How to convert text file with different delimiters into a matrix

|Hi,

My text file contains 60K lines with format:| | |

0 qid:10 1:0.000000 2:0.000201 3:0.10090 4:0.0043000 5:0.1900000.. 46:0.650000 #docid = GX000-00-000000

Now I have to load data from this file onto a matrix with column '1' and column '3' to column '48' values ignoring value before colon.

O/P: [0 0.000000 0.000201 0.10090 0.0043000 0.1900000...0.650000] //this is single row in matrix

I tried this:

fid = fopen('Querylevelnorm.txt');

--for loop for each line begins here
  tline = fgetl(fid)
  C=strsplit(tline)
  vector = [];
  --for loop for each value in line starting from C(1) then C(3) to C(48)
     D=strsplit(char(C(3)),':')
     vector=[vector D(2)]  ##formulating of vector
  --end for
--adding present vector to matrix  ## formulating of matrix
--end for

i.e., I am reading each line and first splitting with space and then with colon to get each value. But I am not getting how to populate all these values into a vector and then adding each vector to a matrix. I am new to this matlab programming. Sorry for any wrong editing. Please help me out.

Answer 1

[dpb]

Use textscan and a specific format string...

>> l='0 qid:10 1:0.000000 2:0.000201 3:0.10090 4:0.0043000 5:0.1900000 46:0.650000 #docid = GX000-00-000000';
>> fmt=['%f %*s' repmat('%*d:%f',1,6) '%*[^\n]'];
>> cell2mat(textscan(l,fmt,'collectoutput',1))
ans =
       0         0    0.0002    0.1009    0.0043    0.1900    0.6500
>>

The format string translates as--

a) read first decimal value (use float to collect all into one array in end)
b) skip first string field    (the 'quid:value' will be a single string)
c) read the next sequence of [skipdecimal:readfloat] N times
d) skip rest of line...

See

doc textscan for the details.

NB: use 46 for the repeat count in repmat for your case; 6 matched the length of the line cleaned up to remove the ellipses you posted.

[SRINIVASAREDDY]

Awhh,perfect answer..Thank you very much..It is even fast in processing all the lines.