Error using linprog

[pierre321]

I installed a new Matlab Version (R 2023a) and wanted to try out Yalmip which I have used successfully for many years on Matlab R2022a.
In order to test it, I used a small linear algorithm adopted from the getting stared example.

yalmip('clear')

% Define variables
x = sdpvar(2,1);

% Define constraints 
Constraints = [sum(x) <= 4, 0 <= x(1) <= 3, 0.5 <= x(2) <= 1.5];

% Define an objective
Objective = -(x(1)+x(2));

% Set some options for YALMIP and solver
options = sdpsettings('verbose',1,'solver','quadprog','quadprog.maxiter',100);

% Solve the problem
sol = optimize(Constraints,Objective,options);

% Analyze error flags
if sol.problem == 0
 % Extract and display value
 solution = value(x)
else
 display('Hmm, something went wrong!');
 sol.info
 yalmiperror(sol.problem)
end

---

While the getting started example works just fine, I'm getting a weird warning for this example:

Hmm, something went wrong!

ans =

'Unknown problem in solver (Turn on 'debug' in sdpsettings) (Error using linprog
 LINPROG(f,A,b,Aeq,beq,LB,UB,X0,OPTIONS) does not accept X0. Use LINPROG(f,A,b,Aeq,beq,LB,UB,OPTIONS) instead.)'

ans =

'Unknown problem in solver (Turn on 'debug' in sdpsettings) '

---

If I turn on debug in the otpions, I'm getting the following error:

Error using linprog
LINPROG(f,A,b,Aeq,beq,LB,UB,X0,OPTIONS) does not accept X0. Use LINPROG(f,A,b,Aeq,beq,LB,UB,OPTIONS) instead.

Error in callquadprog>callsolver (line 56)
[x,fmin,flag,output,lambda] = linprog(model.c, model.A, model.b, model.Aeq, model.beq, model.lb, model.ub, model.x0,model.ops);

Error in callquadprog (line 12)
solveroutput = callsolver(model,options);

Error in solvesdp (line 368)
eval(['output = ' solver.call '(interfacedata);']);

Error in optimize (line 31)
[varargout{1:nargout}] = solvesdp(varargin{:});

Error in test1 (line 18)
sol = optimize(Constraints,Objective,options);

What have I done wrong?

[johanlofberg]

It looks like mathworks have introduced a breaking change in their linprog solver in R2023. To get it to work you ill have to edit calllinprog and remove x0 from the argument list as indicated by the error message

[pierre321]

I deleted it as an input argument for the function:

function [x,fval,exitflag,output,lambda]=linprog(f,A,B,Aeq,Beq,lb,ub,options)

It still appears several time within the function and I don't really have the knowledge to make changes in the function.

I think it's easier if I just use gurobi or stay with Matlab22. Again, thank you very much for your answers.

[johanlofberg]

what file are you editing? The crashing line to edit is is callquadprog>callsolver (line 56) as it still according to your error message calls linprog with too many arguments

[pierre321]

The linprog.m file

[johanlofberg]

You should not touch linprog files. You only edit YALMIPs callquadprog

[pierre321]

Sorry, this was my error. I applied the changes to callquadprod and it works now!!!

[pierre-haessig]

Hello,
Should an issue be opened for this? Issue #1320, which I believe is the exact same problem, was opened but then converted to discussion. Is it better to reopen a new one?

[pierre-haessig]

For the reference, the x0 parameter for warm-starting Matlab linprog isn't listed in the doc of the oldest release I could find (Matlab R2018b linprog). Still, it seems it was accepted until R2022b or so.