Removing unique numbers whilst comparing two structures

I have two structs, one called AllPositions_CA with X and Y fields containing this type of layout:

X Y

1x17051 double 1x17051 double

1x17267 double 1x17267 double

1x17579 double 1x17579 double

1x17971 double 1x17971 double

1x17959 double 1x17959 double

1x17947 double 1x17947 double

1x17854 double 1x17854 double

1x641 double 1x641 double

1x17918 double 1x17918 double

1x17544 double 1x17544 double

This structure has a length 114 (I have only shown 10)

I have a second structure called PositionA which has X and Y fields containing this type of layout:

X Y

1x42 double 1x42 double

1x44 double 1x44 double

1x43 double 1x43 double

1x43 double 1x43 double

1x44 double 1x44 double

1x43 double 1x43 double

1x43 double 1x43 double

1x43 double 1x43 double

1x44 double 1x44 double

1x42 double 1x42 double

This particular structure has length of 18000 (I have only shown 10)

I want to be able to take ALL X and Y values as pairs in AllPositions_CA and compare them to X and Y pair values in PositionA and remove any values which are present in PositionA but are not present in AllPositions_CA and put them into a new structure whilst retaining the layout of the structure PositionA. For example after the removal of some values from PositionA the new structure will look like:

X Y

1x40 double 1x40 double

1x39 double 1x39 double

1x41 double 1x41 double

1x40 double 1x40 double

1x42 double 1x42 double

1x41 double 1x41 double

1x39 double 1x39 double

1x40 double 1x40 double

1x43 double 1x43 double

1x38 double 1x38 double

I have looked into intersect and setdiff but cannot get my head around how to loop one structure through another to find differences and return the desired final layout as above.

Thanks

[Jan]

I've read this sentence mutliple times:

I want to be able to take ALL values in AllPositions_CA and compare them to each value in PositionA and remove any values which are present in PositionA but are not present in AllPositions_CA and put them into a new structure whilst retaining the layout of the structure PositionA

It would be easier to use the names A and B instead of AllPositions_CA and PositionA. But even then it does not get clear to me, what you want to achieve.

Do you want to remove all elements of the fields PositionA.X and positionA.Y which occur anywhere in AllPositions_CA.X and AllPositions_CA.Y? Or do the X/Y pairs play a rule, such that PositionA.X(k)(m) and PositionA.Y(k)(m) should be removed only, if they occur in AllPositions_CA.X(k)(n) and PositionA.Y(k)(n)?

A small example would be useful. Use A and B with 2 or 3 elements to clarify what you want.

[Image Analyst]

Please attach AllPositions_CA in a .mat file with the paper clip icon.

Also, see Stephen's answer below.

[Manny Kins]

Hi Jan, Image Analyst, I apologise for not wording my question clearly. I wanted to remove all unmatched values whilst keeping the X and Y values paired, so the comparison was for both the X and Y values. Stephens example is pretty much spot on with what I wanted:

So if I had structures:

C = struct('X',{[0,1,3],[3,4]}, 'Y',{[5,6,8],[8,9]});

A = struct('X',{[0,2,3],[3,10]},'Y',{[5,7,8],[8,11]});

and I wanted to remove all (X/Y pair matched) values then the final struct would look like:

X Y

[0,3] [5,8]

3 8

keeping only the matched pair values and removing the rest.

Answer 1

[Stephen23]

A loop and ismember makes this clear and easy. An alternative would be to merge the X and Y data, and then call setdiff in the loop with its 'rows' option.

I assume that you wanted to match both the X and Y values simultaneously, i.e. you want to match coordinate pairs, not individual values. (you did not specify this explicitly in your question).

Here is a simple working example with a loop:

C = struct('X',{[0,1,3],[3,4]}, 'Y',{[5,6,8],[8,9]});
A = struct('X',{[0,2,3],[3,10]},'Y',{[5,7,8],[8,11]});
Z = struct('X',{},'Y',{}); % new structure.
CX = [C.X]; % All of C.X.
CY = [C.Y]; % All of C.Y.
for k = 1:numel(C)
	idx = ismember(A(k).X,CX) & ismember(A(k).Y,CY);
	Z(k).X = A(k).X(~idx); % copy non matching to Z.
	Z(k).Y = A(k).Y(~idx); % copy non matching to Z.
	A(k).X(~idx) = []; % optional: remove from A.
	A(k).Y(~idx) = []; % optional: remove from A.
end

Giving (for example):

>> A.X
ans =
     0     3
ans =
     3
>> Z.X
ans =
     2
ans =
    10

[Guillaume]

My understanding is that numel(A) ~= numel(C) (one is 114, the other 1800) and that therefore A(k).X needs to be compared against [C.X] (which begs the question why is C not just one array?)

[Stephen23]

@Guillaume: yes, I noticed that too, and made a few small changes to allow for that.

[Manny Kins]

Hi Stephen, Guillaume, thank you for your help. I know the question I asked was worded quite confusingly but somehow you managed to decipher it and give exactly what I wanted! I did indeed want to match both the X and Y values simultaneously, and the additional option of removing the values from the initial array itself was very useful. Much appreciated.