Why a row of my matrix is a list? R

Question

I got a list of matrices X rows by 2 columns, called list_of_matrices_by2

list_of_matrices_by2[1:3]
[[1]]
[,1]   [,2]
[1,] "7204" "d" 
[2,] "7204" "a" 

[[2]]
[,1]   [,2]
[1,] "2032" "b" 
[2,] "2032" "e" 
[3,] "2032" "a" 

[[3]]
[,1]  [,2]
[1,] "802" "d" 
[2,] "802" "b"

I want to stack all my matrices in a all_pairs matrix, so I did this

all_pairs=do.call(rbind,list_of_matrices_by2)
all_pairs[1:10]
[,1]   [,2]
[1,] "7204" "d" 
[2,] "7204" "a" 
[3,] "2032" "b" 
[4,] "2032" "e" 
[5,] "2032" "a" 
[6,] "802"  "d" 
[7,] "802"  "b" 
[8,] "4674" "c" 
[9,] "4674" "a" 
[10,] "3886" "b"
class(all_pairs)
[1] "matrix"

For some reason, I need the rows of this matrix to be of class matrix. But it shouldn't be a problem since rows of matrix are matrix in R, right. But no!

all_pairs[1,]
[[1]]
[1] "7204"

[[2]]
[1] "d

So here are my questions : 1) Why this? How come a row of a matrix can possibly be a list? 2) What would you do to make it work, i.e. each row of my matrix has to be a matrix?

score 3 · Accepted Answer · edited May 23 '17 at 12:32

I am sure that your list_of_matrices_by2 is like:

x <- list(matrix(list("7204","7204","d","a"), ncol = 2),
          matrix(list("2032","2032","2032","b","e","a"), ncol = 2),
          matrix(list("802","802","d","b"), ncol = 2))

#[[1]]
#     [,1]   [,2]
#[1,] "7204" "d" 
#[2,] "7204" "a" 

#[[2]]
#     [,1]   [,2]
#[1,] "2032" "b" 
#[2,] "2032" "e" 
#[3,] "2032" "a" 

#[[3]]
#     [,1]  [,2]
#[1,] "802" "d" 
#[2,] "802" "b" 

unlist(lapply(x, class))
# [1] "matrix" "matrix" "matrix"

unlist(lapply(x, mode))
# [1] "list" "list" "list"

So you do have a matrix, but it is not a matrix of list instead of numeric. You can perform rbind as usual:

y <- do.call(rbind, x)
#     [,1]   [,2]
#[1,] "7204" "d" 
#[2,] "7204" "a" 
#[3,] "2032" "b" 
#[4,] "2032" "e" 
#[5,] "2032" "a" 
#[6,] "802"  "d" 
#[7,] "802"  "b"

It has class "matrix", but still with mode "list". That is why when you extract the first row you get a list:

y[1, ]
#[[1]]
#[1] "7204"

#[[2]]
#[1] "d"

I don't know how you obtain those matrices. If you can control the generation process, it would be good that you end up with matrices of numeric. If you can not control it, you need convert them manually as below:

x <- lapply(x, function (u) matrix(unlist(u), ncol = 2))
unlist(lapply(x, mode))
# [1] "character" "character" "character"

Then you can do

do.call(rbind, x)

Related:

Ok. Thanks for the link. I am not convinced by it, though. `sapply(DF, f)` is essentially instant for almost any `f` I'm ever facing (including, of course, `mode` and `class`). — Frank, Sep 02 '16 at 13:57
ok thx for the lil tip @Zheyuan Li I finally made it but I was forced to first extract all values of my stack of matrices, and then build a new matrix from thoses extracted values. Do not get exactly why but it works. — hans glick, Sep 02 '16 at 14:01

score 0 · Answer 2 · answered Sep 02 '16 at 14:05

Ok guys, I finally found a solution.

Reminder

all_pairs=do.call(rbind,list_of_matrices_by2)

Extraction of all values of my matrix into a vector

extracted_values=unlist(as.vector(t(all_pairs)))

Build the new matrix

all_pairs_new=t(matrix(data = extracted_values,nrow = 2,ncol = 10000))

Why a row of my matrix is a list? R

2 Answers2

Linked