Performance when creating an array of class objects

Hi

Using Matlab 2010b Service Pack 1. We are having a performance issue when creating an array of 1000 objects.

The array is obj.MyClassArray

The objects are of class

classdef MyClass < handle

properties

  prop1
  prop2
  prop3

  nameValueContainerMap % a container.Map(name, MySeriesObject)

end

The map class is

classdef MySeries < handle

properties

        dates
    values % a cell containing n doubles

end

The cell consists of an array of doubles.

We are pre-allocating memory by initialising obj.MyClassArray with 1000 copies of a fully populated object of MyClass, using repmat.

However, we are still finding a performance issue.

Any ideas?

Thanks

Marty

Answer 1

[Adam]

If your class has a default constructor (or basically your constructor can handle the 0-argument case) then the simplest way to pre-allocate an array of 1000 objects is just to type:

myArray(1,1000) = MyClass;

You say "1000 copies of a fully populated object" though which suggests it isn't infact pre-allocation you are doing, but full array allocation already initialised objects. In that case you can't do what I suggested.