Any approach faster to concatenate multi dimension cell?

Dear all,

The objective was to concatenate multi dimension cell, may I know how to avoid such a FOR loop.

Thanks in advance

load('completedata.mat')
c_x=1;
for f_x=1:4
    myData(c_x,:)= data_lap_th{f_x}.e_subj;
    c_x=c_x+1;
end

Answer 1

[per isakson]

Try

>> S = load('completedata.mat');
>> S.data_lap_th
ans = 
    [1x1 struct]    [1x1 struct]    [1x1 struct]    [1x1 struct]
>>
>> sas = [S.data_lap_th{:}];
>> sas
sas = 
1x4 struct array with fields:
    e_subj

Whether it faster I don't know, but it is Matlabish.

[balandong]

Hi Per,

Thanks for the valuable suggestion. Maybe my fault that my writing make the above question not clear enough.

Running your code above, it produce something in the right figure below. However, I am thinking of shaping all the cell as in the left figure below.

[Cedric]

As Per is saying, there are more concise ways to write your loop, but they may not be more efficient:

 myData = cellfun( @(c)c.e_subj, data_lap_th, 'UniformOutput', false ) ;
 myData = vertcat( myData{:} ) ;

Back to your code, if you were dealing with a much larger data set, you could prealloc myData before the loop:

 n = numel( data_lap_th ) ;
 myData = cell( n, 24 ) ;
 for k = 1 : n
    myData(k,:)= data_lap_th{k}.e_subj ;
 end

You could also make the code more flexible by not hard-coding the 24, but getting it from the first record:

 nr = numel( data_lap_th ) ;
 nc = numel( data_lap_th{1}.e_subj ) ;
 myData = cell( n, nc ) ;
 for k = 1 : nr
    myData(k,:)= data_lap_th{k}.e_subj ;
 end

but none of these approaches should make the code significantly faster given the size of your data set.

[balandong]

Hi Cedric, Thanks for valuable response, really appreciate it.

[Cedric]

My pleasure!