loop through a multi layered structure

I have a dataset that consist of a structure in a structure, etc...,

example: data (1x1 struct) with 5 fields (body parts), consists of: - head(1x1 struct), torso(1x1 struct),arm(1x1 struct), leg(1x1 struct), feet(1x1 struct). each struct consists again of 4 fields(types of bones) and these fields consist of 12 fields (measurements),these 12 fields are the same for all bones and body parts.

I want to be able to do a calculation on a specific measurement field of each bone of each body part. How do I do this? So I want to make a loop that goes for example to head.1bone.angle and does the calculation (mean) there and then go to head.2bone.angle, etc.. then to torso.1bones.angle etc... Is this possible to do this?

Answer 1

[Guillaume]

Use fieldnames and dynamic fields. e.g:

for field = fieldnames(data)'
  for subfield = fieldnames(data.(field{1}))'
     result = mean(data.(field{1}).(subfield{1}).angle)
  end
end

[N]

Thank you ! it worked great! extra question, once I do my calculations of each part I would like to save it, how should I do this best?

my code so far: So i did the for loops as you said and then my calculation (this is the norm(euler) part and the syncsign) so then I end up with euler 3 as the result. Now I would like to save this for each body part and bone in a seperate structure/file how should I do this?

for field = fieldnames(data);
  for i = 1:5;
      subfield = fieldnames(data.(field{1}))';
      for j = 2:4;
          subsubfield = fieldnames(data.(field{i}).(subfield{j}))';
          euler = data.(field{i}).(subfield{j}).euler;
          for k = 1:size(euler,1);
               euler2(k,:) = norm(euler(k,:));
          end
        syncsign = 1;
        euler3 = syncsign*euler2;
        data.euler = euler3
      end
  end
end

[Guillaume]

Firstly, I think you misunderstood part of my code, probably because you're not familiar with for loops over cell arrays:

for field = fieldnames(data)' %note the ' (transpose) is important!
  currentfield = field{1};

is equivalent to:

fields = fieldnames(data);
for index = 1:numel(fields)
   currentfield = field{index}

In your example,the outer loop does nothing (field == fieldnames)

Anyway, to answer your question, this should work:

for part = fieldnames(data)'
   for bone = fieldnames(data.(part{1}))'
      euler = data.(part{1}).(bone{1});
      for row = 1:size(euler, 1)
         eulernorm(row) = norm(euler(row, :));
      end
      out.(part{1}).(bone{1}) = eulernorm;
   end
end

Or, if you define a function called rownorm as such:

function [rn] = rownorm(matrix)
  rn = zeros(size(matrix, 1), 1);
  for row = 1:size(matrix, 1)
     rn(row) = norm(matrix(row, :));
  end
end

You can use structun:

out = structfun(@(part) structfun(@(bone) rownorm(bone.euler), part, 'UniformOutput', false), data, 'UniformOutput', false);

[N]

thank you immensely it worked great !!! And you're right I'm not used to working with loops over cell arrays.

[Tiasa Ghosh]

If suppose I wanted to combine each element at each sub-level of the nested structure into one string, how do I run the loop to output such a string? Any hint or help is welcome. Thanks!

[Stephen23]

"If suppose I wanted to combine each element at each sub-level of the nested structure into one string, how do I run the loop to output such a string?"

Simply split the string at the delimiter and then call getfield / setfield directly.

Answer 2

[Azzi Abdelmalek]

Use structfun function

Answer 3

[Guillaume]

Or using structfun, if you want the result as a structure:

structfun(@(part) structfun(@(bone) mean(bone.angle), part, 'UniformOuput', false), data, 'UniformOutput', false)

You can do away with the first UniformOutput but not the second.