How to keep ordering of found indices when using contains?

Hi Everyone,

I have a cell array called "data" containing a field "data.labels". In addition to that I have the cell "lbls" containing label strings and I would like to find the indices of elements in data.labels that contain lbls.

Using

idx = find(contains(data.labes,lbls));

is doing the job but returns the indices sorted in ascending order. How can I keep the original ordering of the indices?

Thanks in advance and all best,

M.

Answer 1

[dpb]

It's not being sorted, it's returning the matches that are found globally just like you asked it to do, not by individual elements in the searching-for array. The contains operation is completed, then that result is passed to find. The locations in the label array are where they are; there's no way to know which of the elements it was that was the one found at any given location; that's not what you asked for.

If you want them returned individually, you'll have to do the search for the elements in lbls one at a time; don't think of any way to differentiate which match is which, otherwise.

Example (modified from cotains documentation):

str = ["Anne","Elizabeth","Marianne","Tracy"];  % sample data; struct is immaterial to problem
pattern = "anne";                               % find one location
TF = find(contains(str,pattern,'IgnoreCase',true));
TF =
     1     3

returns the two locations expected...nothing to see here, really.

pattern = ["anne";"liz"];  % now find more than one element
TF=find(contains(str,pattern,'IgnoreCase',true));
find(TF)
TF =
     1     2     3

Again, it finds then all but the '2' is for 'Elizabeth'. Nothing was sorted, per se, just that as noted, the search is global.

str(end+1)=str(2); str(2)=[]   % rearrange to put lizzie at the end...
str = 
  1×4 string array
    "Anne"    "Marianne"    "Tracy"    "Elizabeth"
TF = find(contains(str,pattern,'IgnoreCase',true))
TF =
     1     2     4

To find individual locations that can keep track of who,

>> TF=arrayfun(@(pattern) find(contains(str,pattern,'IgnoreCase',true)),pattern,'uni',0)
TF =
  2×1 cell array
    {1×2 double}
    {[       4]}
>> TF{:}
ans =
     1     2
ans =
     4

[dpb]

[Answer moved to comment... dpb]

Hi dbp,

Thanks for your answer. Actually, I made a mistake in asking my question (edited the original post accordingly). So what I actually have is not a struct but a cell array. Let me give you an example to illustrate of what I get and what I would like to get:

data.labels = {'l1';'l2';'l3';'l4';'l5'};
lblsA = {'l4';'l2'};
lblsB = {'l2';'l4'};

>>find(contains(data.labels,lblsA))
ans =
     2
     4

>> find(contains(data.labels,lblsB))
ans =
     2
     4

So even though the order of elements in lblsA and lblsB is different, the output of find is the same, which let me believe find is ordering the indices in ascending order. So what I would like to have is

ans = 
     2
     4

ans = 
     4
     2

Any solution is highly appreciated.

[dpb]

find is operating on data.labels; the positions of the two elements is where they are in the array; find tells you where in that array; it has nothing to do with the order in some other array.

contains is simply the wrong tool for the recast question...

 [liaA,libA]=ismember(labels,lblsA);
 [liaB,libB]=ismember(labels,lblsB)

[liaA libA liaB libB]
ans =
 0  0  0  0
 1  2  1  1
 0  0  0  0
 1  1  1  2
 0  0  0  0

find will still return [2 4] for both because that's where the entries are but the second return argument will tell where in the second array they're located in that order.

You have to recognize you're asking for the location with respect to the second array, not the first to have the other piece of information.

All in all, I'd think the first solution will probably work out more easily used but don't know the actual end application.

[Mario Goldenbaum]

Thanks a lot for clarifying this to me.