I agree that your proposed change is clearer.

I think cellfun can only return an empty matrix when you assign the result to a variable, e.g:

x=cellfun(@f,{})

sets x to [], while a sentence like

cellfun(@f,{})

does not set "ans" to anything. But it's probably better to stay safe.

Making structfun a built-in function would require more work, so maybe for the moment we could take your solution and add BIST tests from file #56743.

The issue from the original post is not fixed, but there is a patch for it in comment #11 (file #56743).

I also get the same with Matlab 2024a. I think it's an error message, even if the word "error" is not actually displayed. You cant catch it with try-catch.

I filed bug #66642 for it.

While testing for this issue, I noticed a case that should produce an error but does not (in current default branch).

octave:> ver octave
----------------------------------------------------------------------
GNU Octave Version: 10.0.0 (hg id: 462e19654ec9)
GNU Octave License: GNU General Public License
Operating System: Linux 6.8.0-51-generic #52-Ubuntu SMP PREEMPT_DYNAMIC Thu Dec  5 13:09:44 UTC 2024 x86_64
----------------------------------------------------------------------
octave:> function [a,b]=f(x);a=x;end
octave:> s = struct ("a", 1, "b", 4);
octave:> [a,b]=structfun(@f,s)
a =
   1
   4
b =
   0
   0

I expected an error, because variable b should not have been given any value. In Matlab, you get an error:

Output argument "b" (and possibly others) not assigned a value in the execution with "ff_ab" function.

It looks like the problem is in cellfun:

octave:13> [a,b]=cellfun(@f,{1,4})
a =
   1   4
b =
   0   0

I can file a bug report for that.

Thanks for the info, @lt1234. The problem with your solution is that the number of outputs of a function is not always known a priori. E.g. this function:

function varargout=ff2(x)
varargout={x,2*x};
end

returns two outputs, but nargout(@ff2) returns -1, which means Octave does not know a priori the number of outputs.

Because of that, your solution throws an error in this case:

s = struct ("a", 1, "b", 4);
[a, b] = structfun (@ff2, s);
error: element number 2 undefined in return list

I updated the patch with some tests which have to do with that case. This is attached here.

@lt1234, can you confirm if the patch fixes the issue in your case? (for convinience, I also attach the file, structfun.m, resulting from applying the patch).

(file #56743, file #56744)

There's another case where Matlab result differs slightly from Octave, and the patch I submitted also eliminates the discrepancy.

In Octave (without the patch), the following two calls to structfun return the same:

octave:> x=struct;
octave:> structfun(@(z)z,x)
ans = [](0x1)
octave:> y=structfun(@(z)z,x)
y = [](0x1)

while in Matlab (or Octave with the patch), the first call returns nothing:

>> x=struct;
>> structfun(@(z)z,x)
>> y=structfun(@(z)z,x)
y = [](0x1)

I updated the patch with a couple of tests for that case.

(file #56737)

Correction for the fix, taking into account non-uniform output.

(file #56733)

Same issue.  The command window returns:

Error using foobar
Too many output arguments.

I think the nargout in structfun is associated with structfun.  It should be redirected according to nargout=nargout(fcn);, plus other fixes.  Thanks.

I can reproduce it with Octave-10.0.0. But I don't get the traceback pointing to l.108 of structfun.m (a regression?)

>> FIG.main=figure;
>> structfun(@delete, FIG)
error: delete: function called with too many outputs
>>

The cause of this bug looks to be that 'structfun.m' unconditionally expects the invoked function to return an output in L.108, while 'delete.m' supplies none.