6
\$\begingroup\$

As an exercise to get better in Python, I decided to implement a Binary Search Tree. This implementation uses two classes: BSTNode and BSTree.

Because most methods were implemented using recursion, I created a public method, for most of them, that wraps a private one, which is in charge of making the real work. I did that, because I didn't wanted to give the user the responsibility of passing the root of the tree to every method call. With that, the user can write my_tree.insert(10) instead of my_tree.insert(my_tree.root, 10).

Is this a reasonable way of solve this problem? Is there a better one? Are private methods and attributes commonly used in Python?

The implementation:

class BSTNode(object):
    """ Node class used by the Binary Search Tree. """
    def __init__(self, value):
        self.value = value
        self.size = 1
        self.left = None
        self.right = None

    def compute_size(self):
        """ Computes the `self` size according to its children sizes. """
        self.size = 1
        if self.left:
            self.size = self.size + self.left.size
        if self.right:
            self.size = self.size + self.right.size


class BSTree(object):
    """
    Binary Search Tree implementation which doesn't allow repeated
    Nodes.
    """
    def __init__(self):
        self.__root = None
        self.__size = 0

    ###############
    # Public API  #
    ###############
    def pre_order_traversal(self, fn=None):
        """
        Traverse tree in pre-order and apply `fn` to every Node value.
        """
        if fn is None:
            def fn(x): return print(x)
        self.__pre_order_traversal(self.__root, fn)

    def in_order_traversal(self, fn=None):
        """
        Traverse tree in in-order and apply `fn` to every Node value.
        """
        if fn is None:
            def fn(x): return print(x)
        self.__in_order_traversal(self.__root, fn)

    def post_order_traversal(self, fn=None):
        """
        Traverse tree in post-order and apply `fn` to every Node value.
        """
        if fn is None:
            def fn(x): return

            print(x)
        self.__pre_order_traversal(self.__root, fn)

    def insert(self, value):
        """ Inserts a Node. Doesn't insert duplicated values."""
        if self.__root is None:
            self.__root = BSTNode(value)
        else:
            self.__root = self.__insert(self.__root, value)
        self.__size = self.__root.size

    def remove(self, value):
        """
        Removes a Node which contains the value `value`.
        To remove a Node, three cases must be handled.
        Case 1: leaf node
                    -> delete it
        Case 2: node has one child
                    -> delete node and put its child in its place
        Case 3: node has two children
                    -> delete node and put its smallest child from its right branch in its place
        """
        if self.__root:
            self.__root = self.__remove(self.__root, value)

    def contains(self, value):
        """ Returns True if `value` is found. """
        return self.__contains(self.__root, value)

    def size(self):
        """ Returns the number of elements inside the BST. """
        return self.__size

    ###############
    # Private API #
    ###############
    def __pre_order_traversal(self, node, fn):
        if node is None:
            return
        fn(node.value)
        if node.left:
            self.__pre_order_traversal(node.left, fn)
        if node.right:
            self.__pre_order_traversal(node.right, fn)

    def __in_order_traversal(self, node, fn):
        if node is None:
            return
        if node.left:
            self.__in_order_traversal(node.left, fn)
        fn(node.value)
        if node.right:
            self.__in_order_traversal(node.right, fn)

    def __post_order_traversal(self, node, fn):
        if node is None:
            return
        if node.left:
            self.__post_order_traversal(node.left, fn)
        if node.right:
            self.__post_order_traversal(node.right, fn)
        fn(node.value)

    def __insert(self, node, value):
        if node is None:
            return BSTNode(value)
        if node.value > value:
            node.left = self.__insert(node.left, value)
        elif node.value < value:
            node.right = self.__insert(node.right, value)
        node.compute_size()
        return node

    def __remove(self, node, value):
        if node.value == value:

            # Case 1
            if node.left is None and node.right is None:
                return None

            # Case 2
            elif node.left and node.right is None:
                return node.left

            # Case 2
            elif node.left is None and node.right:
                return node.right

            # Case 3
            else:
                parent_node = node
                smallest_node = node.right
                while smallest_node.left:
                    parent_node = smallest_node
                    smallest_node = smallest_node.left

                # The right Node is the smallest one
                if parent_node == node:
                    smallest_node.left = node.left

                # The smallest Node was found to the left of its right branch
                else:
                    parent_node.left = smallest_node.right
                    smallest_node.left = node.left
                    smallest_node.right = node.right
                return smallest_node

        elif node.value > value and node.left:
            node.left = self.__remove(node.left, value)

        elif node.value < value and node.right:
            node.right = self.__remove(node.right, value)

        node.compute_size()
        return node

    def __contains(self, node, value):
        if node.value == value:
            return True
        if node.value > value and node.left:
            return self.__contains(node.left, value)
        if node.value < value and node.right:
            return self.__contains(node.right, value)
        return False


if __name__ == "__main__":
    tree = BSTree()

    values = [100, 50, 150, 25, 75, 120, 200, 110, 115]

    for v in values:
        tree.insert(v)

    print("####")
    tree.pre_order_traversal()

    tree.remove(100)

    print("####")
    tree.pre_order_traversal()
\$\endgroup\$
4
  • \$\begingroup\$ Are you targeting Python 2, Python 3, or both? If Python 3, do you need compatibility with Python < 3.3? \$\endgroup\$ Sep 17 '17 at 3:00
  • \$\begingroup\$ I forgot to mention that I am targeting Python 3. I don't think that I am going to ever use this piece of code in production. \$\endgroup\$
    – rcanepa
    Sep 17 '17 at 3:09
  • \$\begingroup\$ I would suggest that your "Private API" methods don't actually use self. They would probably be better as methods on the node class, such that BSTree.foo calls self.root.foo(), etc. \$\endgroup\$
    – aghast
    Sep 17 '17 at 14:22
  • \$\begingroup\$ @AustinHastings Comments are for seeking clarification to the question, and may be deleted. Please put all suggestions for improving the code in answers, even brief remarks. \$\endgroup\$ Sep 17 '17 at 18:40
4
\$\begingroup\$

Bugs

You have a size-tracking bug. Also, the .contains() method crashes on an empty tree.

>>> t = BSTree()
>>> t.size()
0
>>> t.insert(3)
>>> t.size()
1
>>> t.remove(3)
>>> t.size()
1
>>> t.contains(3)
Traceback (most recent call last):
  ... line 170, in __contains
    if node.value == value:
AttributeError: 'NoneType' object has no attribute 'value'

Interface

It would be nice if the return value of the .insert() and .remove() methods indicated whether the operations succeeded or failed.

Furthermore, the size of the tree can either increase by 1 if .insert() succeeds, decrease by 1 if .remove() succeeds, or remain the same if either fails. There is no need for BSTNode to have a .compute_size() method.

I suggest defining .size() as a read-only property, so that users can write t.size instead of t.size().

Use a single underscore prefix to indicate private members. Double underscore is for name mangling.

Be consistent in your docstring wording. PEP 257 says that you should prefer the imperative to the indicative:

  • The docstring is a phrase ending in a period. It prescribes the function or method's effect as a command ("Do this", "Return that"), not as a description; e.g. don't write "Returns the pathname ...".

Traversal API

You mis-wrote the preamble for .post_order_traversal():

if fn is None:
    def fn(x): return

    print(x)

But even when written correctly, the preamble is silly. In Python 3, print is a function, so you could just make it the default value:

def post_order_traversal(self, fn=print):
    …

For even more idiomatic Python, avoid passing fn as a callback. Instead, do the traversal as a generator function, yielding each value. The caller would write instead:

for value in tree.post_order_traversal():
    print(value)

Private helper methods

Each of your private helper method exists to serve one corresponding public method. For situations like this, just write an inner function, to reduce clutter, and to avoid having to jump around when reading the code.

class BSTree:
    …

    def pre_order_traversal(self):
        def pre_order(node):
            if node is None: return
            yield node.value
            yield from pre_order(node.left)
            yield from pre_order(node.right)
        yield from pre_order(self._root)

    …

Suggested solution

For brevity, I have omitted the docstrings.

I've fixed the crash in .contains(), and rewritten its helper method to be more compact.

In __remove(), cases 1 and 2 can be condensed down to two lines. I've also renamed smallest_node to more clearly indicate that it refers to the next larger node in the tree.

class BSTNode:
    def __init__(self, value):
        self.value = value
        self.left = self.right = None

class BSTree:
    def __init__(self):
        self._root = None
        self._size = 0

    def pre_order_traversal(self):
        def pre_order(node):
            if node is None: return
            yield node.value
            yield from pre_order(node.left)
            yield from pre_order(node.right)
        yield from pre_order(self._root)

    def in_order_traversal(self):
        def in_order(node):
            if node is None: return
            yield from in_order(node.left)
            yield node.value
            yield from in_order(node.right)
        yield from in_order(self._root)

    def post_order_traversal(self):
        def post_order(node):
            if node is None: return
            yield from post_order(node.left)
            yield from post_order(node.right)
            yield node.value
        yield from post_order(self._root)

    @property
    def size(self):
        return self._size

    def contains(self, value):
        def _contains(node, value):
            return (
                False if node is None else
                _contains(node.left, value) if value < node.value else
                _contains(node.right, value) if value > node.value else
                True
            )
        return _contains(self._root, value)

    def insert(self, value):
        def _insert(node, value):
            if node is None:
                return BSTNode(value)
            elif value == node.value:
                return None
            elif value < node.value:
                node.left = _insert(node.left, value)
            elif value > node.value:
                node.right = _insert(node.right, value)
            return node
        self._root = _insert(self._root, value)
        if self._root:
            self._size += 1
        return self._root is not None

    def remove(self, value):
        def _remove(node, value):
            if node.value == value:
                if not (node.left and node.right):
                    return node.left or node.right, True
                else:
                    # Replace the node with its next larger successor
                    successor, parent = node.right, node
                    while successor.left:
                        successor, parent = successor.left, successor

                    successor.left = node.left
                    if parent != node:
                        parent.left = successor.right
                        successor.right = node.right
                    return successor, True
            elif value < node.value and node.left:
                node.left, removed = _remove(node.left, value)
                return node, removed
            elif value > node.value and node.right:
                node.right, removed = _remove(node.right, value)
                return node, removed
            return node, False
        if self._root is None:
            return False
        self._root, removed = _remove(self._root, value)
        self._size -= int(removed)
        return removed
\$\endgroup\$
1
  • \$\begingroup\$ Thank you for taking the time to give me such a thoughtful review. I really liked the idea of returning tuples to indicate an operation's result and using generators instead of callback functions. \$\endgroup\$
    – rcanepa
    Sep 17 '17 at 15:20
1
\$\begingroup\$
        self.size = self.size + self.left.size

This is more naturally expressed as self.size += self.left.size

    self.__root = None
    self.__size = 0

The double underscore invokes name mangling, which can be helpful it you anticipate inheritance. Here, a single underscore would probably suffice.

You organized the source code into Public and Private API sections, which clearly makes some sort of sense. Personally I would have found it more helpful, when reading code, to see each public method immediately followed by its private method. For example, your (excellent) docstrings matter a great deal when trying to understand implementations, but one must scroll to find them. Methods like pre_order_traversal() are already using nested functions. Private helpers that start with _ are perfectly nice, but you might consider using the alternative of hiding each helper by making it a nested function.

values = [100, 50, 150, 25, 75, 120, 200, 110, 115]

Thank you for including test code that exercises the target code. That is certainly helpful. It would be more helpful to include a unittest.TestCase with good measured code coverage of the target code.

Overall, it looks solid. Ship it!

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.