how to let 'textscan' stop when read an empty line

I have double quoted csv file:

 "header1"
(descriptions)
"A","B","3","A","B","3"
"C","B","3","T","B","2"
"D","B","3","A","H","3"
"E","B","3","A","B","6"
"F","B","3","Z","J","3"
(empty line here)
 "header2"
(descriptions)
"A","B","3","A","B","3"
"C","B","5","A","B","3"
"D","B","4","A","H","3"
"E","B","3","A","L","1"
"F","B","2","A","I","3"

I tried to use:

 fid=fopen('..\data.csv', 'r')
 fgetl(fid)
 fgetl(fid)

 inputtext = textscan(fid, '%q%q%q%q%q%q', ...
                         'delimiter', ',');

but it reads everything until hits the end of the data file (including header2). How to read data above the empty line? Thanks!!

Answer 1

[Ken]

instead of reading the whole block of data using textscan, I tried to read it into a cell line by line. Textscan will stop when hits the empty line.

Code:

           while (~feof(fid_t))
                temp1 = fgetl(fid_t);
                if isempty(temp1) 
                    break; 
                end
                   inputtext=textscan(temp1, formatstring,...
                'delimiter',',')
                   temp_out=[inputtext{:}];
                  Data(end+1,:)=temp_out;
           end

and it works finally. It is still not an effective way, but it works. any suggestion?

Answer 2

[Walter Roberson]

textscan() is not designed for this.

You can remove newline from the Whitespace list, but to compensate you have to explicitly match newline in your textscan() format at the point you expect newline to be.

[Ken]

Walter. Actually I got the idea from

http://www.mathworks.com/products/matlab/examples.html?file=/products/demos/shipping/matlab/textscanDemo.html

The code there (textscan) could read a block of data into a matrix. Textscan stopped reading data-line when there was an empty line. I really don't know the trick, or maybe they used other commands? I am trying to do the same thing but it failed :(. Textscan reads everything.

I will try your advice for removing newlines from 'empty line', but sounds not an effective way ...

[Walter Roberson]

That particular demo is reading blocks of floating point numbers rather than strings. Different stopping conditions internally, unfortunately.