Documentation for arrays of objects

[cesss]

Hi!

Your library is great! But one thing I couldn't find is how to serialize and deserialize arrays of objects, as well as nested objects. I saw that you have a couple of closed tickets regarding how to work with arrays of objects, so I believe that explaining it in the documentation will help a lot.

My main use for this library would be to dump arrays of C structs (and sometimes a struct can have another array of structs nested inside it)... so documentation about this would be very welcome!

[ibireme]

Thanks. I will add more documentation and demos later.

[kay54068]

hi,

I also have this need. have Do you have any updates?

[VaslD]

I use YYJSON in a Swift project, and this is the minimal code to serialize to (i.e. create, or fill) a yyjson_mut_doc. It's in Swift but should be quite comprehensible to C++ developers:

func encode(in document: UnsafeMutablePointer<yyjson_mut_doc>) -> UnsafeMutablePointer<yyjson_mut_val> {
    switch self {
    case .empty:
        return yyjson_mut_null(document)!
    case let .boolean(bool):
        return yyjson_mut_bool(document, bool)!
    case let .string(string):
        return string.withCString {
            yyjson_mut_strcpy(document, $0)!
        }
    case let .number(decimal):
        return (decimal as NSDecimalNumber).stringValue.withCString {
            yyjson_mut_rawcpy(document, $0)!
        }
    case let .array(array):
        let container = yyjson_mut_arr(document)
        for value in array {
            yyjson_mut_arr_append(container, value.encode(in: document))
        }
        return container!
    case let .dictionary(dictionary):
        let container = yyjson_mut_obj(document)
        for (key, value) in dictionary.sorted(by: { $0.key < $1.key }) {
            yyjson_mut_obj_add(container, key.withCString {
                yyjson_mut_strcpy(document, $0)!
            }, value.encode(in: document))
        }
        return container!
    }
}

Basically you need 3 things to work with arrays (of objects) and objects:

1. A yyjson_mut_doc to allocate memory for new values, which is the input parameter of the function above.
2. A newly allocated yyjson_mut_arr or yyjson_mut_obj as a container, as seen in the two container statements.
3. A yyjson_mut_val (or yyjson_val), either allocated, converted from your C type(s), or obtained somewhere else, to put into the container, which is the output of the function above.

The function above calls itself recursively to process elements in arrays and objects.

[VaslD]

Deserialization is just reversing the input/output of the function:

(Assuming you're switching on the type of a yyjson_val)

case YYJSON_TYPE_ARR:
    var array = [JSON]()
    array.reserveCapacity(unsafe_yyjson_get_len(pointer))
    var iterator = yyjson_arr_iter()
    yyjson_arr_iter_init(pointer, &iterator)
    while let value = yyjson_arr_iter_next(&iterator) {
        array.append(JSON(value))
    }
    self = .array(array)

case YYJSON_TYPE_OBJ:
    var dictionary = [String: JSON](minimumCapacity: unsafe_yyjson_get_len(pointer))
    var iterator = yyjson_obj_iter()
    yyjson_obj_iter_init(pointer, &iterator)
    while let pair = yyjson_obj_iter_next(&iterator) {
        let key = String(cString: unsafe_yyjson_get_str(pair))
        let value = JSON(yyjson_obj_iter_get_val(pair)!)
        dictionary[key] = value
    }
    self = .dictionary(dictionary)

Sorry, still in Swift. The while let syntax checks and breaks the loop if return is null.

[Diego-Hernandez-Moodys]

Let's build the following json

{
    "collection": "embedded_movies",
    "steps": [
        {
            "search": {
                "index": "string",
                "queryVector": [0.0, 0.0, 0.0],
                "limit": 10,
            }
        },
        {
            "output": {
                "_id": 1,
                "score": { "meta": "vectorSearchScore" }
            }
        }
    ]
}

double* vector = malloc( 3 * sizeof(double) );
for (int i = 0; i < 3; i++)
{
    vector[i] = 0.0;
}
int size_vector = 3;
const* collection = "embedded_movies";

// Allocate
yyjson_mut_doc *doc = yyjson_mut_doc_new(NULL);
yyjson_mut_val *root = yyjson_mut_obj(doc);
yyjson_mut_val *root_array = yyjson_mut_arr(doc);

yyjson_mut_val *search = yyjson_mut_obj(doc);
yyjson_mut_val *search_body = yyjson_mut_obj(doc);
yyjson_mut_val *embedding_vector = yyjson_mut_arr_with_double(doc, vector, size_vector);

yyjson_mut_val *output = yyjson_mut_obj(doc);
yyjson_mut_val *output_body = yyjson_mut_obj(doc);
yyjson_mut_val *meta = yyjson_mut_obj(doc);

// Use
yyjson_mut_obj_add_str(doc, root, "collection", collection);

// Build the { search: {} } portion
yyjson_mut_obj_add_val(doc, search, "search", search_body);
yyjson_mut_obj_add_str(doc, search_body, "index", "string");
yyjson_mut_obj_add_int(doc, search_body, "limit", 10);
yyjson_mut_obj_add_val(doc, search_body, "queryVector", embedding_vector);

// Build the { output: {} } portion
yyjson_mut_obj_add_val(doc, output, "output", output_body);
yyjson_mut_obj_add_int(doc, output_body, "_id", 1);
yyjson_mut_obj_add_val(doc, output_body, "score", meta);
yyjson_mut_obj_add_str(doc, meta, "meta", "vectorSearchScore");

// Build the [ search, output ]
yyjson_mut_arr_append(root_array, search);
yyjson_mut_arr_append(root_array, output);

yyjson_mut_obj_add_val(doc, root, "steps", root_array);
yyjson_mut_doc_set_root(doc, root);

// mut_doc -> imutable doc
yyjson_doc *out = yyjson_mut_doc_imut_copy(doc, NULL);

// Free
yyjson_mut_doc_free(doc);
free(vector);
yyjson_doc_free(out);