Cannot supply list of matrices to populate 3 dimensional array int

[prototaxites]

Describe the bug
When trying to supply a list of matrices to an array[x, y, z] int, I get the following error:

“Error in mde(x) : ‘list’ object cannot be coerced to type ‘integer’”

To Reproduce

library(cmdstanr)

stan_file <- write_stan_file(
  "data {
    array[4, 3, 2] int x;
  }
  transformed data {
    print(x);
  }"
)
mod <- cmdstan_model(stan_file)

a <- matrix(1:6, nrow = 3, ncol = 2)
data = list(x = list(a, a, a, a))
fit <- mod$sample(
  data = data,
  fixed_param = TRUE,
  chains = 1
)

Expected behavior
The data should be read into the model. Data is populated correctly if supplying a 3D integer array instead.

Operating system
macOS 13.4.1

CmdStanR version number
0.6.0

[mike-lawrence]

I saw the discussion on this on Slack and appreciate the argument made there that it may be reasonable to consider this a bug given the infrequent use of both the array() container type and integer() data type in R (edit: on further testing, use of the integer() data type by the user isn't necessary), but FYI the way to do this is:

(
	1:6
	%>% as.integer() #edit: this isn't necessary, and even if as.numeric() is here, cmdstanr converts to int internally
	%>% rep(times=4)
	%>% array(
		dim = c(4,3,2)
	)
	%>% list(x=.)
)  ->
	data
glimpse(data)

Note though print(data) will look odd bc the R default for printing is to print things such that the final dimension defines what is split for printing, rather than the first as might be expected.

So, aside from this being possible to express in R, just slightly unfamiliar/unwieldy, I guess this issue reflects a desire to automate conversion of the list-of-matrices to the proper array format? That is achieved via:

(
	x_as_list_of_matrices
	%>% unlist()
	%>% array( dim = c( length(x) , dim(x_as_list_of_matrices[[1]]) ) )
) ->
	x_as_array

[prototaxites]

I saw the discussion on this on Slack and appreciate the argument made there that it may be reasonable to consider this a bug given the infrequent use of both the array() container type and integer() data type in R

It works as expected if you declare an array[x,y,z] real - so I feel the equivalent for int should also work!

but FYI the way to do this is

I found a different way to generate the correct array structure. Starting with a single matrix with the first column representing two groups:

mat <- matrix(rep(1:6, each = 3), nrow = 6, ncol = 3)
arr <- simplify2array(by(mat, mat[,1], as.matrix))[,-1,] # brackets remove grouping col from result

This returns a [3, 2, 2] matrix, where the final index indexes groups.