feat: add Uint8Array support for zero-copy worker_threads transfer

[umeshmore45]

Description

This PR adds comprehensive Uint8Array support to enable zero-copy data transfer to worker_threads, significantly improving performance for multi-threaded applications handling large payloads.

Motivation

Currently, Redis values are returned as Buffer, which requires copying when transferring to worker threads. With Uint8Array support, the underlying ArrayBuffer can be transferred with zero-copy semantics.

Performance Benefits:

• Zero memory copying when transferring to workers
• Reduced memory usage for large payloads
• Better performance in multi-threaded applications

For a 10MB payload, this eliminates 10MB of memory copying when transferring between threads.

Usage Example

import { createClient, RESP_TYPES } from 'redis';
import { Worker } from 'worker_threads';

const client = await createClient().connect();

// Get data as Uint8Array
const data = await client
  .withTypeMapping({
    [RESP_TYPES.BLOB_STRING]: Uint8Array
  })
  .get('large-file');

// Zero-copy transfer to worker
const worker = new Worker('./worker.js');
worker.postMessage(data, [data.buffer]);

// After transfer, data.byteLength === 0 (ownership transferred)

Changes Made

Type System (types.ts):

• Extended BlobStringReply, SimpleStringReply, VerbatimStringReply to support Uint8Array
• Updated UnwrapConstructor and ReplyWithTypeMapping to handle Uint8Array
• Added Uint8Array to RedisArgument type

Decoder (decoder.ts):

• Added Uint8Array conversion in 4 string decoding methods (single/multi-chunk)
• Handles both simple strings and blob strings correctly
• Maintains backward compatibility with Buffer

Encoder (encoder.ts):

• Added Uint8Array handling in command encoding
• Properly ordered Buffer check before Uint8Array (Buffer extends Uint8Array)
• Updated error messages to include Uint8Array

Command Support:

• Updated type guards in HSET, GEOSEARCH, SEARCH, AGGREGATE to accept Uint8Array arguments
• Fixed Buffer/Uint8Array handling in cluster-slots.ts and time-series/helpers.ts

Testing:

• Added 4 decoder tests (simple/blob/verbatim strings)
• Added 1 encoder test
• Added 1 integration test for type mapping
• All existing tests continue to pass

Breaking Changes

None! This is fully backward compatible:

• Buffer remains the default type
• Existing code continues to work unchanged
• Uint8Array is opt-in via withTypeMapping()

Files Changed

• 12 files modified
• 95 lines added, 19 lines removed
• Net addition: 76 lines

Closes #3106

---

Checklist

• Does npm test pass with this change (including linting)?
• Is the new or changed code fully tested?
• Is a documentation update included (if this change modifies existing APIs, or introduces new ones)?

Note: Documentation can be added in a follow-up PR if needed. The code includes inline comments and the PR description provides usage examples.

---

Note

Medium Risk
Touches core request/response encoding/decoding, reply typing, and queue/caching invalidation paths; regressions could affect many commands and pub/sub/monitor behavior. Changes are largely additive and backed by targeted regression tests for Uint8Array mappings.

Overview
Adds first-class Uint8Array support end-to-end: RedisArgument and reply types now include Uint8Array, and the RESP decoder/encoder can emit/accept plain Uint8Array when configured via typeMapping.

Fixes a set of Uint8Array regressions caused by Buffer-only guards and Uint8Array#toString() semantics, including numeric parsing (GEODIST, ZRANK WITHSCORE), tuple/map key decoding (e.g. HGETALL, PUBSUB NUMSUB), cluster slot calculation, and RESP2 pub/sub/monitor/reset handling in commands-queue.

Updates client-side cache invalidation to normalize Redis keys consistently across string/Buffer/Uint8Array and adds comprehensive tests covering encoding/decoding, type-mapping, command argument parsing, and cache/queue behaviors.

^{Written by Cursor Bugbot for commit 64936ec. This will update automatically on new commits. Configure here.}

[jit-ci]

Hi, I’m Jit, a friendly security platform designed to help developers build secure applications from day zero with an MVS (Minimal viable security) mindset.

In case there are security findings, they will be communicated to you as a comment inside the PR.

Hope you’ll enjoy using Jit.

Questions? Comments? Want to learn more? Get in touch with us.

[nkaradzhov]

Thanks @umeshmore45, i will take a look

[nkaradzhov]

Please address the below:

1. Uint8Array keys are misclassified in sorted-set key parsing (ZINTER/ZUNION/ZUNIONSTORE).
   Add parseArgs regression cases in ZINTER.spec.ts, ZUNION.spec.ts, and ZUNIONSTORE.spec.ts for single and array Uint8Array keys.
2. CONFIG SET single-parameter form treats Uint8Array as object entries.
   Add regression in CONFIG_SET.spec.ts verifying single-parameter Uint8Array input uses the correct branch.
3. CMS.MERGE has the same plain-argument discriminator bug for Uint8Array[].
   Add regression in MERGE.spec.ts for plain Uint8Array[] sources (no weights).
4. Client-side cache invalidation key normalization is inconsistent for Uint8Array vs Buffer/string.
   Add regression in cache.spec.ts proving a cached Uint8Array key is invalidated by equivalent server-style invalidation payloads.
5. Tuple/map key transforms use .toString() on Uint8Array, producing byte-list keys ("102,111,111").
   Add regressions in HGETALL.spec.ts and PUBSUB_NUMSUB.spec.ts asserting semantic string keys are preserved under Uint8Array mapping.
6. Numeric transforms parse Uint8Array replies as NaN.
   Add command-level regressions in GEODIST.spec.ts, ZRANK_WITHSCORE.spec.ts, and GET.spec.ts.
7. RESET handling in queue does not recognize Uint8Array SIMPLE_STRING replies and can hang in monitor/type-mapped flows.
   Add a new queue unit test (under packages/client/lib/client, e.g. commands-queue.spec.ts) covering monitor with SIMPLE_STRING -> Uint8Array and ensuring reset() resolves.
8. RESP2 pub/sub PING resolution path ignores Uint8Array mapping behavior.
   Add queue-level regression test proving RESP2 pub/sub PING respects SIMPLE_STRING mapping when mapped to Uint8Array.
9. Test strictness must be tightened so type regressions are detectable.
   Replace permissive deepEqual with deepStrictEqual where binary type identity matters in decoder.spec.ts:66.
   For all new Uint8Array assertions, require both instanceof Uint8Array and !Buffer.isBuffer(value) (including types.spec.ts:54).

For each item, include both the code fix and a regression test that would fail without the fix and pass with it.

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.