What is this?
Tantivy is a full-text search engine library written in Rust, inspired by Apache Lucene, used to build search functionality directly into applications.
Tantivy is a full-text search engine library written in Rust, inspired by Apache Lucene, used to build search functionality directly into applications.Is this a standalone search server like Elasticsearch, or something else?It is a library (crate), not an off-the-shelf server. If you need a ready-made distributed search server, the README points to Quickwit, which is built on top of Tantivy.
It is a library (crate), not an off-the-shelf server. If you need a ready-made distributed search server, the README points to Quickwit, which is built on top of Tantivy.How do I add it to a Rust project?Add tantivy as a dependency via Cargo, the standard Rust crate manager, and use its IndexWriter/IndexReader API to build and query an index.
Add tantivy as a dependency via Cargo, the standard Rust crate manager, and use its IndexWriter/IndexReader API to build and query an index.Do I need Rust toolchain and Cargo to use it?Yes, since it's distributed as a Rust crate, you need Rust and Cargo to build and depend on it.
Yes, since it's distributed as a Rust crate, you need Rust and Cargo to build and depend on it.Can I edit documents once they're indexed?No, data in Tantivy is immutable; to change a document you must delete it and reindex it, and changes only become visible after commit and reloading readers.
No, data in Tantivy is immutable; to change a document you must delete it and reindex it, and changes only become visible after commit and reloading readers.Can I use Tantivy from Python or Ruby instead of Rust?Yes, there are bindings: tantivy-py for Python and tantiny for Ruby, with other community bindings available on GitHub.
Yes, there are bindings: tantivy-py for Python and tantiny for Ruby, with other community bindings available on GitHub.Does Tantivy support non-Latin language tokenization?Yes, besides built-in stemming for 17 Latin languages, there is third-party tokenizer support for Chinese (tantivy-jieba, cang-jie), Japanese (lindera, Vaporetto, tantivy-tokenizer-tiny-segmenter), and Korean (lindera + lindera-ko-dic-builder).
Yes, besides built-in stemming for 17 Latin languages, there is third-party tokenizer support for Chinese (tantivy-jieba, cang-jie), Japanese (lindera, Vaporetto, tantivy-tokenizer-tiny-segmenter), and Korean (lindera + lindera-ko-dic-builder).How does Tantivy's performance compare to Lucene?According to the project's search latency benchmark, Tantivy is approximately 2x faster than Lucene.
According to the project's search latency benchmark, Tantivy is approximately 2x faster than Lucene.Does Tantivy support updating an index after initial creation?Yes, it supports incremental indexing, so new documents can be added to an existing index over time.
Yes, it supports incremental indexing, so new documents can be added to an existing index over time.What kinds of projects have been built with Tantivy?Examples include seshat (a Matrix message database/indexer), tantiny (Ruby search), lnx (a typo-tolerant search engine with a REST API), and Bichon (a Rust email archiver with a web UI).
Examples include seshat (a Matrix message database/indexer), tantiny (Ruby search), lnx (a typo-tolerant search engine with a REST API), and Bichon (a Rust email archiver with a web UI).What are similar projects?bleve, php-fts, fsearch, sonic, xerj, flexsearch, meilisearch, hotdata-cli
bleve, php-fts, fsearch, sonic, xerj, flexsearch, meilisearch, hotdata-cliWhich topics is it in?full-text-search, search-engine, search, multilingual, nlp, database, performance, rust, command-line-tool, api, library, developer-tools
full-text-search, search-engine, search, multilingual, nlp, database, performance, rust, command-line-tool, api, library, developer-tools