Release Notes

Version 1.3.0

📅 Release Date

September 11, 2026

📖 Overview

Two helpers arrive in utils/, and both are here for the same reason: they had already been written twice in the estate's applications, and the copies had started to disagree.

chunk existed in two applications with genuinely different edge-case behaviour — one treated a non-positive size as 1, the other returned the whole input as a single run. Json.parse existed in two other applications, identical down to the doc comment, differing only in the example. Neither divergence had caused an incident. Neither could be caught by a test, because no test could see both copies.

That is the test utils/ sets for itself, stated in its own remarks: a helper that is pure, vendor-neutral, and already written twice in two consumers is one those consumers should share rather than keep diverging copies of. Both qualify. Nothing else in this release changes the surface.

⚠️ Breaking Changes

• None. Both additions are new exports; no existing export is removed, renamed, or narrowed.

🚀 Features

• List.chunk(values, size) splits a list into consecutive runs of at most size. Written for the case where a downstream call has a ceiling on how much it will take at once — a database batch bounded by statements or bound parameters, an API bounded by ids per request — and the caller knows that ceiling while this function does not. The bound is the caller's to name.

  On the reconciliation: a size that is non-positive or non-finite yields a step of 1 rather than looping forever or collapsing the input into a single run. A size read from configuration can arrive as 0 or NaN, and of the two behaviours the estate had, a step of 1 is the one that degrades safely — the alternative turns a concurrency limiter into an unbounded fan-out precisely when its configuration is wrong.

  Both existing call sites already guard against a non-positive size before reaching chunk, so adopting this is behaviour-preserving for them today. The reconciliation matters for the next caller, not the current ones.

• Json.parse<T>(text) parses without throwing, answering null for every way the input fails to be a usable object — malformed syntax, an empty body, or a bare null literal — because none of them is something a handler can act on differently. A falsy primitive that is still a value (0, false) comes back intact.

🔧 Enhancements

• CI follows the estate's current workflow shape. features.yml, main.yml, and release.yml were updated in step with the other packages, so a change to the shared reusable workflows lands the same way here as everywhere else.
• wrangler.json carries its $schema and the reference binding order. Repository configuration only — nothing in package.json#files changes — but this file is what a consumer copies when wiring their own bindings, so its ordering is documentation.
• CODE_OF_CONDUCT.md was brought in line with the Contributor Covenant text the rest of the estate carries.

🐛 Bug Fixes

• None. No prior behaviour changed.

🔐 Security

• The kernel is still dependency-free. dependencies and peerDependencies both remain empty, and neither helper reaches for anything: List.chunk is array arithmetic and Json.parse wraps a global. The no-restricted-imports rule naming no vendor, runtime, or HTTP framework is unchanged and still passing.
• qs bumped in the development tree via Dependabot. A devDependency only — it does not ship, since package.json#files publishes lib/ alone.

🧪 Tests

npm run lint, npm run test:run, npm run build, and npm run build:docs all clean. 164 specs across 12 suites, up from 145 across 10.

• test/utils/List.spec.ts covers the boundaries that matter — an exact multiple, a remainder, an empty input — and the arguments a caller should never pass but eventually does, since a size of 0 or NaN from configuration would otherwise not terminate.
• test/utils/Json.spec.ts covers each way input fails to be usable, and asserts that 0 and false survive, which is the case a naive falsy check gets wrong.

📚 Documentation

• utils/ states its own admission test. The module remarks now name why Signature, List and Json are here although the kernel itself calls none of them, and why an application's own helpers — the ones encoding decisions about its input and its domain — are not.
• docs/changes-log/2026-09-05.md records the working detail behind List.

⬆️ Upgrading

npm install @bayudwiyansatria/core@1.3.0

Nothing to change on upgrade. To adopt the helpers, delete the local copy and re-export in its place, which keeps every existing import path working:

// src/utils/chunk.ts — entire file
export { List } from '@bayudwiyansatria/core'

export const chunk = List.chunk

Check the edge-case behaviour of the copy being deleted before replacing it. Where a caller can pass a size from configuration, confirm it guards against 0 — this release's semantics give a step of 1 there, which is not what every previous copy did.

🚨 Known Issues

• None.

📦 Dependencies

Unchanged. The kernel names no vendor and takes no dependency by design.

👥 Contributors

• Bayu Dwiyan Satria

🙏 Acknowledgments

Thanks to the four applications that carried these helpers independently long enough to make the case for sharing them — and, in chunk's case, long enough for the two copies to disagree, which is the clearest argument this package makes for its own existence.

For more information, visit the project's GitHub repository.