Using Floats

• Almost anything you can do to an int you can do to a float: assignment, addition, subtraction, multiplication, division, and comparison with ==, !=, <=, >=, <, and > are all valid. One exception is that there is no modulus operator %.

• Floating point constants must include a decimal point (or they are assumed to be int) and may use either e or E to express exponents: 6.022e23, for example.

• As we saw before, x is declared to be a float variable by the statement float x;. You can print x by using a "%f" field in printf, and read it by using a "%f" field in scanf.

• "Floating point numbers are like piles of sand. Everytime you use one, you lose a little bit." Here are some things to look out for when using floating point numbers:
  • big number + little number = big number.
  • 0.2 = 0.199999999....
  • 0.2 != 0.1 + 0.1
  • (10^30 + -10^30 ) + 1 != 10^30 + ( -10^30 + 1 )
  • rounded number near x - another rounded number near x = noise. [Or why 2c/(-b+sqrt(b^2-4ac)) is better than (-b - sqrt(b^2-4ac))/2a.]
  • 1/MAX_DOUBLE = 0.0 [underflow]
  • x/(x^2 + 1) versus 1/(x + 1/x). [premature overflow]
  • +-INF, NaN, +-0.