Export Struct (.repr = dat + lab) to Textfile while preserving labels

I have a struct file that consists of a rather large dat file (11444x1000) and a file of labels (11444 x 1) that I'd like to export to a text file. Ideally it would read out so that each line consists of the label followed by the 1000 unit vector. I'm comfortable in python but relatively new to matlab so I'm unclear on what the right steps would be. I assume I need something like

fileID = fopen('output.txt','w'); 

fprintf(fileID,'%f %f\n',y);

% but here all of the online instructions are for individual variables rather than structures.

fclose(fileID);

Answer 1

[Walter Roberson]

[nrow, ncol] = size(YourStruct.data);
data_cell = mat2cell(YourStruct.data, ones(1,nrow), ncol);  %break by rows
labels_cell = YourStruct.labels(:);
maxlen = max( cellfun(@length, labels_cell) );
rowfmt = ['%-', sprintf('%d', maxlen), 's ', repmat('%9.5f ', 1, ncol-1), '%.5f\n'];
labeled_data_cell = [labels_cell, data_cell] .';  %transpose it!
fprintf(fid, rowfmt, labeled_data_cell{:});

There are a few tricks here:

• you can compute a format string for fprintf
• fprintf follows down columns first, not across rows first. So we transpose the data so the rows become columns that are processed and printed as rows
• fprintf() cannot handle struct or cell directly
• you cannot mix text and numeric in a single argument
• but if you create a cell array and if you use {:} notation on the cell array, it becomes as if you had given everything as individual arguments, like everything was written out separated by commas. So with this code arrangement you effectively put in a string argument that is the label, then you put in a numeric argument that is a row, then another string argument, and so on.

You could go as far as calculating the field width on the numbers to align the decimal points (remember to allow for the negative sign if your values can be negative)

[Amanda Van Horne]

Thank you very much! Your explanations are also super helpful since I was trying to have fprintf treat struct/cell like variables and I see now why it wasn't really working.

[Amanda Van Horne]

Actually I have a follow up question... It worked with a simpler data set that actually had something called sData in the workspace (that only contained the first 1000 vectors). When I use the output (MYFILENAME.mat) and open that in the workspace, it opens as repr (and inside repr is repr.dat and repr.lab). But when I use repr as the YourStruc.labels or YourStruc.data I get an error that this does not exist:

Reference to non-existent field 'data'.

Error in OutputDatatoText (line 2) [nrow, ncol] = size(repr.data);

Thoughts?