How to implement text Auto-complete feature using Ternary Search Tree
Given a set of strings S and a string patt the task is to autocomplete the string patt to strings from S that have patt as a prefix, using a Ternary Search Tree. If no string matches the given prefix, print “None”.
Examples:
Input: S = {“wallstreet”, “geeksforgeeks”, “wallmart”, “walmart”, “waldomort”, “word”],
patt = “wall”
Output:
wallstreet
wallmart
Explanation:
Only two strings {“wallstreet”, “wallmart”} from S matches with the given pattern “wall”.Input: S = {“word”, “wallstreet”, “wall”, “wallmart”, “walmart”, “waldo”, “won”}, patt = “geeks”
Output: None
Explanation:
Since none of word of set matches with pattern so empty list will be printed.
Trie Approach: Please refer this article to learn about the implementation using Trie data structure.
Ternary Search Tree Approach Follow the steps below to solve the problem:
- Insert all the characters of the strings in S into the Ternary Search Tree based on the following conditions:
- If the character to be inserted is smaller than current node value, traverse the left subtree.
- If the character to be inserted is greater than current node value, traverse the right subtree to.
- If the character to be inserted is same as that of the current node value, traverse the equal subtree if it is not the end of the word. If so, mark the node as the end of the word.
- Follow a similar approach for extracting suggestions.
- Traverse the tree to search the given prefix patt following a similar traversal technique as mentioned above.
- If the given prefix is not found, print “None”.
- If the given prefix is found, traverse the tree from the node where the prefix ends. Traverse the left subtree and generate suggestions followed by the right and equal subtrees from every node .
- Every time a node is encountered which has the endofWord variable set, it denotes that a suggestion has been obtained. Insert that suggestion into words.
- Return words after generating all possible suggestions.
Below is the implementation of the above approach:
C++
// C++ Program to generate // autocompleted texts from // a given prefix using a // Ternary Search Tree #include <bits/stdc++.h> using namespace std; // Define the Node of the // tree struct Node { // Store the character // of a string char data; // Store the end of // word int end; // Left Subtree struct Node* left; // Equal Subtree struct Node* eq; // Right Subtree struct Node* right; }; // Function to create a Node Node* createNode(char newData) { struct Node* newNode = new Node(); newNode->data = newData; newNode->end = 0; newNode->left = NULL; newNode->eq = NULL; newNode->right = NULL; return newNode; } // Function to insert a word // in the tree void insert(Node** root, string word, int pos = 0) { // Base case if (!(*root)) *root = createNode(word[pos]); // If the current character is // less than root's data, then // it is inserted in the // left subtree if ((*root)->data > word[pos]) insert(&((*root)->left), word, pos); // If current character is // more than root's data, then // it is inserted in the right // subtree else if ((*root)->data < word[pos]) insert(&((*root)->right), word, pos); // If current character is same // as that of the root's data else { // If it is the end of word if (pos + 1 == word.size()) // Mark it as the // end of word (*root)->end = 1; // If it is not the end of // the string, then the // current character is // inserted in the equal subtree else insert(&((*root)->eq), word, pos + 1); } } // Function to traverse the ternary search tree void traverse(Node* root, vector<string>& ret, char* buff, int depth = 0) { // Base case if (!root) return; // The left subtree is // traversed first traverse(root->left, ret, buff, depth); // Store the current character buff[depth] = root->data; // If the end of the string // is detected, store it in // the final ans if (root->end) { buff[depth + 1] = '\0'; ret.push_back(string(buff)); } // Traverse the equal subtree traverse(root->eq, ret, buff, depth + 1); // Traverse the right subtree traverse(root->right, ret, buff, depth); } // Utility function to find // all the words vector<string> util(Node* root, string pattern) { // Stores the words // to suggest char buffer[1001]; vector<string> ret; traverse(root, ret, buffer); if (root->end == 1) ret.push_back(pattern); return ret; } // Function to autocomplete // based on the given prefix // and return the suggestions vector<string> autocomplete(Node* root, string pattern) { vector<string> words; int pos = 0; // If pattern is empty // return an empty list if (pattern.empty()) return words; // Iterating over the characters // of the pattern and find it's // corresponding node in the tree while (root && pos < pattern.length()) { // If current character is smaller if (root->data > pattern[pos]) // Search the left subtree root = root->left; // current character is greater else if (root->data < pattern[pos]) // Search right subtree root = root->right; // If current character is equal else if (root->data == pattern[pos]) // Search equal subtree root = root->eq; // If not found else return words; pos++; } // Search for all the words // from the current node words = util(root, pattern); return words; } // Function to print // suggested words void print(vector<string> sugg, string pat) { for (int i = 0; i < sugg.size(); i++) cout << pat << sugg[i].c_str() << "\n"; } // Driver Code int main() { vector<string> S = { "wallstreet", "geeksforgeeks", "wallmart", "walmart", "waldormort", "word" }; Node* tree = NULL; // Insert the words in the // Ternary Search Tree for (string str : S) insert(&tree, str); string pat = "wall"; vector<string> sugg = autocomplete(tree, pat); if (sugg.size() == 0) cout << "None"; else print(sugg, pat); return 0; } |
chevron_right
filter_none
Output:
wallmart wallstreet
Time Complexity: O(L* log N) where L is length of longest word.
The space is proportional to the length of the string to be stored.
