Loop through structure elements with parfor

I have a structure with programatically generated fieldnames (let's call it myStruct), and an optimization program (let's call it optfunc) needs to run on data contained in each fieldname. I would like to use parfor to accelerate the process, but the normal way to loop through structs creates unclassified variables. Any tips on how this can be fixed? E.g.:

myFields = fieldnames(myStruct);
nFields = size(myFields,1);

parfor iField = 1:nFields
    dataSet = myStruct.(myFields{iField});
    optOut = optfunc( dataSet );
end

[Walter Roberson]

Perhaps loop over contents of struct2cell ?

Answer 1

[Edric Ellis]

Given the following example data

myStruct = struct('one', rand(1), ...
    'two', rand(2), ...
    'three', rand(3));

My slight adaption of your program works correctly:

myFields = fieldnames(myStruct);
nFields = size(myFields,1);
out = NaN(1, nFields);
parfor iField = 1:nFields
    dataSet = myStruct.(myFields{iField});
    out(iField) = max(dataSet(:));
end

But note that myStruct is not sliced in this case - as @Walter suggests, you could use struct2cell to achieve that.

myContents = struct2cell(myStruct);
out2 = NaN(1, numel(myContents));
parfor iField = 1:numel(myContents)
    dataSet = myContents{iField};
    out2(iField) = max(dataSet(:));
end

[Matthew Thompson]

Yes, the struct2cell command was just what I was overlooking. Thank you both.