function p = update_one_eval_bound(p,i);

% This function computes upper and lower bounds on the scalar nonlinear
% nonlinear functions modelled using nonlinear evaluation-based operators.
% Given a lower bound xL and upper bound xU on x, find upper and lower
% bounds on f(x)

arg = p.evalMap{i}.variableIndex;
xL = p.lb(arg);
xU = p.ub(arg);
L = -inf;
U = inf;
if ~isempty(p.evalMap{i}.properties.bounds)
    % A bound generator is available!
    [L,U] = feval(p.evalMap{i}.properties.bounds,xL,xU,p.evalMap{i}.arg{2:end-1});

elseif strcmpi(p.evalMap{i}.properties.monotonicity,'increasing')
    % No generator is available, but bound follows from monotinicity
    arg = p.evalMap{i}.arg;
    if length(arg{1}) > 1
        disp(['The ' p.evalMap{i}.fcn ' operator does not have a bound operator'])
        disp('This is required for multi-input single output operators');
        disp('Sampling approximation does not work in this case.');
        error('Missing bound operator');
    end
    arg{1} = xL;
    L = feval(p.evalMap{i}.fcn,arg{1:end-1});
    arg{1} = xU;
    U = feval(p.evalMap{i}.fcn,arg{1:end-1});

elseif strcmpi(p.evalMap{i}.properties.monotonicity,'decreasing')
    arg = p.evalMap{i}.arg;
    arg{1} = xL;
    U = feval(p.evalMap{i}.fcn,arg{1:end-1});
    arg{1} = xU;
    L = feval(p.evalMap{i}.fcn,arg{1:end-1});

else
    % To get some kind of bounds, we just sample the function
    % and pick the min and max from there. This only works for
    % simple functions with limited variation...
    % We assume it is f(x,parameters)
    if length(xL)>1
        % We can only sample if it is a scalar function
        disp([p.evalMap{i}.fcn ' is not supported in the global solver (only scalar functions support)'])
        error([p.evalMap{i}.fcn ' is not supported in the global solver'])
    end
    if ~isinf(xL) & ~isinf(xU)
        xtest = linspace(xL,xU,100);
        arg = p.evalMap{i}.arg;
        arg{1} = xtest;
        values = feval(p.evalMap{i}.fcn,arg{1:end-1});
        [minval,minpos] = min(values);
        [maxval,maxpos] = max(values);
        xtestmin = linspace(xtest(max([1 minpos-5])),xtest(min([100 minpos+5])),100);
        xtestmax = linspace(xtest(max([1 maxpos-5])),xtest(min([100 maxpos+5])),100);
        arg{1} = xtestmin;
        values1 = feval(p.evalMap{i}.fcn,arg{1:end-1});
        arg{1} = xtestmax;
        values2 = feval(p.evalMap{i}.fcn,arg{1:end-1});
        L = min([values1 values2]);
        U = max([values1 values2]);
    end
end
p.lb(p.evalVariables(i)) = max([p.lb(p.evalVariables(i)) L],[],2);
p.ub(p.evalVariables(i)) = min([p.ub(p.evalVariables(i)) U],[],2);