feat: support to sepcify ef search param (#1844)

Signed-off-by: BubbleCal <bubble-cal@outlook.com>

nodejs/__test__/table.test.ts

@@ -477,6 +477,54 @@ describe("When creating an index", () => {
     expect(rst.numRows).toBe(1);
   });
 
+  it("should create and search IVF_HNSW indices", async () => {
+    await tbl.createIndex("vec", {
+      config: Index.hnswSq(),
+    });
+
+    // check index directory
+    const indexDir = path.join(tmpDir.name, "test.lance", "_indices");
+    expect(fs.readdirSync(indexDir)).toHaveLength(1);
+    const indices = await tbl.listIndices();
+    expect(indices.length).toBe(1);
+    expect(indices[0]).toEqual({
+      name: "vec_idx",
+      indexType: "IvfHnswSq",
+      columns: ["vec"],
+    });
+
+    // Search without specifying the column
+    let rst = await tbl
+      .query()
+      .limit(2)
+      .nearestTo(queryVec)
+      .distanceType("dot")
+      .toArrow();
+    expect(rst.numRows).toBe(2);
+
+    // Search using `vectorSearch`
+    rst = await tbl.vectorSearch(queryVec).limit(2).toArrow();
+    expect(rst.numRows).toBe(2);
+
+    // Search with specifying the column
+    const rst2 = await tbl
+      .query()
+      .limit(2)
+      .nearestTo(queryVec)
+      .column("vec")
+      .toArrow();
+    expect(rst2.numRows).toBe(2);
+    expect(rst.toString()).toEqual(rst2.toString());
+
+    // test offset
+    rst = await tbl.query().limit(2).offset(1).nearestTo(queryVec).toArrow();
+    expect(rst.numRows).toBe(1);
+
+    // test ef
+    rst = await tbl.query().limit(2).nearestTo(queryVec).ef(100).toArrow();
+    expect(rst.numRows).toBe(2);
+  });
+
   it("should be able to query unindexed data", async () => {
     await tbl.createIndex("vec");
     await tbl.add([

nodejs/lancedb/query.ts

@@ -385,6 +385,20 @@ export class VectorQuery extends QueryBase<NativeVectorQuery> {
     return this;
   }
 
+  /**
+   * Set the number of candidates to consider during the search
+   *
+   * This argument is only used when the vector column has an HNSW index.
+   * If there is no index then this value is ignored.
+   *
+   * Increasing this value will increase the recall of your query but will
+   * also increase the latency of your query. The default value is 1.5*limit.
+   */
+  ef(ef: number): VectorQuery {
+    super.doCall((inner) => inner.ef(ef));
+    return this;
+  }
+
   /**
    * Set the vector column to query
    *

nodejs/src/query.rs

@@ -167,6 +167,11 @@ impl VectorQuery {
         self.inner = self.inner.clone().nprobes(nprobe as usize);
     }
 
+    #[napi]
+    pub fn ef(&mut self, ef: u32) {
+        self.inner = self.inner.clone().ef(ef as usize);
+    }
+
     #[napi]
     pub fn bypass_vector_index(&mut self) {
         self.inner = self.inner.clone().bypass_vector_index()