cell array of classes
Show older comments
Hi,
Consider a simple class Aclass:
classdef Aclass
properties
w
end
methods
function a = Aclass(b)
a.w = b;
end
end
end
The only data member is a vector. Now say I have a very large matrix that I'd like to turn into a cell array of Aclass objects:
n = 60000; k = 20;
l = zeros(n,k);
% Convert to a cell array of Aclass objects,
sz = size(l,1);
l2 = mat2cell(l,ones(1,sz),k);
l2 = reshape(cellfun(@Aclass,l2,'UniformOutput',0),[sz 1]);
Here, the matrix l with n rows is converted into a cell array with n members, each of which is an Aclass.
My question is: is this the best way to do this? The problem is that it's quite slow when n becomes large. Is there a way to make a "custom" constructor that would return a cell array of Aclass rapidly, perhaps using a MEX file?
Many thanks,
JLT
Accepted Answer
More Answers (1)
David Young
on 13 Jan 2012
Is there any reason to use a cell array rather than an array of Aclass objects? The latter should be more efficient in both time and storage.
You can make such an array in various ways, but probably the neatest is to provide the functionality in the class constructor. Per isakson's link gives the framework. Here's what it might look like for your case:
classdef Aclass
properties
w
end
methods
function a = Aclass(b)
if nargin > 0
a(size(b,1)) = Aclass; % preallocate array
for k = 1:size(b,1)
a(k).w = b(k,:);
end
end
end
end
end
and to create the array you simply do
arr = Aclass(l);
I think this goes faster than your cell-based code, but I haven't checked very carefully.
9 Comments
Jean-Luc
on 15 Jan 2012
Jean-Luc
on 15 Jan 2012
David Young
on 15 Jan 2012
Yes, I'd go for explicit get.w, essentially reversing the constructor. Make sure to preallocate the output matrix before the loop to fill its rows.
per isakson
on 15 Jan 2012
"cell2mat({arr.w}')" looks strange to me. Why not use [arr.w]'
Jean-Luc
on 16 Jan 2012
David Young
on 16 Jan 2012
What a.w is is a comma-separated list. It's equivalent to a(1).w, a(2).w, a(3).w (in the case where a has length 3). So [a.w] does a horizontal concatenation of the w's. However, to reconstruct your original matrix, you need vertical concatenation.
On reflection, I'd probably go for a normal method rather than get.w, as you might want to retain the possibility of getting the comma-separated list. Such a method just looks like:
function b = getw(a)
b = vertcat(a.w);
end
Jean-Luc
on 17 Jan 2012
Jean-Luc
on 17 Jan 2012
David Young
on 17 Jan 2012
In fact, it looks like get.w does not work, but for a different reason.
There is no recursion, because within get.w a reference to a.w goes directly to the property. (Oddly enough, the documentation doesn't seem to say this for get methods, though it does say it for set methods. However, it has to be the case that recursion is avoided, or get methods just wouldn't be useful.)
The reason get.w fails is that it is called once for each element of the array, rather than being passed the whole array. (I don't understand the rationale for this.) What it boils down to is that it looks like you have to use an ordinary method like getw.
Categories
Find more on Matrix Indexing in Help Center and File Exchange
Products
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!