I have the matrix a=[5 5 10 10 2 4 5],I want the result to be d=[5 1 2 7;10 3 4 0;2 5 0 0;4 6 0 0],

I have the matrix

a=[5 5 10 10 2 4 5],

I want the result to be

d=[ 5 1 2 7;
   10 3 4 0;
    2 5 0 0;
    4 6 0 0],

[the cyclist]

It is not at all clear to me how d relates to a .

[Azzi Abdelmalek]

How?

[yousef Yousef]

the first column indicates the value.The other columns show the indecis for ex. 5 exists in 1,2&7

Answer 1

[Image Analyst]

How about this:

a=[5 5 10 10 2 4 5]
d = transforma(a); % Call transforma() in the main program.

function d = transforma(a)
d=[ 5 1 2 7;
   10 3 4 0;
    2 5 0 0;
    4 6 0 0];

It meets all the criteria you've given.

[Azzi Abdelmalek]

Nice answer!

[yousef Yousef]

There is not such command in Matlab(transforma)

[Image Analyst]

There is once I defined it. Did you overlook the

function d = transforma(a)

statement where I defined it?

[yousef Yousef]

I want it in general case,like any matrix in this form

Answer 2

[Sean de Wolski]

a=[5 5 10 10 2 4 5],
clear d;
[u,~,iy] = unique(a,'stable');
for ii = numel(u):-1:1
  lhs = [u(ii) find(iy==ii).'];
  d(ii,1:numel(lhs)) = lhs;
end

[Image Analyst]

I knew it . They are getting very close to finishing the Crystal Ball Toolbox. Sean, care to say when it will be released, or is it still a closely held secret?

[Walter Roberson]

It can be done without using find() at all, by using bsxfun(@eq) and logical indexing that against 1:length(a)

[yousef Yousef]

Thanks .Its working well ,just its too complicated.I will be so glad if you explain the main steps in 1 or 2 words

[yousef Yousef]

It works well in small matrices but if the matrix is big such as randi(700,10,20),I found a repetitions in one column while in my case this repetition is not desired.

     6     1     2    12    15
    10     3     7    18     0
     4     4     5     9     0
     1     6    16     0     0
     2     8    19    13     0
     7    10    11     0     0
     8    13     0     0     0
     5    14    20     0     0
     3    17    20     0     0

the repetitions are normal in rows but forbidden in columns,so how to solve this problem.

20 is repeated in the third column

[yousef Yousef]

if you want to see the results and what I mean exactly ,just write your email to send you the results.

[Sean de Wolski]

@Walter, I'd be curious to time bsxfun(...) against find. I agree that any opportunity for a bsxfun, it should be used just 'cuz it's awesome.

@IA, you know how silent we are on future features :)

@Yousef, I do see the duplicated 20 in column two above. But I don't know how you got it. Please give me full reproduction steps.

When I run something like the following, it passes as I expect:

a = randi(700,[1 10000]);

% a=[5 5 10 10 2 4 5],
clear d;
[u,~,iy] = unique(a,'stable');
for ii = numel(u):-1:1
lhs = [u(ii) find(iy==ii).'];
d(ii,1:numel(lhs)) = lhs;
end
T = matlab.unittest.TestCase;
import matlab.unittest.constraints.IsEqualTo
T.verifyThat(ismember(1:numel(a),d(:,2:end)),IsEqualTo(true(1,numel(a))));
T.verifyThat(nnz(d(:,2:end)),IsEqualTo(numel(a)));

Interactive verification passed.

Interactive verification passed.

Answer 3

[Andrei Bobrov]

[a1,b,c] = unique(a(:),'first')
[~,ii] = sort(b); 
a1 = a1(ii); 
c = ii(c);

n = max(histc(c,(1:max(c))'));
f1 = @(x){[x(:)',zeros(1,n-numel(x))]};
p = accumarray(c,(1:numel(c))',[],f1)
out = [a1, cat(1,p{:})];

or with Nan

[a1,b,c] = unique(a(:),'first')
[~,ii] = sort(b); 
a1 = a1(ii); 
c = ii(c);

idx = bsxfun(@times,1:numel(a),bsxfun(@eq,a1,a));
idx(idx==0) = nan;
ii = sort(idx,2);
out = [a1,ii(:,~all(isnan(ii)))];

Answer 4

[Roger Stafford]

Here's another version. Let a be your array and d be the desired result.

 [u,t,q] = unique(a(:),'first','stable');
 [q,p] = sort(q);
 r = p;
 n = size(p,1);
 r(p) = (1:n)';
 t = r(t);
 r = cumsum(1-accumarray(t(2:end),diff(t),[n,1]));  
 d = [i.accumarray([q,r],p)];