Parsing an array A based on trigger data in array B ?

Hi everyone!

I have what should be a simply question but I cannot seem to find an effective solution. I have two columns of data in Excel files that I read into Matlab, at the moment as two separate arrays (column A and B).

Column A contains numeric data and Column B is a trigger stamp for the data in A. Example:

What I need to do is to parse A and split it into a number of (30) arrays depending on the trigger point in B i.e. the very large A would become A1, A2, A3 ... A30, split according to when there is a data-stamp in B.

[Jan]

Please post how you "read the Excel file". Then creating a matching answer is easier.

[Katie]

Hi Jan,

I was just crudely reading from the excel file and writing the columns to vectors, but I have switched to reading a cdv instead as Chad suggested, much neater. My trouble is how to split the matrix based on the second column (or B in my earlier version) that is causing the difficulty. Thanks.

[OLD version, reading from Excel] A=xlsread('file.xls','A:B'); rows=sizem(1); % Assigning the columns to vectors for i=1:1:rows if(~isnan(A(i,1))) signal(i)=A(i,1); end end etc.

Answer 1

[Chad Gilbert]

If I save similar data out as a .csv file:

   A,B
   1.303,1
   1.61,
   4,2
   1.213,

and then read them in with:

   >> a = csvread('a.csv',1,0)

I get a matrix the same shape as your data, with 0's filling all the empty spots:

   a =

    1.3030    1.0000
    1.6100         0
    4.0000    2.0000
    1.2130         0

This will be pretty easy to parse into a cell array of arrays.

[Katie]

Thanks Chad. That might be more efficient than reading the excel file alright, but either way I could make the A and B arrays into a matrix. It's the next step of parsing or splicing A (or the A element of the matrix) based on B that doesn't seem to be working out.

E.g. based on your data - A1 [1.3030 1.6100] A2 [4.0000 1.2130] etc.

Thanks again

[Chad Gilbert]

Ah, I see. I thought your trouble was with reaading. To split it up, Rob's answer ought to work. Or a shorter version might be:

   ind = [find(a(:,2)>0); size(a,1)+1];
   b = arrayfun(@(i)a(ind(i):ind(i+1)-1),1:length(ind)-1,'uniformoutput',false)

[Katie]

Hi Chad, and Rob!

I tried Chad's first as I had already switched to using a matrix 'a'. This works brilliantly, thank you both. The only thing that is wrong at the moment with it is that it populates b with 29 vectors, not 30. It misses the first one every time. I might just be missing something simple.

I can see in the workspace that 'b' is a 1 x 29 matrix now. If I want to write this out to a new file I'm guessing I have to do more with it?

Thank you both again, much appreciated!

Answer 2

[Rob]

I'm sure there are more elegant solutions, but here's one way:

idx1 = find(~isnan(B));
idx2 = [idx1(2:end)-1; length(A)];
C = cell(length(idx1), 1);
for k = 1:length(idx1)
   C{k} = A(idx1(k):idx2(k));
end

If the non trigger values in B are zeros then:

idx1 = find(B>0);

-Rob