Removing top layer of struct

Hi,

When I load a struct I've previously saved to a .mat file, I've specified a generic name for the struct before, say fileStruct. I'd like to open the struct via load, and have the name of the struct be more specific, like "fileStructA", but when I try fileStructA=load(.....,fileStruct), I get a struct back whose name is fileStructA, with the next level down still named "fileStruct". Can I avoid this happening, and just get a struct fileStructA without that second redundant level? If not, what's the easiest way to remove the top layer of a struct? Do I just have to simply define a new struct?

Cheers

[Stephen23]

"When I load a struct I've previously saved to a .mat file, I've specified a generic name for the struct before, say fileStruct. I'd like to open the struct via load, and have the name of the struct be more specific, like "fileStructA""

So far no one has mentioned the likely cause of this all: forcing meta-data into the structure name.

This situation commonly occurs when beginners import MAT files in a loop and overwrite the imported data on each iteration... so they imagine that the only solution is to rename the imported data variable on each loop iteration, usually wanting "the same name as the file" or some such similar idea. This approach is https://en.wikipedia.org/wiki/Anti-pattern

In fact code will be simpler, more efficient, and much more robust when the names (either in the files or in the workspace) do not change on each loop iteration, and instead use simple indexing on each iteration, just like the MATLAB documentation shows:

https://www.mathworks.com/help/matlab/import_export/process-a-sequence-of-files.html

Answer 1

[Geoff Hayes]

Jonathan - the behaviour that you are observing seems reasonable (and intended) given that there could be more than one variable in the mat file. I don't think that you can get around this issue.

The easiest way to "strip off" the top layer struct is to do something like

 fileStructA = load('myData.mat');
 fileStructA = fileStructA.fileStruct;

Answer 2

[Chieh-Hsun Wu]

Hello Janathan,

Here is my way of doing it:

load('myData.mat');%Load the structure file 'MyStructure'
Names = fieldnames(MyStructure);%List all variables under MyStructure
for nn = 1:length(Names)
  eval([Names{nn},' = MyStructure.',Names{nn},';']);% Assign data to original names
end

Hope this works. CHW

[Jan]

@Chieh-Hsun Wu: This contains 2 bad ideas: 1. loading the variables directly into the workspace. 2. Creating a list of variables by eval. Please read https://www.mathworks.com/matlabcentral/answers/304528-tutorial-why-variables-should-not-be-named-dynamically-eval .

[Rik]

Roberto, please use the flags only to attract attention of site admins.

Also, Jan pointed out two important reasons for why you shouldn't use this, even if it works.

[BINGXIN YAN]

I think the reason is like this. Consider the scenario, we want to define a function with a large number of input variables. We may want to save the input variables in a struct for convenience. We then set the struct_input as input of our function. The code might be like this:

struct_input = struct()
struct_input.x = x
struct_input.y = y
struct_input.z = z
struct_input.a = a
struct_input.b = b
struct_input.c = c

Under the function, we may need to "release" the input variables from the struct_input one-by-one, like this

[val] = function(struct_input):
x = struct_input.x
y = struct_input.y
z = struct_input.z
a = struct_input.a
b = struct_input.b
c = struct_input.c
val = x*y+a*z+exp(b*c)
end

The deficiency is obvious, we need to code 2*length(struct_input) times for the input variables.

BUT, with the above code, we can change the code to this

[val] = function(struct_input):
Names = fieldnames(struct_input);
for nn = 1:length(Names)
  eval([Names{nn},' = struct_input.',Names{nn},';']);
end
end

That is my understand.

[Stephen23]

"The deficiency is obvious"

Yes: very inefficient, obfuscated code.

"we need to code 2*length(struct_input)"

Indicating confusion between the size of a structure and how many fields it has.

See also:

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