Inside the Algolia Engine: Index w/ Instant Search | Algolia

Search engine usage has evolved significantly in recent years, with users now expecting instant feedback and tolerance for typing mistakes without validation proposals. This has led to changes in the indexing process of search engines like Algolia, which focuses on optimizing indexing speed and search performance. The goal of indexing is to store data in a specific data-structure that is efficient for search operations. One approach to instant search involves creating an inverted list for all prefixes of a word instead of just the word itself. Another approach keeps the indexing process light by keeping only one inverted list per word but requires processing at query time. Algolia's unique approach involves building a radix tree on-disk directly from the list of words, using depth-first search and flushing nodes as soon as possible to save memory. This method allows for efficient prefix search and has minimal impact on indexing performance.