How can I read block of data using textscan?

Hi guys, I am working with a block of data decribing some features of an element.

for each loop i need to read a block. My problem is that the code could only read the first variable,Lpp but could not read correctly lines 2,3 and 4.

Here is my attempt;

function[inp]= ReadInput
%%To read input file containing offset points
inp = struct;                           %Structure containing variable in the input file
inp.nstrp=10;                           

Input ='strpoffset1.txt';               
fid=fopen(Input);                      
a=textscan(fid,'%f',1,'headerlines',4);                  
inp.Lpp=a{1}(1);                        
inp.xnstrp = zeros(inp.nstrp,1);        
inp.nptstrp = zeros(inp.nstrp,1);        
inp.offsets = cell(inp.nstrp,1);        

for aa=1:inp.nstrp                                  
    b = textscan(fid,'%f',3);                       % Read next three line
    inp.xnstrp(aa,1)= b{1}(2);                      % content of line 2 in the offset file
    inp.nptstrp(aa,1)= b{1}(3);                     % content of line 3
    c = textscan(fid,'%f %f',inp.nptstrp(aa,1));    % Read offset points
    inp.offsets{aa}(:,1)= c{1}(:);                  % offset in y axis
    inp.offsets{aa}(:,2)= c{2}(:);                  % offset in z axis
end
fclose(fid);                                        %Close input file
end

Answer 1

[dpb]

Basics are:

fnames={'Lpp','xnstrp','nptstrp','offsets','c'};                % define struct fieldnames
inp=cell2struct(cell(5,1),fnames,1);                            % build empty input struct
fid=fopen('strpoffset1.txt','r');                               % open file
cleaner=onCleanup(@()fclose(fid));                              % create cleanup fncn to close open file
hdr=num2cell(cell2mat(textscan(fid,'%f',4,'headerlines',4)));   % read first header
c=textscan(fid,'%f %f\n',hdr{4},'CollectOutput',1);             % read data array to cell array
inp=cell2struct([hdr;c],fieldnames(inp),1);                     % save first input struct array
i=1;                                                            % initialize counter
while ~feof(fid)
  i=i+1;                                                        % increment counter
  hdr=num2cell(cell2mat(textscan(fid,'%f',4)));                 % read next header
  c=textscan(fid,'%f %f\n',hdr{4},'CollectOutput',1);           % read data array to cell array
  inp(i)=cell2struct([hdr;c],fieldnames(inp),1);                % save into input struct array
  %ADDENDUM -- ERROR CHECK FOR MISMATCHED ARRAY SIZE
  if size(c{1},1)~=hdr{4}
    error('Mismatched size: \nOffsets=%d, found %d elements',inp(i).offsets,size(c{1},1))
  end  
end
fid=fclose(fid);

Above doesn't preallocate for the struct array; if file isn't too big probably won't be a terrible issue...

Worked here for beginning of the file; the file attached, however, isn't regular -- the third begins like

3		
24	
14.5000	
11	
12   7.9567	
1.462   7.9387	

which doesn't look consistent -- altho will still be able to read. the "12" looks out of place but may be right.

The real problem will arise later because there are blank records and the number of header records turns to three instead of four.

If this is real, you'll have to read the file record-by-record and parse it unless it is known a priori what and where these changes occur so can code for them.

[Idris jae]

many thanks for your time. let me be more elaborate for you to understand what I am trying to achieve.

you may assume a sliced bread, my code attempt to call the properties of each slice.

lets say I have 10 slice (nstrp), distance of each slice from the end (xnstrp), and on each slice, i have some points i want to analyse (nptstrp), each of those points have coordinates (offset) equivalent to the number of points.

what i am trying to achieve is that, for every slice I call the loop should read;

nstrp=1

Lpp=24,

xnstrp=13.5

nptstrp=10,

offset= a b

What you did above is far better than what I was having. How bad is my data table?

[Idris jae]

I am planning of using say, 20 slice.

thank you in anticipation of your input

[dpb]

"How bad is my data table?"

All depends on whether what the attached file shows is real or not.

The missing data for symmetric number of header data is a killer as you don't have an entry for some of those values that you say you want; the synchronization is totally lost after that point.

Unless there's some way to know where those are from some a priori source that can be used to code logic around those problems, all you can (try to) do is to read a record at a time and try to figure out what it is and parse it.

Can't read what isn't there.

[Idris jae]

I think thats where my problem begins. I actually formulate the data to reflect waht i have in mind. my brain is blocked.

In my small knowledge, what I described above is what I expect to achieve.

What format will you advice to adopt to format the table better,pls.?

[dpb]

We can't solve that for you...the file format that you start with in the example works ok as long as the whole file follows the rules -- but what you attached didn't:

...
2	
24	
14.00	
12	
1.461   7.9567	
1.462   7.9387	
1.046   7.7721	
...
0.000   2.3055	
0.000   2.1666	
0.000   2.0718		
3		
24	
14.5000	
11	
12   7.9567	
1.462   7.9387	
1.046   7.7721	
...
0.000   2.3055	
0.000   2.1666	
4	
24	
15.00	
13	
1.461   7.9567	
1.462   7.9387	
...
0.000   2.1666	
0.000   2.0718	
0.000   1.9724	
	
	
	
	
24	
15.5000	
10	
1.461   7.9567	
1.462   7.9387	
1.046   7.7721	
0.755   7.3652	
0.62    7.2689	
0.364   7.2024	
0.241   7.1824	
0.12    7.1299	
0.000   7.0023	
0.000   2.3055	
	
	

The above section from the file that I shortened by eliding part of arrays for brevity is consistent for first two but as noted earlier the 3rd looks suspicious even though number of elements is consistent--the "12" in the first row of the array data looks very much like that's an error.

After that section, then group 4 looks ok again, but from then on it's a train wreck. There are a bunch of blank records and no section number so only three header records before the floating array.

Whatever created the input file looks to be broke -- you need to fix it first.

Answer 2

[Idris jae]

Many thanks for your effort.

Attached is the amended table. surprisingly it wasnnt able to read from any neither your code nor mine

Answer 3

[Idris jae]

but without the serial numbers, 1, 2... and introduce 4 spaces after the 1st block i generate the result. b{1}, instead of returning '-13.5' as the second variable, first 3 variable of the 3rd block were returned as 2nd

Input =

    'strpoffset1.txt'
ans = 
  struct with fields:
      nstrp: 8
        Lpp: 24
     xnstrp: [8×1 double]
    nptstrp: [8×1 double]
    offsets: {8×1 cell}
    
K>> b{1}
ans =
    3.0000
   24.0000
  -14.5000

K>> 

[dpb]

My code worked just fine until got to serial group 3 which has

>> inp(3)
ans = 
  struct with fields:

        Lpp: 3
     xnstrp: 24
    nptstrp: -14.5000
    offsets: 11
          c: [9×2 double]
>> 

So after it the synchronization is screwed up going forward.

Another problem in the data file -- you told it you had 11 pairs to read but are only 9 in the file.

Again, whatever it is that is building the file isn't being consistent in either what it's being given or is generating or however the file is being produced...

You can find this stuff yourself if you just look carefully...

[Idris jae]

Many thanks dpb, i will on that

[dpb]

I added a sanity check for each section that those two are in synch -- if not, it now tells you and aborts.

Exercise for Student: I didn't put the section number that failed in the error message output...