Error when assigning categories to a tall array

I ran into a wiered problem for me as when I tried to assign categories to a tall array, it failed while it went through if I used the gathered version. See attached a small dataset as a test dataset (only 100 strings), and below a working example:
% test
load('testSmall.mat');
code = "1";
meaning = "test";
res = categorical(testSmall, code, meaning); % ok
testTall = tall(testSmall);
res2 = categorical(testTall, code, meaning); % error occurred
The error message:
Error using categorical (line 528)
VALUESET must be numeric when DATA is numeric.
Error in tall/categorical>iGetOutputAdaptor (line 89)
localData = categorical([], extraArgs{:}, varargin{idx:end});
Error in tall/categorical (line 44)
outAdaptor = iGetOutputAdaptor(tdata,varargin{:});
It happened that when I nevigate into the categorical function, the first parameter (inputData) was always empty when I use the tall array testTall as the input. Why this happens?

 Accepted Answer

That also appears to be a bug in how tall handles scalar strings for cats and values. If you make code and meaning into vectors it works fine. Here I give them a dummy value:
code = ["1" "2"];
meaning = ["test" "answer"];
res = categorical(testSmall, code, meaning); % ok
testTall = tall(testSmall);
res2 = categorical(testTall, code, meaning);
I'll report this one to development as it's completely reproducible.

11 Comments

yeah Thanks Sean. as long as there are more than two elements, it works perfectly. I can do this but it is just annoying for me given this code bug.
So all my three questions are related. I've kind of solved the second and the third. My first is still hanging there. Any suggestion please? Without solving this, I just couldn't get my hands on this dataset. Instead, I'm working on a subset these days to get the program for other processings moved on.
I (we) really need a few rows of representative data to help you any further with the original question.
understand. I'm wondering if it's fine if I sent you a subset privately?
Sure. Message me >>char([115 100 101 119 111 108 115 107 64 109 97 116 104 119 111 114 107 115 46 99 111 109]) and I can send you a secure file transfer link.
yeah great! messaged you with a dropbox link (hope this is secure as well lol).
I haven't recieved it; I think my spam filter blocks dropbox links!
Sorry Sean. Just saw your reply. I'll drop you an email again.
Sean, any chance that you received my new email?
I have not. I just checked my spam filter as well and verified that the email from the MATLAB command above works.

Sign in to comment.

More Answers (0)

Categories

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!