round function fails when used in subscript index

i'm puzzled. the following code runs without a problem:

x=[1 2 3]
x(round(end/2))

but this:

x=3;
y=round(x/2);
x=[1 2 3]
x(round(end/2)

gives the following error: "error using round. Not enough input arguments."

so round fails only if had has been called before here. anyone who can explain this?

[David Sanchez]

Here is what I get:

>> x=[1 2 3]
x(round(end/2))

x=3;
y=round(x/2);
x=[1 2 3]
x(round(end/2))
x =
     1     2     3
ans =

       2
  x =
       1     2     3
  ans =
       2

no error displayed. Are you sure you are doing it all right?

[Peter]

yes I'm sure. using MATLAB R2011b.

[dpb]

...so round fails only if had has been called before here

Doesn't really imply that's the root cause...what if the test case doesn't use end but another constant or other expression?

Since R2011b has been superceded, ir can reproduce the above, go ahead and submit a bug report to official TMW support but likely they'll tell you to upgrade.

Oh--and see if there are any patches available to you for your license.

Answer 1

[Jan]

The implicit usage of end has been affected by many reported bugs in the past:

x = [1 2 3]
x(round(end/2))

See e.g. "end"-related the bugs existing in 2011b: Bug 693663, Bug 783026, Bug 754525, Bug 666701 (this could be your problem). Therefore I suggest to avoid this in general but specify the length you want to measure explicitly:

x = [1 2 3]
x(round(length(x) / 2))

As nice side effect this is even slightly faster.

Btw., reading the list of known bugs is a good programming practice. Then you do not have to be surprised again and dig in the code to find the problems.

[Peter]

thanks all, avoiding the implicit usage of end indeed solves the problem. still weird to me why implicit end within a call to round only fails in this case if it has been preceded by another call to round. but no reason anymore to use implicit end so case closed.