Results of math with integer

Dear community,

I recently had to work with uint8 and did some simple calculations over it. I tought using integer math tells that if we woukd result in a non integer result, the non integer part gets truncated:

a=uint8(4);
b=uint8(7);
median([a b])

ans = uint8 6

shouldn't this be 5, since the median of 4 and 7 would be the mean = 5.5 and then truncated to 5. Or is this just a matlab convention that results are rounded like doubles?

similarly

mean([a b])

ans = 5.5000

the result is a double, but shouldn't it be a uint8 (and rounded according to whatever convention)? The data type is explicitly supported in the documentation of mean.

I was just wondering.

best regards

Jonas

Answer 1

[Stephen23]

"I tought using integer math tells that if we woukd result in a non integer result, the non integer part gets truncated:"

What you were taught is true for some other languages, but note that "MATLAB rounds the result, when necessary, according to the default rounding algorithm."

source: https://www.mathworks.com/help/matlab/matlab_prog/integers.html

The MEDIAN() documentation states "median computes natively in the numeric class of A, such that class(M) = class(A).", which together with the rounding information above explains the output that you show (assuming that for an odd number of elements the middle two are averaged).

The MEAN() output class is given in this table:

https://www.mathworks.com/help/matlab/ref/mean.html#bt5b82t-1-outtype

mean(uint8([4,7]),'native')

ans = uint8 6

[Jonas]

thanks for clearing this up.