What is the Matlab equivalent of the yield keyword in Python?

Question

I need to generate multiple results but one at a time, as opposed to everything at once in an array.

How do I do that in Matlab with a generator like syntax as in Python?

Answer 1

When executing functions that use the yield keyword, they actually return a generator. Generators are a type of iterators. While MATLAB does not provide the syntax for either, you can implement the "iterator interface" yourself. Here is an example similar to xrange function in python:

classdef rangeIterator < handle
    properties (Access = private)
        i
        n
    end

    methods
        function obj = rangeIterator(n)
            obj.i = 0;
            obj.n = n;
        end

        function val = next(obj)
            if obj.i < obj.n
                val = obj.i;
                obj.i = obj.i + 1;
            else
                error('Iterator:StopIteration', 'Stop iteration')
            end
        end

        function reset(obj)
            obj.i = 0;
        end
    end
end

Here is how we use the iterator:

r = rangeIterator(10);
try
    % keep call next() method until it throws StopIteration
    while true
        x = r.next();
        disp(x);
    end
catch ME
    % if it is not the "stop iteration" exception, rethrow it as an error
    if ~strcmp(ME.identifier,'Iterator:StopIteration')
        rethrow(ME);
    end
end

Note the when using the construct for .. in .. in Python on iterators, it internally does a similar thing.

You could write something similar using regular functions instead of classes, by using either persistent variables or a closure to store the local state of the function, and return "intermediate results" each time it is called.

Answer 2

In MATLAB (not yet? in Octave), you can use closures (nested, scoped functions):

function iterator = MyTimeStampedValues(values)

    index = 1;

    function [value, timestamp, done] = next()
        if index <= length(values)
            value = values(index);
            timestamp = datestr(now);
            done = (index == length(values));
            index = index + 1;
        else
            error('Values exhausted');
        end
    end

    iterator = @next;
end

and then

iterator = MyTimeStampedValues([1 2 3 4 5]);
[v, ts, done] = iterator();    % [1, '13-Jan-2014 23:30:45', false]
[v, ts, done] = iterator();    % ...