Search all strings occuring in structure

I have a structure called A (as attached) with the fields Mother, SampleName and Value. Every eight variables SampleNames belong to the same Mother, which is coded in the name (e.g. Mother: 'cutTDM100_111_111_111_111_000' and SampleName: 'cutTDM050_111_111_111_111_122'). I would like to sum up the eight values that each belong to the same Mother. I tried a loop using strcmp, but there is something wrong and it keeps on adding even after the eight value.

sum_value = zeros (3,3)

for k=1:length(A)
 while strcmp(A(k+1).Mother,A(k).Mother)
    value_add= (A(k).value)
    sum_value= sum_value + value_add
 end 
end

This is probably not the most elegant solution anyways and some index for the output would have to be added. Another thing I could think of, is extracting all the occuring Mother Names and write a loop based on this. However, I could not figure out how to search all the strings occuring withing the structure. Getfield only returns one name.

   getfield(A(n), 'Mother')

Answer 1

[Jos (10584)]

So, I assume:

• in your array A, A(1:8) belong to the same number, A(9:16) to another mother, etc..
• There are N mothers and, hence, N*8 elements in A.
• A(k).Value is a scalar

AllValues = [A.Value]
AllValues = reshape(AllValues,8,[]) % a 8-by-N array
SumValue = sum(AllValues,1) % sum each row

Note that this is not very flexible ...

[Julia]

I can't check it right now, but I think this does not work, since A(k).value is a 3x3 matrix.

[Jos (10584)]

Aha, we can concatenate and sum in higher dimensions! This might then work:

AllValues = cat(3,A.Value) % a 3x3x(N*8) array
AllValues = reshape(AllValues,3,3,[],8) % a 3-3-N-by-8 array
SumValue = sum(AllValues,4) % a 3x3xN matrix

It could be that I switch the N and the 8, so you have to check!

[Julia]

I switched N and 8 and now it works. Combined with the input of Guillaume this is sorted now, looking the following. Thanks guys.

AllValues = cat(3,A.value) % a 3x3x(N*8) array
AllValues = reshape(AllValues,3,3,8,[]) % a 3-3-N-by-8 array
SumValue = sum(AllValues,3) % a 3x3xN matrix

[mother, ~, subs] = unique({A.Mother}); %identify identical mothers

for n=1:(length(mother))
    family(n).mother = mother(1,n)
end

for n=1:(length(mother))
    family(n).sum_50s_Daughters = SumValue(:,:,1,n)
end

Answer 2

[Guillaume]

If I understood correctly this should do what you want:

mothers = {A.Mother}; %concatenate all Mother fields into a cell array of string
[mother, ~, subs] = unique(mothers); %identify identical mothers
sum_value = accumarray(subs, [A.value]); %sum (default function of accumarray) all values belonging to identical mothers

Note that sum_value is in the same order as mother which is alphabetical. If you prefer them to be in the same order as the values occur in the structure, add the 'stable' option to unique.

Also note that sum_value is the sum of all the values for an identical mother, if you want to restrict it to at most 8 values (and discard the rest?) then:

sum_value = accumarray(subs, [A.value], [], @(v) sum(v(1:min(8, end))));

edited for incorrect order in the arguments of accumarray

[Julia]

This is what I was looking for. However, accumarray does not accept the 3x3 matrices as second argument. The size of ([A.value is]) 3072, while the subs list is only 1024 long.

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

[Guillaume]

For some reason, I've swapped first and second argument of accumarray. It should have read

accumarray(subs, [A.value]);

Note that it assumes that each value field is scalar.

[Guillaume]

Since you've confirmed that the value field is not scalar, it's probably simpler to forget about accumarray. You'd still start with unique:

[mothers, ~, subs] = unique({A.Mother});
sum_value = cell(size(mothers));
for mother_idx = 1:numel(mothers)
   values = {A(subs == mother_idx)};
   sum_value{mother_idx} = sum(cat(3, values{:}), 3);
end

If you only want to sum a maximum of 8 values per unique mothers, then replace the sum line with:

sum_value{mother_idx} = sum(cat(3, values{1:min(end, 8}), 3);

[Julia]

Neither accumarray nor sum seem to work for the 3x3 matrix

Undefined function 'sum' for input arguments of type 'struct'.

Answer 3

[Walter Roberson]

Inside your while loop, you are not changing any of the values you are strcmp()'ing on. Perhaps you just want an "if" instead of a "while".

[Julia]

The idea was that the fieldvalue of Mother changes between the eight and ninth iteration. Changing to if does not help.

[Guillaume]

Inside the while loop, k never changes, so if you enter the while loop (because the statement is true) then you'll never exit (since it's always the same true statement that is evaluated)

[Julia]

I have rewritten the loop based on the list of unique mother names. Now it is jumping the 9th value, since the condition is false, and keeps on going on with the 10th value. What do I need to change so I add up values 1-8, 9-16 and so on? I am doing it the complicated way since I don't know how to use accumarray for the 3x3 matrix, see above. Thanks

%build structure
for n=1:(length(motherlist))
    family(n).mother = motherlist(n,1)
end

l=1
sum_value=zeros(3,3)
for k=1:length(A)
    if strcmp(A(k).Mother,family(l).mother)
        value_add= (A(k).value)
        sum_value = sum_value + value_add
    else
        family(l).sum_50s_Daughters = sum_value
        l=l+1
        value_add=zeros(3,3) 
        sum_value=zeros(3,3)
    end 
end

[Julia]

I would still like to know, how to deal with a case like this. Does the if condition have to be outside the for-loop? How do I prevent jumping values?