Uploading .mat files that contain nested structs into a datastore

I would like to upload .mat file into a datastore that contains nested structs. Like so...

TestData.Unit1.Parameter 1

TestData.Unit1.Parameter 2

TestData.Unit2.Parameter 1

TestData.Unit2.Parameter 2

When I try to upload this into a datastore I cannot read the content of the fields inside the nested struct. The size of the .mat files are enormous and take a really long time to load one parameter. How do I accomplish this while still using datastore or something of the like? I know we are supposed to use fileDataStore to create the custom read function but the read function I created only displays the first level of the structs.

[Matt J]

When you say "upload" do you truly mean that the datastore reads from files stored on a remote computer? If so, maybe the reading time is predominantly due to the remote connection, rather than the file size.

[Dennan]

It is files stored on a network of computers. So there is no way to "point" to the data and call specific parameters as needed?

Answer 1

[Catalytic]

I think the wisest course would be to reorganize your data and assign each "Unit" to its own separate .mat file.

[Dennan]

Normally I would agree but I was given specific instructions to not create anymore .mat files because I am loading from a specific store of data

[Dennan]

The data is write protected

[Matt J]

It wouldn't matter if the data is write-protected. You could read the data, reorganize it, and copy the reorganized data somewhere controlled by you and where you would have full read/write acess.

If for some reason you can't do that, so be it. However, one of the upsides of @Catalytic's suggestion is that if you split the data this way, you could shuffle the Parameter instances. In the form you have things now, the shuffle() command can only vary the the read-out order of the files, not the order of the Parameters within the files.

Answer 2

[Matt J]

This example shows how to use a fileDataStore to read partial data from a .mat file

https://www.mathworks.com/help/matlab/ref/matlab.io.datastore.filedatastore.html#mw_77a0bae2-8cf5-4875-84ba-1c9918ec1263

I might modify their example and use a matfile object to read the data without actually loading the entire file.

[Dennan]

Yes I tried that specific example. I tried using the matfile but it doesn't see any variables.

fds = fileDatastore("C:\Users\******\Desktop\datafile.mat","ReadMode","partialfile","ReadFcn",@load_variable)

function [data,variables,done] = load_variable(filename,variables)
    
    % If variable list is empty, 
    % create list of variables from the file
    if isempty(variables) 
        variables = who('-file', filename);
    end
    
    % Load a variable from the list of variables
    data = matfile(filename, variables{1});
    
    % Remove the newly-read variable from the list
    variables(1) = []; 
    
    % Move on to the next file if this file is done reading.
    done = isempty(variables); 
    
end

read(fds)

This is the error back:

Error using matlab.io.datastore.FileDatastore/read (line 29)

Error using ReadFcn @load_variable for file:

C:\Users\********\Desktop\datafile.mat

Error using matlab.io.MatFile (line 422)

No value was given for 'Data'. Name-value pair arguments require a name followed by a value.

Error in matfile (line 75)

mf = matlab.io.MatFile(varargin{:});

Error in load_variable (line 10)

data = matfile(filename, variables{1});

[Walter Roberson]

    data = matfile(filename, variables{1});

That is the wrong way to use matfile. The proper way to use matfile is

    matObj = matfile(filename);
    data = matObj.(variables{1});

[Matt J]

If the instances to be read are the nested Parameterx fields, then I think you would need something more elaborate than the literal example from the doc:

function [data,tasks,done] = load_variable(filename,tasks)
    
    % Initialize tasks
    if isempty(tasks) 
        tasks.unitList = who('-file', filename);
    end
    
    %Initialize current Unit
    if ~isfield(task,'currentUnit')
        tasks.currentUnit=load(filename, tasks.unitList{1}).(tasks.unitList{1});
        tasks.paramList=fieldnames(tasks.currentUnit);
    end
    
    % Load a piece of data
    data = tasks.currentUnit.(tasks.paramList{1});
    
    % Remove the read-out parameter from paramList
    tasks.paramList(1) = []; 
    
    if isempty(tasks.paramList) %Check if we are finished processing a Unit
       tasks=rmfield(tasks,{'currentUnit','paramList'});
       tasks.unitList(1)=[];
    end
    
    done =isempty(tasks.unitList); %Check if all Units in the file have been processed
    if done
       tasks=[];
    end
    
end

[Dennan]

>> read(fds)

ans =

struct with fields:

Para1: [1×1 struct]

Para2: [1×1 struct]

That worked. How do you get to the actual data within the struct?

[Matt J]

See above.

[Stephen23]

"How do you get to the actual data within the struct?"

https://www.mathworks.com/help/matlab/matlab_prog/create-a-structure-array.html

Based on what you show in your comment, perhaps something like this:

S = read(fds);
S.Para1
S.Para2

[Matt J]

Based on what you show in your comment, perhaps something like this:

Yes, but the OP would like the read() operation to return this directly.