Measure class, precision, significant digits, and divmod()

Ethan Furman ethan at
Tue Jul 15 18:56:59 CEST 2008

Hey all.

My thanks to all who have responded so far with my other questions.  It 
is much appreciated.

Some background on what I'm doing (a good explanation can be found at  When 
measuring, there is some uncertainty as to the *exact* value; so, when 
doing calculations with these measured numbers, one should not present 
an answer that seems more accurate than the source.  For addition and 
subtraction this means the final answer cannot be more precise than the 
least precise input:  5.0 + 21.23 = 26.2 (first number only measured to 
the tenths, so answer only good to the tenths); for multiplication and 
division this means the answer cannot have more significant digits than 
the input with the fewest significant digits:  231 * 0.25 = 58 (0.25 
only has two significant digits, so answer only has two significant 
digits).  Important point -- counting is not subject to precision error.

As I have mentioned before, I am making this Measure class for two 
reasons:  experience with unit testing, I like playing with numbers, I 
am unaware of anything like this having yet been done (okay, three 
reasons ;).

So far, I have the init, add, sub, mul, div, cmp, neg, pos, and abs 
done, and I'm ready to tackle floordiv, mod, and divmod... okay, maybe 
'ready' is too strong a word -- one of the obstacle's I'm facing is that 
I, myself, don't have any real use for this class, so I'm stuck on 
deciding how floordiv, mod, and divmod should function.  As mentioned 
above, counting is precise... so should divmod (a // b, a % b), which 
returns the count of times b goes into a, and the remainder, be an exact 
number, or should it behave like floor(a / b), and preserve the inherent 
inaccury present in the measured value a?  Should I have floordiv 
preserve inaccuracy, and divmod be exact?

Any and all feedback welcome, particularly from anyone who might 
actually use the Measure class.  ;)


More information about the Python-list mailing list