Who maintains tree-sitter-typescript now?

[inth3shadows]

tree-sitter-typescript hasn't had a commit since 75b3874 (2025-01-30). It has 24 open PRs (the oldest is tree-sitter/tree-sitter-typescript#302 from June 2024) and 34 open issues. None of the PRs opened since then has a maintainer review.

The backlog now includes duplicate PRs, so contributors are fixing the same bugs twice:

• export type *: fix: parse export type * tree-sitter-typescript#358 and feat: support export type * and export type * as Ns re-exports (TypeScript 5.0) tree-sitter-typescript#360
• variance annotations (<in T>, <out T>, TS 4.7): feat: support variance annotations on type parameters (TypeScript 4.7) tree-sitter-typescript#361 and feat: support variance annotations on type parameters tree-sitter-typescript#364. Both currently break type F<out> = out[], which master parses (details in the PR comments). Another user has confirmed the bug on current master in Type-parameter variance annotations (in / out / in out, TS 4.7) are not in the grammar — every use fails to parse tree-sitter-typescript#370, but no maintainer has replied.
• type arguments on tagged template calls: fix: add optional type arguments to template call tree-sitter-typescript#342 and fix: parse type arguments on tagged template call tree-sitter-typescript#350

Downstream tools are starting to carry their own patches or forks instead. At least one fork already carries a fix for tree-sitter/tree-sitter-typescript#364.

Two questions:

1. Is anyone currently responsible for reviewing PRs here? If so, is there a better place to bring them to their attention?
2. If not, could the repo get another maintainer, or move to tree-sitter-grammars like other community-maintained grammars?

I'm glad to help review or consolidate the duplicate PRs if that would help.

[clason]

Tree-sitter grammars under the org have their own individual maintainers (for historical reasons; they are in the org partly because they are test fixtures, and partly for historical reasons); maintainers on tree-sitter (this project) don't automatically have privileges there.

I have no interest in maintaining a typescript grammar (and, to be frank, little trust in people I don't know), so moving this to tree-sitter-grammars would not essentially change the situation.

But they're open source, so anybody is free -- and welcome! -- to fork and maintain the grammar; downstream tools will switch to the active one.

[inth3shadows]

A data point for whoever picks this up, since forking is the suggested path.

I built a differential check that treats tsc (syntax diagnostics only) as ground truth and diff-parses pinned real-world corpora with a grammar's wasm. The corpora are Effect, Zod, Excalidraw and VS Code src/ (about 14.5k files), plus one-construct fixtures. Results at the pinned SHAs:

grammar	files tsc accepts that fail to parse
upstream 75b3874 (v0.23.2)	259
v0.23.2 + a variance patch (based on #364 plus the out-as-name alias from brajevicm@9fcaa13)	201
SegaraRai@82b3609 (most-ahead fork, 15 commits)	206

SegaraRai fixes using/await using in for-of heads, static accessor, and 8 VS Code files. It also regresses 15 files that master parses:

• reserved words as import/export aliases (export { x as null }): 7 Effect/Zod files
• typeof import('x').default in a type position: 6 VS Code files
• out as a type-parameter name (<out = string>, tsx <out,>() =>), the same regression as Don't use clock_gettime on macOS #361/Incorrect parsing behavior when NDEBUG macro is defined #364

Still open on every grammar I tested: generic call signatures on their own line (#335, the main cause of whole-file ERRORs in Effect) (see edit) and import attributes on export … from.

The check is a single script that loads any grammar's wasm: scripts/ts-differential.mjs (docs: docs/grammars/ts-differential.md). Anyone maintaining a fork is welcome to run it against their branch. A fork that passes it with no regressions is one downstream tools could switch to with some confidence.

Edit (2026-11-05): I ran three more forks through the same check, and the paragraph above was wrong about #335. Files tsc accepts that fail to parse, across the same 4 corpora:

grammar	failing files
marionebl/tree-sitter@53071c9 (7 commits)	45 (Effect 185 → 29)
dokimasia@839d520	153
SegaraRai@82b3609	206
BrokkAi@bb76a11	237

• marionebl fixes How to package for electron #335 with a 16-line scanner change (4cbff7a). That clears 15 of Effect's 16 whole-file ERRORs, including Schema.ts, Stream.ts and Channel.ts, with no file made worse. Its one regression is out as a type-parameter name (type A<out> = out[]), the same as Don't use clock_gettime on macOS #361/Incorrect parsing behavior when NDEBUG macro is defined #364.
• dokimasia also fixes How to package for electron #335 and clears every Excalidraw and VS Code failure (import types, typeof import()), but it has no variance support, so 59 files regress.
• BrokkAi has neither fix.