Read with subsampling in MATLAB - matlab

Question

Is there anyway to read with subsampling in MATLAB? The input data look like:
id=3,age=25, 0.5 0.5 0.2 0.6 0.6 0.5
id=1,age=15, 0.5 0.8 0.2 0.9 0.6 0.9
id=7,age=24, 0.5 0.2 0.9 0.6 0.1 0.5
(Edited) For the LAST SIX columns, I only want columns that are multiple of three (i.e. the 3rd column and the 6th column in the LAST SIX COLUMNS, which is equivalent to the 5th column and the 8th column in the whole data file) be read. That is, a matrix like:
0.2 0.5
0.2 0.9
0.9 0.5
Ideally, the code looks like:
for line=1:maxLine
header(line,:) = fscanf(fid,'id=%d,age=%d,',[1,2]);
content(line,:) = fscanf(fid,'only read columns multiple of three');
end;
I know that I can read the whole line and sub-sample, the problem is, the array I'm dealing with has large scale, 10k+ columns, I do not want to consume too much memory.

Answer 1

There is the way:
if you use fopen you can get access to file only one time, so you can't do it in a loop. So you need to load all necessary data in one array and then transpose it.
Your desire to save free memory is achieved by ignoring some elements:
fid = fopen('new.txt','r');
A = fscanf(fid, 'id=%d,age=%d, %*f %*f %f %*f %*f %f\n', [4 inf])
I used your data and get this result:
A =
3.0000 1.0000 7.0000
25.0000 15.0000 24.0000
0.2000 0.2000 0.9000
0.5000 0.9000 0.5000
as you can read here http://www.mathworks.com/help/matlab/ref/fscanf.html?searchHighlight=fscanf fscanf reads data into columns, that is why we need to transpose it.
So using A=A' gives result you want:
A =
3.0000 25.0000 0.2000 0.5000
1.0000 15.0000 0.2000 0.9000
7.0000 24.0000 0.9000 0.5000
Now you can make two different matrices if needed.