In this session we'll take a quick detour from the general topic of working with lists by introducing a new class of data structures, namely pairs and tuples. You'll see how pairs and tuples can help in program composition and decomposition, and you'll see that by using a somewhat larger example. We're going to demonstrate the material in this session with a somewhat larger example. The task is to define a function to sort lists that's more efficient than insertion sort. One good algorithm that's particular suitable for a functional list implementation is merge sort. The idea behind merge sort is as follows: if a list consists of zero or one elements, it's obviously already sorted so there's nothing to do. Otherwise, we separate the lists into two sub lists each containing around half of the elements of the original list. An easy way to do that would be to simply take the first n elements of the list and then the second n element of the lists if the lengths of the list is 2n. Once we have the two sub lists we sort them each in turn and then we merge the two sorted sub lists into a single sorted list. To see the algorithm in code, look at this function here. So we would have a function msort for merge sort. It takes a list of Ints for the moment we restrict ourselves to lists of a single type, and it returns a list of Ints The first thing is the splitting, so we take the length of the list divided by two, that's our n. If n is zero than, than the original length was zero or one, because division truncates towards zero. In both of these cases, the list is already sorted so we can simply return it. If n is not zero, Then what we do is we split the list at point N so we'll get to split that function in a moment. It returns essentially the first half of the list and the second half of the list. Then we sort both of these halfs with the recursive call to msort and finally we merge the two sorted lists. The definition of merge is here, is a function that takes two lists of Ints and we have left out its implementation so far. So if we look at an implementation of merge, then here's a possible one. We're going to improve that later. So merge takes two lists of Ints and it would proceed by a pattern match. First, on the left list, so if the left list is nil. Then the merge must consist of all the elements of the right list so we return ys. If the left list is not Nil so let's say consists of head element x followed by a tail xs1, Then we do a pattern match on the right hand list the ys. If that is nil then we can simply return xs. If that is not nil then we have two head elements x and y and two tail lists x is one and y is one. And what we do is we compare the head elements with each other. So if x is smaller than y, then obviously x must be the first element of our sorted lists, so we take x followed by a merge of all the remaining elements are from the excess list, that's excess one and all the elements of the ys list. If, on the other hand, x is not less than y, then we can take y as the first element of the sorted list, and we follow that by a merge of all the x elements followed by all the other y elements that the ones that follow y. So, that would be y is one here. So we've seen that the M sort algorithm made use of the split-end function. Let's take a closer look at this one. So the split it function on lists returns two sub lists, namely the elements of a given list up to the given index and the elements that start from that index. And those two lists are in fact returned in a pair. Now lets take a little time to look at pairs and their generalization tupples. A pair in Scala. Is written just x and y, where x and y are the elements of the pair. So here's an example, you could form a pair from the string answer and the number. 22. Call it pair, and the worksheet would respond that we have just formed a pair of type string comma int in parentheses. So, that's the type of this pair. And the value of that pair is obviously the string answer, and the value 42. So that was a way to form pairs. We can also decompose pairs using pattern matching, so that you see down here. We can type pair and we can match it against the pattern that contains two variables, label and value in a pair. And the worksheet would answer that we have just defined a label which is a string and its value is answer. And we have defined a value which is an int and it's value is 42. This, of course, works analogously also for tuples with more than two elements. So, you can have triples, quadruples, and so on. So far, all the types we've encountered in Scala were really abbreviations for some instance of a class type. And tuples are no exception. In fact, the tuple type of t1 to tn in parens is just an abbreviation of the parameterized type, Scala.tuple n of t1 to tn, with, as type parameters. If we look at tuple expressions then E1 to EN and parens is equivalent to the function application skeletal tuple N of E1 to EN. And finally, a tuple pattern, p1 to pn, is equivalent to the constructor pattern. Again, scala.tuple n p1 to pn. To make that work, we have to have a look at the tuple class. So all the tuple classes are modeled after the following pattern that you see here for tuple two. It's a case class it takes t1 and t2 as parameters. And then tuple two would have two fields which are named underscore one and underscore two of types t1 and t2 respectively. That's almost all. The only other thing we need to do is work on the ways tuples are represented. We do not want to print them with tuple two so we override the two string function to give us back the usual tuple syntax, so open parentheses followed by the field separated by commas followed by a closing parentheses. So what this definition of the case class shows is that the fields of a tuple can in fact be accessed with names underscore one, underscore two, and so om up to the number of elements in the pattern. That means that instead of the pattern matching that we've seen. Say, now label value equals pair. One could also have written label equals pair.<u>11.</u> So the first element of the pair and value equals pair.<u>2.2.</u> But the pattern matching form is generally preferred, because it's both shorter and clearer. So, let's do an exercise. The merge function we've seen so far uses a nested pattern match and that was rather long and it was also not so nice because it didn't reflect the inherent symmetry of the merge algorithm. We had, first had to do at pattern match on the left-hand side and then a nested pattern match on the right-hand side but, for merge, it doesn't really matter what is left-hand and what is right-hand side. So. Let's rewrite merge using a pattern matching over pairs. So what I want to do here is, I want to have the same signature as merge, but I want to start off with saying, well, let's form a pair of xa, ys, and then match on the pair. So, How would I do that? To develop an answer for this quiz let's use the worksheet again. I've already opened a worksheet for merge sort and I've put the outline of the merge sort function as we've seen it so far in it. The definition of merge here is still missing, so let's work on that. So the suggestion was, let's do a pattern match on the pair xs ys. What kind of patterns would apply? Well, the first pattern that's interesting is if either one of the argument lists is nil. So the first one would be nil nys. In that case, if the first list x-s is nil, and then the second list y-s, is just arbitrary, then we always return y-s, because there's nothing to merge. Analogously, if the first list is arbitrary and the second list is nil, we always return x-s. So that leaves us with the third case, where we say that both lists are non-nil. Let's write it like that. So X followed by XX1 and Y followed by YS1. And then in that case what do we do. Well, we now need to compare X and Y. If it's less than then we know that X is the first element of our result and we follow that with merge XX1 and YS. And if that's not the case and Y is the first element. We then follow it with merge of XX and YS1. And what we have here is, it complains that we need to give a result type, which is correct. So now it should compile without problems, and we can now apply merge sort to some test data, so let's define nums equals list of two minus four, five, seven, one. Just to get some data here and we do msort of nums. And it did sort them, the way it was supposed to.
