Could someone please show me how to add time data?

I have eleven files that each have a column with time data. They all start at zero seconds, but are not all exactly the same in length. I would like to put these files into one large file with the time in sequence starting with zero and becoming larger with each file, not starting back at zero after each file. What is an efficient way of doing this? I have tried a few things but nothing will work for files with irregular lengths. Any help would be greatly appreciated.

[Tony Pate]

If you need me to elaborate more or provide more information just let me know. I could really use some help.

[dpb]

Is the interval fixed or variable? If fixed, simply concatenate files and build a time vector of appropriate overall length.

If they're variable, loop thru the files reading the last time of each sequentially and then add that to the subsequent file's values (with whatever is the appropriate delta desired between the end of the one and beginning of the next; if they all are zero-based then there will be two identical time points w/o this shift).

Depending on release, use datenum (Matlab Classic) or the datetime class with current or recent release; I don't recall just when was introduced.

[Tony Pate]

The interval is fixed with a time-step of 0.002 between points. I've been trying concatenation techniques, but I'm not having any luck. Would you mind explaining how to properly do such an operation? Maybe just a short example or something. Oh, and there isn't a time-stamp on this data. I can't convert the time to a standard time and read it through that way. It definitely would be much easier if I could do that because I've already created a function to concatenate a list of files with a given time-stamp. I just don't really know how to do it without one. I put a dummy example of one of the files to look at what I mean.

[Tony Pate]

Thank you for trying to help me with me question. I really appreciate it.

[dpb]

Unfortunately, can't read the above file, only have 32-bit R2012b here.

But, what's the content besides the time vector; other variable(s) or and array of some number of columns?

I'll make a presumption of the latter and write some aircode that shows the idea for concatenating a set of files...

d=dir('*.mat');  % all the .mat files, use appropriate pattern
xall=[];            % a null array to collect into
for i=1:length(d)
  s=load(char(d(i).name));     % load each file in sequence
  xall=[xall;s.data(:,2:end)]; % concatenate except for time 
end
nEl=size(xall,1);   % how long in the end is it?
T=nEl*dt;
t=[0:dt:T].';       % create the fixed dt time vector
xall=[t xall];      % and add as first column
save xall           % done; make new file

Obviously change variables to those or your choosing that match and you'll want to ensure that the time vector thus created is the same length as the array (not one short or one long owing to inevitable floating point rounding) and fixup the result so the meld into the resulting array won't fail.

[per isakson]

"The interval is fixed with a time-step of 0.002 between points" &nbsp Not exactly

>> whos X
  Name         Size             Bytes  Class     Attributes
  X         3000x13            312000  double              

>> format long 
>> X(1:4,1)
ans =
                   0
   0.001953000000000
   0.003906000000000
   0.005859000000000
>>
>> dX = diff(X(:,1));
>> max(dX), min(dX)
ans =
   0.001953000000000
ans =
   0.001952999999999

[Tony Pate]

How can I fix my code knowing that there is a slight difference with the time-step? How can I read in the last time value for each file to put them all together?

[dpb]

That dT is fixed (not 0.002, but constant) within machine roundoff. If that's so for each file, then I'd revert to the previous technique just using the actual difference instead of the approximate 0.002; how can it possibly matter if you're concatenating files that didn't come sequentially in the first place?

Answer 1

[per isakson]

Here is one way of doing it. I use four copies of you example file for this test.

tic
folderspec = 'h:\m\cssm';
glob = 'Example_DROID*.mat';
X = cssm( folderspec, glob );
toc
dX = diff(X(:,1));
disp( [ max(dX), min(dX)] )

outputs

Elapsed time is 0.109893 seconds.
   0.001953000000000   0.001952999999997
>>

where

function    X = cssm( folderspec, glob )

      sad = dir( fullfile( folderspec, glob ) );

      len = length( sad );
      cac = cell( 1, len );
      dt  = 0.001953; % magic number
      dX1 = 0;

      for jj = 1 : len
          S   = load( fullfile( folderspec, sad(jj).name ) );
          S.X(:,1) = S.X(:,1) + dX1; 
          cac{jj} = S.X;
          dX1 = S.X( end, 1 ) + dt;    
      end
      X = cat( 1, cac{:} );
  end

[Tony Pate]

Thank you both for your help!

[Tony Pate]

Could you explain to me what folderspec and glob are? Does glob mean that they all have to have the same name?

[Tony Pate]

Also, to get a consistent dt of the entire group of files, would you just use: dX = diff(X(:,1)); dt = ((max(dX)+min(dX))/2);

if the dt changes a good bit from file to file?

[per isakson]

• folderspec is a variable name, which I use for the full name of a directory path. In Matlab the name, path, is taken.
• glob is a variable name, which is often used for filenames with wildcard characters. See glob (programming)

[per isakson]

"to get a consistent dt of the entire group of files" &nbsp Why do you want that?

"if the dt changes a good bit from file to file?" &nbsp If so, it's for some reason.