This question continues your implementation of the time series class you started last week. Then, you stored your time series as a python list.
Add methods __iter__ to your project Time Series class to iterate over values, a method itertimes to iterate over times, and a method iteritems to iterate over time-value pairs. (This is a similar interface to python dictionaries). To test these, check both the types of the results and the answers you expect.
Next, you will store the data as a numpy 1-D array.
Please implement a new class ArrayTimeSeries which inherits your TimeSeries class and uses numpy.array to store the data internally.
At this point you will notice a fundamental mismatch: unlike python lists or array.arrays which are dynamic, numpy expects you to provide a length of the sequence. This means that you cant process each element of the time series as it comes in, and must calculate the length of the input sequence to allocate space for the time series in your constructor. (This makes it hard to write a "direct from file" without intermediate storage implementation; but we shall worry about this later).
Because your class inherits from your TimeSeries class, you'll notice you get some functionality automatically. You may choose to reimplement __getitem__, __setitem__, __len__ , and the iteration functions at your discretion. Please do NOT implement a __str__ or __repr__ function. Note that this means the __str__ and __repr__ functions will need to work with both classes, and you may need to fix your original versions.
Make sure that any doctests and tests you wrote before to test all kinds of sequences as input still pass. (If you didnt write any, how do you know your time series Class from last time is any good? We'll be testing your code by running our own tests as well!)
Next weeks we'll refactor these classes and add support for synthetic time series, and incoming data that is streaming!