jsondecode — Parse UTF-8 JSON text into MATLAB-compatible values in RunMat.

jsondecode(text) converts UTF-8 JSON text into MATLAB-compatible data values. Mapping of numbers, booleans, strings, objects, and arrays follows MATLAB semantics, including struct/cell expansion rules.

Syntax

value = jsondecode(text)

Inputs

Returns

Errors

How jsondecode works

• Accepts a character vector or string scalar containing UTF-8 JSON data; leading and trailing whitespace is ignored.
• JSON numbers decode to double-precision scalars or dense tensors that follow MATLAB's column-major layout.
• JSON booleans decode to logical scalars or logical arrays with the same shape as the source JSON array.
• JSON strings decode to char row vectors; homogeneous JSON arrays of strings become string arrays with matching dimensions.
• JSON objects decode to scalar structs whose field names match the JSON keys; arrays of objects become cell arrays of structs that preserve element order.
• JSON arrays decode to numeric, logical, or string arrays when every element shares the same type and the array is perfectly rectangular. Rectangular heterogeneous arrays (including those that mix null with other types) become cell arrays that preserve the original row/column layout, while irregular arrays fall back to 1-by-N cell vectors that keep element order.
• JSON null decodes to the empty double []. When it appears inside a heterogeneous array it becomes a cell element containing [].
• Invalid JSON raises jsondecode: invalid JSON text (...) with the underlying parser message.

Does RunMat run jsondecode on the GPU?

jsondecode does not execute on the GPU and has no provider hook. A numeric gpuArray is not JSON text and is rejected without downloading its payload.

GPU memory and residency

jsondecode is a host-text parser. RunMat does not represent text in numeric GPU buffers, so resident numeric handles are invalid input and reject before provider access; ordinary host text always produces host values.

Examples

Parsing a scalar number

value = jsondecode("42")

Expected output:

value = 42

Decoding a matrix of doubles

text = "[[1,2,3],[4,5,6]]";
matrix = jsondecode(text)

Expected output:

matrix =
     1     2     3
     4     5     6

Converting a JSON object into a struct

person = jsondecode('{"name":"Ada","age":37}');
disp(person.name);
disp(person.age)

Expected output:

Ada
    37

Decoding an array of objects into a cell array

text = '[{"name":"Ada","role":"Researcher"},{"name":"Grace","role":"Engineer"}]';
people = jsondecode(text);
people{1}.name
people{2}.role

Expected output:

people =
  1x2 cell array
    {1x1 struct}    {1x1 struct}

ans = 'Ada'
ans = 'Engineer'

Handling heterogeneous arrays with cell arrays

text = '["RunMat", 42, true]';
data = jsondecode(text)

Expected output:

data =
  1x3 cell array
    {"RunMat"}    {[42]}    {[1]}

Working with null and empty JSON values

value = jsondecode("null");
disp(value);
isempty(value)

Expected output:

[]
ans = logical 1

Using jsondecode with coding agents

Open a RunMat example with live inputs, then ask the agent to explain how jsondecode changes the result.

FAQ

Related Io functions

Open-source implementation

Unlike proprietary runtimes, every RunMat function is open-source. Read exactly how jsondecode is executed, line by line, in Rust.

• View the source for jsondecode in Rust on GitHub
• Learn how the RunMat runtime works
• Found a bug? Open an issue with a minimal reproduction.

About RunMat

RunMat is an open-source runtime that executes MATLAB-syntax code blazing on any GPU. It is licensed under the Apache 2.0 license.

• RunMat automatically optimizes your math for GPU execution on Apple, Nvidia, and AMD hardware. No code changes needed. Simulations that took hours now take minutes.
• Start running code in seconds. RunMat runs in the browser, on the desktop, or from the CLI. No license server, no IT ticket.

Getting started · Benchmarks · Pricing