how to use unique in a loop

I'm trying to use unique as a for loop control variable and getting an error.

In code I used about two years ago, I used this format:

___________________________________________________________

for nTrDop = unique(trDp); % cycle unique detection Doppler locations

indRng = (trDp == nTrDop); % all range indices for given Doppler

rngPerDop = trRg(indRng); % range values for range indices

tmpTrPwr = trPCRDMap(rngPerDop, nTrDop); % amps for range locations

............ more processing on set of range-Doppler signal amplitudes .........

end

(I've abbreviated the code for two reasons: 1) some is proprietary; 2) it's

irrelevant to my question.)

__________________________________________________________

Today, I tried to use what seems to be the same format but get an error with

this code:

___________________________________________________________

for nCPI = unique(E1);

id1 = (E1 == nCPI);

F1 = E1(id1); % get MofN1 rows for CPI #nCPI

.............. again, more processing

___________________________________________________________

The error occurs on ' id1 = (E1 == nCPI)' : "Matrix dimensions must agree"

and I see that, generally, E1 and nCPI ARE different sizes. However, in the

earlier code, it seems I was able to compare individual elements of 'trDp' to

individual values 'nTrDop' using this format. Has MATLAB code changed?

If not, can someone identify the problem?

[Rik]

This code will only work as expected when it is a vector, and then maybe only either a row or a column vector (I'm on mobile so I can't check).

Why don't you use ismember? It can both provide the unique values, and the locations. Or even the second output of unique, which could also be helpful here.

[Dean Ranmar]

Thanks, Rik. The variables in questions are vectors. I will see what I can do with ismember. I did write a workaround but I don't yet understand why the new code fails. Even if I find a better or just-as-good solution with ismember, I've found it always pays to understand WHY something doesn't work ... Again, thanks.