How can I use textscan without delimiter?

Let's say some numeric data is not delimited, buth we can delimit it based on their location. For example, I would like to read the following data using textscan:

 950 99810251026 999 496
 951 999102610271000 785
 9521000102710281001 497
 9531001102810291002 498
10001061108410851062 438
10011062108510861063 439

my code is:

cell2mat(textscan(fid, '%4f%4f%4f%4f%4f%4f'));

and the result is:

 950 9981  251   26  999 496
 951 9991  261  271    0 785
9521    1  271  281    1 497
9531   11  281  291    2 498
1000 1061 1084 1085 1062 438
1001 1062 1085 1086 1063 439

but what I wanted is:

 950  998 1025 1026  999  496
 951  999 1026 1027 1000  785
 952 1000 1027 1028 1001  497
 953 1001 1028 1029 1002  498
1000 1061 1084 1085 1062  438
1001 1062 1085 1086 1063  439

What's wrong with my code?

Answer 1

[per isakson]

Must be done in two steps

• read as character to avoid the white-space-magic
• convert to numerical

fid = fopen( 'cssm.txt' );
cac = textscan( fid, '%4c%4c%4c%4c%4c%4c', 'Whitespace', '' );
fclose( fid );
buf = cellfun( @(str) str2num( str ), cac, 'uni', false );
cell2mat( buf )

outputs

ans =
         950         998        1025        1026         999         496
         951         999        1026        1027        1000         785
         952        1000        1027        1028        1001         497
         953        1001        1028        1029        1002         498
        1000        1061        1084        1085        1062         438
        1001        1062        1085        1086        1063         439

The last line of the text file needs to be terminated by a new-line-character(?)

[Jung-Woo]

Thank you so much. It works.

Answer 2

[Walter Roberson]

str2double( textscan(fid, repmat('%4c',1,6)) )

The mistake in your code is that the counts for %4f do not start until the first non-blank character of the field. ' 1234567' with '%4f' format skips the blank and counts the 4 from '1' grabbing '1234'.

It is tempting to try to use 'Whitespace', '' so that it doesn't skip blanks. Unfortunately, when reading from a file, the numeric formats fail if they are required to start parsing from a blank instead of starting from a number-forming character (digit, + or -, decimal).

[Jung-Woo]

From your code, I got the following answer:

 NaN 9.9001e+23 1.1222e+19 NaN NaN 4.7444e+17

Please check it again.

[Walter Roberson]

Dang I didn't realize that %c returned char array instead of cell arrays.

cols = cellfun(@(C) str2double(cellstr(C)),'Uniform',0);
horzcat(cols{:})

[Jung-Woo]

Thank you. I will try it later.

[Jung-Woo]

Yes, I confirmed it works. Thanks a lot.