You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We are working on the next version of our Lance file format. The key improvements are null support in all types, columnar compression, and better performance with large blobs.
We want to make table maintenance automatic, if opted into. This would make it much easier to manage tables, since you would not need to manually call operations like compact, re-index, and cleanup.
We will work on native full text search indices for Lance. LanceDB's current implementation of FTS doesn't support incremental updates or working in Node or Rust. Nor does it work on object storage. Implementing FTS at the Lance level will remove these limitations.
Finally, there are smaller features we'd like to implement to improve queries. Providing a general SQL front-end for SELECT queries will provide native support for aggregations as well as provide an easy way to implement table views. Bitmap scalar indices will provide much better filter performance on columns with few distinct values.
Add a sql method to execute arbitrary SELECT statements
Register user-defined functions for use in projections and filters
Vector search
We continue to work at making vector search performant at very high scales (10 billion vectors).
V2 File Format
We are working on the next version of our Lance file format. The key improvements are null support in all types, columnar compression, and better performance with large blobs.
Table format
We want to make table maintenance automatic, if opted into. This would make it much easier to manage tables, since you would not need to manually call operations like compact, re-index, and cleanup.
Full text search
We will work on native full text search indices for Lance. LanceDB's current implementation of FTS doesn't support incremental updates or working in Node or Rust. Nor does it work on object storage. Implementing FTS at the Lance level will remove these limitations.
Read / write integration
To make Lance a great choice for AI data lakes, we will be adding connectors to two of the most popular distributed frameworks, Spark and Ray.
Query
Finally, there are smaller features we'd like to implement to improve queries. Providing a general SQL front-end for SELECT queries will provide native support for aggregations as well as provide an easy way to implement table views. Bitmap scalar indices will provide much better filter performance on columns with few distinct values.
sql
method to execute arbitrary SELECT statements_distance
column _distance filter pushdown #2096The text was updated successfully, but these errors were encountered: