6.2 Errors and exceptions

Exceptions and response codes

The functions of API for MATLAB will throw a MATLAB exception in case of errors. There are two types of exception messages:

• Simple calls to functions setting up the model will throw an exception with a single string describing the error, for example missing or incorrect data.

• Calls to functions such as mosekmodel.solve which invoke the MOSEK library will throw exceptions with a standardized string containing a MOSEK response code and an explanation (see below). The response code can be cross-referenced against the list of MOSEK response codes in Sec. 13.6 (Response codes).

For this reason it is a good idea to call MOSEK functions in a try-catch block. The one case where it is extremely important to check for an exception is when calling mosekmodel.solve. For more information see Sec. 6.1 (Accessing the solution).

As an example, consider:

model = mosekmodel(objective = [NaN], numvar = 1);
try
  model.solve();
catch ME
  fprintf("An error during solve(). Message:\n%s\n", ME.message);
end

This will produce as output:

An error during solve(). Message:
MSK_RES_ERR_NAN_IN_C(1470):The objective vector c contains an invalid value for variable '' (0).

In many cases (especially related to licensing) a much more verbose error message will be printed to the log.

Parsing the exception message.

An exception message produced by MOSEK during a call to mosekmodel.solve has the format

ERROR_NAME(ERROR_CODE):MESSAGE

where the ERROR_NAME is a string and ERROR_CODE is an integer, both of which can be cross-referenced against Sec. 13.6 (Response codes). The integer value is not guaranteed to remain constant between MOSEK versions, so it is recommended to test for equality of strings in ERROR_NAME.

The message can be parsed into individual components as follows:

model = mosekmodel(objective = [NaN], numvar = 1);
try
  model.solve();
catch ME
  expr = "([A-Z_]*)\(([0-9]*)\):(.*)";
  [tokens, matches] = regexp(ME.message, expr, "tokens", "match");
  fprintf("Error name: %s\n",   tokens{1}{1});
  fprintf("Code:       %.0f\n", str2double(tokens{1}{2}));
  fprintf("Message:    %s\n",   tokens{1}{3});
end

which leads to an output such as

Error name: MSK_RES_ERR_NAN_IN_C
Code:       1470
Message:    The objective vector c contains an invalid value for variable '' (0).

Optimizer errors and warnings

The optimizer may also produce warning messages. They indicate non-critical but important events, that will not prevent solver execution, but may be an indication that something in the optimization problem might be improved. Warning messages are normally printed to a log stream (see Sec. 6.3 (Input/Output)). A typical warning is, for example:

MOSEK warning 53: A numerically large upper bound value  6.6e+09 is specified for constraint 'C69200' (46020).

Error and solution status handling example

Below is a source code example with a simple framework for handling major errors when assessing and retrieving the solution to a conic optimization problem.

Listing 6.2 Sample framework for checking optimization result. Click here to download.

    % Solve the model
    try
        model.solve();
    catch ME
        warning("An error during optimization; handle it here.");
        rethrow(ME);
    end
    
    % We check if the interior-point solution exists and what status it has
    [exists, prosta, solsta] = model.hassolution("interior");

    if exists
        disp("Solved the problem with statuses:");
        disp(prosta);
        disp(solsta);

        switch solsta
            case "OPTIMAL"
                disp("Optimal solution found:");
                x = model.getsolution("interior");
                disp(x);
            case "PRIM_INFEAS_CER"
                disp("The problem is primal infeasible.");
            case "DUAL_INFEAS_CER"
                disp("The problem is dual infeasible.");
            case "UNKNOWN"
                disp("Solution status UNKNOWN. This could indicate numerical issues");
            default
                disp("Another solution status:")
                disk(solsta)
        end
    else
        warning("Solution does not exists");
    end