My if x ~= y statement is not working

Dear all,

I have a really simple annoying problem which I can't wrap my head around.

In my code:

for k = time
  if E_load2(k) ~= Allowed_winter
    disp(k)
    overload = overload+1;
  else
    overload = overload+0;
  end
end

I can see that the if loop returns the logical true even though I know that E_load2(k) == to Allowed_winter at that the certain k, and thus "overload" becomes larger than it should.

Allowed_winter = -0.5 if that helps.

Please help!

Answer 1

[dpb]

== is precisely that to the LSB; undoubtedly the values in E_load2 in question are close but not identically equal. See the <FAQ> for the "why" in some detail.

For the k in question try

delt=E_load2-Allowed_winter

and observe the result; you'll find it will be something very small but not zero.

Use a tolerance on the comparison or later releases have the function ismembertol that makes writing the expression a little easier.

overload=sum(~ismembertol(E_load2,Allowed_winter));

should do the trick (note no loop needed here).

[Nils Norlander]

Thanks a lot mate! Do I have to define "tol" in "ismembertol" to use this syntax?

[dpb]

Short answer, "no", the two argument version uses a default value; if that isn't ok, then may need to specify one but if it's normal FPP roundoff then the default should work well. See

doc ismembertol  % for details

[Nils Norlander]

OK, seems like ~ismembertol worked out perfectly. Thank you for taking your time to help me!

[dpb]

So why wasn't this the Answer you Accepted, then? :)

[Nils Norlander]

Sorry, I'm pretty new to this community as I learned Matlab for my masters thesis.

[dpb]

No problem; just wondered about the inconsistency if ismembertol did work out (as I figured it would)...

Answer 2

[Adam]

Never compare floating point numbers with == or ~=

Floats are not 100% accurate so you will run into these kinds of confusions. If you get a value that is 0.50000000000001 then it will return false even though it may have been the result of a floating point rounding error.

http://matlab.wikia.com/wiki/FAQ#Why_is_0.3_-_0.2_-_0.1_.28or_similar.29_not_equal_to_zero.3F

[Nils Norlander]

Thank you so much! You've saved me a lot of head ache.

Answer 3

[John D'Errico]

Why are you bothering with this line?

overload = overload+0;

The last time I checked, adding zero to something does nothing but waste CPU cycles. I sincerely doubt that your goal is code that runs more slowly than necessary.

Anyway, it is high time to learn how to use MATLAB.

overload = sum(E_Load2 ~= Allowed_winter);

No loop required.

Or, if time is just the indices of a subset of the elements of E_load2, then use

overload = sum(E_Load2(time) ~= Allowed_winter);

Finally, your problem with the test. This is likely due to the fact that E_load2 or Allowed_winter is not exactly -0.5 in some cases.

[Nils Norlander]

Thank you for your input! I knew the overload = overload+0; was unnecessary, and was simply there in a desperate way to bugtest