Bump version: 0.3.3 → 0.3.4

expose index remap options in `compact_files`

.bumpversion.cfg

@@ -1,5 +1,5 @@
 [bumpversion]
-current_version = 0.3.3
+current_version = 0.3.4
 commit = True
 message = Bump version: {current_version} → {new_version}
 tag = True

Cargo.toml

@@ -5,9 +5,9 @@ exclude = ["python"]
 resolver = "2"
 
 [workspace.dependencies]
-lance = { "version" = "=0.8.7", "features" = ["dynamodb"] }
-lance-linalg = { "version" = "=0.8.7" }
-lance-testing = { "version" = "=0.8.7" }
+lance = { "version" = "=0.8.8", "features" = ["dynamodb"] }
+lance-linalg = { "version" = "=0.8.8" }
+lance-testing = { "version" = "=0.8.8" }
 # Note that this one does not include pyarrow
 arrow = { version = "47.0.0", optional = false }
 arrow-array = "47.0"
@@ -19,7 +19,7 @@ arrow-arith = "47.0"
 arrow-cast = "47.0"
 chrono = "0.4.23"
 half = { "version" = "=2.3.1", default-features = false, features = [
-    "num-traits"
+    "num-traits",
 ] }
 log = "0.4"
 object_store = "0.7.1"

node/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vectordb",
-  "version": "0.3.3",
+  "version": "0.3.4",
   "description": " Serverless, low-latency vector database for AI applications",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
@@ -81,10 +81,10 @@
     }
   },
   "optionalDependencies": {
-    "@lancedb/vectordb-darwin-arm64": "0.3.3",
-    "@lancedb/vectordb-darwin-x64": "0.3.3",
-    "@lancedb/vectordb-linux-arm64-gnu": "0.3.3",
-    "@lancedb/vectordb-linux-x64-gnu": "0.3.3",
-    "@lancedb/vectordb-win32-x64-msvc": "0.3.3"
+    "@lancedb/vectordb-darwin-arm64": "0.3.4",
+    "@lancedb/vectordb-darwin-x64": "0.3.4",
+    "@lancedb/vectordb-linux-arm64-gnu": "0.3.4",
+    "@lancedb/vectordb-linux-x64-gnu": "0.3.4",
+    "@lancedb/vectordb-win32-x64-msvc": "0.3.4"
   }
 }

node/src/index.ts

@@ -23,7 +23,7 @@ import { Query } from './query'
 import { isEmbeddingFunction } from './embedding/embedding_function'
 
 // eslint-disable-next-line @typescript-eslint/no-var-requires
-const { databaseNew, databaseTableNames, databaseOpenTable, databaseDropTable, tableCreate, tableAdd, tableCreateVectorIndex, tableCountRows, tableDelete, tableCleanupOldVersions, tableCompactFiles } = require('../native.js')
+const { databaseNew, databaseTableNames, databaseOpenTable, databaseDropTable, tableCreate, tableAdd, tableCreateVectorIndex, tableCountRows, tableDelete, tableCleanupOldVersions, tableCompactFiles, tableListIndices, tableIndexStats } = require('../native.js')
 
 export { Query }
 export type { EmbeddingFunction }
@@ -260,6 +260,27 @@ export interface Table<T = number[]> {
    * ```
    */
   delete: (filter: string) => Promise<void>
+
+  /**
+   * List the indicies on this table.
+   */
+  listIndices: () => Promise<VectorIndex[]>
+
+  /**
+   * Get statistics about an index.
+   */
+  indexStats: (indexUuid: string) => Promise<IndexStats>
+}
+
+export interface VectorIndex {
+  columns: string[]
+  name: string
+  uuid: string
+}
+
+export interface IndexStats {
+  numIndexedRows: number | null
+  numUnindexedRows: number | null
 }
 
 /**
@@ -502,6 +523,14 @@ export class LocalTable<T = number[]> implements Table<T> {
         return res.metrics
       })
   }
+
+  async listIndices (): Promise<VectorIndex[]> {
+    return tableListIndices.call(this._tbl)
+  }
+
+  async indexStats (indexUuid: string): Promise<IndexStats> {
+    return tableIndexStats.call(this._tbl, indexUuid)
+  }
 }
 
 export interface CleanupStats {

node/src/remote/index.ts

@@ -14,7 +14,9 @@
 
 import {
   type EmbeddingFunction, type Table, type VectorIndexParams, type Connection,
-  type ConnectionOptions, type CreateTableOptions, type WriteOptions
+  type ConnectionOptions, type CreateTableOptions, type VectorIndex,
+  type WriteOptions,
+  type IndexStats
 } from '../index'
 import { Query } from '../query'
 
@@ -241,4 +243,21 @@ export class RemoteTable<T = number[]> implements Table<T> {
   async delete (filter: string): Promise<void> {
     await this._client.post(`/v1/table/${this._name}/delete/`, { predicate: filter })
   }
+
+  async listIndices (): Promise<VectorIndex[]> {
+    const results = await this._client.post(`/v1/table/${this._name}/index/list/`)
+    return results.data.indexes?.map((index: any) => ({
+      columns: index.columns,
+      name: index.index_name,
+      uuid: index.index_uuid
+    }))
+  }
+
+  async indexStats (indexUuid: string): Promise<IndexStats> {
+    const results = await this._client.post(`/v1/table/${this._name}/index/${indexUuid}/stats/`)
+    return {
+      numIndexedRows: results.data.num_indexed_rows,
+      numUnindexedRows: results.data.num_unindexed_rows
+    }
+  }
 }

node/src/test/test.ts

@@ -328,6 +328,24 @@ describe('LanceDB client', function () {
       const createIndex = table.createIndex({ type: 'ivf_pq', column: 'name', num_partitions: -1, max_iters: 2, num_sub_vectors: 2 })
       await expect(createIndex).to.be.rejectedWith('num_partitions: must be > 0')
     })
+
+    it('should be able to list index and stats', async function () {
+      const uri = await createTestDB(32, 300)
+      const con = await lancedb.connect(uri)
+      const table = await con.openTable('vectors')
+      await table.createIndex({ type: 'ivf_pq', column: 'vector', num_partitions: 2, max_iters: 2, num_sub_vectors: 2 })
+
+      const indices = await table.listIndices()
+      expect(indices).to.have.lengthOf(1)
+      expect(indices[0].name).to.equal('vector_idx')
+      expect(indices[0].uuid).to.not.be.equal(undefined)
+      expect(indices[0].columns).to.have.lengthOf(1)
+      expect(indices[0].columns[0]).to.equal('vector')
+
+      const stats = await table.indexStats(indices[0].uuid)
+      expect(stats.numIndexedRows).to.equal(300)
+      expect(stats.numUnindexedRows).to.equal(0)
+    }).timeout(50_000)
   })
 
   describe('when using a custom embedding function', function () {

rust/ffi/node/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "vectordb-node"
-version = "0.3.3"
+version = "0.3.4"
 description = "Serverless, low-latency vector database for AI applications"
 license = "Apache-2.0"
 edition = "2018"
@@ -14,7 +14,7 @@ arrow-array = { workspace = true }
 arrow-ipc = { workspace = true }
 arrow-schema = { workspace = true }
 chrono = { workspace = true }
-conv = "0.3.3"
+conv = "0.3.4"
 once_cell = "1"
 futures = "0.3"
 half = { workspace = true }

rust/ffi/node/src/lib.rs

@@ -239,6 +239,8 @@ fn main(mut cx: ModuleContext) -> NeonResult<()> {
     cx.export_function("tableDelete", JsTable::js_delete)?;
     cx.export_function("tableCleanupOldVersions", JsTable::js_cleanup)?;
     cx.export_function("tableCompactFiles", JsTable::js_compact)?;
+    cx.export_function("tableListIndices", JsTable::js_list_indices)?;
+    cx.export_function("tableIndexStats", JsTable::js_index_stats)?;
     cx.export_function(
         "tableCreateVectorIndex",
         index::vector::table_create_vector_index,

rust/ffi/node/src/table.rs

@@ -247,7 +247,7 @@ impl JsTable {
         }
 
         rt.spawn(async move {
-            let stats = table.compact_files(options).await;
+            let stats = table.compact_files(options, None).await;
 
             deferred.settle_with(&channel, move |mut cx| {
                 let stats = stats.or_throw(&mut cx)?;
@@ -276,4 +276,91 @@ impl JsTable {
         });
         Ok(promise)
     }
+
+    pub(crate) fn js_list_indices(mut cx: FunctionContext) -> JsResult<JsPromise> {
+        let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
+        let rt = runtime(&mut cx)?;
+        let (deferred, promise) = cx.promise();
+        // let predicate = cx.argument::<JsString>(0)?.value(&mut cx);
+        let channel = cx.channel();
+        let table = js_table.table.clone();
+
+        rt.spawn(async move {
+            let indices = table.load_indices().await;
+
+            deferred.settle_with(&channel, move |mut cx| {
+                let indices = indices.or_throw(&mut cx)?;
+
+                let output = JsArray::new(&mut cx, indices.len() as u32);
+                for (i, index) in indices.iter().enumerate() {
+                    let js_index = JsObject::new(&mut cx);
+                    let index_name = cx.string(index.index_name.clone());
+                    js_index.set(&mut cx, "name", index_name)?;
+
+                    let index_uuid = cx.string(index.index_uuid.clone());
+                    js_index.set(&mut cx, "uuid", index_uuid)?;
+
+                    let js_index_columns = JsArray::new(&mut cx, index.columns.len() as u32);
+                    for (j, column) in index.columns.iter().enumerate() {
+                        let js_column = cx.string(column.clone());
+                        js_index_columns.set(&mut cx, j as u32, js_column)?;
+                    }
+                    js_index.set(&mut cx, "columns", js_index_columns)?;
+
+                    output.set(&mut cx, i as u32, js_index)?;
+                }
+
+                Ok(output)
+            })
+        });
+        Ok(promise)
+    }
+
+    pub(crate) fn js_index_stats(mut cx: FunctionContext) -> JsResult<JsPromise> {
+        let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
+        let rt = runtime(&mut cx)?;
+        let (deferred, promise) = cx.promise();
+        let index_uuid = cx.argument::<JsString>(0)?.value(&mut cx);
+        let channel = cx.channel();
+        let table = js_table.table.clone();
+
+        rt.spawn(async move {
+            let load_stats = futures::try_join!(
+                table.count_indexed_rows(&index_uuid),
+                table.count_unindexed_rows(&index_uuid)
+            );
+
+            deferred.settle_with(&channel, move |mut cx| {
+                let (indexed_rows, unindexed_rows) = load_stats.or_throw(&mut cx)?;
+
+                let output = JsObject::new(&mut cx);
+
+                match indexed_rows {
+                    Some(x) => {
+                        let i = cx.number(x as f64);
+                        output.set(&mut cx, "numIndexedRows", i)?;
+                    }
+                    None => {
+                        let null = cx.null();
+                        output.set(&mut cx, "numIndexedRows", null)?;
+                    }
+                };
+
+                match unindexed_rows {
+                    Some(x) => {
+                        let i = cx.number(x as f64);
+                        output.set(&mut cx, "numUnindexedRows", i)?;
+                    }
+                    None => {
+                        let null = cx.null();
+                        output.set(&mut cx, "numUnindexedRows", null)?;
+                    }
+                };
+
+                Ok(output)
+            })
+        });
+
+        Ok(promise)
+    }
 }

rust/vectordb/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "vectordb"
-version = "0.3.3"
+version = "0.3.4"
 edition = "2021"
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license = "Apache-2.0"

rust/vectordb/src/index/vector.rs

@@ -12,6 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use lance::format::{Index, Manifest};
 use lance::index::vector::ivf::IvfBuildParams;
 use lance::index::vector::pq::PQBuildParams;
 use lance::index::vector::VectorIndexParams;
@@ -106,6 +107,27 @@ impl VectorIndexBuilder for IvfPQIndexBuilder {
     }
 }
 
+pub struct VectorIndex {
+    pub columns: Vec<String>,
+    pub index_name: String,
+    pub index_uuid: String,
+}
+
+impl VectorIndex {
+    pub fn new_from_format(manifest: &Manifest, index: &Index) -> VectorIndex {
+        let fields = index
+            .fields
+            .iter()
+            .map(|i| manifest.schema.fields[*i as usize].name.clone())
+            .collect();
+        VectorIndex {
+            columns: fields,
+            index_name: index.name.clone(),
+            index_uuid: index.uuid.to_string(),
+        }
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

rust/vectordb/src/table.rs

@@ -18,14 +18,16 @@ use std::sync::Arc;
 use arrow_array::{Float32Array, RecordBatchReader};
 use arrow_schema::SchemaRef;
 use lance::dataset::cleanup::RemovalStats;
-use lance::dataset::optimize::{compact_files, CompactionMetrics, CompactionOptions};
+use lance::dataset::optimize::{
+    compact_files, CompactionMetrics, CompactionOptions, IndexRemapperOptions,
+};
 use lance::dataset::{Dataset, WriteParams};
-use lance::index::IndexType;
+use lance::index::{DatasetIndexExt, IndexType};
 use lance::io::object_store::WrappingObjectStore;
 use std::path::Path;
 
 use crate::error::{Error, Result};
-use crate::index::vector::VectorIndexBuilder;
+use crate::index::vector::{VectorIndexBuilder, VectorIndex};
 use crate::query::Query;
 use crate::utils::{PatchReadParam, PatchWriteParam};
 use crate::WriteMode;
@@ -238,8 +240,6 @@ impl Table {
 
     /// Create index on the table.
     pub async fn create_index(&mut self, index_builder: &impl VectorIndexBuilder) -> Result<()> {
-        use lance::index::DatasetIndexExt;
-
         let mut dataset = self.dataset.as_ref().clone();
         dataset
             .create_index(
@@ -257,6 +257,14 @@ impl Table {
         Ok(())
     }
 
+    pub async fn optimize_indices(&mut self) -> Result<()> {
+        let mut dataset = self.dataset.as_ref().clone();
+
+        dataset.optimize_indices().await?;
+
+        Ok(())
+    }
+
     /// Insert records into this Table
     ///
     /// # Arguments
@@ -353,12 +361,44 @@ impl Table {
     /// for faster reads.
     ///
     /// This calls into [lance::dataset::optimize::compact_files].
-    pub async fn compact_files(&mut self, options: CompactionOptions) -> Result<CompactionMetrics> {
+    pub async fn compact_files(
+        &mut self,
+        options: CompactionOptions,
+        remap_options: Option<Arc<dyn IndexRemapperOptions>>,
+    ) -> Result<CompactionMetrics> {
         let mut dataset = self.dataset.as_ref().clone();
-        let metrics = compact_files(&mut dataset, options, None).await?;
+        let metrics = compact_files(&mut dataset, options, remap_options).await?;
         self.dataset = Arc::new(dataset);
         Ok(metrics)
     }
+
+    pub fn count_fragments(&self) -> usize {
+        self.dataset.count_fragments()
+    }
+
+    pub fn count_deleted_rows(&self) -> usize {
+        self.dataset.count_deleted_rows()
+    }
+
+    pub fn num_small_files(&self, max_rows_per_group: usize) -> usize {
+        self.dataset.num_small_files(max_rows_per_group)
+    }
+
+    pub async fn count_indexed_rows(&self, index_uuid: &str) -> Result<Option<usize>> {
+        Ok(self.dataset.count_indexed_rows(index_uuid).await?)
+    }
+
+    pub async fn count_unindexed_rows(&self, index_uuid: &str) -> Result<Option<usize>> {
+        Ok(self.dataset.count_unindexed_rows(index_uuid).await?)
+    }
+
+    pub async fn load_indices(&self) -> Result<Vec<VectorIndex>> {
+        let (indices, mf) = futures::try_join!(
+            self.dataset.load_indices(),
+            self.dataset.latest_manifest()
+        )?;
+        Ok(indices.iter().map(|i| VectorIndex::new_from_format(&mf, i)).collect())
+    }
 }
 
 #[cfg(test)]