vercel/next.js

Why this matters

next.js is the most-watched JS / React framework on GitHub (~140k stars). It is a hybrid pnpm + Cargo monorepo: a pnpm workspace (packages/*, apps/*, bench/*, crates/*/js, turbopack/packages/*, turbopack/crates/*/js) running alongside a Cargo workspace (crates/* + turbopack/crates/*

• scripts/send-trace-to-jaeger).

This is a distinct shape from microsoft/typescript (which IS the TS compiler — flat src/+tests/), from facebook/react (Yarn classic single-language workspace), and from vercel/turbo (modern Rust+TS hybrid with smaller surface). next.js is the canonical hybrid pnpm + Cargo mega-monorepo — and the canonical demonstration that no per-language linter sees the whole tree.

Headline catch

next.js is the first hybrid pnpm + Cargo dual-workspace win. alint’s polyglot bundle composition (monorepo/cargo-workspace@v1

• monorepo/pnpm-workspace@v1 layered together) is the tightest fit in the case-study catalogue, and surfaces drift no per-language linter catches because each linter only sees half the tree:

• 3 of 19 npm packages missing license fields
• 4 of 63 crates missing the standard MIT/MPL license
• gitattributes EOL pin sliding into drift would silently break Windows test runs
• The validate-externals-doc.js cross-file registry sync is a hand-rolled 600-LOC scripts/check-*.{js,mjs,sh} script family — alint’s v0.10 cross_file_value_equals ship-target collapses it to one declarative rule

prettier sees formatting. eslint sees TS/JS AST. cargo clippy sees Rust AST. None of them sees the cross-language conventions — license uniformity, EOL pin, husky-hook content integrity, the 19 canary-pinned package versions, the [patch.crates-io] ↔ [workspace] members consistency.

Where alint earns its keep here

• Bundled monorepo/cargo-workspace@v1 + monorepo/pnpm-workspace@v1 layered together cover both halves of the tree in one declarative file. No other tool composes ecosystem rules at this layer.
• The 11-bundled-ruleset extends + 59 next.js-specific rules (130 rules total post-resolution) cover every structural assertion the existing tooling makes about repo state, plus several next.js doesn’t enforce today (per-crate license uniformity, husky-hook content integrity, gitattributes EOL pin).
• The 7 hand-rolled scripts/check-*.{js,mjs,sh} (~600 LOC) collapse to 2 v0.10 rule kinds + a few command: shellouts. validate-externals-doc.js becomes cross_file_value_equals (8+ confirmations: airflow + tokio + next.js + clap + uv + react + pnpm + pytorch + tensorflow). check-manifests.js becomes registry_paths_resolve (6+ confirmations). Codegen + git-state-mutation scripts (check-examples.sh, check-pre-compiled.sh) stay shellouts — alint reads files; it doesn’t regenerate and diff.
• Pitfall #16 was discovered authoring this config. JSONPath rules can’t regex-match against JSON booleans (silent runtime error: “value at path is not a string (got bool), can’t apply regex”). Surfaced two existing case studies with silently-broken rules (microsoft-typescript + vercel-turbo). Now in the published CONFIG-AUTHORING.md catalogue. Strengthens the smoke-test fixture audit as a v0.10+ DX item.
• dir_name_matches_field extension with unscoping is a next.js-specific demand: packages/font/ → name: "@next/font", packages/next-mdx/ → name: "@next/mdx". The bare dir_name_matches_field from vercel/turbo would fire on every scoped package; the unscoping transform is the v0.10+ extension this case study motivates.

Future story angles

• Position as the fourth tile on alint.org/examples (after kubernetes, airflow, microsoft/typescript), with the angle: “for hybrid monorepos that span multiple ecosystems, no per-language linter sees the whole tree — alint is the layer that does.” The 11-bundled-ruleset compose-them-together story is the headline hook that no per-language linter (eslint / clippy / typos / ast-grep / alex) can reach.
• Hybrid-monorepo launch narrative — pair next.js with vercel/turbo (Rust monorepo orchestrator with zero hand-rolled scripts) for a “modern monorepo tier” launch tile pair. Same Vercel ecosystem; very different shapes; both have structural drift no per-language linter catches.
• scope_filter evolution refactor — the dual pnpm + Cargo workspace shape benefits sharply from named scopes (rust-workspace, js-workspace, js-bench); cuts ~40 lines
  • makes per-subtree intent explicit. Canonical demonstration of the v0.9.17 named-scope payoff.
• compliance/reuse@v1 + agent-hygiene@v1 adoption ladder rung — alint suggest against the live tree flagged agent-hygiene (medium); compliance/reuse is a deliberate override (next.js uses MIT directly, no per-file SPDX headers). Both worth documented decisions.