Downsizing a struct with different data type fields

I have a 66x1 struct and want to transform into a 1x1. The struct contains logical, numerical, datetime and string fields. I used

for fieldCtr = 1:length(myfields)
    if isnumeric(getfield(oldstruct,myfields{fieldCtr}))
        eval(['newStruct.' myfields{fieldCtr} ' = [oldstruct.' myfields{fieldCtr} '];'])
  else
       isstring(getfield(oldstruct,myfields{fieldCtr}))
       evalc(['newStruct.' myfields{fieldCtr} ' = [oldstruct.' myfields{fieldCtr} '];'])
    end
end

It works for every data type. except that the strings are now concatenated and all the fields that cointain strings became one huge string.

How can I avoid that? Or is there any smarter way to downsize the entire struct?

[Stephen23]

Using eval to access fields of a structure has been outdated for fifteen years:

https://blogs.mathworks.com/loren/2005/12/13/use-dynamic-field-references/

You should use dynamic fields:

https://www.mathworks.com/help/matlab/matlab_prog/generate-field-names-from-variables.html

"Or is there any smarter way to downsize the entire struct?"

struct2cell, reshape and/or permute, cell2struct

[Walter Roberson]

The only difference between eval() and evalc() is that evalc() "captures" any text that results from executing the command (such as by a disp() or fprintf()) and makes that text available. There is no difference in how eval() or evalc() treat numeric vs non-numeric fields.

Answer 1

[Walter Roberson]

nfield = length(myfields);
newStruct = cell2stuct(cell(nfield,1), myfields);
for fieldCtr = 1 : nfield
    thisfield = myfields{fieldCtr};
    sample = oldstruct(1).(thisfield);
    if isrow(sample)
        newStruct.(thisfield) = vertcat(oldstruct.(thisfield));
    elseif iscolumn(sample)
        newStruct.(thisfield) = horzcat(oldstruct.(thisfield));
    else
        nd = ndims(sample);
        newStruct.(thisfield) = cat(nd+1, oldstruct.(thisfield));
    end
end

This code tries to be intelligent about how to arrange the results. It turns rows into 2D arrays by vertical concatenation; it turns columns into 2D arrays by horizontal concatenation; it turns other arrays into one higher dimension by concatenating one dimension higher than it already is (so for example, two 32 x 64 arrays would not turn into 64 x 64 or 32 x 128, and would instead turn into 32 x 64 x 2)

You might want to make an exception for character vectors: instead of creating a character array, you might want to create a cell array of character vectors, so as to allow for the possibility that the character vectors are different lengths.

[Matheus Pacifici]

Thank you Roberson,

It works perfectly for every struct I tried.