MATLAB syntax (parantheses without intermediate steps)

Why is following intermediate calculation step necessary?

temp = abs(rand(10)-eye(10));
result = mean(temp(:));

In Octave is it possible to write the same on one single line:

result = mean(abs(rand(10)-eye(10))(:));

Answer 1

[dpb]

Because to date TMW has chosen to not implement post-addressing expressions on results.

"WHY?" you'd have to ask TMW and it's unlikely they will discuss such internals design decisions/plans publicly.

You can write the expression on one line in MATLAB, too, just more explicitly...

result = mean(mean(abs(rand(10)-eye(10))));  % is one common idiom for 2D arrays

or

result = mean(reshape(abs(rand(10)-eye(10))),:,1);  % is generic

Answer 2

[Fangjun Jiang]

mean(mean(abs(rand(10)-eye(10))))

[Jan]

Or easier:

mean(mean(abs(rand(10) - eye(10))))
% Or:
mean(abs(rand(10) - eye(10)), 'all')

[EDITED] I have confused eye(10) with ones(10). Fixed now.

[Christian Huggler]

My code is just an example. You're right, it works with calling "mean" twice. But let's change the example to "median" - matematical not the same on using this function twice.

[Steven Lord]

That fails to behave the same way as the original code if the inputs have more than 2 dimensions.

x = rand(3, 4, 5);
mean(mean(abs(x - 0.5)))

Answer 3

[Steven Lord]

That intermediate step isn't necessary in this case, since you're using a release that supports the 'all' dimension input argument to mean. I'll create the data as separate variables so you can check that the two-step and one-step approaches give the same result.

x = rand(10);
temp = abs(x-eye(10));
result1 = mean(temp(:));

result2 = mean(abs(x-eye(10)), 'all');

isequal(result1, result2) % true