chore: merge main into blob concurrency repair

Preserve descriptor-based blob sizing and version pinning together with shared request permits and finite cursor ranges. Adapt probe-only regression tests to the upstream probe helper.

nodejs/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.40.0-beta.11"
+version = "0.40.0-beta.12"
 publish = false
 license.workspace = true
 description.workspace = true

nodejs/NODEJS_THIRD_PARTY_LICENSES.md

@@ -275,6 +275,7 @@
 [@types/node@20.16.10](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
 [@types/node@20.17.9](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
 [@types/node@22.7.4](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/node@25.9.8](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
 [@types/semver@7.5.6](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
 [@types/stack-utils@2.0.3](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
 [@types/tmp@0.2.6](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
@@ -306,6 +307,9 @@
 [apache-arrow@16.0.0](https://github.com/apache/arrow) - Apache-2.0
 [apache-arrow@17.0.0](https://github.com/apache/arrow) - Apache-2.0
 [apache-arrow@18.0.0](https://github.com/apache/arrow) - Apache-2.0
+[apache-arrow@19.0.1](https://github.com/apache/arrow) - Apache-2.0
+[apache-arrow@20.0.0](https://github.com/apache/arrow) - Apache-2.0
+[apache-arrow@21.2.0](https://github.com/apache/arrow-js) - Apache-2.0
 [argparse@1.0.10](https://github.com/nodeca/argparse) - MIT
 [argparse@2.0.1](https://github.com/nodeca/argparse) - Python-2.0
 [array-back@3.1.0](https://github.com/75lb/array-back) - MIT
@@ -348,6 +352,7 @@
 [color@4.2.3](https://github.com/Qix-/color) - MIT
 [combined-stream@1.0.8](https://github.com/felixge/node-combined-stream) - MIT
 [command-line-args@5.2.1](https://github.com/75lb/command-line-args) - MIT
+[command-line-args@6.0.2](https://github.com/75lb/command-line-args) - MIT
 [command-line-usage@7.0.1](https://github.com/75lb/command-line-usage) - MIT
 [concat-map@0.0.1](https://github.com/substack/node-concat-map) - MIT
 [convert-source-map@2.0.0](https://github.com/thlorenz/convert-source-map) - MIT
@@ -398,12 +403,14 @@
 [file-entry-cache@6.0.1](https://github.com/royriojas/file-entry-cache) - MIT
 [fill-range@7.1.1](https://github.com/jonschlinkert/fill-range) - MIT
 [find-replace@3.0.0](https://github.com/75lb/find-replace) - MIT
+[find-replace@5.0.2](https://github.com/75lb/find-replace) - MIT
 [find-up@4.1.0](https://github.com/sindresorhus/find-up) - MIT
 [find-up@5.0.0](https://github.com/sindresorhus/find-up) - MIT
 [flat-cache@3.2.0](https://github.com/jaredwray/flat-cache) - MIT
 [flatbuffers@1.12.0](https://github.com/google/flatbuffers) - Apache*
 [flatbuffers@23.5.26](https://github.com/google/flatbuffers) - Apache*
 [flatbuffers@24.3.25](https://github.com/google/flatbuffers) - Apache-2.0
+[flatbuffers@25.9.23](https://github.com/google/flatbuffers) - Apache-2.0
 [flatted@3.2.9](https://github.com/WebReflection/flatted) - ISC
 [follow-redirects@1.15.6](https://github.com/follow-redirects/follow-redirects) - MIT
 [foreground-child@3.3.0](https://github.com/tapjs/foreground-child) - ISC
@@ -644,6 +651,7 @@
 [uc.micro@2.1.0](https://github.com/markdown-it/uc.micro) - MIT
 [undici-types@5.26.5](https://github.com/nodejs/undici) - MIT
 [undici-types@6.19.8](https://github.com/nodejs/undici) - MIT
+[undici-types@7.24.6](https://github.com/nodejs/undici) - MIT
 [update-browserslist-db@1.0.13](https://github.com/browserslist/update-db) - MIT
 [uri-js@4.4.1](https://github.com/garycourt/uri-js) - BSD-2-Clause
 [uuid@9.0.1](https://github.com/uuidjs/uuid) - MIT

nodejs/__test__/arrow_versions.ts

@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+// Every Apache Arrow release the package accepts as a peer dependency, as
+// separately installed copies so the tests can feed objects built by one
+// Arrow version into a LanceDB running on another.  When a new Arrow major
+// is released, add the alias to `package.json` and extend this list; the
+// test files pick the change up from here.
+import * as arrow15 from "apache-arrow-15";
+import * as arrow16 from "apache-arrow-16";
+import * as arrow17 from "apache-arrow-17";
+import * as arrow18 from "apache-arrow-18";
+import * as arrow19 from "apache-arrow-19";
+import * as arrow20 from "apache-arrow-20";
+import * as arrow21 from "apache-arrow-21";
+
+export { arrow15, arrow16, arrow17, arrow18, arrow19, arrow20, arrow21 };
+
+export type ApacheArrow =
+  | typeof arrow15
+  | typeof arrow16
+  | typeof arrow17
+  | typeof arrow18
+  | typeof arrow19
+  | typeof arrow20
+  | typeof arrow21;
+
+export const arrowVersions: ApacheArrow[] = [
+  arrow15,
+  arrow16,
+  arrow17,
+  arrow18,
+  arrow19,
+  arrow20,
+  arrow21,
+];
+
+/** The newest supported Arrow release, for tests that only need one. */
+export const latestArrow = arrow21;

nodejs/__test__/registry.test.ts

@@ -3,13 +3,10 @@
 
 import * as apiArrow from "apache-arrow";
 
-import * as arrow15 from "apache-arrow-15";
-import * as arrow16 from "apache-arrow-16";
-import * as arrow17 from "apache-arrow-17";
-import * as arrow18 from "apache-arrow-18";
-
 import * as tmp from "tmp";
 
+import { arrowVersions } from "./arrow_versions";
+
 import { connect } from "../lancedb";
 import {
   EmbeddingFunction,
@@ -18,7 +15,7 @@ import {
 } from "../lancedb/embedding";
 import { getRegistry, register } from "../lancedb/embedding/registry";
 
-describe.each([arrow15, arrow16, arrow17, arrow18])("LanceSchema", (arrow) => {
+describe.each(arrowVersions)("LanceSchema", (arrow) => {
   test("should preserve input order", async () => {
     const schema = LanceSchema({
       id: new arrow.Int32(),
@@ -29,7 +26,7 @@ describe.each([arrow15, arrow16, arrow17, arrow18])("LanceSchema", (arrow) => {
   });
 });
 
-describe.each([arrow15, arrow16, arrow17, arrow18])("Registry", (arrow) => {
+describe.each(arrowVersions)("Registry", (arrow) => {
   let tmpDir: tmp.DirResult;
   beforeEach(() => {
     tmpDir = tmp.dirSync({ unsafeCleanup: true });
@@ -240,7 +237,7 @@ describe("Registry.setVar", () => {
         return data.map(() => [1, 2, 3]);
       }
       embeddingDataType() {
-        return new arrow18.Float32() as apiArrow.Float;
+        return new apiArrow.Float32();
       }
       protected getSensitiveKeys() {
         return ["secretKey"];

nodejs/__test__/sanitize.test.ts

@@ -2,7 +2,13 @@
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 import * as arrow from "../lancedb/arrow";
-import { sanitizeField, sanitizeMap, sanitizeType } from "../lancedb/sanitize";
+import {
+  sanitizeField,
+  sanitizeMap,
+  sanitizeSchema,
+  sanitizeType,
+} from "../lancedb/sanitize";
+import { arrow21 } from "./arrow_versions";
 
 describe("sanitize", function () {
   describe("sanitizeType function", function () {
@@ -59,6 +65,8 @@ describe("sanitize", function () {
       ["durationmicrosecond", new arrow.DurationMicrosecond()],
       ["durationmillisecond", new arrow.DurationMillisecond()],
       ["durationsecond", new arrow.DurationSecond()],
+      ["largebinary", new arrow.LargeBinary()],
+      ["largeutf8", new arrow.LargeUtf8()],
     ] as const;
 
     it.each(allTypeNameTestCases)(
@@ -89,6 +97,110 @@ describe("sanitize", function () {
         "Unrecognized type name in schema: invalid_type",
       );
     });
+
+    // These types only exist in Arrow 20 (IntervalMonthDayNano) and 21 (the
+    // rest), so the objects come from a second copy of the newest release.  The
+    // library itself may be running on an older Arrow (see the CI matrix), in
+    // which case sanitizing them has to fail with a clear message instead.
+    const hostArrow = arrow as unknown as Record<string, unknown>;
+    const hostSupportsNewerTypes = typeof hostArrow.Utf8View === "function";
+    const whenSupported = hostSupportsNewerTypes ? describe : describe.skip;
+    const whenUnsupported = hostSupportsNewerTypes ? describe.skip : describe;
+
+    const newerTypeTestCases = [
+      ["Utf8View", () => new arrow21.Utf8View()],
+      ["BinaryView", () => new arrow21.BinaryView()],
+      ["IntervalMonthDayNano", () => new arrow21.IntervalMonthDayNano()],
+      [
+        "LargeList",
+        () =>
+          new arrow21.LargeList(new arrow21.Field("item", new arrow21.Utf8())),
+      ],
+    ] as const;
+
+    whenSupported("types added in Arrow 20 and 21", function () {
+      it.each(newerTypeTestCases)(
+        "should sanitize a %s type from another copy of Arrow",
+        function (name, makeForeign) {
+          const foreign = makeForeign();
+          const result = sanitizeType(foreign);
+          // Interval subclasses share `Type.Interval`, so the sanitized type
+          // is the generic `Interval` carrying the same unit, as for the
+          // other units.
+          const expected = name === "IntervalMonthDayNano" ? "Interval" : name;
+          expect(result).toBeInstanceOf(hostArrow[expected] as () => unknown);
+          expect(result.typeId).toBe(foreign.typeId);
+          expect(result.toString()).toBe(foreign.toString());
+        },
+      );
+
+      it.each([
+        ["binaryview", "BinaryView"],
+        ["utf8view", "Utf8View"],
+        ["intervalmonthdaynano", "Interval"],
+      ])('should map type name "%s" to %s', function (name, expected) {
+        expect(sanitizeType(name)).toBeInstanceOf(
+          hostArrow[expected] as () => unknown,
+        );
+      });
+
+      it("should sanitize the child field of a LargeList", function () {
+        const foreign = new arrow21.LargeList(
+          new arrow21.Field("item", new arrow21.Int32(), false),
+        );
+        const result = sanitizeType(foreign) as arrow.List;
+        expect(result.children[0]).toBeInstanceOf(arrow.Field);
+        expect(result.children[0].name).toBe("item");
+        expect(result.children[0].nullable).toBe(false);
+        const child = result.children[0].type;
+        expect(arrow.DataType.isInt(child)).toBe(true);
+        expect((child as arrow.Int).bitWidth).toBe(32);
+        expect((child as arrow.Int).isSigned).toBe(true);
+      });
+    });
+
+    whenUnsupported("types newer than the installed Arrow", function () {
+      it.each(
+        newerTypeTestCases.filter(([name]) => name !== "IntervalMonthDayNano"),
+      )(
+        "should explain that %s needs a newer apache-arrow",
+        function (name, makeForeign) {
+          expect(() => sanitizeType(makeForeign())).toThrow(
+            `The Arrow type ${name} requires apache-arrow`,
+          );
+        },
+      );
+
+      it("should sanitize IntervalMonthDayNano to a generic Interval", () => {
+        // The subclass shares `Type.Interval`, which every supported Arrow
+        // release has; only the unit is new.
+        const result = sanitizeType(new arrow21.IntervalMonthDayNano());
+        expect(result).toBeInstanceOf(arrow.Interval);
+        expect((result as arrow.Interval).unit).toBe(
+          arrow21.IntervalUnit.MONTH_DAY_NANO,
+        );
+      });
+    });
+
+    it("should rebuild a schema from a second copy of the same Arrow version", function () {
+      // Arrow 21 makes `instanceof Schema` succeed across copies of the library
+      // but not `instanceof <DataType>`; the schema must still be rebuilt.
+      const foreign = new arrow21.Schema([
+        new arrow21.Field("x", new arrow21.Float32(), true),
+      ]);
+      // Only Arrow 21+ overrides `Symbol.hasInstance` this way.
+      const crossCopyInstanceof = Object.getOwnPropertySymbols(
+        arrow.Schema,
+      ).includes(Symbol.hasInstance);
+      expect(foreign instanceof arrow.Schema).toBe(crossCopyInstanceof);
+      const result = sanitizeSchema(foreign);
+      expect(result).not.toBe(foreign);
+      const type = result.fields[0].type;
+      expect(type).toBeInstanceOf(arrow.DataType);
+      expect(arrow.DataType.isFloat(type)).toBe(true);
+      expect((type as arrow.Float).precision).toBe(arrow.Precision.SINGLE);
+      expect(() => arrow.makeBuilder({ type })).not.toThrow();
+    });
   });
 
   describe("sanitizeField function", function () {

nodejs/__test__/table.test.ts

@@ -2,15 +2,11 @@
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 import * as fs from "fs";
+import { getEventListeners } from "node:events";
 import { pathToFileURL } from "node:url";
 import * as path from "path";
 import * as tmp from "tmp";
 
-import * as arrow15 from "apache-arrow-15";
-import * as arrow16 from "apache-arrow-16";
-import * as arrow17 from "apache-arrow-17";
-import * as arrow18 from "apache-arrow-18";
-
 import {
   AutoQuery,
   Connection,
@@ -55,19 +51,21 @@ import {
   instanceOfFullTextQuery,
 } from "../lancedb/query";
 import { LocalTable } from "../lancedb/table";
+import {
+  type ApacheArrow,
+  arrow18,
+  arrowVersions,
+  latestArrow,
+} from "./arrow_versions";
 
-describe.each([arrow15, arrow16, arrow17, arrow18])(
+describe.each(arrowVersions)(
   "Given a table",
   // biome-ignore lint/suspicious/noExplicitAny: <explanation>
   (arrow: any) => {
     let tmpDir: tmp.DirResult;
     let table: Table;
 
-    const schema:
-      | import("apache-arrow-15").Schema
-      | import("apache-arrow-16").Schema
-      | import("apache-arrow-17").Schema
-      | import("apache-arrow-18").Schema = new arrow.Schema([
+    const schema: InstanceType<ApacheArrow["Schema"]> = new arrow.Schema([
       new arrow.Field("id", new arrow.Float64(), true),
     ]);
 
@@ -92,6 +90,38 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       await expect(table.countRows()).resolves.toBe(3);
     });
 
+    it("creates and adds Arrow tables with large string and binary columns", async () => {
+      const data = new arrow.Table({
+        text: arrow.vectorFromArray(["alpha", "beta"], new arrow.LargeUtf8()),
+        bytes: arrow.vectorFromArray(
+          [Buffer.from("one"), Buffer.from("two")],
+          new arrow.LargeBinary(),
+        ),
+      });
+      const conn = await connect(tmpDir.name);
+      const largeTable = await conn.createTable("large_columns", data);
+      await largeTable.add(data);
+
+      const rows = await largeTable.query().toArray();
+      expect(rows.map((row) => row.text)).toEqual([
+        "alpha",
+        "beta",
+        "alpha",
+        "beta",
+      ]);
+      expect(rows.map((row) => Buffer.from(row.bytes).toString())).toEqual([
+        "one",
+        "two",
+        "one",
+        "two",
+      ]);
+      const fields = (await largeTable.schema()).fields;
+      expect(fields.map((field) => field.type.typeId)).toEqual([
+        arrow.Type.LargeUtf8,
+        arrow.Type.LargeBinary,
+      ]);
+    });
+
     it("should support a foreign Float64 vector schema end to end", async () => {
       const conn = await connect(tmpDir.name);
       const schema = new arrow.Schema([
@@ -709,6 +739,114 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
 );
 
 // https://github.com/lancedb/lancedb/issues/1963
+// These types were added in Arrow 21, so they are built with a second copy of
+// that release rather than the per-version matrix above, and reading them back
+// needs the library's own Arrow to know them too.
+const hostArrowSupportsViewTypes =
+  typeof (arrow as unknown as Record<string, unknown>).Utf8View === "function";
+(hostArrowSupportsViewTypes ? it : it.skip)(
+  "creates and adds tables with LargeList, Utf8View and BinaryView columns",
+  async () => {
+    const tmpDir = tmp.dirSync({ unsafeCleanup: true });
+    const db = await connect(tmpDir.name);
+    const makeData = () =>
+      new latestArrow.Table({
+        id: latestArrow.vectorFromArray([1, 2], new latestArrow.Int32()),
+        tags: latestArrow.vectorFromArray(
+          [["a", "b"], ["c"]],
+          new latestArrow.LargeList(
+            new latestArrow.Field("item", new latestArrow.Utf8(), true),
+          ),
+        ),
+        text: latestArrow.vectorFromArray(
+          ["short", "a string that is longer than the twelve inline bytes"],
+          new latestArrow.Utf8View(),
+        ),
+        bytes: latestArrow.vectorFromArray(
+          [new Uint8Array([1, 2, 3]), new Uint8Array(20).fill(7)],
+          new latestArrow.BinaryView(),
+        ),
+      });
+    const table = await db.createTable("views", makeData());
+    await table.add(makeData());
+
+    const schema = await table.schema();
+    // Take the enum from the Arrow 21 copy: the host Arrow may predate
+    // `Type.LargeList`, and this file still has to type-check there.
+    expect(schema.fields.map((f) => f.type.typeId)).toEqual([
+      latestArrow.Type.Int,
+      latestArrow.Type.LargeList,
+      // Lance stores the view types as their offset-based equivalents.
+      latestArrow.Type.Utf8,
+      latestArrow.Type.Binary,
+    ]);
+    const rows = await table.query().toArrow();
+    expect(rows.numRows).toBe(4);
+    const row = rows.get(1)!.toJSON();
+    expect(row.id).toBe(2);
+    expect(row.tags.toJSON()).toEqual(["c"]);
+    expect(row.text).toBe(
+      "a string that is longer than the twelve inline bytes",
+    );
+    expect(Array.from(row.bytes)).toEqual(Array(20).fill(7));
+
+    // Arrow JS 21 serializes a *sliced* view column with a truncated views
+    // buffer; the library rebuilds view columns as Utf8/Binary so this works.
+    const sliced = makeData().slice(1, 2);
+    expect(sliced.getChild("text")!.data[0].offset).toBe(1);
+    const slicedTable = await db.createTable("views_sliced", sliced);
+    await slicedTable.add(sliced);
+    const slicedRows = await slicedTable.query().toArrow();
+    expect(slicedRows.numRows).toBe(2);
+    for (const i of [0, 1]) {
+      const r = slicedRows.get(i)!.toJSON();
+      expect(r.id).toBe(2);
+      expect(r.text).toBe(
+        "a string that is longer than the twelve inline bytes",
+      );
+      expect(Array.from(r.bytes)).toEqual(Array(20).fill(7));
+    }
+
+    // Regression: a field with an integer-like name. JavaScript enumerates
+    // such keys first, so a key-ordered rebuild would pair "1" with the text
+    // column's values. The batches are built explicitly to pin field order.
+    const numericSchema = new latestArrow.Schema([
+      new latestArrow.Field("text", new latestArrow.Utf8View(), true),
+      new latestArrow.Field("1", new latestArrow.Utf8View(), true),
+      new latestArrow.Field("n", new latestArrow.Int32(), true),
+    ]);
+    const column = (
+      values: unknown[],
+      type: InstanceType<typeof latestArrow.DataType>,
+    ) => latestArrow.vectorFromArray(values, type).data[0];
+    const numericBatch = new latestArrow.RecordBatch(
+      numericSchema,
+      latestArrow.makeData({
+        type: new latestArrow.Struct(numericSchema.fields),
+        length: 3,
+        nullCount: 0,
+        children: [
+          column(["t0", "t1", "t2"], new latestArrow.Utf8View()),
+          column(["one0", "one1", "one2"], new latestArrow.Utf8View()),
+          column([0, 1, 2], new latestArrow.Int32()),
+        ],
+      }),
+    );
+    const numericData = new latestArrow.Table(numericSchema, [numericBatch]);
+    const numericTable = await db.createTable("views_numeric", numericData);
+    await numericTable.add(numericData.slice(1));
+    const numericRows = (await numericTable.query().toArrow()).toArray();
+    expect(numericRows.map((r) => r.toJSON())).toEqual([
+      { text: "t0", "1": "one0", n: 0 },
+      { text: "t1", "1": "one1", n: 1 },
+      { text: "t2", "1": "one2", n: 2 },
+      { text: "t1", "1": "one1", n: 1 },
+      { text: "t2", "1": "one2", n: 2 },
+    ]);
+    tmpDir.removeCallback();
+  },
+);
+
 it("should query documents with LangChain PDF metadata", async () => {
   const tmpDir = tmp.dirSync({ unsafeCleanup: true });
   try {
@@ -2601,6 +2739,27 @@ describe("when dealing with blob columns", () => {
     expect(Buffer.from(await files[1]!.read()).toString()).toBe("beta");
   });
 
+  it("adds a LargeBinary Arrow column to a blob table", async () => {
+    const db = await connect(tmpDir.name);
+    const schema = new Schema([new Field("id", new Int64()), blob("image")]);
+    const table = await db.createTable(
+      "blobs",
+      [{ id: 1n, image: Buffer.from("alpha") }],
+      { schema },
+    );
+    const payload = Buffer.from("beta");
+    const data = new arrow18.Table({
+      id: arrow18.vectorFromArray([2n], new arrow18.Int64()),
+      image: arrow18.vectorFromArray([payload], new arrow18.LargeBinary()),
+    });
+    await table.add(data);
+
+    const rows = await table.query().withRowId().toArray();
+    const rowId = rows.find((row) => row.id === 2n)!._rowid as bigint;
+    const [actual] = await table.fetchBlobs("image", [rowId]);
+    expect(actual).toEqual(payload);
+  });
+
   it("reads a half-open range", async () => {
     const { table, rowIds } = await openBlobTable();
     const files = await table.fetchBlobFiles("image", rowIds);
@@ -2625,6 +2784,148 @@ describe("when dealing with blob columns", () => {
     ).rejects.toThrow(/exceeds blob size/);
   });
 
+  it("reads at most maxBytes and advances the cursor", async () => {
+    const { table, rowIds } = await openBlobTable();
+    const [handle] = await table.fetchBlobFiles("image", rowIds);
+    expect((await handle!.read(2n)).toString()).toBe("al");
+    expect(await handle!.tell()).toBe(2n);
+    expect((await handle!.read(2n)).toString()).toBe("ph");
+    expect((await handle!.read(10n)).toString()).toBe("a");
+    expect(await handle!.tell()).toBe(5n);
+    expect(await handle!.read(1n)).toEqual(Buffer.alloc(0));
+  });
+
+  it("seeks the cursor", async () => {
+    const { table, rowIds, alpha } = await openBlobTable();
+    const [handle] = await table.fetchBlobFiles("image", rowIds);
+    await handle!.seek(3n);
+    expect(await handle!.tell()).toBe(3n);
+    expect((await handle!.read()).toString()).toBe("ha");
+    await handle!.seek(0n);
+    expect(await handle!.read()).toEqual(alpha);
+  });
+
+  it("rejects a negative seek position", async () => {
+    const { table, rowIds } = await openBlobTable();
+    const [handle] = await table.fetchBlobFiles("image", rowIds);
+    await expect(handle!.seek(-1n)).rejects.toThrow(/cannot be negative/);
+  });
+
+  it("readRanges returns one buffer per range in order", async () => {
+    const { table, rowIds } = await openBlobTable();
+    const [handle] = await table.fetchBlobFiles("image", rowIds);
+    const buffers = await handle!.readRanges([
+      { start: 3n, end: 5n },
+      { start: 0n, end: 2n },
+      { start: 3n, end: 5n },
+    ]);
+    expect(buffers.map((b) => b.toString())).toEqual(["ha", "al", "ha"]);
+    expect(await handle!.tell()).toBe(0n);
+  });
+
+  it("fails when a readRanges end is past the blob size", async () => {
+    const { table, rowIds, alpha } = await openBlobTable();
+    const [handle] = await table.fetchBlobFiles("image", rowIds);
+    await expect(
+      handle!.readRanges([
+        { start: 0n, end: 1n },
+        { start: 0n, end: BigInt(alpha.length + 1) },
+      ]),
+    ).rejects.toThrow(/exceeds blob size/);
+  });
+
+  it("fails reads after close and closes once", async () => {
+    const { table, rowIds } = await openBlobTable();
+    const [handle] = await table.fetchBlobFiles("image", rowIds);
+    expect(handle!.isClosed()).toBe(false);
+    await handle!.close();
+    expect(handle!.isClosed()).toBe(true);
+    await expect(handle!.close()).resolves.toBeUndefined();
+    await expect(handle!.read()).rejects.toThrow(/already closed/);
+    await expect(handle!.read(1n)).rejects.toThrow(/already closed/);
+    await expect(handle!.readRange(0n, 1n)).rejects.toThrow(/already closed/);
+    await expect(handle!.readRanges([{ start: 0n, end: 1n }])).rejects.toThrow(
+      /already closed/,
+    );
+    await expect(handle!.seek(0n)).rejects.toThrow(/already closed/);
+    await expect(handle!.tell()).rejects.toThrow(/already closed/);
+    expect(handle!.size()).toBe(5n);
+  });
+
+  it("rejects reads with an already aborted signal", async () => {
+    const { table, rowIds, alpha } = await openBlobTable();
+    const [handle] = await table.fetchBlobFiles("image", rowIds);
+    const signal = AbortSignal.abort();
+    const abortError = { name: "AbortError" };
+    await expect(handle!.read(undefined, { signal })).rejects.toMatchObject(
+      abortError,
+    );
+    await expect(handle!.read(2n, { signal })).rejects.toMatchObject(
+      abortError,
+    );
+    await expect(handle!.readRange(0n, 1n, { signal })).rejects.toMatchObject(
+      abortError,
+    );
+    await expect(
+      handle!.readRanges([{ start: 0n, end: 1n }], { signal }),
+    ).rejects.toMatchObject(abortError);
+    await expect(
+      table.fetchBlobs("image", rowIds, { signal }),
+    ).rejects.toMatchObject(abortError);
+    await expect(
+      table.fetchBlobFiles("image", rowIds, { signal }),
+    ).rejects.toMatchObject(abortError);
+    expect(await handle!.read()).toEqual(alpha);
+  });
+
+  it("rejects with the abort reason", async () => {
+    const { table, rowIds } = await openBlobTable();
+    const [handle] = await table.fetchBlobFiles("image", rowIds);
+    const reason = new Error("caller gave up");
+    await expect(
+      handle!.readRange(0n, 1n, { signal: AbortSignal.abort(reason) }),
+    ).rejects.toBe(reason);
+  });
+
+  it("reads with a signal and removes its abort listener", async () => {
+    const { table, rowIds, alpha, beta } = await openBlobTable();
+    const controller = new AbortController();
+    const { signal } = controller;
+    const [handle] = await table.fetchBlobFiles("image", rowIds, { signal });
+    expect((await handle!.readRange(0n, 2n, { signal })).toString()).toBe("al");
+    expect(await table.fetchBlobs("image", rowIds, { signal })).toEqual([
+      alpha,
+      beta,
+      null,
+    ]);
+    expect(getEventListeners(signal, "abort")).toHaveLength(0);
+    controller.abort();
+    expect(await handle!.read()).toEqual(alpha);
+  });
+
+  it("cancels an in-flight read", async () => {
+    const db = await connect(tmpDir.name);
+    const schema = new Schema([
+      new Field("id", new Int64(), true),
+      blob("payload", { inlineSizeThreshold: 0, dedicatedSizeThreshold: 1 }),
+    ]);
+    const payload = Buffer.alloc(32 * 1024 * 1024, 7);
+    const table = await db.createTable("large_blobs", [{ id: 1n, payload }], {
+      schema,
+    });
+    const [row] = await table.query().withRowId().toArray();
+    const [handle] = await table.fetchBlobFiles("payload", [
+      row._rowid as bigint,
+    ]);
+    const controller = new AbortController();
+    const pending = handle!.readRange(0n, handle!.size(), {
+      signal: controller.signal,
+    });
+    controller.abort();
+    await expect(pending).rejects.toMatchObject({ name: "AbortError" });
+    expect(await handle!.readRange(0n, 4n)).toEqual(payload.subarray(0, 4));
+  });
+
   it("rejects fetchBlobs on a non-blob column", async () => {
     const { table, rowIds } = await openBlobTable();
     await expect(table.fetchBlobs("id", rowIds)).rejects.toThrow(/blob/i);
@@ -3007,7 +3308,7 @@ it("passes cleanupOlderThan to the native binding as an absolute timestamp", asy
   expect(optimize).toHaveBeenCalledWith(cutoff.getTime(), true);
 });
 
-describe.each([arrow15, arrow16, arrow17, arrow18])(
+describe.each(arrowVersions)(
   "when optimizing a dataset",
   // biome-ignore lint/suspicious/noExplicitAny: <explanation>
   (arrow: any) => {

nodejs/lancedb/arrow.ts

@@ -66,11 +66,20 @@ export type SchemaLike =
 export type FieldLike =
   | Field
   | {
-      type: string;
+      type: string | DataTypeLike;
       name: string;
       nullable: boolean;
       metadata?: Map<string, string>;
     };
+/**
+ * A `DataType` from any copy or version of apache-arrow.
+ *
+ * Arrow 21 brands its classes with `unique symbol` properties, so a type
+ * object from a second copy of the library no longer satisfies the `DataType`
+ * type of this one even though it is structurally identical. Inputs that only
+ * need to be sanitized accept this looser shape instead.
+ */
+export type DataTypeLike = DataType | { readonly typeId: number };
 
 /**
  * Create an Arrow field backed by LanceDB's JSON extension type.
@@ -1059,6 +1068,71 @@ export async function fromRecordsToStreamBuffer(
   return Buffer.from(await writer.toUint8Array());
 }
 
+// `Type.Utf8View` / `Type.BinaryView` as numbers: the enum members only exist
+// in Arrow 21+, and this module compiles against every supported release.
+const UTF8_VIEW_TYPE_ID = 24;
+const BINARY_VIEW_TYPE_ID = 23;
+
+/**
+ * Copy a Utf8View / BinaryView `Data` into a single `Data` of `type`.
+ */
+function materializeViewData(child: ArrowData, type: DataType): ArrowData {
+  const builder = makeBuilder({ type, nullValues: [null] });
+  for (const value of new Vector([child])) {
+    builder.append(value);
+  }
+  return builder.finish().flush();
+}
+
+/**
+ * Rebuild any top-level Utf8View / BinaryView column as Utf8 / Binary.
+ *
+ * Lance stores the view types as their offset-based equivalents anyway, so
+ * nothing is lost. Doing it here also sidesteps an Arrow JS 21 bug: its IPC
+ * writer emits a truncated views buffer for a *sliced* view array, which the
+ * Rust reader rejects with "Need at least N bytes in buffers[0]".
+ *
+ * The record batches are rebuilt positionally rather than through a
+ * `Record<string, Vector>`: JavaScript enumerates integer-like keys first, so
+ * a field named e.g. `"1"` would otherwise be paired with the wrong column.
+ *
+ * Tables without view columns are returned as-is.
+ */
+function materializeViewColumns(table: ArrowTable): ArrowTable {
+  const replacements = new Map<number, DataType>();
+  table.schema.fields.forEach((field, i) => {
+    if (field.type.typeId === UTF8_VIEW_TYPE_ID) {
+      replacements.set(i, new Utf8());
+    } else if (field.type.typeId === BINARY_VIEW_TYPE_ID) {
+      replacements.set(i, new Binary());
+    }
+  });
+  if (replacements.size === 0) {
+    return table;
+  }
+  const fields = table.schema.fields.map((field, i) => {
+    const type = replacements.get(i);
+    return type === undefined
+      ? field
+      : new Field(field.name, type, field.nullable, field.metadata);
+  });
+  const schema = new Schema(fields, table.schema.metadata);
+  const batches = table.batches.map((batch) => {
+    const children = batch.data.children.map((child, i) => {
+      const type = replacements.get(i);
+      return type === undefined ? child : materializeViewData(child, type);
+    });
+    const data = makeData({
+      type: new Struct(fields),
+      length: batch.numRows,
+      nullCount: 0,
+      children,
+    });
+    return new RecordBatch(schema, data);
+  });
+  return new ArrowTable(schema, batches);
+}
+
 /**
  * Serialize an Arrow Table into a buffer using the Arrow IPC File serialization
  *
@@ -1075,7 +1149,9 @@ export async function fromTableToBuffer(
   if (schema !== undefined && schema !== null) {
     schema = sanitizeSchema(schema);
   }
-  const tableWithEmbeddings = await applyEmbeddings(table, embeddings, schema);
+  const tableWithEmbeddings = materializeViewColumns(
+    await applyEmbeddings(table, embeddings, schema),
+  );
   validateBlobSchema(tableWithEmbeddings.schema);
   const writer = RecordBatchFileWriter.writeAll(tableWithEmbeddings);
   return Buffer.from(await writer.toUint8Array());
@@ -1146,7 +1222,8 @@ export async function fromRecordBatchToBuffer(
 export async function fromRecordBatchToStreamBuffer(
   batch: RecordBatch,
 ): Promise<Buffer> {
-  const writer = RecordBatchStreamWriter.writeAll([batch]);
+  const table = materializeViewColumns(new ArrowTable([batch]));
+  const writer = RecordBatchStreamWriter.writeAll(table);
   return Buffer.from(await writer.toUint8Array());
 }
 
@@ -1163,7 +1240,9 @@ export async function fromTableToStreamBuffer(
   embeddings?: EmbeddingFunctionConfig,
   schema?: SchemaLike,
 ): Promise<Buffer> {
-  const tableWithEmbeddings = await applyEmbeddings(table, embeddings, schema);
+  const tableWithEmbeddings = materializeViewColumns(
+    await applyEmbeddings(table, embeddings, schema),
+  );
   const writer = RecordBatchStreamWriter.writeAll(tableWithEmbeddings);
   return Buffer.from(await writer.toUint8Array());
 }

nodejs/lancedb/blob.ts

@@ -141,13 +141,16 @@ export class BlobFile {
   }
 
   /**
-   * Reads from the cursor to the end and advances the cursor.
+   * Reads from the cursor and advances the cursor.
    *
-   * A second call returns an empty buffer. {@link BlobFile.readRange} does
-   * not move the cursor.
+   * Reads to the end when `maxBytes` is omitted, or at most `maxBytes` bytes
+   * otherwise. Returns an empty buffer at the end of the blob.
+   * {@link BlobFile.readRange} does not move the cursor.
    */
-  read(): Promise<Buffer> {
-    return this.inner.read();
+  read(maxBytes?: bigint, options?: BlobReadOptions): Promise<Buffer> {
+    return runWithSignal(options?.signal, (signal) =>
+      this.inner.read(maxBytes, signal),
+    );
   }
 
   /**
@@ -155,11 +158,96 @@ export class BlobFile {
    *
    * Fails when `end` is past the blob size. Does not move the cursor.
    */
-  readRange(start: bigint, end: bigint): Promise<Buffer> {
-    return this.inner.readRange(start, end);
+  readRange(
+    start: bigint,
+    end: bigint,
+    options?: BlobReadOptions,
+  ): Promise<Buffer> {
+    return runWithSignal(options?.signal, (signal) =>
+      this.inner.readRange(start, end, signal),
+    );
+  }
+
+  /**
+   * Reads several half-open byte ranges. Returns one buffer per range, in
+   * the order given.
+   *
+   * Fails when any `end` is past the blob size. Does not move the cursor.
+   */
+  readRanges(
+    ranges: BlobRange[],
+    options?: BlobReadOptions,
+  ): Promise<Buffer[]> {
+    return runWithSignal(options?.signal, (signal) =>
+      this.inner.readRanges(ranges, signal),
+    );
+  }
+
+  /** Moves the cursor to `position`, in bytes from the start of the blob. */
+  seek(position: bigint): Promise<void> {
+    return this.inner.seek(position);
+  }
+
+  /** Returns the cursor position in bytes. */
+  tell(): Promise<bigint> {
+    return this.inner.tell();
+  }
+
+  /**
+   * Releases the handle. Reads after `close()` fail. Calling it again does
+   * nothing.
+   */
+  close(): Promise<void> {
+    return this.inner.close();
+  }
+
+  /** Returns true after {@link BlobFile.close}. */
+  isClosed(): boolean {
+    return this.inner.isClosed();
   }
 }
 
+/** Options for blob reads. */
+export type BlobReadOptions = {
+  /**
+   * Cancels the read. The call rejects with `signal.reason`, and the native
+   * read stops, including in-flight requests to a remote table.
+   */
+  signal?: AbortSignal;
+};
+
+/** @ignore */
+export async function runWithSignal<T>(
+  signal: AbortSignal | undefined,
+  run: (signal: AbortSignal | undefined) => Promise<T>,
+): Promise<T> {
+  if (signal === undefined) {
+    return run(undefined);
+  }
+  signal.throwIfAborted();
+  // napi-rs replaces `onabort` on the signal it receives, so pass a private
+  // signal and forward the caller's abort to it.
+  const controller = new AbortController();
+  const onAbort = () => controller.abort();
+  signal.addEventListener("abort", onAbort, { once: true });
+  try {
+    return await run(controller.signal);
+  } catch (err) {
+    if (signal.aborted) {
+      throw signal.reason;
+    }
+    throw err;
+  } finally {
+    signal.removeEventListener("abort", onAbort);
+  }
+}
+
+/** A half-open byte range `[start, end)` for {@link BlobFile.readRanges}. */
+export type BlobRange = {
+  start: bigint;
+  end: bigint;
+};
+
 export function coerceBlobValue(value: unknown): BlobInput | null {
   if (value == null) {
     return null;

nodejs/lancedb/index.ts

@@ -84,7 +84,7 @@ export {
 } from "./arrow";
 
 export { blob, isBlobField, BlobFile } from "./blob";
-export type { BlobOptions } from "./blob";
+export type { BlobOptions, BlobRange, BlobReadOptions } from "./blob";
 
 export {
   Connection,
@@ -195,6 +195,7 @@ export * as rerankers from "./rerankers";
 export {
   SchemaLike,
   TableLike,
+  DataTypeLike,
   FieldLike,
   RecordBatchLike,
   DataLike,

nodejs/lancedb/sanitize.ts

@@ -9,6 +9,7 @@
 // comes from the exact same library instance.  This is not always the case
 // and so we must sanitize the input to ensure that it is compatible.
 
+import * as arrowModule from "apache-arrow";
 import { BufferType, Data, Vector } from "apache-arrow";
 import type { IntBitWidth, TKeys, TimeBitWidth } from "apache-arrow/type";
 import {
@@ -43,6 +44,8 @@ import {
   Interval,
   IntervalDayTime,
   IntervalYearMonth,
+  LargeBinary,
+  LargeUtf8,
   List,
   Map_,
   Null,
@@ -74,6 +77,66 @@ import {
   Utf8,
 } from "./arrow";
 
+/**
+ * Whether `value` was created by *this* copy of apache-arrow's `ctor`.
+ *
+ * Arrow 21 overrides `Symbol.hasInstance` on `Schema`, `Field`, `Table`,
+ * `RecordBatch`, `Vector` and `Data` so that `instanceof` also matches objects
+ * built by a different copy of the library.  Those are exactly the objects
+ * this module has to rebuild: the `DataType` instances nested inside them are
+ * deliberately *not* given that treatment upstream, so Arrow's own builders and
+ * writers reject them.  Walk the real prototype chain instead of trusting
+ * `instanceof`.
+ */
+function isOwnInstance<T>(
+  value: unknown,
+  // biome-ignore lint/suspicious/noExplicitAny: matches any class constructor
+  ctor: abstract new (...args: any[]) => T,
+): value is T {
+  return (
+    typeof value === "object" &&
+    value !== null &&
+    Object.prototype.isPrototypeOf.call(ctor.prototype, value)
+  );
+}
+
+/**
+ * Types added to Arrow JS after release 18, keyed by the `Type` enum value
+ * they use.  The library compiles and runs against every Arrow release in the
+ * peer range, so these are looked up by name at runtime instead of imported:
+ * an older Arrow simply does not have the class.
+ */
+type NewerTypeName =
+  | "IntervalMonthDayNano"
+  | "LargeList"
+  | "BinaryView"
+  | "Utf8View";
+const NEWER_TYPES: ReadonlyArray<{
+  name: NewerTypeName;
+  typeId: number;
+  since: number;
+}> = [
+  { name: "IntervalMonthDayNano", typeId: -31, since: 20 },
+  { name: "LargeList", typeId: 21, since: 21 },
+  { name: "BinaryView", typeId: 23, since: 21 },
+  { name: "Utf8View", typeId: 24, since: 21 },
+];
+
+// biome-ignore lint/suspicious/noExplicitAny: constructor arity varies by type
+type DataTypeConstructor = new (...args: any[]) => DataType;
+
+function newerTypeConstructor(name: NewerTypeName): DataTypeConstructor {
+  const ctor = (arrowModule as unknown as Record<string, unknown>)[name];
+  if (typeof ctor !== "function") {
+    const since = NEWER_TYPES.find((t) => t.name === name)?.since;
+    throw new Error(
+      `The Arrow type ${name} requires apache-arrow ${since} or newer, ` +
+        "but the installed apache-arrow does not provide it",
+    );
+  }
+  return ctor as DataTypeConstructor;
+}
+
 type SanitizationContext = {
   types: WeakMap<object, DataType>;
   vectors: WeakMap<object, Vector>;
@@ -210,19 +273,37 @@ export function sanitizeList(typeLike: object) {
   return sanitizeListWithContext(typeLike, createSanitizationContext());
 }
 
+function sanitizeListChildWithContext(
+  typeLike: object,
+  context: SanitizationContext,
+  typeName: string,
+): Field {
+  if (!("children" in typeLike) || !Array.isArray(typeLike.children)) {
+    throw Error(
+      `Expected a ${typeName} type to have an array-like \`children\` property`,
+    );
+  }
+  if (typeLike.children.length !== 1) {
+    throw Error(`Expected a ${typeName} type to have exactly one child`);
+  }
+  return sanitizeFieldWithContext(typeLike.children[0], context);
+}
+
 function sanitizeListWithContext(
   typeLike: object,
   context: SanitizationContext,
 ) {
-  if (!("children" in typeLike) || !Array.isArray(typeLike.children)) {
-    throw Error(
-      "Expected a List type to have an array-like `children` property",
-    );
-  }
-  if (typeLike.children.length !== 1) {
-    throw Error("Expected a List type to have exactly one child");
-  }
-  return new List(sanitizeFieldWithContext(typeLike.children[0], context));
+  return new List(sanitizeListChildWithContext(typeLike, context, "List"));
+}
+
+function sanitizeLargeListWithContext(
+  typeLike: object,
+  context: SanitizationContext,
+) {
+  const LargeList = newerTypeConstructor("LargeList");
+  return new LargeList(
+    sanitizeListChildWithContext(typeLike, context, "LargeList"),
+  );
 }
 
 export function sanitizeStruct(typeLike: object) {
@@ -449,11 +530,30 @@ function sanitizeTypeWithContext(
   return type;
 }
 
+function sanitizeNewerTypeById(
+  typeLike: object,
+  typeId: number,
+  context: SanitizationContext,
+): DataType | undefined {
+  const entry = NEWER_TYPES.find((t) => t.typeId === typeId);
+  if (entry === undefined) {
+    return undefined;
+  }
+  if (entry.name === "LargeList") {
+    return sanitizeLargeListWithContext(typeLike, context);
+  }
+  return new (newerTypeConstructor(entry.name))();
+}
+
 function sanitizeTypeById(
   typeLike: object,
   typeId: Type,
   context: SanitizationContext,
 ): DataType {
+  const newer = sanitizeNewerTypeById(typeLike, typeId, context);
+  if (newer !== undefined) {
+    return newer;
+  }
   switch (typeId) {
     case Type.NONE:
       throw Error("Received a Type with a typeId of NONE");
@@ -465,8 +565,12 @@ function sanitizeTypeById(
       return sanitizeFloat(typeLike);
     case Type.Binary:
       return new Binary();
+    case Type.LargeBinary:
+      return new LargeBinary();
     case Type.Utf8:
       return new Utf8();
+    case Type.LargeUtf8:
+      return new LargeUtf8();
     case Type.Bool:
       return new Bool();
     case Type.Decimal:
@@ -566,7 +670,7 @@ function sanitizeFieldWithContext(
   fieldLike: unknown,
   context: SanitizationContext,
 ): Field {
-  if (fieldLike instanceof Field) {
+  if (isOwnInstance(fieldLike, Field)) {
     return fieldLike;
   }
   if (typeof fieldLike !== "object" || fieldLike === null) {
@@ -620,7 +724,7 @@ function sanitizeSchemaWithContext(
   schemaLike: SchemaLike,
   context: SanitizationContext,
 ): Schema {
-  if (schemaLike instanceof Schema) {
+  if (isOwnInstance(schemaLike, Schema)) {
     return schemaLike;
   }
   if (typeof schemaLike !== "object" || schemaLike === null) {
@@ -647,7 +751,7 @@ function sanitizeSchemaWithContext(
 }
 
 export function sanitizeTable(tableLike: TableLike): Table {
-  if (tableLike instanceof Table) {
+  if (isOwnInstance(tableLike, Table)) {
     return tableLike;
   }
   if (typeof tableLike !== "object" || tableLike === null) {
@@ -675,7 +779,7 @@ function sanitizeRecordBatch(
   batchLike: RecordBatchLike,
   context: SanitizationContext,
 ): RecordBatch {
-  if (batchLike instanceof RecordBatch) {
+  if (isOwnInstance(batchLike, RecordBatch)) {
     return batchLike;
   }
   if (typeof batchLike !== "object" || batchLike === null) {
@@ -704,11 +808,17 @@ type DictionaryDataLike = DataLike & {
   dictionary?: DictionaryVectorLike;
 };
 
+// Arrow 21 added `variadicBuffers` to `Data` for the Utf8View / BinaryView
+// types.  Older Arrow versions do not have the property, so it is optional.
+type VariadicDataLike = DataLike & {
+  variadicBuffers?: ReadonlyArray<Uint8Array>;
+};
+
 function sanitizeData(
   dataLike: DataLike,
   context: SanitizationContext,
 ): Data<DataType> {
-  if (dataLike instanceof Data) {
+  if (isOwnInstance(dataLike, Data)) {
     return dataLike;
   }
   const cachedData = context.data.get(dataLike);
@@ -726,7 +836,12 @@ function sanitizeData(
       context.vectors.set(dictionaryLike, dictionary);
     }
   }
-  const data = new Data(
+  // Arrow 21 added an eighth constructor parameter (`variadicBuffers`); older
+  // releases ignore the extra argument but their typings do not declare it.
+  const DataCtor = Data as unknown as new (
+    ...args: unknown[]
+  ) => Data<DataType>;
+  const data = new DataCtor(
     sanitizeTypeWithContext(dataLike.type, context),
     dataLike.offset,
     dataLike.length,
@@ -739,6 +854,7 @@ function sanitizeData(
     },
     dataLike.children.map((child) => sanitizeData(child, context)),
     dictionary,
+    (dataLike as VariadicDataLike).variadicBuffers,
   );
   context.data.set(dataLike, data);
   return data;
@@ -747,7 +863,11 @@ function sanitizeData(
 const constructorsByTypeName = {
   null: () => new Null(),
   binary: () => new Binary(),
+  largebinary: () => new LargeBinary(),
+  binaryview: () => new (newerTypeConstructor("BinaryView"))(),
   utf8: () => new Utf8(),
+  largeutf8: () => new LargeUtf8(),
+  utf8view: () => new (newerTypeConstructor("Utf8View"))(),
   bool: () => new Bool(),
   int8: () => new Int8(),
   int16: () => new Int16(),
@@ -768,6 +888,8 @@ const constructorsByTypeName = {
   timesecond: () => new TimeSecond(),
   intervaldaytime: () => new IntervalDayTime(),
   intervalyearmonth: () => new IntervalYearMonth(),
+  intervalmonthdaynano: () =>
+    new (newerTypeConstructor("IntervalMonthDayNano"))(),
   durationnanosecond: () => new DurationNanosecond(),
   durationmicrosecond: () => new DurationMicrosecond(),
   durationmillisecond: () => new DurationMillisecond(),

nodejs/lancedb/table.ts

@@ -17,7 +17,7 @@ import {
   tableFromIPC,
 } from "./arrow";
 
-import { BlobFile } from "./blob";
+import { BlobFile, BlobReadOptions, runWithSignal } from "./blob";
 import { EmbeddingFunctionConfig, getRegistry } from "./embedding/registry";
 import { IndexOptions } from "./indices";
 import { Job } from "./job";
@@ -545,6 +545,7 @@ export abstract class Table {
   abstract fetchBlobs(
     column: string,
     rowIds: readonly (bigint | number)[],
+    options?: BlobReadOptions,
   ): Promise<(Buffer | null)[]>;
 
   /**
@@ -557,6 +558,7 @@ export abstract class Table {
   abstract fetchBlobFiles(
     column: string,
     rowIds: readonly (bigint | number)[],
+    options?: BlobReadOptions,
   ): Promise<(BlobFile | null)[]>;
 
   /**
@@ -1248,8 +1250,11 @@ export class LocalTable extends Table {
   async fetchBlobs(
     column: string,
     rowIds: readonly (bigint | number)[],
+    options?: BlobReadOptions,
   ): Promise<(Buffer | null)[]> {
-    const values = await this.inner.fetchBlobs(column, rowIdsToBigInts(rowIds));
+    const values = await runWithSignal(options?.signal, (signal) =>
+      this.inner.fetchBlobs(column, rowIdsToBigInts(rowIds), signal),
+    );
     // N-API Option maps missing values to undefined. Collapse those to null.
     return values.map((value) => value ?? null);
   }
@@ -1257,10 +1262,10 @@ export class LocalTable extends Table {
   async fetchBlobFiles(
     column: string,
     rowIds: readonly (bigint | number)[],
+    options?: BlobReadOptions,
   ): Promise<(BlobFile | null)[]> {
-    const files = await this.inner.fetchBlobFiles(
-      column,
-      rowIdsToBigInts(rowIds),
+    const files = await runWithSignal(options?.signal, (signal) =>
+      this.inner.fetchBlobFiles(column, rowIdsToBigInts(rowIds), signal),
     );
     // N-API Option maps missing values to undefined. Collapse those to null.
     return files.map((file) =>

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.40.0-beta.11",
+	"version": "0.40.0-beta.12",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",

nodejs/npm/linux-arm64-gnu/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.40.0-beta.11",
+	"version": "0.40.0-beta.12",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-gnu.node",

nodejs/npm/linux-arm64-musl/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-musl",
-	"version": "0.40.0-beta.11",
+	"version": "0.40.0-beta.12",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-musl.node",

nodejs/npm/linux-x64-gnu/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-gnu",
-	"version": "0.40.0-beta.11",
+	"version": "0.40.0-beta.12",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-gnu.node",

nodejs/npm/linux-x64-musl/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-musl",
-	"version": "0.40.0-beta.11",
+	"version": "0.40.0-beta.12",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-musl.node",

nodejs/npm/win32-arm64-msvc/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lancedb/lancedb-win32-arm64-msvc",
-  "version": "0.40.0-beta.11",
+  "version": "0.40.0-beta.12",
   "os": [
     "win32"
   ],

nodejs/npm/win32-x64-msvc/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.40.0-beta.11",
+	"version": "0.40.0-beta.12",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package.json

@@ -11,7 +11,7 @@
     "ann"
   ],
   "private": false,
-  "version": "0.40.0-beta.11",
+  "version": "0.40.0-beta.12",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",
@@ -49,10 +49,14 @@
     "@types/jest": "^29.1.2",
     "@types/node": "22.7.4",
     "@types/tmp": "^0.2.6",
+    "apache-arrow": "21.2.0",
     "apache-arrow-15": "npm:apache-arrow@15.0.0",
     "apache-arrow-16": "npm:apache-arrow@16.0.0",
     "apache-arrow-17": "npm:apache-arrow@17.0.0",
     "apache-arrow-18": "npm:apache-arrow@18.0.0",
+    "apache-arrow-19": "npm:apache-arrow@19.0.1",
+    "apache-arrow-20": "npm:apache-arrow@20.0.0",
+    "apache-arrow-21": "npm:apache-arrow@21.2.0",
     "eslint": "^8.57.0",
     "jest": "^29.7.0",
     "shx": "^0.3.4",
@@ -102,7 +106,7 @@
   },
   "peerDependencies": {
     "@types/node": ">=22",
-    "apache-arrow": ">=15.0.0 <=18.1.0"
+    "apache-arrow": ">=15.0.0 <22.0.0"
   },
   "peerDependenciesMeta": {
     "@types/node": {

nodejs/pnpm-lock.yaml

@@ -14,9 +14,6 @@ importers:
       '@opentelemetry/api':
         specifier: ^1.9.0
         version: 1.9.1
-      apache-arrow:
-        specifier: '>=15.0.0 <=18.1.0'
-        version: 18.1.0
       reflect-metadata:
         specifier: ^0.2.2
         version: 0.2.2
@@ -54,6 +51,9 @@ importers:
       '@types/tmp':
         specifier: ^0.2.6
         version: 0.2.6
+      apache-arrow:
+        specifier: 21.2.0
+        version: 21.2.0
       apache-arrow-15:
         specifier: npm:apache-arrow@15.0.0
         version: apache-arrow@15.0.0
@@ -66,6 +66,15 @@ importers:
       apache-arrow-18:
         specifier: npm:apache-arrow@18.0.0
         version: apache-arrow@18.0.0
+      apache-arrow-19:
+        specifier: npm:apache-arrow@19.0.1
+        version: apache-arrow@19.0.1
+      apache-arrow-20:
+        specifier: npm:apache-arrow@20.0.0
+        version: apache-arrow@20.0.0
+      apache-arrow-21:
+        specifier: npm:apache-arrow@21.2.0
+        version: apache-arrow@21.2.0
       eslint:
         specifier: ^8.57.0
         version: 8.57.1
@@ -1703,6 +1712,9 @@ packages:
   '@types/node@22.7.4':
     resolution: {integrity: sha512-y+NPi1rFzDs1NdQHHToqeiX2TIS79SWEAw9GYhkkx8bD0ChpfqC+n2j5OXOCpzfojBEBt6DnEnnG9MY0zk1XLg==}
 
+  '@types/node@25.9.8':
+    resolution: {integrity: sha512-VfMrScDmMhUJQmd5hArdQnFvK0OIeD36uN2Va1FcpYaLB/BgkgM8Ulc50XtYISPMz7APJ30+N0T5EM0jlcdfRw==}
+
   '@types/stack-utils@2.0.3':
     resolution: {integrity: sha512-9aEbYZ3TbYMznPdcdr3SmIrLXwC/AKZXQeCf9Pgao5CKb8CyHuEX5jzWPTkvregvhRJHcpRO6BFoGW9ycaOkYw==}
 
@@ -1840,8 +1852,16 @@ packages:
     resolution: {integrity: sha512-gFlPaqN9osetbB83zC29AbbZqGiCuFH1vyyPseJ+B7SIbfBtESV62mMT/CkiIt77W6ykC/nTWFzTXFs0Uldg4g==}
     hasBin: true
 
-  apache-arrow@18.1.0:
-    resolution: {integrity: sha512-v/ShMp57iBnBp4lDgV8Jx3d3Q5/Hac25FWmQ98eMahUiHPXcvwIMKJD0hBIgclm/FCG+LwPkAKtkRO1O/W0YGg==}
+  apache-arrow@19.0.1:
+    resolution: {integrity: sha512-APmMLzS4qbTivLrPdQXexGM4JRr+0g62QDaobzEvip/FdQIrv2qLy0mD5Qdmw4buydtVJgbFeKR8f59I6PPGDg==}
+    hasBin: true
+
+  apache-arrow@20.0.0:
+    resolution: {integrity: sha512-JUeK0jFRUd7rbmrhhzR3O2KXjLaZ4YYYFOptyUfxOsMIoZCPi6bZR58gVi/xi3HTBMPseXm9PXyQ2V916930pA==}
+    hasBin: true
+
+  apache-arrow@21.2.0:
+    resolution: {integrity: sha512-Hxe6Agq26gQOM954qpzYSllJBPJl+e16U5CkfuMUhLrNba+5nKkttIVlflaovN6oaTratqMGAO8H5u/aNhmHWQ==}
     hasBin: true
 
   argparse@1.0.10:
@@ -2033,6 +2053,15 @@ packages:
     resolution: {integrity: sha512-H4UfQhZyakIjC74I9d34fGYDwk3XpSr17QhEd0Q3I9Xq1CETHo4Hcuo87WyWHpAF1aSLjLRf5lD9ZGX2qStUvg==}
     engines: {node: '>=4.0.0'}
 
+  command-line-args@6.0.2:
+    resolution: {integrity: sha512-AIjYVxrV9X752LmPDLbVYv8aMCuHPSLZJXEo2qo/xJfv+NYhaZ4sMSF01rM+gHPaMgvPM0l5D/F+Qx+i2WfSmQ==}
+    engines: {node: '>=12.20'}
+    peerDependencies:
+      '@75lb/nature': latest
+    peerDependenciesMeta:
+      '@75lb/nature':
+        optional: true
+
   command-line-usage@7.0.4:
     resolution: {integrity: sha512-85UdvzTNx/+s5CkSgBm/0hzP80RFHAa7PsfeADE5ezZF3uHz3/Tqj9gIKGT9PTtpycc3Ua64T0oVulGfKxzfqg==}
     engines: {node: '>=12.20.0'}
@@ -2281,6 +2310,15 @@ packages:
     resolution: {integrity: sha512-6Tb2myMioCAgv5kfvP5/PkZZ/ntTpVK39fHY7WkWBgvbeE+VHd/tZuZ4mrC+bxh4cfOZeYKVPaJIZtZXV7GNCQ==}
     engines: {node: '>=4.0.0'}
 
+  find-replace@5.0.2:
+    resolution: {integrity: sha512-Y45BAiE3mz2QsrN2fb5QEtO4qb44NcS7en/0y9PEVsg351HsLeVclP8QPMH79Le9sH3rs5RSwJu99W0WPZO43Q==}
+    engines: {node: '>=14'}
+    peerDependencies:
+      '@75lb/nature': latest
+    peerDependenciesMeta:
+      '@75lb/nature':
+        optional: true
+
   find-up@4.1.0:
     resolution: {integrity: sha512-PpOwAdQ/YlXQ2vj8a3h8IipDuYRi3wceVQQGYWxNINccq40Anw7BlsEXCMbt1Zt+OLA6Fq9suIpIWD0OsnISlw==}
     engines: {node: '>=8'}
@@ -2302,6 +2340,9 @@ packages:
   flatbuffers@24.12.23:
     resolution: {integrity: sha512-dLVCAISd5mhls514keQzmEG6QHmUUsNuWsb4tFafIUwvvgDjXhtfAYSKOzt5SWOy+qByV5pbsDZ+Vb7HUOBEdA==}
 
+  flatbuffers@25.9.23:
+    resolution: {integrity: sha512-MI1qs7Lo4Syw0EOzUl0xjs2lsoeqFku44KpngfIduHBYvzm8h2+7K8YMQh1JtVVVrUvhLpNwqVi4DERegUJhPQ==}
+
   flatted@3.4.2:
     resolution: {integrity: sha512-PjDse7RzhcPkIJwy5t7KPWQSZ9cAbzQXcafsetQoD7sOJRQlGikNbx7yZp2OotDnJyrDcbyRq3Ttb18iYOqkxA==}
 
@@ -3345,6 +3386,9 @@ packages:
   undici-types@6.21.0:
     resolution: {integrity: sha512-iwDZqg0QAGrg9Rav5H4n0M64c3mkR59cJ6wQp+7C4nI0gsmExaedaYLNO44eT4AtBBwjbTiGPMlt2Md0T9H9JQ==}
 
+  undici-types@7.24.6:
+    resolution: {integrity: sha512-WRNW+sJgj5OBN4/0JpHFqtqzhpbnV0GuB+OozA9gCL7a993SmU+1JBZCzLNxYsbMfIeDL+lTsphD5jN5N+n0zg==}
+
   unist-util-is@6.0.1:
     resolution: {integrity: sha512-LsiILbtBETkDz8I9p1dQ0uyRUWuaQzd/cuEeS1hoRSyW5E5XGmTzlwY1OrNzzakGowI9Dr/I8HVaw4hTtnxy8g==}
 
@@ -5505,6 +5549,10 @@ snapshots:
     dependencies:
       undici-types: 6.19.8
 
+  '@types/node@25.9.8':
+    dependencies:
+      undici-types: 7.24.6
+
   '@types/stack-utils@2.0.3': {}
 
   '@types/tmp@0.2.6': {}
@@ -5694,17 +5742,40 @@ snapshots:
       json-bignum: 0.0.3
       tslib: 2.8.1
 
-  apache-arrow@18.1.0:
+  apache-arrow@19.0.1:
     dependencies:
       '@swc/helpers': 0.5.21
       '@types/command-line-args': 5.2.3
       '@types/command-line-usage': 5.0.4
       '@types/node': 20.19.41
-      command-line-args: 5.2.1
+      command-line-args: 6.0.2
       command-line-usage: 7.0.4
       flatbuffers: 24.12.23
       json-bignum: 0.0.3
       tslib: 2.8.1
+    transitivePeerDependencies:
+      - '@75lb/nature'
+
+  apache-arrow@20.0.0:
+    dependencies:
+      '@swc/helpers': 0.5.21
+      '@types/command-line-args': 5.2.3
+      '@types/command-line-usage': 5.0.4
+      '@types/node': 20.19.41
+      command-line-args: 6.0.2
+      command-line-usage: 7.0.4
+      flatbuffers: 25.9.23
+      json-bignum: 0.0.3
+      tslib: 2.8.1
+    transitivePeerDependencies:
+      - '@75lb/nature'
+
+  apache-arrow@21.2.0:
+    dependencies:
+      '@types/node': 25.9.8
+      flatbuffers: 25.9.23
+      json-with-bigint: 3.5.8
+      tslib: 2.8.1
 
   argparse@1.0.10:
     dependencies:
@@ -5906,6 +5977,13 @@ snapshots:
       lodash.camelcase: 4.3.0
       typical: 4.0.0
 
+  command-line-args@6.0.2:
+    dependencies:
+      array-back: 6.2.3
+      find-replace: 5.0.2
+      lodash.camelcase: 4.3.0
+      typical: 7.3.0
+
   command-line-usage@7.0.4:
     dependencies:
       array-back: 6.2.3
@@ -6178,6 +6256,8 @@ snapshots:
     dependencies:
       array-back: 3.1.0
 
+  find-replace@5.0.2: {}
+
   find-up@4.1.0:
     dependencies:
       locate-path: 5.0.0
@@ -6201,6 +6281,8 @@ snapshots:
 
   flatbuffers@24.12.23: {}
 
+  flatbuffers@25.9.23: {}
+
   flatted@3.4.2: {}
 
   follow-redirects@1.16.0: {}
@@ -7405,6 +7487,8 @@ snapshots:
 
   undici-types@6.21.0: {}
 
+  undici-types@7.24.6: {}
+
   unist-util-is@6.0.1:
     dependencies:
       '@types/unist': 3.0.3

nodejs/src/blob.rs

@@ -1,25 +1,45 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
+use std::future::Future;
 use std::ops::Range;
 use std::sync::Arc;
+use std::sync::atomic::{AtomicBool, Ordering};
 
 use arrow_array::{Array, LargeBinaryArray};
+use futures::future::abortable;
 use lancedb::blob::BlobFile as LanceBlobFile;
+use napi::Env;
 use napi::bindgen_prelude::*;
 use napi_derive::napi;
 
 use crate::error::convert_error;
 
+#[napi(object)]
+pub struct BlobRange {
+    pub start: BigInt,
+    pub end: BigInt,
+}
+
 #[napi]
 pub struct BlobFile {
     inner: Arc<LanceBlobFile>,
+    closed: AtomicBool,
 }
 
 impl BlobFile {
     pub(crate) fn new(inner: LanceBlobFile) -> Self {
         Self {
             inner: Arc::new(inner),
+            closed: AtomicBool::new(false),
+        }
+    }
+
+    fn ensure_open(&self) -> napi::Result<()> {
+        if self.closed.load(Ordering::Acquire) {
+            Err(napi::Error::from_reason("blob file is already closed"))
+        } else {
+            Ok(())
         }
     }
 }
@@ -32,21 +52,124 @@ impl BlobFile {
     }
 
     #[napi]
-    pub async fn read(&self) -> napi::Result<Buffer> {
-        let bytes = self.inner.read().await.map_err(|err| convert_error(&err))?;
-        Ok(Buffer::from(bytes.as_ref()))
+    pub fn read<'env>(
+        &self,
+        env: &'env Env,
+        max_bytes: Option<BigInt>,
+        signal: Option<AbortSignal>,
+    ) -> napi::Result<PromiseRaw<'env, Buffer>> {
+        self.ensure_open()?;
+        let max_bytes = max_bytes
+            .map(|max_bytes| {
+                usize::try_from(parse_u64(max_bytes, "maxBytes")?)
+                    .map_err(|_| napi::Error::from_reason("maxBytes is too large"))
+            })
+            .transpose()?;
+        let inner = self.inner.clone();
+        spawn_abortable(env, signal, async move {
+            let bytes = match max_bytes {
+                None => inner.read().await,
+                Some(max_bytes) => inner.read_up_to(max_bytes).await,
+            }
+            .map_err(|err| convert_error(&err))?;
+            Ok(Buffer::from(bytes.as_ref()))
+        })
     }
 
     #[napi]
-    pub async fn read_range(&self, start: BigInt, end: BigInt) -> napi::Result<Buffer> {
+    pub fn read_range<'env>(
+        &self,
+        env: &'env Env,
+        start: BigInt,
+        end: BigInt,
+        signal: Option<AbortSignal>,
+    ) -> napi::Result<PromiseRaw<'env, Buffer>> {
+        self.ensure_open()?;
         let range = bigint_range(start, end)?;
-        let bytes = self
-            .inner
-            .read_range(range)
-            .await
-            .map_err(|err| convert_error(&err))?;
-        Ok(Buffer::from(bytes.as_ref()))
+        let inner = self.inner.clone();
+        spawn_abortable(env, signal, async move {
+            let bytes = inner
+                .read_range(range)
+                .await
+                .map_err(|err| convert_error(&err))?;
+            Ok(Buffer::from(bytes.as_ref()))
+        })
     }
+
+    #[napi]
+    pub fn read_ranges<'env>(
+        &self,
+        env: &'env Env,
+        ranges: Vec<BlobRange>,
+        signal: Option<AbortSignal>,
+    ) -> napi::Result<PromiseRaw<'env, Vec<Buffer>>> {
+        self.ensure_open()?;
+        let ranges = ranges
+            .into_iter()
+            .map(|range| bigint_range(range.start, range.end))
+            .collect::<napi::Result<Vec<_>>>()?;
+        let inner = self.inner.clone();
+        spawn_abortable(env, signal, async move {
+            let buffers = inner
+                .read_ranges(&ranges)
+                .await
+                .map_err(|err| convert_error(&err))?;
+            Ok(buffers
+                .iter()
+                .map(|bytes| Buffer::from(bytes.as_ref()))
+                .collect())
+        })
+    }
+
+    #[napi]
+    pub async fn seek(&self, position: BigInt) -> napi::Result<()> {
+        self.ensure_open()?;
+        let position = parse_u64(position, "position")?;
+        self.inner
+            .seek(position)
+            .await
+            .map_err(|err| convert_error(&err))
+    }
+
+    #[napi]
+    pub async fn tell(&self) -> napi::Result<BigInt> {
+        self.ensure_open()?;
+        let position = self.inner.tell().await.map_err(|err| convert_error(&err))?;
+        Ok(BigInt::from(position))
+    }
+
+    #[napi]
+    pub async fn close(&self) -> napi::Result<()> {
+        if self.closed.swap(true, Ordering::AcqRel) {
+            return Ok(());
+        }
+        self.inner.close().await.map_err(|err| convert_error(&err))
+    }
+
+    #[napi]
+    pub fn is_closed(&self) -> bool {
+        self.closed.load(Ordering::Acquire)
+    }
+}
+
+/// Runs `fut` on the napi runtime. Aborting `signal` drops the future, which
+/// cancels in-flight reads.
+pub fn spawn_abortable<'env, T>(
+    env: &'env Env,
+    signal: Option<AbortSignal>,
+    fut: impl Future<Output = napi::Result<T>> + Send + 'static,
+) -> napi::Result<PromiseRaw<'env, T>>
+where
+    T: ToNapiValue + Send + 'static,
+{
+    let (fut, handle) = abortable(fut);
+    if let Some(signal) = signal {
+        signal.on_abort(move || handle.abort());
+    }
+    env.spawn_future(async move {
+        fut.await
+            .unwrap_or_else(|_| Err(napi::Error::from_reason("blob read aborted")))
+    })
 }
 
 fn bigint_range(start: BigInt, end: BigInt) -> napi::Result<Range<u64>> {

nodejs/src/query.rs

@@ -658,6 +658,7 @@ impl JsFullTextQuery {
             FtsQuery::Phrase(_) => "phrase".to_string(),
             FtsQuery::Boost(_) => "boost".to_string(),
             FtsQuery::MultiMatch(_) => "multi_match".to_string(),
+            FtsQuery::CombinedFields(_) => "combined_fields".to_string(),
             FtsQuery::Boolean(_) => "boolean".to_string(),
         }
     }

nodejs/src/table.rs

@@ -11,11 +11,12 @@ use lancedb::table::{
     FieldMetadataUpdate as LanceFieldMetadataUpdate, FtsToken as LanceDbFtsToken,
     NewColumnTransform, OptimizeAction, OptimizeOptions, Ref, Table as LanceDbTable,
 };
+use napi::Env;
 use napi::bindgen_prelude::*;
 use napi::threadsafe_function::{ThreadsafeFunction, ThreadsafeFunctionCallMode};
 use napi_derive::napi;
 
-use crate::blob::{BlobFile, copy_blob_buffers, parse_row_ids};
+use crate::blob::{BlobFile, copy_blob_buffers, parse_row_ids, spawn_abortable};
 use crate::error::NapiErrorExt;
 use crate::index::Index;
 use crate::merge::NativeMergeInsertBuilder;
@@ -340,36 +341,44 @@ impl Table {
     }
 
     #[napi(catch_unwind)]
-    pub async fn fetch_blobs(
+    pub fn fetch_blobs<'env>(
         &self,
+        env: &'env Env,
         column: String,
         row_ids: Vec<BigInt>,
-    ) -> napi::Result<Vec<Option<Buffer>>> {
+        signal: Option<AbortSignal>,
+    ) -> napi::Result<PromiseRaw<'env, Vec<Option<Buffer>>>> {
         let row_ids = parse_row_ids(row_ids)?;
-        let array = self
-            .inner_ref()?
-            .fetch_blobs(column.as_str(), &row_ids)
-            .await
-            .default_error()?;
-        Ok(copy_blob_buffers(array))
+        let table = self.inner_ref()?.clone();
+        spawn_abortable(env, signal, async move {
+            let array = table
+                .fetch_blobs(column.as_str(), &row_ids)
+                .await
+                .default_error()?;
+            Ok(copy_blob_buffers(array))
+        })
     }
 
     #[napi(catch_unwind)]
-    pub async fn fetch_blob_files(
+    pub fn fetch_blob_files<'env>(
         &self,
+        env: &'env Env,
         column: String,
         row_ids: Vec<BigInt>,
-    ) -> napi::Result<Vec<Option<BlobFile>>> {
+        signal: Option<AbortSignal>,
+    ) -> napi::Result<PromiseRaw<'env, Vec<Option<BlobFile>>>> {
         let row_ids = parse_row_ids(row_ids)?;
-        let files = self
-            .inner_ref()?
-            .fetch_blob_files(column.as_str(), &row_ids)
-            .await
-            .default_error()?;
-        Ok(files
-            .into_iter()
-            .map(|file| file.map(BlobFile::new))
-            .collect())
+        let table = self.inner_ref()?.clone();
+        spawn_abortable(env, signal, async move {
+            let files = table
+                .fetch_blob_files(column.as_str(), &row_ids)
+                .await
+                .default_error()?;
+            Ok(files
+                .into_iter()
+                .map(|file| file.map(BlobFile::new))
+                .collect())
+        })
     }
 
     #[napi(catch_unwind)]

nodejs/tsconfig.json

@@ -10,7 +10,10 @@
     "resolveJsonModule": true,
     "emitDecoratorMetadata": true,
     "experimentalDecorators": true,
-    "moduleResolution": "Node"
+    "moduleResolution": "Node",
+    // flatbuffers 25 (via apache-arrow 21) ships declarations that need
+    // TypeScript 5.7+ generic typed arrays; skip checking third-party .d.ts.
+    "skipLibCheck": true
   },
   "exclude": ["./dist/*", "./examples/*"],
   "typedocOptions": {