OpenEuphoria

Re: Bug in math.e?

• Posted by DerekParnell (admin) Nov 30, 2010
• 1279 views

 
atom a 
a = 183.4126   
puts(1,to_string(round(a)))      -- got 183, expected 183 
puts(1,to_string(round(a,10)))   -- got 183.400000000000006, expected 183.4 
puts(1,to_string(round(a,100)))  -- got 183.409999999999997, expected 183.42 
puts(1,to_string(round(a,1000))) -- got 183.413000000000011, expected 183.413 

Try this ...

 
atom a 
a = 183.4126   
? round(a,1) 
? round(a,10) 
? round(a,100) 
? round(a,1000) 

The issue you are seeing is that the "to_string" function outputs the number to the maximum number of decimals it can but the "?" statement only outputs a few decimal places.

The problem is that the rounded values cannot be exactly implemented in the IEEE floating point system that Euphoria uses. This can be demonstrated by this code ...

? atom_to_float64( round(a, 1000) ) 
? atom_to_float64( 183.413 ) 

In both cases above, you should get {240,167,198,75,55,237,102,64} which is the IEEE encoding for the nearest number to 183.413.

The rule-of-thumb is only use floating point numbers if you are prepared to accept a margin-of-error in your calculations. If you are working with currency, this is obviously not acceptable, in which case you need to only work with currency values as 'cents' and format them for output using dollars AND cents.