open text file constructed in a specified manner

Hi all,

I have a big text file in this construction:

    blablblala
    blablabla
    1 2 3
    4 5 6
    2 5 8
    blablblala
    blablabla
    5 8 9 
    5 2 3
    2 5 6
    blablblala
    blablabla
    2 5 8 
    98 56 42
    5 8 7
    blablblala
    blablabla
    ...
    ...
    ...
    blablblala
    blablabla
    ...
    ...
    ...
    blablblala
    blablabla
    .
    .
    .

the structure is the same , only the numbers do change. ho could open this kind of text file and get the data in a cell:

    C{1}=
    1 2 3
    4 5 6
    2 5 8
    C{2}=
    5 8 9 
    5 2 3
    2 5 6
    C{3}=...

and so on.

Thank you.

[per isakson]

The strings

    blablblala
    blablabla

do they contain substrings, which can be used as indicators of

    START OF NUMERICAL DATA

and

    END OF NUMERICAL DATA

Answer 1

[dpb]

textscan

Answer 2

[per isakson]

I have answered similar questions a few times, e.g.

• How to delete rows of characters in Text files?
• Regexp text-block matching challenges
• Can REGEXP be utilized in searching M x N matrix values?

[dpb]

And I've done it countless times with textscan and friends...

Load parts of VERY LARGE text file content and create a smaller matrix

PS Per: How did you get just the titles to show up as the links instead of the whole shootin' match when paste the link as mine?

[Guillaume]

@dpb, I've edited your link so it just show the title.

Basically, the markup for a link is <url text_to_display>. If you don't put the text_to_display part , you just get the url displayed as you had.

[per isakson]

dpb, I just tested your use of textscan. It's the first time I use textscan this way. A mental hurdle to me has been that I've regarded formatSpec not matching and textscan terminating as kind of an "error". Your construct is based on these two excerpts from the documentation.

    throughout the entire file and stops when it cannot match formatSpec 
    to the data.    

    textscan automatically resumes reading at the point where it terminated 
    the last read

Next I had to try to replace 'Headerlines',5 by

    [cac,pos] = textscan( fid, '%[^Start]');

i.e. advance the pointer to the position left of Start. Either I've run into an "issue" or I've misinterpreted the documentation. I'll try again later.

&nbsp

"just the titles to show up" &nbsp The example below will display MyTitle

    <http://myweb.mypage.com/ MyTitle>

[dpb]

This one is almost trivial, Per...altho one thing I dislike about textscan is that it won't figure out the number/line automagically so if you don't know that you have to do some work to discern that for the format string. But, if it is known a priori, you can do this one by inspection on the fly in one try...

>> type blabla.txt
  blablblala
  blablabla
  1 2 3
  4 5 6
  2 5 8
  blablblala
  blablabla
  5 8 9 
  5 2 3
  2 5 6
  blablblala
  blablabla
  2 5 8 
  98 56 42
  5 8 7

>> fid=fopen('blabla.txt');
>> fgetl(fid);
>> while ~feof(fid)
  cell2mat(textscan(fid,'%f%f%f','headerlines',2', ...
                  'collectoutput',1))
 end
ans =
   1     2     3
   4     5     6
   2     5     8
ans =
   5     8     9
   5     2     3
   2     5     6
ans =
   2     5     8
  98    56    42
   5     8     7
>> fid=fclose(fid);
>>

The only "trick" used here is fgetl to throw away the '%%' line. You would think that 'commentstyle','%' would take care of that, but no; the 'headerlines',2 option takes over first and "eats" those first two lines so then there are no comment lines in the file after that. I think that's a debatable implementation decision, but it's hard to argue it's really a bug.

@Per--I think you're overthinking textscan. For files like the above simply look for the regularity and what it takes to

1. get past any unique header information first, and then
2. write a format expression (or group thereof) that handles the repetitive block as if it were the only one in the file and finally,
3. repeat that block until done.

In the above he '%f' field will repeat until it fails; using three simply causes it to return the result as an array of three columns instead of as a vector; the file pointer will be at the beginning of the next header line where the failure to convert to a numeric value occurred so then the repeat of skipping the two headerlines picks up and "lather, rinse and repeat"...