Use of arrayfun and sprintf to order data

Hi, I have a 1x4cell (files) with 1x1struct in each cell. Each structure contains several variables. I would like to extract from each structure only the variables called: SL_X_Y, where X can be numbers from 1 to 7 and Y could be numbers from 1 to 3 i.e. SL_5_2. Not all the combinations are found in each structure. The resulting 1x4cell will only contain structures with the variables SL_X_Y that are found.

Thanks

[Andrei Bobrov]

Attach your data.  

 

[Isma_gp]

An example file attached

Answer 1

[Guillaume]

1. Create a function to filter the relevant fields of the structure:

function filtered = filterstruct(unfiltered)
   %FILTERSTRUCT: keep only the fields of the structure which match SL_X_Y with X integer between 1 and 7 and Y integer between 1 and 3
   fn = fieldnames(unfiltered);   %list of all fields
   toremove = cellfun(@isempty, regexp(fn, '^SL_[1-7]_[1-3]$', 'once'));  %which fields do not match?
   filtered = rmfield(unfiltered, fn(toremove)); %get rid of non matching fields
end

2. Apply cellfun:

cellfun(@filterstruct, yourcellarray, 'UniformOutput', false)

Note however, that it appears you use the variable names to store metadata about the variable content, which is not a good idea and the reason you're struggling. Better would have been to have just one SL field which is a 7x3 cell array / matrix. Extraction would have been trivial then.

edit: typo in the regex

[Isma_gp]

The data is coming in this format. It would be good to have a SL 7x3 cell array. However, I would need to organize the data in this array according to the variable names i.e. SL_2_3 data would go into row 2 column 3 cell in my 7x3 array. Can you help with that?

Thanks

[Guillaume]

This function would convert a struct with SL_X_Y fields into a 2D cell array:

function out = convert2cell(in)
   %Convert input structure with fields SL_X_Y into 2D cell array
   %X is an integer between 1 and 7 and indicates the cell row
   %Y is an integer between 1 and 3 and indicates the cell column
   validateattributes(in, {'struct'}, {'scalar', 'nonempty'}, 1);  %only support scalar structures. Could be extended to structure arrays
   fn = fieldnames(in);
   rc = cellfun(@str2double, regexp(fn, '^SL_([1-7])_([1-3])$', 'tokens', 'once'), 'UniformOutput', false);
   assert(~any(cellfun(@isempty, rc)), 'structure contains non SL_X_Y fields'); 
   rc = vertcat(rc{:});
   in = struct2cell(in); %convert struct into cell array
   out = cell(7, 3);
   out(sub2ind(size(out), rc(:, 1), rc(:, 2))) = in;
end

You can then pass this to cellfun:

newc = cellfun(@convert2cell, yourfilteredcellarray, 'UniformOutput', false)

And convert the cell array of cell arrays into a 3D array:

newc = cat(3, newc{:})

[Isma_gp]

I have some troubles to make it work. I'm attaching an example of the input file. I would like a 7x3 cell array containing the SLAM_SH10_X_Y variables in the right cell (X=row Y=column), for each of the four structures.

Thanks

[Guillaume]

Of course, it doesn't work. In each function, the regular expression is looking for the pattern 'SL_X_Y' (as you originally stated), not 'SLAM_SH10_X_Y'. In addition, the X and Y limits don't match what you said (X goes up to 5, Y goes up to 7).

You need to change the regular expression accordingly.

This will do the filtering and transformation into a cell array. I've removed any restriction on the magnitude of X and Y. The cell array will be as big as necessary.

function filtered = filtertocell(unfiltered)
   %FILTERTOCELL: filter a structure to only keep fields SLAM_SH10_X_Y and convert fields to 2D cell array
   %X and Y are integers indicating the row, column respectively of the destination cell
   validateattributes(unfiltered, {'struct'}, {'scalar', 'nonempty'});
   fn = fieldnames(unfiltered);
   rc = cellfun(@str2double, regexp(fn, '^SLAM_SH10_(\d+)_(\d+)$', 'tokens', 'once'), 'UniformOutput', false);
   unshapedfiltered = struct2cell(rmfield(unfiltered, fn(cellfun(@isempty, rc))));
   rc = vertcat(rc{:});
   filtered = cell(max(rc));
   filtered(sub2ind(size(filtered), rc(:, 1), rc(:, 2))) = unshapedfiltered;
end
cellfun(@filtertocell, files, 'UniformOutput', false)