[Patrick-SCH03]

Summary

Fixes es-toolkit/compat's groupBy, maxBy, and minBy to call iteratees with only the current element, matching Lodash. Passing (element, index, array) instead can produce different results for callbacks such as parseInt, even though the equivalent Lodash calls compile in strict mode with @types/lodash. For groupBy, this mismatch was introduced when #1561 added index and array arguments to the main implementation used by compat.

import { groupBy, maxBy, minBy } from 'es-toolkit/compat';

groupBy(['10', '11', '12'], parseInt);
// lodash:            { '10': ['10'], '11': ['11'], '12': ['12'] }
// es-toolkit/compat:  { '1': ['12'], '10': ['10'], 'NaN': ['11'] }

maxBy(['10', '9', '11'], parseInt);
// lodash: '11'    es-toolkit/compat: '10'

minBy(['10', '9', '11'], parseInt);
// lodash: '9'     es-toolkit/compat: '11'

Changes

• Call getValue(element) instead of getValue(element, i, array) in compat maxBy and minBy.
• Use a local loop in compat groupBy to pass only the element to the iteratee, rather than delegating to the main implementation or adding a wrapper callback.
• Leave the main es-toolkit implementations unchanged, including their index and array arguments.
• Add regression tests for single-argument iteratee calls and parseInt behavior in all three functions. These tests fail on main and pass on this branch.

Validation

All 3,301 compat and array tests pass. Type checking, Prettier, and ESLint checks pass; the two ESLint warnings also exist on main.

Benchmarks

Measured locally on Windows with Node.js 22, comparing main (d6f79606) against this branch (0666b800). The benchmarks workspace failed with ERR_UNSUPPORTED_ESM_URL_SCHEME on Windows, so equivalent benchmarks were run with vitest bench, importing directly from src/compat.

Results are in ops/sec; higher is better. Each groupBy cell contains results from two separate runs.

Function	Case	main	This branch
groupBy	5 items, using the data from benchmarks/performance/groupBy.bench.ts	4,973,898 / 4,749,554	4,725,273 / 5,008,803
groupBy	10,000 items	5,528 / 4,787	5,749 / 6,027
maxBy	10,000 items	127,462	127,408
minBy	10,000 items	128,306	126,796

The final implementation showed no consistent slowdown across these local runs. An initial wrapper-based groupBy implementation reached 3,882,366–4,494,805 ops/sec in the 5-item case, so it was replaced with the local loop.

[vercel]

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

Project	Deployment	Actions	Updated
es-toolkit	Ready	Preview	Sep 27, 2026 6:09am UTC