Struct contents reference from a non-struct array object

Hi,

I have ds containing cell arrrays (1 by 3571) and each cell index points to a table of 4 by 7 elements. When any cell is empty while concatenating, the above error is shown. Basically, I am trying to access the contents of the cells contained within the cell array "Table" and plot different fields on a 1 by 3571 grid where each index refers to a point on a lat/lon grid.

Here is the code: Any ideas how to fix the error and plot the content of the data set.

for n = 1:3571; 
    
    Unc{n}=Table{1,n}.UncA

end;

whos Table
  Name       Size                Bytes  Class    Attributes
  Table      1x3571            9326972  cell      

Here is the sample content of one of the cells contained in teh cell array.

Var A UncA B UncB C UncC

'1' -0.0186917327837646 -0.0266242514128150 0.00420612704074986 -0.00200002085968102 0.107411895593770 0.0187751225715014

'2' -0.0141854422900873 -0.0316765707830826 0.00382987363234866 -0.00237934723997174 0.0345770815773565 0.0224174109442981

'3' -0.0180850135958575 -0.0258819602652245 0.00416311313467285 -0.00194411629488122 NaN NaN

'4' -0.0167188467826965 -0.0315984164923636 0.00407529091905538 -0.00236624020568344 NaN NaN

Answer 1

[Greg]

If I'm understanding correctly, some indices of your cell array contain tables, and others are empty? What do you want to happen with the empty cells?

blnEmptyCells = cellfun(@isempty,Table);
%%% If throwing them out isn't ok, see additional posts below
Table(blnEmptyCells) = [];
Unc = cell(length(Table),1);
for ij = 1:length(Table)
    Unc{ij} = Table{1,ij}.UncA;
end

[skanwal]

Hi Greg,

This code work but deletes the cells with empty values []. I do not want to delete these values becuase i will lose the index values to lat/lon and won't be able to know the location of these points. So, I applied this function to replace the empty cells with NaN:

ds(cellfun(@isempty, ds)) = {nan}

But in this case, it gives me the same error I mentioned above

[Greg]

blnEmptyCells = cellfun(@isempty,Table);
Unc = num2cell(nan(length(Table),1));
for ij = 1:length(Table)
    if blnEmptyCells(ij)
        continue;
    end
    Unc{ij} = Table{1,ij}.UncA;
end

[Greg]

Your original error message is quite clear: you're trying to treat a non-structure as a struct (in later releases the message changes to "dot indexing is not available for this data type"). You can't do Table{1,ij}.UncA if Table{1,ij} = nan. The nan is a double, not a struct, and has no dot indexing ability. You could put a fake struct in Table{1,ij} to "make your code work as is," but that isn't the right solution. The better answer is to use simple isempty checking (see post above) and skip the dot-indexing entirely.

[skanwal]

Thank you Greg. That has solved the problem now. I have the same no of cells in the output as in the original data set.

[Greg]

Happy to help!

Answer 2

[skanwal]

From the origical data (1 by 3071), I have the indices for both the empty and non-empty cells. is there any way that I can fill the newly generated cells array which is 1 by 2076 cell array (with empty cells removed ) with elements such as empty cells or NaN based on the indices in MATLAB where there were empty cells in the original data to keep the dimensions of the data same?