accumarray not working! unique values

 pre-formatted
Owner Area ID Date&Time     Loc1 Loc2   AvSpeed Loc  Loc   Loc   Loc    L M
1     1    1  734747.9197  418.2 424.9  36.53  414.2 422.5              0 0
1     1    1  734747.9273  424.9 426.3  26.77  422.5 422.5 426.3        0 0
1     1    1  734747.9295  426.3 436.5  45.03  422.5 426.3 422.5 426.3  0 0
1     1    1  734747.9389  436.5 436.5  0.027  422.5 426.3 422.5 426.3  0 0
1     1    1  734747.9669  436.5 440.7  45.23  422.5 426.3 440.7 440.7  0 0
1     1    1  734747.9707  440.7 440.7  0.007  440.7 440.7 440.7 440.7  0 0
1     1    1  734747.9787  440.7 444.2  33.46  440.7 440.7 440.7 440.7  0 0
1     1    1  734747.9831  444.2 444.2  0.004  440.7 440.7 440.7 440.7  1 8
1     1    2  734748.9002  418.2 424.9  36.53    414.2  0    422.5 0      0 0
1     1    2  734748.9079  424.9 426.3  26.77  422.5  0    422.5 426.3  0 0

What I want to do is, for each unique Owner,Area &ID, find unique values form 4 Loc columns(the total number of unique values varies for different combinations of Owner, Area & ID).

And the expected output will look sth like:

 pre-formatted
Owner Area ID StartD&T   StartLoc Loc1  Loc2  Loc3  Loc4 EndLoc EndD&T
1    1     1 734747.9197 418.2  422.5  426.3 440.7      444.2  734747.9831

Start loc is the smallest value in Loc1 & EndLoc is the largest value in Loc2, everything in between comes from the four Loc columns above. Usually, there will be 2-4 different of these for each unique Owner, Area, ID set. In above example, all but last two is a unique set of 1,1,1 so, I need to find the three Loc's in between from among all four Loc cloumns.

I tried accumarray, but can't get it to work. I am hoping some guidance will help me get to at least get started. Thank you so much.

[Walter Roberson]

Please show your code attempt.

[MountainKing]

Q=xlsread('accumarraytry.xlsx');

 pre-formatted
[a1, b1, c1] = unique( Q(:,1) );
[a2, b2, c2]  = unique(Q(:,2) );
[a3, b3, c3]  = unique(Q(:,3) );
tB = accumarray( [c1(:), c2(:),c3(:)], Q(:,8:11), [], @unique, NaN);

"Second input VAL must be a vector with one element for each row in SUBS, or a scalar."

Answer 1

[Sean de Wolski]

Newest

It sounds like what you really want to do is create a matrix of yoru subs as you have and then do a unique 'rows' on this:

M = [c1(:) c2(:) c3(:) c4(:)];
[uv,~,idxM] = unique(M,'rows');

This will tell you what rows are the same (via idxM). From here you can decide how to continue and if accumarray is a good idea or if you want to merge the rows of Q, based on their idxM.

even newer

Even if you were callling it with a vector you are still going to run into the error of unique outputing more than one result. Thus you will need to wrap the results in a cell array in order to have a a scalar output.

new per comment clarification

Ahh. Yes, it means exactly what it says!

You are passing val in as an nx4 matrix. It expects a scalar or an nx1 vector. Thus you need to either call accumarray four times with each vector or not use it.

older

Since you have double precision non-integer values you will have to bin them and find their indices with either histc or unique. Here is a two-dimensional example:

x = [1.1; 1.1; 1.1; 2.2; 2.2; 4.7; 8.9; 11; 11; 8.9]; %unique this
y = rand(1,10)*5; %histc this
data = 1:10; %some data
[u,~,idxx] = unique(x); %where are the unique values in x?
[~,idxy] = histc(y,0:5); %where do the elements of y fall in 0:5

accumarray([idxx(:) idxy(:)],data)

Answer 2

[Walter Roberson]

[a1, b1, c1] = unique( Q(:,1) );
[a2, b2, c2] = unique(Q(:,2) );
[a3, b3, c3] = unique(Q(:,3) );
tB = accumarray( repmat([c1(:), c2(:),c3(:)], 4, 1), reshape( Q(:,8:11), [], 1), [], @(V) {unique(V)}, NaN);

I am a bit uncomfortable about you wanting to apply unique() to the elements, as that is going to return a variable number of results. I have adjusted your code so that the results will be stored as a cell (needed to have accumarray have multiple values per index combination), but I do not understand from your sample output how you plan to output those unique values.

How are the empty cells in your input being represented?