Functional Data Structures

Functional Programming in Java

Created by Mark Perry, @mprry, G+, Blog, LinkedIn, GitHub, maperry78@yahoo.com.au

Introduction

• FP emphasises not updating data
• Alternative?
• Pure functions

Referential Transparency

    // updating is not referentially transparent
    List<Integer> list = new LinkedList<>();
    list.add(1);
    List<Integer> list2 = list; // [1]

    new LinkedList<Integer>().add(1);
    List<Integer> list2 = new LinkedList<>(); // []

Functional Data Structures

• Values are immutable!
• Use pure functions
• Data copying issue?

Immutable Structures

• Copy?
• Reuse list due to immutability
• Data sharing

List Sharing

List<A> xs = list(0, 1, 2);
List<A> ys = list(3, 4, 5);
List<A> zs = xs.append(ys) // [0, 1, 2, 3, 4, 5]
					

Linked List

abstract class List<A> {
    abstract A head();
    abstract List<A> tail();

    static List<A> nil() {...}
    static List<A> cons(A h, List<A> t) {...}
    static List<A> list(A...as) {...}
    }

    class Nil<A> extends List<A> {...}
    class Cons<A> extends List<A> {...}

    List<Integer> a = nil();
    List<Integer> b = cons(1, nil());
    List<Integer> c = cons(1, cons(2, nil()));
    List<Integer> d = cons(1, cons(2, cons(3, nil())));
    List<Integer> e = list(1, 2, 3);

Tree Insertion

Tree<A> xs = ??? // a tree
Tree<A> ys = xs.insert("e")

Tree Insertion (2)

Tree<A> xs = ??? // a tree
Tree<A> ys = xs.insert("e")

Data Sharing

• Efficiency from data sharing
• Often needs tree structures
• Scala Vector, O(1)
  • random access
  • update
  • prepend, append
  • head, tail
• Vector Trie

Exercises

• Assume tail
• Generalise tail

						List<A> tail(List<A> list)
						List<A> drop(List<A> list, int n)
						List<A> dropWhile(List<A> list, F<A, Boolean> f)
					

Naive Implementation

List<Integer> sum(List<Integer> list) {
  if (list.isEmpty()) {
    return 0;
  } else {
    return list.head() + sum(list.tail());
  }
}

List<Integer> product(List<Integer> list) {
  if (list.isEmpty()) {
    return 1;
  } else {
    return list.head() * product(list.tail());
  }
}
					

Fold

<A, B> B fold(List<A> list, B acc, F2<B, A, B> f)

List<Integer> sum2(List<Integer> list) {
    return fold(list, 0, (acc, i) -> acc + i);
}

List<Integer> product2(List<Integer> list) {
    return fold(list, 1, (acc, i) -> acc * i);
}

Fold Left

<B> B foldLeft(List<A> list, B acc, F2<B, A, B> f) {
  if (list.isEmpty()) {
    return acc;
  } else {
    return foldLeft(list.tail(), f.f(acc, list.head()), f);
  }
}

foldLeft(list(1, 4, 9), 0, sum)
foldLeft(list(4, 9), 1 + 0, sum)
foldLeft(list(9), 4 + 1, sum)
foldLeft(nil(), 9 + 5, sum)
14

Fold Right

<B> B foldRight(List<A> list, B acc, F2<B, A, B> f) {
  if (list.isEmpty()) {
    return acc;
  } else {
      return f.f(foldRight(list.tail(), acc, f), list.head());
  }
}

foldRight(list(1, 4, 9), 0, sum)
foldRight(list(4, 9), 0, sum) + 1
(foldRight(list(9), 0, sum) + 4) + 1
((foldRight(nil(), 0, sum) + 9) + 4) + 1
(((0) + 9) + 4) + 1

Exercises

List<A> reverse(List<A> list)
List<B> map(List<A> list, F<A, B> f)
List<A> flatten(List<List<A>> list) // aka join
List<B> flatMap(List<A> list, F<A, List<B>> f) // aka bind, >>=

Trees

abstract class Tree<A>
class Empty<A> extends Tree<A>
class Node<A> extends Tree<A>

Growing Trees

Tree<Integer> t1 = Tree.<Integer>empty();
Tree<Integer> t2 = Tree.leaf(1);
Tree<Integer> t3 = Tree.tree(leaf(3), 5, leaf(7));

Exercises

B foldLeft(Tree<A> tree, B acc, F2<B, A, B> f)
int size(Tree<A> tree)
int depth(Tree<A> tree)
Tree<B> map(Tree<A> t, F<A, B> f)
List<A> traverseLeft(Tree<A> t)

Summary

• Data Sharing
• Folds
• Practiced Pure Functions

Afterword

Functional Programming in Scala, Chiusano and Bjarnason, Chapter 3, Functional Data Structures

Created by Mark Perry, @mprry, G+, Blog, LinkedIn, GitHub, maperry78@yahoo.com.au