Generate CLI output schemas from command output types
メンテナーはふだん 1 日以内に返信
関連するプルリクエストがすでにマージされています。
- #1830 @sentry-junior による — マージ済み
評価
- 難易度
- 5/5
- 見積もり時間
- 1週間以上
- 初心者へのやさしさ
- 12/100
- issue の種類
- リファクタリング
- 明瞭さ
- おおむね明確
- 活発さ
- 停滞
- 技術スタック
- typescript
調査の方向性
Start with packages/cli/src/commands/alert/output-schemas.ts to see how API fields and hand-written CLI fields are declared alongside the TypeScript output types. The existing completion-target and command-api generators are the model for any build step, and packages/cli/AGENTS.md is where the new declaration convention would be documented. This is design work, and PR #1830 is already open against it, so pick it up only after confirming with maintainers which option is chosen.
索引モデルが issue の本文から書いたものです。
説明
Problem
Command output schemas (OutputConfig.schema) drive --fields completion, the JSON field tables in --help, and the generated skill references. Since #1783, the fields that come from the Sentry API are taken from the generated @sentry/api/valibot response schemas (for example packages/cli/src/commands/alert/output-schemas.ts). Fields the CLI adds itself are still written by hand, next to a TypeScript output type that describes the same data:
- target context:
org,project,orgSlug,targetProject - status and result flags: the alert
statuslabel,deleted,dryRun(DryRunOutputField) - computed values:
dsn,url, dashboardperiod/fetchedAt, widgetdata
So every command declares its JSON output twice: once as the TypeScript type its func yields and output.human consumes, and once as a valibot schema. Nothing checks that the two agree. The schema is documentation only (nothing validates output against it), so drift goes unnoticed. A field added to the type silently stops completing, and a renamed field keeps completing a path that no longer exists. Raised by @BYK in https://github.com/getsentry/toolkit/pull/1783#discussion_r4448809546 ("Shouldn't this be auto-generated?").
Goal
One source of truth per command's JSON output: the API part from @sentry/api, and the CLI-added part derived from the command's output type instead of written by hand.
Options
- Generate schemas from TypeScript types at build time. Add a
generate:*step, like the existing completion-target and command-api generators, that reads each command's declared output type and emits a valibot (or plain field-path) description. Candidates: the TS compiler API, or a types-to-schema tool. Descriptions would come from TSDoc on the type's properties.- Pros: types stay the source of truth, and descriptions sit next to the fields.
- Cons: output types are often inferred or built from
Pick/intersections over SDK types. The generator has to resolve those, and it has to know which type belongs to which command.
- Invert it: schema first, types derived. Write the CLI-added entries once as valibot and derive the TS type with
InferOutput<typeof Schema>, sofunc's return type andoutput.human's parameter come from the schema.- Pros: no new generator, and the compiler enforces that the two agree.
- Cons: valibot becomes the authoring format for output types, which means touching many commands.
- Keep both, add a type-level check. Keep hand-written schemas, but add a compile-time assertion per command, e.g.
Expect<Equal<keyof InferOutput<typeof Schema>, keyof Output>>, or a test that compares field paths.- Pros: smallest change.
- Cons: still two copies to maintain, and only drift is caught.
Option 2 looks like the best fit for buildCommand. Option 1 is worth it only if descriptions can come from TSDoc reliably.
Dependencies
- Descriptions for API fields need the upstream change from the #1783 plan:
@sentry/api's generated valibot has nov.description()actions today, so the CLI wraps generated fields inpipe(field, description(...))by hand. Until that lands, this issue only covers the CLI-added fields. - Related: #1387 (migrate fully to
@sentry/apiSDK types), and #1808 / thesentry-cli/json-transform-uses-fieldslint rule (jsonTransformmust applyfields).
Done when
- No command's output schema writes out by hand a field that also exists in its TypeScript output type, or a test/compile error fails when they diverge.
--fieldscompletion,--helpfield tables and the skill references keep working unchanged, and the generated docs don't change unexpectedly.packages/cli/AGENTS.mddocuments how to declare output for a new command.
via @BYK.
--
- 主要言語
- TypeScript
- スター
- 949
- フォーク
- 171
- 平均マージ
- 23時間 5分
- マージ済み PR(30日)
- 320
環境構築
このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。
はじめの一歩
- issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
- 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
- リポジトリをフォークし、ブランチを切って変更します。
- issue 番号を参照したプルリクエストを送ります。
getsentry/toolkit のほかの issue
-
CLI enhancement MCP
難易度 2/5 1〜3時間 初心者へのやさしさ 63/100
getsentry/toolkit#1908 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信
-
Docs MCP URL normalization retains query strings and fragments対応中かも @pcbeingused333 が 4 日前に担当しました。 オープンbug MCP
難易度 2/5 1〜3時間 初心者へのやさしさ 73/100
getsentry/toolkit#1618 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 72/100
getsentry/toolkit#1471 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信
-
難易度 4/5 3〜5日 初心者へのやさしさ 35/100
getsentry/toolkit#1925 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信
-
enhancement MCP
難易度 5/5 1週間以上 初心者へのやさしさ 25/100
getsentry/toolkit#1911 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信
getsentry/toolkit の issue をすべて見る
似ている issue
-
needs-area-label
難易度 2/5 1〜3時間 初心者へのやさしさ 66/100
メンテナーはふだん 2 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 76/100
liberation-data/drivine#615 ·
メンテナーはふだん 1 日以内に返信
-
backend bug ready-for-agent
難易度 2/5 1〜3時間 初心者へのやさしさ 78/100
geolens-io/geolens#3639 ·
メンテナーはふだん 1 日以内に返信
-
bug
難易度 2/5 1〜3時間 初心者へのやさしさ 76/100
radius-project/ai-extensions#1196 ·
メンテナーはふだん 1 日以内に返信
-
good first issue
難易度 1/5 1〜3時間 初心者へのやさしさ 88/100