fix(isBuffer): add browser export condition to avoid 44KB Buffer polyfill

[jantimon]

• adds a browser-safe isBuffer stub (() => false) and a rollup alias plugin that swaps it in during the browser build
• adds browser export condition to every subpath in package.json so bundlers like webpack/vite pick up the polyfill-free build automatically
• refactors all inline typeof Buffer / Buffer.isBuffer() checks across mergeWith, isEmpty, cloneDeepWith, and isEqualWith to use the centralized isBuffer function

closes #1670

Build size comparison (Next.js test app)

Webpack

	Raw	Gzipped
Before	458 KB	143 KB
After	438 KB	136 KB
Diff	-19.3 KB	-6.5 KB (4.5%)

Turbopack

	Raw	Gzipped
Before	460 KB	134 KB
After	416 KB	116 KB
Diff	-44.3 KB	-18.2 KB (13.6%)

Test plan

• yarn vitest run — all existing tests pass (isBuffer tests run against the real Node.js implementation)
• yarn lint / tsc --noEmit — clean
• Verified with a Next.js app: .next/static drops from 492K to 464K after the change
• grep -r "Buffer" dist/browser/ confirms no Buffer references in the browser build

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
es-toolkit	Ready	Preview, Comment	Apr 18, 2026 10:14am

[Copilot]

Pull request overview

Adds a browser-safe build path for isBuffer to prevent bundlers (e.g. webpack/Next.js, Vite) from injecting large Buffer polyfills, and refactors internal call sites to use the centralized predicate.

Changes:

• Introduces isBuffer.browser.ts stub and a Rollup alias to swap it into the browser ESM/UMD builds.
• Adds browser export conditions for top-level subpath exports in publishConfig.exports.
• Replaces inline Buffer checks in several utilities (isEqualWith, cloneDeepWith, isEmpty, mergeWith) with isBuffer().

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/predicate/isEqualWith.ts	Refactors Buffer detection in typed-array comparison to use isBuffer().
src/predicate/isBuffer.browser.ts	Adds browser-safe isBuffer stub used by the browser build.
src/object/cloneDeepWith.ts	Switches Buffer cloning branch to use centralized isBuffer().
src/compat/predicate/isEmpty.ts	Replaces inline Buffer check with isBuffer() in compat isEmpty.
src/compat/object/mergeWith.ts	Uses isBuffer() when deciding whether to deep-clone Buffer source values.
rollup.config.mjs	Adds browser ESM build output + alias plugin wiring for isBuffer stub.
package.json	Adds browser conditional export mappings under publishConfig.exports.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated no new comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[kms0219kms]

LGTM

[raon0211]

Hello, thanks for your pull request!

Your change definitely fixes the bundle size bloat. That said, I'm a bit concerned that having multiple builds could lead to larger install sizes over time, and it would make adding additional subpath exports more complex down the line.

As an alternative, I took a different approach and used globalThis.Buffer instead of Buffer so that the Buffer polyfill isn't pulled in automatically. Could you take a look and test it on your end?

[jantimon]

when fixing this bug claude also suggested the globalThis hack to trick bundlers in not seeing that Buffer is used
so it is a bet that bundlers will never improve global object detection

to be more future proof and to ensure it works with every bundler I removed the Buffer part entirely so a bundle update won't silently add this bug again

also targeted builds are always smaller than isomorphic builds

but I can totally understand that you don't want to maintain the more complex build setup

I will try the globalThis hack in webpack turbopack and tanstack start and let you know - rspack2 is hard to test right now as it is still alpha

[jantimon]

hey @raon0211 tested the globalThis.Buffer route, tested it across 3 bundlers

just realised you already changed it on this branch

numbers on a minimal next 16 + tanstack start (vite 8) app that imports cloneDeep client-side:

bundler	delta raw	delta gz
webpack	-21.97	-6.68
turbopack	-22.12	-6.69
vite	±0	±0

turbopack number is smaller than what I got here originally (-44 KB)

it already rewrites bare Buffer → globalThis.Buffer itself, so no polyfill ever got injected there.

sanity check on the next/webpack client chunk: pages/index-*.js drops from 26 KB → 3.7 KB, and _isBuffer /
INSPECT_MAX_BYTES symbols disappear entirely.

#1691 is just the globalThis change

[jantimon]

I also measured your approach is ~250 bytes larger in the browser after minification - I guess that's not much just wanted to state it because es-toolkit is build for optimal size and speed

[raon0211]

Thanks for your help!

[raon0211]

I also measured your approach is ~250 bytes larger in the browser after minification - I guess that's not much just wanted to state it because es-toolkit is build for optimal size and speed

Yup, thanks for pointing that out. I think that's an acceptable tradeoff to keep the overall library complexity down.

[jantimon]

@raon0211 thanks for releasing!