Finding rows where the first column values are equal

Array A:

13453	10359	9955
5257	5299	5258
5849	3644	5848
5397	7230	5396
17132	17130	17118
4767	4768	4770
8291	8292	8316
3191	3190	15970
13389	14409	14410
14840	13022	13021

Array B:

13453	12292	9955	10359
6805	3039	3012	3011
2005	8087	2394	17740
5257	5258	5299	2012
6309	6290	6310	6289
5849	5811	5848	3644
5397	7230	8425	5396
4775	4760	6529	1118
17132	17130	13938	17118
4767	4791	4770	4768
8291	8292	5145	8316
3191	3190	14432	15970
10539	14895	9170	10540
13389	14409	9927	14410
14840	14387	13021	13022

Expected output: Indices of rows in Array B where the first column value is the same as Array A:

1
4
6
7
9
10
11
12
14
15

When I try to use ismember, I get an error that "arrays have incompatible sizes".

x = find (B(:,1) == A (:,1));

Answer 1

[Arif Hoq]

A=[13453	10359	9955
5257	5299	5258
5849	3644	5848
5397	7230	5396
17132	17130	17118
4767	4768	4770
8291	8292	8316
3191	3190	15970
13389	14409	14410
14840	13022	13021];
B=[13453	12292	9955	10359
6805	3039	3012	3011
2005	8087	2394	17740
5257	5258	5299	2012
6309	6290	6310	6289
5849	5811	5848	3644
5397	7230	8425	5396
4775	4760	6529	1118
17132	17130	13938	17118
4767	4791	4770	4768
8291	8292	5145	8316
3191	3190	14432	15970
10539	14895	9170	10540
13389	14409	9927	14410
14840	14387	13021	13022];

Output=find(ismember(B(:,1),A(:,1)))

Output = 10×1

1 4 6 7 9 10 11 12 14 15

[Pelajar UM]

Ah Great! Thank you so much.

Now what if I want to find rows of B that contain the same 3 numbers as in A, but in no particular order (could be column 1,2,3 or 1,2,4 or 2,3,4, or so on).

[Arif Hoq]

would you please explain more. i dont get you

[Arif Hoq]

i guess you are looking for this

[output2 B]=ismember(B,A)

[Pelajar UM]

For example, in array A, we have 13453,10359,9955

The current algorithm now only checks the first value, 13453.

I want it to check so that all the 3 values appear in array B but in no particular order. For example, it could be

10359,9955,X,13453

or

9955,X,13453,10359

As long as there is a combination of 4 numbers, 3 of which are these (in no particular order), it should return the index of that row.

[Arif Hoq]

try this:

[idx C]=ismember(B,A)
Output2=B.*idx

[Pelajar UM]

Thanks but this returns a Nx4 output.

The output should be Nx1 (because it is the index of rows in B that contain a combination of 3 numbers from each row of A).

I guess I can write a loop for this, but that's what I am trying to avoid (it woiuld be slow).

The loop goes like this:

Check cell 1:1 in A against rows in B. Once you find the first occurance, check all the other cells in that row for equality to cell 1:2. and if you find it, then also for cell 1:3.

If all 3 cells are found in that row, return the index.,go to cell 2:1 of A, repeat.

If not, proceed to search in the next row of B until a row is found that has cell 1:1, 1:2 and 1:3 in it.

I have 60,000+ rows in B, so this is going to be heavy.

[Arif Hoq]

check the previous variable 'C' or try this

output=B(idx)

[Pelajar UM]

Thanks, C is Nx4 as well.

idx already has the same dimensions as B. And"output" would have 4 times more rows. The correct answer should have (B rows - A rows)X1.

[Arif Hoq]

please format your question from this link

https://www.mathworks.com/matlabcentral/answers/help/rtc#rtc_summary

Then, can you show me your expected result in a proper format from the above mentioned link? atleast 2/3 rows and columns. please look at below.

13453           0        9955       10359
   0           0           0           0
   0           0           0           0
5257        5258        5299           0
   0           0           0           0
5849           0        5848        3644

if you want a column vector result. can you show me your expected result ? atleast 5/6 columns.

13453
5257
5849
5397
17132
4767
8291
3191

[Pelajar UM]

Thanks a lot for your time. To begin with, I noticed that when I apply this to the whole dataset, the results are not correct:

A=readmatrix ("A.txt");
B=readmatrix ("B.txt");

Output=find(ismember(B(:,1),A(:,1)))

[Arif Hoq]

that's obivious. if you check carefully you will get all the elements of A(:,1) exists in B(:,1). before we worked with only some definite rows, maybe 10 to 20 rows. but B.txt and A.txt are big data file that contains severals thousands similar data to each other.

x=14990
A=readmatrix('A.txt');
B=readmatrix('B.txt');
A1=A(1:x,1);
B1=B(1:x,1);
% Output=find(ismember(B(:,1),A(:,1)));
Output2=find(ismember(B1,A1));
Output2(1:10)

try this code. until A(1:14890) and B(1:14990) the result is

1
4
6
7
9
10
11
12
13
14

but whenever its 14990, then you can get index 3. because A(14990,1)=2005 that is exists in B(3,1)

1
3
4
6
7
9
10
11
12
13

[Arif Hoq]

you can use ismember function to check the element of A in B

[Lia output4]=ismember(A(:,1),B(:,1));

[Pelajar UM]

Thank you so much. At first glance, output4 seems like the correct answer, until you sort it and see there are many zeros. For example, there is a zero in row 34 of output4. Then you go to row 34 of A and find this combination:

11191	17441	17442

This corresponds to row 74 in B:

17440  11191  17441  17442 

or row 46 of output4 is zero. Combination in A:

12012	10422	14680

This corresponds to row 88 in B:

12013  12012  10422  14680 

[Arif Hoq]

because A(34,1)=11191 does not exist in B.

A=readmatrix('A.txt');
B=readmatrix('B.txt');
% [Lia output4]=ismember(A(:,1),B(:,1));
[Lia1 output5]=ismember(B(:,1),11191);
any(output5)

ans = logical
   0

see for A(33,1)=13088. it exists on B

A=readmatrix('A.txt');
B=readmatrix('B.txt');
% [Lia output4]=ismember(A(:,1),B(:,1));
[Lia1 output5]=ismember(B(:,1),13088);
any(output5)

ans = logical
   1

[Pelajar UM]

Right. But is there a way that I could look for a combination of 3 numbers at once. If I want to look column by column, that would need a loop.

More importantly, the number in column 2 of A may not appear in column 2 of B, it could be in column 3. Meaning that I have to look at 3x4 possibilities.

I think one shorcut would be to only look at column 2 if the index was 0 for column 1. There are 719 zeroes right now, this is more manageable.

[Pelajar UM]

Something like this:

[Lia1, output1]=ismember(A(:,1),B(:,1));

idx1=find(all(Lia1==0,2));
[Lia2, output2]=ismember(A(:,1),B(:,2));

idx2=output2(idx1,:);

But I cannot figure out why there are sitll some zero values in idx2. For example row 13 is zero, this corresponds to 492 in idx1.

Row 492 in A is

3177  3325  3178 

which should correspond to row 836 in B:

7137  3177  3325  3178 

This only happens for 13 rows. The rest looks good.