Subscripted assignment dimension mismatch

Dear All,

I have recently upgraded to R2016b (from R2013b). When I run my code, I get the following error message:

Subscripted assignment dimension mismatch

Since I have not modified my code, I do not see any logic in it. Do you have any idea about what happened?

Thank you in advance, Antonio Sechi

P.S.: I use a MacBook Pro Retina running 10.12.6.

[KL]

Post the complete error message and the part of the code that's causing trouble!

[Antonio Sechi]

OK. Here is a screenshot.

[KL]

What is the value and size of m? what's the size of objects?

Another tip to get faster answers here, always "post" your code in the proper way (minimal working example that recreates your problem), instead of a screen shot.

[Antonio Sechi]

Thank you all for your suggestions. To reply to Walter Robertson. The code that you suggested (objects(:, :, i, :) = permute(m, [1 2 4 3]); did not eliminate the bug. Since I am a cell biologist, I cannot fully understand what I have to do. For instance, how can I start the debugger? Regarding the comment of KL, I can send you the entire code and test image, however I must be sure that it is kept confidential.

[Walter Roberson]

At the command line give the command

dbstop if error

then run the code. When it stops telling you about the error, go to the command line and command

size(m)
size(objects)

and let us know what it says.

[Antonio Sechi]

m is 331 324 3 (uint8 as indicated in the workspace) objects is 662 647 50 3 (these figures, expect the 3, correspond to the original multi-TIFF image). Objects is 4-D uint8

[Guillaume]

Well, clearly the sizes don't match. objects is (nearly) twice as big as m in the first two dimensions. How is the size of objects determined?

[Antonio Sechi]

at least the first three dimensions of objects correspond to the original data, which is a multiTIFF 662x647 containing 50 frames (attached is one frame). I do not know how objects size is determined, nor I have the chance to get in contact with the persons that wrote the code. Thank you for now.

Answer 1

[Walter Roberson]

What is size(m) and size(Objects) at the time of the failure?

R2016b introduced implicit expansion, similar to bsxfun. For example (1:2).' + (3:5) would previously have generated an error because of trying to add a 2 x 1 to a 1 x 3, but it is now equivalent to bsxfun(@plus, (1:2).', (3:5)) . This change tended to uncover some subtle bugs in code.

But other than that, you might possibly need

objects(:, :, i, :) = permute(m, [1 2 4 3]);

because your 3rd dimension of m is not a scalar to match against the scalar i.

Answer 2

[Guillaume]

In R2014 (a or b, can't remember) mathworks completely rewrote the graphics engine of matlab. With that rewrite came a lot of changes of behaviour (e.g. graphics handle are no longer numeric). It looks like your code is grabbing some figure frame (with your mygetframe function). Possibly the frame is now a different size than you expect (since we don't see how you declared object we don't know if you hardcoded the size) or possibly your function does not even return a frame anymore.

In any case, since your code involves graphics, you will have to use the debugger to check that all these functions that deal with graphics still work as expected.