How to avoid eval when eval seems unavoidable

Everybody on this forum agrees that eval is a terrible terible thing to use. But often it seems to me to be unavoidable. The one below is only the most recent of these that I have encountered. Suppose I want to "unpack" a struct, using the fields of the struct as variable names.

For example

S.a = 1
S.b = 2
etc

Now I want to create variables a=1 and b=2, etc. The only way I know how to do this goes something like this;

Fields	= fieldnames(S);
for ii=1:numel(Fields);
    field = Fields{ii};
    value = getfield(S,field);
    eval([ field '=' num2str(value) ';']);
end
a 
b

How can I do a loop like this, without using eval?

[Steven Lord]

Do you have to unpack the struct? Is there a reason you can't or don't want to just index into the struct?

[Leo Simon]

definitely wnat (though don't necessarily need) to unpack the struct.

Answer 1

[Matt J]

One way to avoid eval in this case is to use an automatic code-writing tool instead of unpacking at run time. I created the following precisely for this purpose,

https://www.mathworks.com/matlabcentral/fileexchange/26216-structure-fields-to-variables

[Leo Simon]

This is really nice but isn't there some kind of way to actually make it work at runtime? Eg., it looks like this *should* work, (using the maligned eval)

eval(structvars(S))

but of course it doesn't. Would be truly wonderful if it did though.

[Matt J]

I don't see what advantage you anticipate in having the unpacking done at runtime. The expressed purpose of your post was to avoid the disadvantages of eval(). However, you can make it work with eval() by doing,

eval(structvars(S).')

Answer 2

[Walter Roberson]

orderfields, struct2cell, cell expansion in the context of a multiple variable output.

Assuming that the list of fields is fixed but not necessarily their order.

[Leo Simon]

Thanks, Walter, but I need the kind of output that @Matt J 's program generates.

[Steven Lord]

Why? What's your use case where you need the fields as independent variables?

[Walter Roberson]

It is not clear to me why my suggestion is not suitable for your purposes? Do you need to have struct with unpredictable field names that have to be moved into variables?