fix(node): support Apache Arrow 19 through 21 and run the library on each

Builds on #4359 (which added the Arrow 19-21 test aliases) and makes the
package genuinely work across the whole peer range:

* Resolve the library's own `apache-arrow` to 21.2.0 (devDependency) so
  the main CI jobs build, type-check and test on the newest release, and
  widen the peer range to `>=15.0.0 <22.0.0` instead of an exact pin.
* Add a CI job that re-installs with `apache-arrow` resolved to 19.0.1 and
  20.0.0 and runs the suite, covering what a consumer on those versions
  gets.
* Arrow 21 overrides `Symbol.hasInstance` on Schema/Field/Table/RecordBatch/
  Data so `instanceof` matches objects from another copy of the library,
  but not on DataType. The sanitizer's short-circuits therefore returned
  foreign schemas unchanged and Arrow's builders then failed with
  "Unrecognized type 'NONE'". Check the real prototype chain instead.
* Sanitize the types added in Arrow 20/21 (LargeList, Utf8View, BinaryView,
  IntervalMonthDayNano), looked up dynamically so the library still compiles
  and runs on older Arrow releases, with a clear error when the installed
  Arrow lacks the type. Carry `variadicBuffers` through `Data`.
* Loosen `FieldLike` so a branded DataType from another Arrow copy still
  type-checks; export the new `DataTypeLike`.
* Share the Arrow version list between test files, exercise embedding
  function types per version in the registry tests, and add sanitize and
  end-to-end coverage for the new column types.
* Enable `skipLibCheck`: flatbuffers 25 (via Arrow 21) ships declarations
  that need TypeScript 5.7+.
* Record the new dev dependencies in the third-party license inventory.

Co-Authored-By: Vincent Koc <vincentkoc@ieee.org>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

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

@@ -3,16 +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 arrow19 from "apache-arrow-19";
-import * as arrow20 from "apache-arrow-20";
-import * as arrow21 from "apache-arrow-21";
-
 import * as tmp from "tmp";
 
+import { arrowVersions } from "./arrow_versions";
+
 import { connect } from "../lancedb";
 import {
   EmbeddingFunction,
@@ -21,16 +15,6 @@ import {
 } from "../lancedb/embedding";
 import { getRegistry, register } from "../lancedb/embedding/registry";
 
-const arrowVersions = [
-  arrow15,
-  arrow16,
-  arrow17,
-  arrow18,
-  arrow19,
-  arrow20,
-  arrow21,
-];
-
 describe.each(arrowVersions)("LanceSchema", (arrow) => {
   test("should preserve input order", async () => {
     const schema = LanceSchema({
@@ -253,7 +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

@@ -7,14 +7,6 @@ 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 * as arrow19 from "apache-arrow-19";
-import * as arrow20 from "apache-arrow-20";
-import * as arrow21 from "apache-arrow-21";
-
 import {
   AutoQuery,
   Connection,
@@ -59,22 +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, arrow19, arrow20, arrow21])(
+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
-      | import("apache-arrow-19").Schema
-      | import("apache-arrow-20").Schema
-      | import("apache-arrow-21").Schema = new arrow.Schema([
+    const schema: InstanceType<ApacheArrow["Schema"]> = new arrow.Schema([
       new arrow.Field("id", new arrow.Float64(), true),
     ]);
 
@@ -748,6 +739,58 @@ describe.each([arrow15, arrow16, arrow17, arrow18, arrow19, arrow20, arrow21])(
 );
 
 // 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();
+    expect(schema.fields.map((f) => f.type.typeId)).toEqual([
+      arrow.Type.Int,
+      arrow.Type.LargeList,
+      // Lance stores the view types as their offset-based equivalents.
+      arrow.Type.Utf8,
+      arrow.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));
+    tmpDir.removeCallback();
+  },
+);
+
 it("should query documents with LangChain PDF metadata", async () => {
   const tmpDir = tmp.dirSync({ unsafeCleanup: true });
   try {
@@ -3209,7 +3252,7 @@ it("passes cleanupOlderThan to the native binding as an absolute timestamp", asy
   expect(optimize).toHaveBeenCalledWith(cutoff.getTime(), true);
 });
 
-describe.each([arrow15, arrow16, arrow17, arrow18, arrow19, arrow20, arrow21])(
+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: unknown };
 
 /**
  * Create an Arrow field backed by LanceDB's JSON extension type.

nodejs/lancedb/index.ts

@@ -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 {
@@ -76,6 +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>;
@@ -212,19 +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) {
@@ -451,11 +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");
@@ -572,7 +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) {
@@ -626,7 +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) {
@@ -653,7 +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) {
@@ -681,7 +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) {
@@ -710,11 +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);
@@ -732,7 +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,
@@ -745,6 +854,7 @@ function sanitizeData(
     },
     dataLike.children.map((child) => sanitizeData(child, context)),
     dictionary,
+    (dataLike as VariadicDataLike).variadicBuffers,
   );
   context.data.set(dataLike, data);
   return data;
@@ -753,7 +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(),
@@ -774,6 +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/package.json

@@ -49,6 +49,7 @@
     "@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",
@@ -105,7 +106,7 @@
   },
   "peerDependencies": {
     "@types/node": ">=22",
-    "apache-arrow": ">=15.0.0 <=21.2.0"
+    "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 <=21.2.0'
-        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
@@ -1852,10 +1852,6 @@ packages:
     resolution: {integrity: sha512-gFlPaqN9osetbB83zC29AbbZqGiCuFH1vyyPseJ+B7SIbfBtESV62mMT/CkiIt77W6ykC/nTWFzTXFs0Uldg4g==}
     hasBin: true
 
-  apache-arrow@18.1.0:
-    resolution: {integrity: sha512-v/ShMp57iBnBp4lDgV8Jx3d3Q5/Hac25FWmQ98eMahUiHPXcvwIMKJD0hBIgclm/FCG+LwPkAKtkRO1O/W0YGg==}
-    hasBin: true
-
   apache-arrow@19.0.1:
     resolution: {integrity: sha512-APmMLzS4qbTivLrPdQXexGM4JRr+0g62QDaobzEvip/FdQIrv2qLy0mD5Qdmw4buydtVJgbFeKR8f59I6PPGDg==}
     hasBin: true
@@ -5746,18 +5742,6 @@ snapshots:
       json-bignum: 0.0.3
       tslib: 2.8.1
 
-  apache-arrow@18.1.0:
-    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: 5.2.1
-      command-line-usage: 7.0.4
-      flatbuffers: 24.12.23
-      json-bignum: 0.0.3
-      tslib: 2.8.1
-
   apache-arrow@19.0.1:
     dependencies:
       '@swc/helpers': 0.5.21

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": {