Skip to content

i18n:report 与 dataset 完全没有 translator —— dataset 看板的表头/度量标签永远是英文(#4854 的范围外发现,拆单跟踪) #5199

Closed as not planned
Closed as not planned
i18n:report 与 dataset 完全没有 translator —— dataset 看板的表头/度量标签永远是英文(#4854 的范围外发现,拆单跟踪)#5199
Assignees
os-project-manager
Labels
domain:servicesenhancementNew feature or request

Description

@xuyushun441-sys
xuyushun441-sys
opened on Aug 4, 2026

从 #4854(已随 PR #5165 修复关闭)拆出的独立范围,原文「Worth deciding together with」一节,记录不认领。

事实(#4854 报告方实测)

已有的线索(可能不需要新翻译面)

同一 dashboard 实测:dimension name 与真实字段名一致的(current_grade)已渲染中文字段标签;不一致的(account_name ≠ 字段 name)显示英文。⇒ 若 resolver 在 dimension 命中真实字段名时回落到字段标签,大多数场景零新面解决。

两个方向(裁决位,先测量后动手)

  1. dimension→字段标签回落(小):resolveDatasetLabel 类比 fix(spec): 视图标签按运行时真实下发的文档形状解析(#4854) #5165 的修法,命中字段名即用字段翻译;photo 面不新增。剩余不命中的(改名 dimension)维持英文,记台账。
  2. 完整 translator 面(大):TranslationDataSchema 增 _datasets/_reports 键 + translator 注册 —— 新授权面,须按 未知键静默剥离仍是全仓默认:把 #3405 的 strict 收紧从一个 schema 推广到整个可授权面(ADR-0078 完整性闸门) #4001 战役标准 strict + 三处锚定,且要回答「谁产出这些翻译」。

按创业专注原则倾向方向 1 先行、方向 2 挂 v18 地图;实测(translator 管线的实际结构)可能推翻,届时回报。

Refs:#4854、PR #5165、packages/spec/src/system/i18n-resolver.ts。

Activity

  1. xuyushun441-sys
    added
    enhancementNew feature or request
    domain:spec
    on Aug 4, 2026
  2. os-zhuang
    added
    pm:queue
    on Aug 6, 2026

  3. os-zhuang commented on Aug 6, 2026

    @os-zhuang
    os-zhuang
    on Aug 6, 2026 – with Claude
    Contributor

    分诊（spec 车道欠账代扫，维护者 2026-08-05 指令）：挂 pm:queue，附 PM 方向裁定（留否决窗口）——方向 1 先行：resolveDatasetLabel 在 dimension 命中真实字段名时回落字段翻译（类比 #5165 已合修法，零新授权面，符合维护者「聚焦不铺大」口径与正文自己的倾向）；剩余改名 dimension 维持英文记台账。方向 2（_datasets/_reports 新翻译面）不在本单——那是新授权面，须按 #4001 战役标准另立单挂 v18。派发词写明：先测 translator 管线实际结构，实测推翻方向 1 时按 needs_decision 返回勿硬做。不构成认领。会话：session_01N3uGFF8teXbpgtbEJ1aYXu

    Generated by Claude Code

  4. os-zhuang
    added
    pm:queue
    domain:services
    and removed
    domain:spec
    on Aug 8, 2026

  5. os-zhuang commented on Aug 8, 2026

    @os-zhuang
    os-zhuang
    on Aug 8, 2026 – with Claude
    Contributor

    Re-triage after the dispatch handback (comments 5224290471 / 5224300457): domain:spec → domain:services, pm:queue restored. This seat is the routing producer, so the relabel is applied here rather than debated.

    Operative scope (route A from the dev report, which supersedes the title/body's refuted translator framing): populate label in buildFieldMeta from the compiled cube, in both strategies. Verified on origin/main @ a36db28: native-sql-strategy.ts:790 declares Array({ name: string; type: string }) and pushes { name, type } only; the objectql strategy's buildFieldMeta is the same shape; dataset-executor.ts:1041 already documents the adjacent #5537 consequence. The dataset contract declares dimensions[].label / measures[].label (I18nLabelSchema), and three consumers already read fields[].label first — declared-but-never-delivered with readers in place is a restore-invariant, zero new authorization face, which is why this goes to the queue and not the maintainer's inbox.

    Boundaries carried forward unchanged: direction 2 (_datasets/_reports translator face) stays out, parked for v18 per the 2026-08-06 triage ruling. Escalation trigger for the dev: if delivering the label turns out to require a new wire key rather than populating the already-declared one, stop and return needs_decision.

    Adjacent, not duplicates (cross-links): #5141 (report renderer ignores the measure label it is handed — renderer side), #5076 (i18n bundle never applied on analytics surfaces). Different mechanisms; this card is the producer side.

    本评论来自分诊座位 Routine（#5474 试点），不构成认领。

    Generated by Claude Code

  6. os-project-manager
    added
    pm:dispatched
    on Aug 8, 2026
  7. os-project-manager
    self-assigned this
    on Aug 8, 2026

  8. os-project-manager commented on Aug 8, 2026

    @os-project-manager
    os-project-manager
    on Aug 8, 2026 – with Claude
    Collaborator

    CLAIM — services-lane PM session session_01USNUyHEr7uaU6MoEWXitei, branch claude/issue-5199-dataset-field-label.

    Claim history re-read before assigning: the earlier claim (session_011M7UwH25Unfi73UHim7ajY, i18n sweep) was released by its own handback (comments 5224290471 / 5224300457) after its dev falsified the card's original premise, and the re-triage at 05:58Z routed the card here with an operative scope. No live competing claim.

    Dispatching route A exactly as the re-triage framed it: populate label in buildFieldMeta from the compiled cube in both strategies (native-sql-strategy.ts, objectql-strategy.ts), so the dimensions[].label / measures[].label the dataset contract already declares reach the three consumers that already read fields[].label first. Boundaries carried into the dispatch verbatim: direction 2 (_datasets/_reports translator face) stays out and parked for v18; escalation trigger — if delivering the label turns out to need a new wire key rather than populating the already-declared one, the dev stops and returns needs_decision rather than inventing surface. The refuted translator framing in the title/body is superseded by route A; the title will be corrected at merge.

    Generated by Claude Code

  9. os-project-manager commented on Aug 8, 2026

    @os-project-manager
    os-project-manager
    on Aug 8, 2026 – with Claude
    Collaborator

    关闭:两次范围先后被实测证伪;存活的那个缺口已另立 #6761(决策箱)。 派发不产出 PR —— dev 按派发令的「实测推翻即停」条款停手,未硬做。

    第二次证伪(本轮,route A)

    派发范围来自 05:58Z 重分诊:「DatasetResultField.label 从未被填充,在两个 strategy 的 buildFieldMeta 从 compiled cube 补上」。该前提为假:

    照 route A 硬做会有害,不只是冗余:compiled cube 的 label 在 dataset-compiler.ts:374/406 缺省回落为机器名(typeof d.label === 'string' ? d.label : d.name),而下游富化带 if (f.label == null) 守卫 —— 一个没声明 label 的 dataset 会被发出 label:'owner',并永久顶掉本该由文档提供的标签。

    反向验证也倒过来做了(没有「我的修复」可撤,就去撤已发货的富化):摘掉那两行 ⇒ 17 条 pin 转红,其中两条的用例名直接点题(enriches measure fields with their display label + format、enriches dimension columns with their dataset display label)。这 17 条就是「route A 的前提不成立」的最硬证据。

    第一次证伪(2026-08-08 03:2xZ,i18n sweep 座位)

    方向 1(resolveDatasetLabel 回落字段翻译)的前提同样已死:该函数两仓都不存在,而它描述的行为早在 objectui 14f6999(08-01,早于本单开单三天)就已实现。详见评论 5224290471 / 5224300457。

    为什么关闭而不是第三次重划范围

    标题与正文写的是 translator 缺口,两次重划的范围又都被测量推翻;第三次把一张尚未拍板的决策塞进这个标题,只会让下一个 dev 继续读到错的题面。存活的缺口——内联 locale map 形式的 label 被生产端整个丢弃(typeof … === 'string' 守卫 + compiler 的 d.name 替换),而三个渲染端都在等——已带完整测量另立 #6761,挂 needs-user-decision:修法必须先回答「谁解析内联 map、解析器住哪」,而后端今天零个这样的解析器,第一个由那张卡种下,属契约取向,不由车道自选。

    state_reason: not_planned —— 本卡描述的工作没有被执行,因为它不该被执行;不是「已完成」。

    经办:services 车道 PM session_01USNUyHEr7uaU6MoEWXitei。

    Generated by Claude Code

  10. os-project-manager
    removed
    pm:queue
    pm:dispatched
    on Aug 8, 2026
  11. os-project-manager
    closed this as not plannedon Aug 8, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

domain:servicesenhancementNew feature or request

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions