Convert list of values to a list of list indexes for each value

Question

I have a list 'cats.list' with 6 elements. There are 9 unique integers that are members of one or more elements. E.g.

cats.list <- list(c(1, 2, 6), c(1, 8, 9), c(3, 4, 5, 7),
                  c(3, 6, 7), c(1, 3, 7, 8, 9), c(4, 5, 9))

I want to create a list with one element for each of the 9 integers in 'cats.list'. Each element in the new list should contain the list indexes in 'cat.list' for a given integer.

For example, 1 occurs in the list elements 1, 2, 5 in 'cat.list'. 2 occurs in element 1 only. 3 occurs in element 3, 4, 5. So the first three element in the new list would be:

el.list <- list(c(1, 2, 5), 1, c(3, 4, 5)...) 

How can I create such a list of indexes for any 'cats.list'?

Answer 1

1) reshape2 Use melt in reshape2 to convert cats.list into a data frame whose first column value is the element and whose second column L1 is the corresponding component number in cats.list that that element belongs to. Then unstack that with the indicated formula.

library(reshape2)

unstack(melt(cats.list), L1 ~ value)

giving:

$`1`
[1] 1 2 5

$`2`
[1] 1

$`3`
[1] 3 4 5

$`4`
[1] 3 6

$`5`
[1] 3 6

$`6`
[1] 1 4

$`7`
[1] 3 4 5

$`8`
[1] 2 5

$`9`
[1] 2 5 6

2) split We could also do it this way without any packages. rep(seq_along(L), L) equals m$L1 from (1) and unlist(cats.list) equals m$value from (1).

L <- lengths(cats.list)
split(rep(seq_along(L), L), unlist(cats.list))

3) stack/unstack We can also do this using only base R and stack/unstack if we name the cats.list components.

cats.named <- setNames(cats.list, seq_along(cats.list))
unstack(stack(cats.named), ind ~ values)

Note

We can plot this as a bipartite graph like this:

library(igraph)
library(reshape2)

m <- melt(cats.list)
M <- table(m)
g <- graph_from_incidence_matrix(M)
plot(g, layout = layout_as_bipartite)

Answer 2

Use -

cats.list <- list(c(1, 2, 6), c(1, 8, 9), c(3, 4, 5, 7), c(3, 6, 7), 
                  c(1, 3, 7, 8, 9), c(4, 5, 9))
output <- c()
for(i in sort(unique(unlist(cats.list)))){
  output <- c(output, list(grep(i,cats.list)))
}

Output

[[1]]
[1] 1 2 5

[[2]]
[1] 1

[[3]]
[1] 3 4 5

[[4]]
[1] 3 6

[[5]]
[1] 3 6

[[6]]
[1] 1 4

[[7]]
[1] 3 4 5

[[8]]
[1] 2 5

[[9]]
[1] 2 5 6

Explanation

unlist(cats.list) flattens the existing list, wrapping it with unique and sort creates the search list with which you can iterate for the search

The magic lies in grep(i,cats.list), which readily gives what you want for each search.

Putting it together in an output list is trivial. Hope that helps!

EDIT

Thanks to @ G. Grothendieck, this can be shortened to --

output <- lapply(sort(unique(unlist(cats.list))), grep, cats.list)