Description 
CATPAD  concatenate matrices with different sizes by padding with NaN.
M = CATPAD(dim,A1, A2, A3, ..., AN) concatenates along the dimension
DIM the arrays A1 through AN into one large matrix. The vectors do
not need to have the same size, nor number of dimensions. The size
of the output M is determined by the dimmention of concatenation,
and the size of the inputs. Any inputs that are not the correct
size will be padded with NaNs if the inputs are numeric. If they
are strings, they will be padded with a space " ".
[M TF] = CATPAD(...,'padval',padval) pads the input data with the value
specified by PADVAL. The default is NaN for numeric inputs, and a
space " " for string inputs.
[M TF] = CATPAD(...) outputs ana array of logicals TF having true
values when the values in that position of M were from the original
data (i.e. not padded).
Examples:
a = 1:4; b = 1:5; c = []; d = 1:3; dim = 1;
M = catpad(dim,a,b,c,d)
M =
1 2 3 4 NaN
1 2 3 4 5
1 2 3 NaN NaN
% Note: The input "c" was empty, and therefore doesn't require it's
% own row.
a = rand(3); b = magic(5); c = rand([4 5 4]); dim = 3;
M = catpad(dim,a,b,c);
size(M)
5 5 6
str1 = 'What do you think of this function?';
str2 = 'I like it!';
dim = 1;
M = catpad(1,str1,str2)
M =
What do you think of this function?
I like it!
Example: Find original NaNs
a = 1:3; b = [2:5 NaN]; c = [1 NaN]; dim = 1;
[M,tf] = catpad(dim,a,b,c)
% find the original NaN
[row,col] = find(tf & isnan(M))
% > row = [3 2] , col = [2 5]
a = 1:3; b = 1; c = 1:4; dim = 1; padval = inf;
M = catpad(dim,a,b,c,'padval',padval)
M =
1 2 3 Inf
1 Inf Inf Inf
1 2 3 4
See also CAT, RESHAPE, STRVCAT, CHAR, HORZCAT, VERTCAT, ISEMPTY
By Jonathan Sullian  October 2011
Many thanks to Michael VĂ¶lker for his function sub2allind which has made
this function possible. Also, I would like to aknowledge Jos (10584)
whose submission padcat has inspired this one.
