iset/isetcam/matlab-environment-setup

MATLAB Environment Setup

Covers getting a working MATLAB session with the right repository paths, whether started from the macOS Desktop app or via the VS Code MATLAB extension.

Repository Path Setup

After starting MATLAB, paste these commands into the MATLAB Command Window for the session you intend to use:

addpath(genpath(fullfile(getenv('HOME'),'Documents','MATLAB','isetcam')));
 
which ieInit
 
addpath(genpath(fullfile(getenv('HOME'),'Documents','MATLAB','isetvalidate')));
 
addpath(genpath(fullfile(getenv('HOME'),'Documents','MATLAB','tools','UnitTestToolbox')));
 
addpath(genpath(fullfile(getenv('HOME'),'Documents','MATLAB','isetbio')));

For sessions where you plan to run several parfor-heavy tutorials or examples, you can also start the pool explicitly after path setup, or place this in startup.m:

ieParallelPoolWarmUp('config','conservative','runSilent',true);

Configuring the VS Code MATLAB Extension

To ensure the VS Code MATLAB extension finds your specific MATLAB installation, point it at the MATLAB app bundle:

1. Open Settings in VS Code (Cmd + ,).
2. Search for matlab.installPath.
3. Set it to your MATLAB installation, for example:

/Applications/MATLAB_R2025b.app (or your specific version).

Alternative: edit settings JSON directly

Global user settings (~/Library/Application Support/Code/User/settings.json) apply across all VS Code workspaces unless overridden by a project-specific .vscode/settings.json. Either file can set:

"MATLAB.installPath": "/Applications/MATLAB_R2025b.app",
"MATLAB.matlabConnectionTiming": "onStart",
"files.associations": {
    "*.m": "matlab"
}

VS Code precedence: workspace settings (.vscode/settings.json) override user settings.

Recommended Startup Order

When you need the MATLAB Desktop or Live Editor, start MATLAB from the macOS Desktop first, then open VS Code. With MATLAB.matlabConnectionTiming set to onStart, the VS Code MATLAB extension starts a second background MATLAB process for language-server features after you open a .m file.

This is expected. The Desktop MATLAB owns the GUI session, while the VS Code MATLAB process supports editor integration. Starting the Desktop first avoids a stale VS Code-started MATLAB/MathWorks ServiceHost session blocking the Desktop app from opening later.

Recommended routine:

1. Launch MATLAB from /Applications/MATLAB_R2025b.app.
2. Wait until the MATLAB Desktop is fully open and responsive.
3. Open VS Code and this repository.
4. Open a .m file so the MATLAB extension starts its background session.
5. Paste the repository path commands above into the MATLAB session you

intend to use.

If the Desktop later fails to open after VS Code MATLAB work, quit VS Code and MATLAB, then use Activity Monitor to stop lingering MATLAB, MathWorksServiceHost, MATLABConnector, and related MathWorks helper processes before relaunching the Desktop.

Guarding startup.m

The VS Code MATLAB extension may launch MATLAB differently from the Desktop application. If your personal startup.m does substantial desktop-specific initialization, guard that code so VS Code sessions can start cleanly:

% Detect if MATLAB is being launched by VS Code
isVSCode = ~usejava('desktop') || ...
    ~isempty(getenv('VSCODE_PID')) || ...
    ~isempty(getenv('VSCODE_IPC_HOOK_CLI'));
 
if isVSCode
    disp('Set up ISETCam paths manually — see the matlab-environment-setup skill')
 
    % Share the engine so the VS Code extension can connect for debugging
    matlab.engine.shareEngine;
 
    fprintf('MATLAB initialized for VS Code.\n');
    return
elseif isdeployed
    % Skip initialization for compiled apps
else
    % Standard Desktop initialization
    reset(groot);
    % Your usual plotting/graphics defaults here
end

Verification

• Open a folder: Open your GitHub repository folder in VS Code.
• Start MATLAB: Click the MATLAB icon in the Activity Bar or open a .m

file. The extension should start a MATLAB session in the integrated terminal.

• Path check: In the VS Code MATLAB terminal, run:

``matlab path ``

Verify that the ISETCam directory is included.

Non-Interactive / CLI MATLAB

For non-interactive checks (for example, from an agent or CI-style shell), run the app-bundle executable with -batch; the matlab-evaluation skill covers locating the installed release and the batch command form.

If launching MATLAB from a sandboxed shell fails silently or exits with status 1, retry unsandboxed or escalated — MATLAB may need to write preferences or cache files outside the repository.