessary packages
library( plyr )

# splits df by the 'id' and 'key' variables
# and return the 'value' for each pairing
df.split <-
    dlply( 
        .data = df
        , .variables = c( "id", "key" )
        , .fun = function(i) i[["value"]]
    )

# view results
df.split
# $x.a
# [1] 1
# 
# $x.b
# [1] 2
# 
# $y.a
# [1] 3
# 
# $y.b
# [1] 4
# 
# attr(,"split_type")
# [1] "data.frame"
# attr(,"split_labels")
# id key
# 1  x   a
# 2  x   b
# 3  y   a
# 4  y   b

# end of script #

@Colonel Beauvel's answer to the SO post Emulate split() with dplyr group_by: return a list of data frames was helpful in answering this question.

Answer 2

One solution with limited number of split and nested *apply :

lapply(split(df, df$id), function(x) setNames(apply(x, 1L, function(x) as.list(x["value"])), x[["key"]]))

Nested lapply and split alternative :

lapply(split(df, df$id), function(x) lapply(split(x["value"], x$key), as.list))

Improvments are welcome !