The chosen data structure demands for loops. There is no "efficient" solution, which extracts the fields in a vectorized way. You can at least avoid some pitfalls:
fileList = {};
nVeh = numel(veh);
for k = 1:nVeh
aVeh = veh(k);
nMeas = numel(aVeh.meas);
for kk = 1:nMeas
fileList{end+1} = aVeh.meas(kk).data.fileName;
end
end
Here the output grows iteratively. If you know a maximum number of elements, you can and should pre-allocate the output:
nMax = 5000;
fileList = cell(1, nMax);
iList = 0;
nVeh = numel(veh);
for k = 1:nVeh
aVeh = veh(k);
nMeas = numel(aVeh.meas);
for kk = 1:nMeas
iList = iList + 1;
fileList{iList} = aVeh.meas(kk).data.fileName;
end
end
fileList = fileList(1:iList); % Crop ununsed elements
Choosing nMax too large is very cheap, while a too small value is extremely expensive. Even 1e6 does not require a remarkable amount of time.
numel is safer than length. The temporary variable aVeh safes the time for locating veh(k) in the inner loop repeatedly.
If you do not have any information about the maximum number of output elements, collect the data in pieces at first:
nVeh = numel(veh);
List1 = cell(1, nMax);
for k = 1:nVeh
aVeh = veh(k);
nMeas = numel(aVeh.meas);
List2 = cell(1, nMeas);
for kk = 1:nMeas
List2{kk} = aVeh.meas(kk).data.fileName;
end
List1{k} = List2;
end
fileList = cat(2, List1{:});
