EBSD data become unreadable when switching from 5.11.2 to 6.1.1

[Remsdel]

I can not longer read or plot the ebsd variable when using 6.1.1 while it was fine with 5.11.2. Any known issues with MATLAB 2024a for the 6.1.1 version?

[kilir]

Hi,
please describe the steps you were trying to do and if there are any error messages, post those as well.
Cheers, Rüdiger

[Remsdel]

Yes, I am glad to describe it for you.
It is in fact rather strange.
When opening the ebsd variable under mtex 6.1.1, it appears to be empty without the usual properties visible (see screenshot below). Any operations with it give an error message because it is empty. Strangely it does not apply to the grains variable (see screenshot).
The same variables were opened with mtex-5.11.2 without a problem (I have been using the later for a long time without issues).
I running matlab 2024a on windows 11.

[DriptD]

Hi,

Although I could use the EBSD variable, I could not check the contents. I wrote more about the problem here - #2521 (comment). For the time being, I moved back to v5.11.2.

Sincerely,
Dripta

[ralfHielscher]

Hi everybody,

I found the reason for this behavior. There are at least three ways to fix this, each of which has some disadvantages. Therefore it is very important for me to understand, which variable type you want to click on: EBSD, grains, orientations, vectors, slipSystems .... please list them all.

Now the reason and the potential solutions. Reason: EBSD/numel reports Matlab the ebsd variable length 1000000 -> matlab wants to display ebsd(1) .... ebsd(1000000) in one column.
Why did this not appear in MTEX 5.X? Because EBSD/numel always reported "1"

Solutions:

1. change EBSD/numel back to returning "1" Drawback: Inconsistent with the rest of Matlab / MTEX as numel (number of elements) is for any vector / matrix the number of its elements.
2. change the Matlab function "arrayviewfunc" This would be the most consistent and efficient method. However, it might be version dependent and unstable for future Matlab version.
3. Check in EBSD/numel who is calling this function. If it is called by "arrayviewfunc" return "1" otherwise the correct result. Drawback: slow when used in a loop.

Now it depends on you. How important is the feature of clicking on a variable? Why are you doing it? For which variables you are doing it? The above applies to almost all variable types in MTEX.

Ralf

PS: Not being able to display the variable in the variable editor does mean it is not loaded.

Ralf.

[DriptD]

Hi Ralf,

I am sorry if these sound naive. I am unable to locate the script needed to make the change listed in (1). Also, what change should I make to the Matlab function arrayviewfunc?

Sincerely,
Dripta

[Remsdel]

Thanks for looking this up. Concerning your question, my specific issue is unfortunately not limited to the feature of clicking on the EBSD variable to see its properties. More importantly, it is also that I cannot do ANY operation with the ebsd variable, such as plotting, calcgrains…. It is basically unusable. Note that it is not the case for the grains variable. What is confounding to me is why most people do not seem to experience this issue. There must be something specific to our setup? Would an upgrade to a more recent Matlab version potentially solve it? Cordially, Rémi

[ralfHielscher]

ok, lets investigate this.

1. What gives you

mtexdata twins

do you see an output in the command line?
2. Now lets try to plot

plot(ebsd,ebsd.orientations)

do you get a figure with an EBSD map?
3. Next compute grains

grains = calcGrains(ebsd)

does this work? Do you get any output?

I'm very sorry you have those problems. I really want to figure out where they orgininate from.

All the best, Ralf.

[Remsdel]

Issue seems to be centered on gridified esbd data. Non-gridified ebsd data can be handled (plotted, grains calculated). Ebsd data gridified in 5.11 cannot be handled in 6.1 Trying to gridify in 6.1 ebsdS=gridify(ebsd) crashed Matlab

[ralfHielscher]

Can you provide some data that reproduce this problem?

[Remsdel]

Dear Ralf, I can send some data, buy I still need to figure out which one would be really useful for you. In the meantime, I notice the following warning when loading in 6.1 a gridified ebsd dataset generated in 5.11 ***@***.*** There is apparently an unrecognized ‘rot’ field. This variable is unusable in 6.1. Note also that if I generate the gridified data in 6.1, it is then OK. So an obvious solution would be to gridify data in 6.1 but I have a lot of past data for which I mostly work on the gridified dataset so I might have not saved the ungridified one. Would there still be a fix to be able to use gridified data generated in 5.11 with the 6.X environment, that would be awesome. Cordially, Rémi

[vtvivian]

An easy workaround for this could be to un-gridify the variable in 5.11 then resave.

% in MTEX 5.11
ebsd = EBSD(ebsdGrid);
save(ebsd, "file.mat");
% now go to MTEX 6.1
startup_mtex;
%load file.mat
ebsdGrid6 = ebsd.gridify;

[ralfHielscher]

I have fixed this issue in 7e20a3d . Replacing EBSDAnalysis/@EBSD/EBSD.m by the file of this commit should resolve your issue.

Ralf.

[Remsdel]

Yes indeed! It does work now. Thanks a lot I still cannot see the content of the ebsd data like I used to but that is a lesser worry for the moment. I will look again at your reply concerning this issue. Note that I immediately tried the isOuterBoundary function that was the initial reason for me to switch to the 6.1 version. It does kind of work but gives me a 2D array instead of a column like I get normally with isBoundary (logical column array of the same length at the number of grains). I noticed that one of the 2D array dimension was the same as the expected number of grains. By doing the rather awkward manipulation borderGrains = logical(sum(borderGrains))’, I do fall back on the expected logical column array but I do not understand why isOuterBoundary would give me a 2D array. Is this normal? Cordially, Rémi