No, there is no stable method. Consider this:
x = [2, 5.891, 7.000000000000001, 6.999999999999999];
format short
disp(x)
format long g
disp(x)
fprintf('%.16g ', x);
fprintf('\n');
Some decimal values cannot be represented exactly in binary format. Then it is a question of taste to decide how many decimals are existing. Example:
x = 0:0.1:1;
disp(x(4))
disp(x(4) - 0.3)
fprintf('%.16f\n', x(4));
A kind of solution: Round to a suiting number of decimals at first ( x = round(x / 1e5)*1e5 ), then use fprintf to create a string and count the trailing zeros. Of couse you will still get some repeated .9999... for some input. There is no way to avoid this.
