Rounding number
Blitz3D Forums/Blitz3D Beginners Area/Rounding number
| ||
| Hello! Please help me. i have absolutely other value in result. float value (a) sometime can be integer, but I have rounding number. Why?  ;;;;;; a_data# = 1471996 b_data = 1471996 a = CreateBank(4) b = CreateBank(4) PokeFloat (a,0,a_data) PokeFloat (b,0,b_data) Print PeekFloat (a,0) Print PeekFloat (b,0) WaitKey() |
| ||
| Floating point numbers are always an approximation, and only a certain number of digits are accurate. Note that that's not a blitzbasic limitation, but inherent in how floating point numbers work in any language. If you need exact numbers, use a different datatype, like double... |
| ||
| "...and only a certain number of digits are accurate." Yes, i found exact numbers for float (in b3d): "0.....999999" by visual test :) Thank you, xlsior! |
| ||
| A good rule of thumb is to assume all floating point values are approximate. But in fact some are exact, including the one in your example. Single precision floats have 24-bit accuracy. Integer values up to 2^24 = 16777216 are exact, including your a_data# = 1471996. It LOOKS inexact because the display, and float-to-string conversion, rounds to six digits. For example 12345.6789 will display as 12345.7 with the rounding. Incidentally, you can always determine at a glance whether a displayed number is Float or Int by whether or not it has a decimal point. Try this to see 6-digit rounded display, and 24-bit exact values, in action. test 1471996 test 1471997 test 2^24 ; Can be represented exactly. test 1 + 2^24 ; Can't be represented exactly, would require 25 bits. WaitKey Function test( x# ) ; Examine integer-valued float. Local i% = x Print Print x Print i End Function |
| ||
| I understood, thank you very much, Floyd! |
   |