Search
 
SCRIPT & CODE EXAMPLE
 

PYTHON

code to printing a binary search tree in python



class BSTNode:
    def __init__(self, key=None):
        self.left = None
        self.right = None
        self.key = key
# Insert method can add a list of nodes to the BST
    def insert(self, keyList):
       for i in keyList:
          self.insertKey(i)
# This insertKey 
    def insertKey(self, key):
        if not self.key:
            self.key = key
            return
        if self.key == key:
            return
        if key < self.key:
            if self.left:
                self.left.insertKey(key)
                return
            self.left = BSTNode(key)
            return
        if self.right:
            self.right.insertKey(key)
            return
        self.right = BSTNode(key)
    def display(self):
        lines, *_ = self._display_aux()
        for line in lines:
            print(line)

    def _display_aux(self):
        """Returns list of strings, width, height, and horizontal coordinate of the root."""
        # No child.
        if self.right is None and self.left is None:
            line = '%s' % self.key
            width = len(line)
            height = 1
            middle = width // 2
            return [line], width, height, middle

        # Only left child.
        if self.right is None:
            lines, n, p, x = self.left._display_aux()
            s = '%s' % self.key
            u = len(s)
            first_line = (x + 1) * ' ' + (n - x - 1) * '_' + s
            second_line = x * ' ' + '/' + (n - x - 1 + u) * ' '
            shifted_lines = [line + u * ' ' for line in lines]
            return [first_line, second_line] + shifted_lines, n + u, p + 2, n + u // 2

        # Only right child.
        if self.left is None:
            lines, n, p, x = self.right._display_aux()
            s = '%s' % self.key
            u = len(s)
            first_line = s + x * '_' + (n - x) * ' '
            second_line = (u + x) * ' ' + '' + (n - x - 1) * ' '
            shifted_lines = [u * ' ' + line for line in lines]
            return [first_line, second_line] + shifted_lines, n + u, p + 2, u // 2

        # Two children.
        left, n, p, x = self.left._display_aux()
        right, m, q, y = self.right._display_aux()
        s = '%s' % self.key
        u = len(s)
        first_line = (x + 1) * ' ' + (n - x - 1) * '_' + s + y * '_' + (m - y) * ' '
        second_line = x * ' ' + '/' + (n - x - 1 + u + y) * ' ' + '' + (m - y - 1) * ' '
        if p < q:
            left += [n * ' '] * (q - p)
        elif q < p:
            right += [m * ' '] * (p - q)
        zipped_lines = zip(left, right)
        lines = [first_line, second_line] + [a + u * ' ' + b for a, b in zipped_lines]
        return lines, n + m + u, max(p, q) + 2, n + u // 2
    #Inorder Walk    
    def inorder(self):
        if self.left:
            self.left.inorder()
        print(self.key)
        if self.right:
            self.right.inorder()
a = BSTNode()
a.insert([5,7,4,3,5,1,3,6]) #inserting some random numbers in the form of list
a.inorder()
a.display()
Comment

Binary Search Tree python

"""Binary Search Tree

Included:
	- Insert
    - Find data/ minimum/ maximum
    - Size
    - Height
    - Traversals (Inorder, preorder, postorder)
    - Test Code for it
    
Not included:
	- Delete
    - Display (as 2D BST)
	and others...
"""

class Node: # Node
    def __init__(self, data, start=None):
        self.data = data
        self.left = self.right = start

class BinarySearchTree:
    def __init__(self):
        self.root = None

    # Insert a node
    def put(self, data):
        if self.root is None: # BST empty
            self.root = Node(data)
        else:
            curr = self.root # initialise curr pointer
            while curr:
                if data < curr.data: # data put at left
                    if curr.left:
                        curr = curr.left
                    else: # no left child
                        curr.left = Node(data)
                        break
                else: # data put at right
                    if curr.right:
                        curr = curr.right
                    else:  # no right child
                        curr.right = Node(data)
                        break   
                        
    def find(self, data):
        curr = self.root   # start from root
        while curr:
            if data < curr.data:  # data at left
                curr = curr.left
            elif data > curr.data:    # data at right
                curr = curr.right
            else:
                return True # found
        return False
    
    def min_of(self):
        curr = self.root # start from root
        while curr.left:
            curr = curr.left # keep going left
        return curr.data
        
    def max_of(self):
        curr = self.root # start from root
        while curr.right:
            curr = curr.right # keep going right
        return curr.data
    
    def size(self):
        def go(node): # helper
            if node:
                return 1 + go(node.left) + go(node.right)
            return 0
        return go(self.root)   
    
    def height(self):
        def go(node): # helper
            if node:
                return 1 + max(go(node.left), go(node.right))
            return -1 # it has -1 if empty
        return go(self.root)

        
    # In_order
    def in_order(self):
        lst = [] # results
        def go(node): # helper
            nonlocal lst
            if node: # If node exists
                go(node.left) # left
                lst.append(node.data) # Node
                go(node.right) # Right
        go(self.root)
        return lst
    
    # Pre_order
    def pre_order(self):
        lst = [] # results
        def go(node): # helper
            nonlocal lst
            if node: # If node exists
                lst.append(node.data) # Node
                go(node.left) # left
                go(node.right) # Right
        go(self.root)
        return lst
    
    # Post_order
    def post_order(self):
        lst = [] # results
        def go(node): # helper
            nonlocal lst
            if node: # If node exists
                go(node.left) # left
                go(node.right) # Right
                lst.append(node.data) # Node
        go(self.root)
        return lst

# Test
from random import randint, sample
def BST_tester(BST):
    r = randint(5, 10)
    lst = [randint(10, 100) for _ in range(r)]
    print(f"List: {lst}", 
          f"Sorted: {sorted(lst)}",
          sep = "
", end = "

")
    
    B = BST()
    for item in lst:
        B.put(item)
    lst.sort()
    print("AFTER INSERTION:",
          f"Size: {B.size()} | {B.size() == len(lst)}",
          f"Height: {B.height()} | I can't code the display for this",
          f"Min: {B.min_()} | {B.min_() == min(lst)}",
          f"Max: {B.max_()} | {B.max_() == max(lst)}",
          sep = "
", end = "

")
    
    items = sample(lst, 5) + [randint(10, 100) for _ in range(5)] # 5 confirm in list, 5 might be in list
    found = [B.find(_) for _ in items]
    zipped = list(zip(items, found))
    inlst, notinlst = zipped[:5], zipped[5:] 
    print(f"Sampled from lst: {inlst} | All True: {all(found[:5])}",
          f"Random: {notinlst}",
          sep = "
", end = "

")
    
    print(f"Inord: {B.in_ord()} | Correct Insertion: {lst == B.in_ord()}",
          f"Preord: {B.pre_ord()}",
          f"Postord: {B.post_ord()}",
          sep = "
")
BST_tester(BST)
Comment

Binary Search tree implementation in python

"""
	              17
		    /    
		   /      
		  /	   		  
		 4         20
		/	   /	
	       /         /  
	      /         /    
	     1      9   18    23
			        
				 
				   
				   34
							 
"""							 

# Binary Search Tree implementation in Python

class BinaryTreeNode():
    def __init__(self, data):
        self.data = data
        self.left = None
        self.right = None
        
    def add_child(self, data):
        if data == self.data: # check if the of new data exist already in the tree, if yes don't add
            return
        
        if data < self.data:
            # Add to left subtree
            if self.left:
                self.left.add_child(data) # Recursively call the add_child method to add the data to an appropriate place
            else:
                self.left = BinaryTreeNode(data)
        else:
            # Add to right subtree
            if self.right:
                self.right.add_child(data) # Recursively call the add_child method to add the data to an appropriate place
            else:
                self.right = BinaryTreeNode(data)
    
    # Visit Left subtree, then Root node and finaly Right subtree
    def in_order_traversal(self):  # Left - Root - Right
        elements = []
        
        # Getting all elements of the Left Subtree    
        if self.left:
            elements += self.left.in_order_traversal() # Recursively get all the elements of the left subtree and add them into the list
        elements.append(self.data) # Adding the root node to the list
        
        # Getting all elements of the Right Subtree    
        if self.right:
            elements += self.right.in_order_traversal() # Recursively get all the elements of the right subtree and add them into the list
        return elements
        
    # Get all elements from the Root node then the left subtree and finanally the Right subtree 
    def pre_order_traversal(self): # Root - Left - Right
        elements = []
        
        elements.append(self.data)
        
        if self.left:
            elements += self.left.pre_order_traversal()  # Recursively get all the elements of the left subtree and add them into the list
        
        if self.right:
            elements += self.right.pre_order_traversal()  # Recursively get all the elements of the right subtree and add them into the list

        
        return elements # get the Root node element
        
    # Get all elements from the Right subtree then the left subtree and finally the Root node    
    def post_order_traversal(self):
        elements = []
        
        if self.left:
            elements += self.left.post_order_traversal()  # Recursively get all the elements of the left subtree and add them into the list
        
        if self.right:
            elements += self.right.post_order_traversal()  # Recursively get all the elements of the right subtree and add them into the list
            
        elements.append(self.data) # Get the Root node element
        
        return elements
        
        
    def search_element(self, elem): # complexity of log n O(log n)
        if self.data == elem:
            return True
        elif elem < self.data:
            # This means if present, element would be on the left 
            if self.left:
               return self.left.search_element(elem)  
            else:
                return False
            
        else:
            # This means if present, element would be on the right
            if self.right:
                return self.right.search_element(elem)  
            else:
                return False
    
    
    def sum_of_all_elements_in_tree(self):
        return sum(self.in_order_traversal())
        
    def max_element_in_tree(self):
        return max(self.in_order_traversal())    
    
    def min_element_in_tree(self):
        return min(self.in_order_traversal())    
    
    
# Tree Builder helper method
def build_binary_tree(lst_elem: list):
    if len(lst_elem) >1:
        root_node = BinaryTreeNode(lst_elem[0])
        for i in lst_elem[1:]:
            root_node.add_child(i)
       
        #root_node.search_element(20)
        #print(root_node.in_order_traversal())
        return root_node
    else:
        return print("Insufficient number of elements")
        

if __name__ == '__main__':
   mt = build_binary_tree([17, -5, 4, 1, 20, 9, -1, 23, 18, 0, 34])
   print("In Order Traversal", mt.in_order_traversal())
   print("Post Order Traversal", mt.post_order_traversal())
   print("Pre Order Traversal", mt.pre_order_traversal())
   print(mt.search_element(20))
   print("Sum of all elemnts in tree", mt.sum_of_all_elements_in_tree())
   print("Max element in tree is", mt.max_element_in_tree())
   print("Min element in tree is", mt.min_element_in_tree())
Comment

Python code for binary search tree

# বাইনারি সার্চ ট্রি-এর ক্ষেত্রে ২ টি বিষয় মাথায় রাখতে হবে:
'''
১. যদি ট্রি-তে আগে থেকে কোনো নোড না থাকে (অর্থাৎ বর্তমান root নোড none থাকবে),
            তাহলে নতুন যোগ করা নোডটিই হবে ট্রি- এর root নোড । আবার-
২. নতুন নোডটি যদি root নোডের সরাসরি চাইল্ড হয়, তাহলেও root  নোডের  পরিবর্তন ঘটবে।
            এ কারণেই আমরা root নোডকে রিটার্ন করি।
'''
# There are two things to keep in mind when it comes to binary search trees:
'''
1. If the tree does not already have a node (ie the existing root node will have none), 
                then the newly added node will be the root node of the tree. Again
2. If the new node is a direct child of the root node, the root node will also change. 
                This is why we return to the root node.
'''
# 1st system:

class TreeNode:
    def __init__(self,data):
        self.data = data
        self.parent = None
        self.left = None
        self.right = None

    def __repr__(self):
        return repr(self.data)

    def add_left(self, node):
        self.left = node
        if node is not None:
            node.parent = self

    def add_right(self, node):
        self.right = node
        node.parent = self

# now bst_insert:
def bst_insert(root,node):
    last_node = None
    current_node = root
    while current_node is not None:
        last_node = current_node
        if node.data < current_node.data:
            current_node = current_node.left
        else:
            current_node = current_node.right
    if last_node is None:
        # tree was empty. node is the only node, hence root
        root = node # new node add
    elif node.data < last_node.data:
        last_node.add_left(node)
    else:
        last_node.add_right(node)
    return root
'''
              _10_
             /    
            5      17
           /      /  
          3      12  19
         /              
        1   4         
   
'''
# now create_bst:
def create_bst():
    li = list(map(int,input().split()))
    root = TreeNode(li)
    for item in root:
        node = TreeNode(item)
        root = bst_insert(root, node)
    return root

# In_order tree traverse:
def in_order(node):
    if node.left:
        in_order(node.left)
    print(node)
    if node.right:
        in_order(node.right)

# bst- tree minimum node:
def bst_minimum(root):
    while root.left is not None:
        root = root.left
    print(root)

# bst_tree maximum node:
def bst_maximum(root):
    while root.right is not None:
        root = root.right
    print(root)


# এখন আমরা BST-তে কোনো ডেটা খুঁজে বের করার ফাংশন টি লিখে ফেলি।
# Now we write the function to find any data in BST.
def best_search(node,key):
    while node is not None:
        if node.data == key:
            return node
        if key < node.data:
            node = node.left
        else:
            node = node.right
    return node

# Now check to your code:
if __name__ == "__main__":
    root = create_bst()
    print("Tree is root =",root)
    print()
    # In_order tree traverse:
    print("In_order Tree:")
    in_order(root)
    print()

    # bst- tree minimum node:
    print("Minimum node:")
    bst_minimum(root)
    print()

    # bst_tree maximum node:
    print("Maximum node:")
    bst_maximum(root)
    print()

    # input with searching:
    print("bst_search:")
    for key in [int(input("Please Enter the search key: ")),int(input("Please Enter the search key: "))]:
        print("Searching =",key)
        result = best_search(root, key)
        print("Result =",result)


Comment

python binary tree search

def search(node, search_item):

    # Base case for recursion:
    # The recursion will stop if the search item has been found
    if search_item == node.data:
        return True

    # Check if the search item is greater than the node data
    # and there is another node to the right to check
    elif search_item > node.data and node.right is not None:
        return search(node.right, search_item)

    # Check if the search item is less than the node data
    # and there is another node to the left to check
    elif search_item < node.data and node.left is not None:
        return search(node.left, search_item)

    # Otherwise the search item does not exist
    else:
        return False
Comment

binary search tree in python

Binary Search Tree at this link:
  
https://github.com/shreyasvedpathak/Data-Structure-Python/tree/master/BinaryTrees
Comment

implementation of binary search tree in python

# Python program to demonstrate
# insert operation in binary search tree
 
# A utility class that represents
# an individual node in a BST
 
 
class Node:
    def __init__(self, key):
        self.left = None
        self.right = None
        self.val = key
 
# A utility function to insert
# a new node with the given key
 
 
def insert(root, key):
    if root is None:
        return Node(key)
    else:
        if root.val == key:
            return root
        elif root.val < key:
            root.right = insert(root.right, key)
        else:
            root.left = insert(root.left, key)
    return root
 
# A utility function to do inorder tree traversal
 
 
def inorder(root):
    if root:
        inorder(root.left)
        print(root.val)
        inorder(root.right)
 
 
# Driver program to test the above functions
# Let us create the following BST
#    50
#  /     
# 30     70
#  /  / 
# 20 40 60 80
 
r = Node(50)
r = insert(r, 30)
r = insert(r, 20)
r = insert(r, 40)
r = insert(r, 70)
r = insert(r, 60)
r = insert(r, 80)
 
# Print inoder traversal of the BST
inorder(r)
Comment

Binary search tree in python

# বাইনারি সার্চ ট্রি-এর ক্ষেত্রে ২ টি বিষয় মাথায় রাখতে হবে:
'''
১. যদি ট্রি-তে আগে থেকে কোনো নোড না থাকে (অর্থাৎ বর্তমান root নোড none থাকবে),
            তাহলে নতুন যোগ করা নোডটিই হবে ট্রি- এর root নোড । আবার-
২. নতুন নোডটি যদি root নোডের সরাসরি চাইল্ড হয়, তাহলেও root  নোডের  পরিবর্তন ঘটবে।
            এ কারণেই আমরা root নোডকে রিটার্ন করি।
'''
# There are two things to keep in mind when it comes to binary search trees:
'''
1. If the tree does not already have a node (ie the existing root node will have none), 
                then the newly added node will be the root node of the tree. Again
2. If the new node is a direct child of the root node, the root node will also change. 
                This is why we return to the root node.
'''

# second system:

class TreeNode:
    def __init__(self,data):
        self.data = data
        self.parent = None
        self.left = None
        self.right = None

    def __repr__(self):
        return repr(self.data)

    def add_left(self, node):
        self.left = node
        if node is not None:
            node.parent = self

    def add_right(self, node):
        self.right = node
        node.parent = self

# now bst_insert:
def bst_insert(root,node):
    last_node = None
    current_node = root
    while current_node is not None:
        last_node = current_node
        if node.data < current_node.data:
            current_node = current_node.left
        else:
            current_node = current_node.right
    if last_node is None:
        # tree was empty. node is the only node, hence root
        root = node # new node add
    elif node.data < last_node.data:
        last_node.add_left(node)
    else:
        last_node.add_right(node)
    return root
'''
              _10_
             /    
            5      17
           /      /  
          3      12  19
         /              
        1   4         
   
'''
# now create_bst:
def create_bst():
    root = TreeNode(10)
    for item in [5,17,3,7,12,19,1,4]:
        node = TreeNode(item)
        root = bst_insert(root, node)
    return root

# In_order tree traverse:
def in_order(node):
    if node.left:
        in_order(node.left)
    print(node)
    if node.right:
        in_order(node.right)

# bst- tree minimum node:
def bst_minimum(root):
    while root.left is not None:
        root = root.left
    print(root)

# Node transfer:
def bst_transfer(root, current_node, new_node):
    if current_node.parent is None:
        root = new_node
    elif current_node == current_node.parent.left:
        current_node.parent.add_left(new_node)
    else:
        current_node.parent.add_right(new_node)
    return root

# Node delete:
def bst_delete(root,node):
    if node.left is None:
        root = bst_transfer(root,node,node.right)
    elif node.right is None:
        root = bst_transfer(root,node,node.left)

    else:
        min_node = bst_minimum(node.right)
        if min_node.parent != node:
            root = bst_transfer(root,min_node,min_node.right)
            min_node.add_right(node.right)
        root = bst_transfer(root,node,min_node)
        min_node.add_left(node.left)
    return root



# এখন আমরা BST-তে কোনো ডেটা খুঁজে বের করার ফাংশন টি লিখে ফেলি।
# Now we write the function to find any data in BST.
def best_search(node,key):
    while node is not None:
        if node.data == key:
            return node
        if key < node.data:
            node = node.left
        else:
            node = node.right
    return node

# Now check to your code:
if __name__ == "__main__":
    root = create_bst()
    print("Tree is root =",root)
    print()

    print("BST:")
    in_order(root)

    for key in [1,5,10]:
        node = best_search(root,key)
        print("will delete =", node)
        root = bst_delete(root,node)
        print("BST:")
        in_order(root)
Comment

PREVIOUS NEXT
Code Example
Python :: SciPy Convex Hull 
Python :: mulitplication symbo for unpacking in python 
Python :: numpy array from list 
Python :: tkinter delete toplevel 
Python :: python jwt 
Python :: python sys.argv 
Python :: matplotlive y axis 
Python :: readlines replace  
Python :: get country from city python 
Python :: python substring from end 
Python :: flask session timeout 
Python :: login required django 
Python :: training linear model sklearn 
Python :: python index max list 
Python :: counter +1 python 
Python :: flask flash 
Python :: create numpy array with ones 
Python :: estimate time to run a chunk of code in python 
Python :: wikipedia python module 
Python :: how to check encoding of csv 
Python :: tkinter toplevel 
Python :: python casting 
Python :: how to cut image python 
Python :: BURGERS2 
Python :: numpy timedelta object has no attribute days 
Python :: tkinter background image python 3 
Python :: record audio with processing python 
Python :: read parquet from s3 and convert to dataframe 
Python :: connect mongodb with python 
Python :: group by 2 columns pandas 
ADD CONTENT
Topic
Content
Source link
Name
7+1 =