Indexes
Singdata Lakehouse supports multiple index types for accelerating query filtering in different scenarios and reducing data scan volume.
Index Type Comparison
| Index Type | Applicable Queries | Applicable Fields | Typical Scenario |
|---|---|---|---|
| Bloomfilter Index | Equality queries (=, IN) | High-cardinality fields, such as user ID, order number | Quickly skip data files that do not contain the target value |
| Inverted Index | Full-text search (MATCH), keyword search | Text fields, JSON fields | Log search, document search, multi-keyword filtering |
| Vector Index | Approximate Nearest Neighbor search (ANN) | VECTOR type fields | Semantic search, image similarity, RAG retrieval |
Selection Advice
- The field is a high-cardinality ID field with frequent equality queries --> Bloomfilter Index
- The field is text content requiring keyword or phrase search --> Inverted Index
- The field is a vector embedding requiring similarity search --> Vector Index
- Unsure which to use --> Refer to Index Best Practices