readstruct json seems to wrongly combine structs into a vector

I was reading the following json file with the readstruct function:

{
    "demo": [
        {
            "field1": 198.0,
            "field2": 70.0,
        },
        {
            "field3": 1024,
        }
    ]
}

and, to my surprise, I am not getting a cell array with structs, but a struct vector:

>> d = readstruct("demo.json")

d = 

  struct with fields:

    demo: [1×2 struct]

where each of the elements in the struct vector has the following fields:

>> d.demo(1)

ans = 

  struct with fields:

    field1: 198
    field2: 70
    field3: <missing>

>> d.demo(2)

ans = 

  struct with fields:

    field1: <missing>
    field2: <missing>
    field3: 1024

Is this a bug? I would expect the import to return a cell array here.

[Stephen23]

"Is this a bug?"

Not a bug.

It is documented in the section entitled "Get Information on Elements in Nonuniform Structure":

https://www.mathworks.com/help/matlab/ref/readstruct.html#mw_6d5e904a-65dd-411c-be7c-f30f00a765a2

[Adam]

Thanks for pointing this out

Answer 1

[Adam]

Stephen23's comment on the original question contains the correct answer to the question: it's not a bug, but a well documented feature.

I wrote this small function to "fix" my struct vectors. Maybe it could be useful for other people stumbling onto this question

function data = fix_struct_vectors(data)
% this function recurses through a data structure and 
%   + casts all struct vectors to cell arrays 
%   + removes all struct fields for which values are <missing>
if isstruct(data)
    if ~isscalar(data)
        data = num2cell(data);
        data = fix_struct_vectors(data);
        return
    end
    for field = string(fieldnames(data))'
        val = data.(field);
        if ismissing(val)
            data = rmfield(data,field);
        else
            data.(field) = fix_struct_vectors(val);
        end
    end
elseif iscell(data) 
    for ind = 1 : numel(data)
        data{ind} = fix_struct_vectors(data{ind});
    end
end
end

Answer 2

[Vandit]

Hello Adam,

No, this is not a bug. The "readstruct" function in MATLAB reads a JSON file and returns a struct array. In your case, the JSON file contains an array of objects, so it is imported as a struct array. Each element in the struct array corresponds to an object in the JSON array.

If you want to convert the struct array into a cell array of structs, you can use the "struct2cell" function. Here's an example:

cellArray = struct2cell(d.demo)

This will give you a cell array where each cell contains a struct with the fields 'field1', 'field2', and 'field3'.

For more information on "readstruct" and "struct2cell" functions, please refer to the following documentations:

https://www.mathworks.com/help/matlab/ref/readstruct.html

https://www.mathworks.com/help/matlab/ref/struct2cell.html

Hope this helps.

[Adam]

struct2cell does not do what you say here. num2cell does this