Magic binary numbers

(or why floating point numbers are bad)

Now in theory 65.49*100-6549 should equal zero. In practice...

  % echo | awk '{ a=65.49 ; b=100*a-6549; print b }'
  -9.09495e-13

(same happens in perl as well)

I guess 65.49 is an irrational binary. What fun!

I hit this 'cos my money tracking programs told me one credit card had "-0.00" on it. Hmmph.

For anyone who cares the solution is:

  % echo | awk '{ a=65.49 ; b=int(100*a+0.5)-6549; print b }'
  0


I haven't had to do this for 20 years. I guess I don't deal with floating point numbers much!