SmartSearch — E-Commerce Product Search & Ranking Engine

A high-performance product search engine powered by a custom C++ DSA engine bridged to a Flask REST API with PostgreSQL persistence. Built for SDE portfolio demonstrations, tackling classic backend engineering challenges using custom data structures.

📌 Project Overview

SmartSearch solves three classic backend engineering challenges using custom data structures implemented in C++:

• Real-time autocomplete: Trie (O(L) per lookup)
• Top-10 best products ranking: Min-Heap (O(N log K))
• Avoid repeating expensive queries: LRU Cache - HashMap + Doubly Linked List (O(1) get/put)
• Filter products by price range: Binary Search on sorted array (O(log N))

🏗️ Architecture

Browser (HTML + CSS + Vanilla JS) → Flask REST API (Python) → C++ DSA Engine (engine2.dll) & PostgreSQL (SQLAlchemy)

Data flow for a search request:

1. User types "nike shoe" + price ₹2000–₹10000
2. Flask receives GET /api/search?q=nike+shoe&min_price=2000&max_price=10000
3. C++ Engine checks LRU Cache (HashMap lookup O(1))
   • CACHE HIT: return cached result instantly
   • CACHE MISS: Binary Search on price-sorted array O(log N) → substring filter on price-range slice → Min-Heap Top-10 by ranking score O(M log K) → store result in LRU Cache
4. Flask fetches full product details from PostgreSQL by IDs
5. Browser renders Top-10 cards + DSA Performance Panel

🧠 DSA Components Deep Dive

1. 🔍 Trie — Autocomplete

Each TrieNode stores up to 10 product IDs at every prefix. O(L) lookup where L = length of the typed prefix. Used for real-time autocomplete dropdown.

2. 🏆 Min-Heap — Top-K Ranking

Ranking Score = (Rating × 0.5) + (Popularity × 0.3) + (Relevance × 0.2) normalized to /10. Min-Heap of size K=10 keeps only the top-K candidates. Result: O(N log K) — far faster than sorting all N results.

3. ⚡ LRU Cache — O(1) Repeated Queries

HashMap (O(1) lookup) + Doubly Linked List (O(1) eviction) with a capacity of 100 cached queries. On every search: check HashMap → O(1) hit or miss. Cache key includes query + price range.

4. 🔢 Binary Search — Price Range Filter

Products pre-sorted by price at engine load time. Binary search finds the slice in O(log N). Min-Heap then ranks only the M products in that slice — O(M log K). When price filter is ON: complexity drops from O(N log K) → O(M log K) where M << N.

🔧 C++ ↔ Python Bridge (ctypes)

The C++ engine compiles to a Windows DLL (engine2.dll). Python loads it via ctypes. Products are passed from Python → C++ as a pipe-delimited CSV string on engine startup to avoid JSON parsing overhead when loading 1000+ products into C++.

🧪 Try It Live & View Code

Click here to launch the live Platform

View the complete codebase on GitHub

🔗 Connect with Me
🌐 www.tauqueeralam.com
📱 LinkedIn | GitHub

View the live demo below: