Archived
xcsandClaude Opus 5
59afcbfdbd
推荐理由 / 摘要 / 风险 / 验证要求都是规则算出来的:让人直接改它们, 等于让他给自己没做的判断盖章。要改就改上面那一条四态,系统重算。 同批修一条与本次设计矛盾的既有断言:「差异审核与落账确认只读」原把 difference-review 也算只读,而 spec §4.4/§6.2 已把 S3 改成可就地编辑 ——改窄成只管 landing-confirm,differerence-review 的只读面改由新增的 「件②」测试组精确覆盖。 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LKMZDwMbyKzjw6Hembwvet
828 lines
51 KiB
TypeScript
828 lines
51 KiB
TypeScript
/**
|
||
* ACT-WORK-05 差异评估的包内自测 —— **只准 import @sdk / @contracts / @sfeel**。
|
||
* 案例活动没有 `run`,自测不跑回合;测的是清单本身与授权面对齐,结构照 E04 / I01。
|
||
*
|
||
* 七组:清单本身合规 / 台账声明 / 案例模型与清单的对齐 / 报告模版:编辑面与只读面各就各位 /
|
||
* 徽标别名与值域相关的坑(2026-08-27 实测账,回归网)/ 可编辑件的主键与必填列
|
||
* (2026-08-27 复审,回归网)/ 活动入口。
|
||
*/
|
||
import { describe, expect, it } from 'vitest';
|
||
import { activityManifestSchema } from '@contracts';
|
||
import { createMockHost } from '@sdk';
|
||
import { manifest, D } from '../src/manifest.ts';
|
||
import activity from '../src/index.ts';
|
||
|
||
/*
|
||
* 2026-08-29 Task 3 整体重写(spec §4):九步砍成四步,案例模型换成 CMMN + `case-model.ts` 里的
|
||
* `ASSESSMENT_STEPS` 四条剧本。这份包内自测原是照旧九步写的,`案例模型与清单的对齐` 这组的多数
|
||
* 断言在结构上仍然成立(CMMN humanTask id 与剧本 item 一一对应、除放行外每项恰属一个阶段…),
|
||
* 只是数字与名单要随新模型换 —— 下面这几处常量与断言标题已按 `case-model.ts` 的新四步同步更新。
|
||
*/
|
||
const ITEMS = ['S1_inputs', 'S2_assess', 'S3_review', 'T_release'];
|
||
/*
|
||
* 2026-08-29 Task 3:四步里只有 `S3_review`(审核,可循环)与 `T_release`(放行)不给
|
||
* `doneWhen`——判据不许摆在还欠一张卡的那一步眼前(2026-08-28 用户实测逼出的立墙,
|
||
* `humanWrites`/`acceptanceChecklist` 同一课)。`S1_inputs` / `S2_assess` 全 auto,
|
||
* 入库事实 / 分析结论一落地引擎就自己收步,不再需要人工激活等待。
|
||
*/
|
||
const NO_DONE_WHEN = new Set(['S3_review', 'T_release']);
|
||
/*
|
||
* 2026-08-29 Task 7:五张旧局面收成三张 —— `p08-master-create` / `p10-document-create` /
|
||
* `eval-input-list` 三张合并成一张 `input-supplement`(Task 7 已交付);
|
||
* `difference-review` / `landing-confirm` 由 Task 8 / 9 各自重做(同名)。
|
||
*/
|
||
const VIEWS = ['input-supplement', 'difference-review', 'landing-confirm'];
|
||
/** 运营中心 L2 沙箱里那张关键台账数据看板 —— 不是工单件,只在「都声明了」这一条里出现;自己那组断言在文件尾 */
|
||
const BOARD_VIEW = 'activity-ledger';
|
||
const byAddress = (domain: string, object: string) => manifest.declarations.ledger.find((l) => l.domain === domain && l.object === object);
|
||
const viewHtml = (name: string) => manifest.assets.viewTemplates.find((t) => t.name === name)?.html ?? '';
|
||
/* 「台账声明」组用的短写(2026-08-29 重写引入):地址是 `域.对象` 一整串,`byAddress` 拆两段 */
|
||
const ledgerOf = (address: string) => manifest.declarations.ledger.find((l) => `${l.domain}.${l.object}` === address);
|
||
const colOf = (address: string, name: string) => ledgerOf(address)?.fields.find((f) => f.name === name);
|
||
|
||
describe('清单本身合规', () => {
|
||
it('过 @contracts 的 schema', () => {
|
||
expect(() => activityManifestSchema.parse(manifest)).not.toThrow();
|
||
});
|
||
|
||
it('没有静态监听(等外部回音一律落在执行工单上,listen 在本仓不可用)', () => {
|
||
expect(manifest.declarations.listeners).toEqual([]);
|
||
});
|
||
|
||
/* 2026-08-29 重写:spec §7 的判据表从九条改成八条(五条机检 L1/L2 + 两条 L4 裁量 + 一条 L5),L4 从三条收到两条 —— 证据链与硬约束两条留 L4,「推荐等级与明细一致」已升级成 L2 机检(level-by-rule) */
|
||
it('L1/L2 判据都给了谓词;L4 两条不给(裁量项);L5 恰一条', () => {
|
||
for (const c of manifest.contract.acceptanceCriteria) {
|
||
if (c.level === 'L1' || c.level === 'L2') expect(c.predicate, c.key).toBeDefined();
|
||
if (c.level === 'L4') expect(c.predicate, c.key).toBeUndefined();
|
||
}
|
||
expect(manifest.contract.acceptanceCriteria.filter((c) => c.level === 'L4')).toHaveLength(2);
|
||
expect(manifest.contract.acceptanceCriteria.filter((c) => c.level === 'L5')).toHaveLength(1);
|
||
});
|
||
|
||
it('机检谓词只读基座读数与记账旗:不出现活动自报的汇总列名', () => {
|
||
const forbidden = ['relation_total', 'candidate_total', 'recommended_total', 'not_recommended_total'];
|
||
for (const c of manifest.contract.acceptanceCriteria) {
|
||
for (const name of forbidden) expect(c.predicate ?? '', `${c.key} 读了自报数 ${name}`).not.toContain(name);
|
||
}
|
||
});
|
||
|
||
/*
|
||
* 2026-08-29 重写:必交付物从三件(评估输入清单 / 差异评估结论 / 拟落账内容)收到两件 ——
|
||
* 「评估输入清单」那件连同它绑的三张旧单据(pn_check/input_gap/verdict)一起退休。
|
||
* 人环预算从 16(候选放弃/创建决策等岔路)收到 8(补数据 / 审核 / 放行三次调人,零工具、
|
||
* AI 只出场一次)—— 见「契约 > 预算」那条把 8 钉死;这里只需两个数上下一致。
|
||
*/
|
||
it('两件必交付物都有台账地址;人环预算恰为 8', () => {
|
||
expect(manifest.contract.requiredDeliverables).toHaveLength(2);
|
||
for (const d of manifest.contract.requiredDeliverables) expect(d.ledger, d.key).toBeDefined();
|
||
expect(manifest.contract.authorization.humanCallBudget).toBe(8);
|
||
expect(manifest.contract.resources.budget.humanCalls).toBe(8);
|
||
});
|
||
|
||
/*
|
||
* 2026-08-29 重写:启动表单从「需求 + 原PN + 候选数组 + datasheet」四件收到「原PN + 候选」两件
|
||
* (单数,一次评估恰一对一候选)—— `requestDescription` / `candidateInternalPns`(数组)/
|
||
* `datasheets` 三键随 P-01 需求主记录一起撤下(本活动不再碰它)。逐条断言见「契约 > 启动两参」。
|
||
*/
|
||
it('起单一格必填都不设(2026-08-28 用户定);客户与需求两类字段已随重写从启动表单撤下', () => {
|
||
expect(manifest.contract.input.data['required']).toEqual([]);
|
||
const launchProps = Object.keys(manifest.contract.input.data['properties'] as Record<string, unknown>);
|
||
expect(launchProps).not.toContain('customerCode');
|
||
expect(launchProps).not.toContain('customerName');
|
||
expect(launchProps).not.toContain('requestDescription');
|
||
expect(launchProps).not.toContain('datasheets');
|
||
});
|
||
|
||
it('五条样例:第一条满载走数据工单那条路,名字 ≤14 字,形状各不相同', () => {
|
||
expect(manifest.samples).toHaveLength(5);
|
||
for (const s of manifest.samples) expect(s.name.length, s.name).toBeLessThanOrEqual(14);
|
||
expect(new Set(manifest.samples.map((s) => JSON.stringify(s.input))).size).toBe(5);
|
||
const first = manifest.samples[0]?.input as { candidateInternalPns?: string[] };
|
||
expect(first.candidateInternalPns).toContain('XL2596S-5.0');
|
||
const newCustomer = manifest.samples.find((s) => (s.input as { customerName?: string }).customerName !== undefined);
|
||
expect(newCustomer).toBeDefined();
|
||
});
|
||
|
||
/*
|
||
* SKIP(2026-08-29,Task 2 → Task 7 部分交付):`input-supplement` 已随 Task 7 补回
|
||
* `assets.viewTemplates`(`VIEWS` 已更新),但 `difference-review` / `landing-confirm`
|
||
* 两张还在等 Task 8 / 9 —— `new Set(...)` 的整体相等要三张都在才成立,继续跳过。
|
||
*/
|
||
it.skip('三张工单视图 + 一张运营侧看板都声明了,剧本每一步引用的 view 都在其中', () => {
|
||
expect(new Set(manifest.assets.viewTemplates.map((v) => v.name))).toEqual(new Set([...VIEWS, BOARD_VIEW]));
|
||
for (const step of manifest.assets.businessModel.steps as ReadonlyArray<{ view?: string }>) {
|
||
if (step.view !== undefined) expect(VIEWS, step.view).toContain(step.view);
|
||
}
|
||
});
|
||
});
|
||
|
||
describe('台账声明', () => {
|
||
it('恰好九条声明(七个业务对象,P-08 与 P-11 各是主表 + 子表)', () => {
|
||
expect(manifest.declarations.ledger.map((l) => `${l.domain}.${l.object}`).sort()).toEqual([
|
||
'parts.chip_type', 'parts.parameter', 'parts.profile', 'parts.type_param',
|
||
'sub_eval.case', 'sub_eval.part',
|
||
'substitution.difference_item', 'substitution.relation', 'validation.record',
|
||
].sort());
|
||
});
|
||
|
||
it('类型字典:type_id 是自增主键 —— 人在件上新建类型时留空,由基座续号', () => {
|
||
expect(colOf('parts.chip_type', 'type_id')).toMatchObject({ primaryKey: true, type: 'integer', auto: true });
|
||
expect(colOf('parts.chip_type', 'type_name')?.required).toBe(true);
|
||
});
|
||
|
||
it('字段字典:复合主键 + 约束类型长在这里(不在每颗料上)', () => {
|
||
expect(ledgerOf('parts.type_param')?.fields.filter((f) => f.primaryKey).map((f) => f.name)).toEqual(['type_id', 'param_name']);
|
||
expect(colOf('parts.type_param', 'type_id')?.refersTo).toBe('parts.chip_type.type_id');
|
||
expect(colOf('parts.type_param', 'param_desc')?.required).toBe(true);
|
||
expect(colOf('parts.type_param', 'constraint_type')).toMatchObject({ required: true, tally: ['hard', 'soft'] });
|
||
});
|
||
|
||
it('P-08:类型必填并引用字典;createOnly 撤掉 —— 在库老料要能补上类型这一格', () => {
|
||
expect(colOf('parts.profile', 'chip_type_id')).toMatchObject({ required: true, type: 'integer', refersTo: 'parts.chip_type.type_id' });
|
||
expect(ledgerOf('parts.profile')?.createOnly).toBeUndefined();
|
||
/* 封装与工艺本来就是「核心参数」—— 它们该以类型字段的身份出现在字典里,不在主记录上另开一列 */
|
||
for (const gone of ['package', 'process_technology', 'eol_pcn_status', 'application_scenarios']) {
|
||
expect(colOf('parts.profile', gone)).toBeUndefined();
|
||
}
|
||
});
|
||
|
||
it('参数子表:约束类型与自指引用墙都退休了', () => {
|
||
expect(colOf('parts.parameter', 'constraint_type')).toBeUndefined();
|
||
expect(colOf('parts.parameter', 'param_name')?.refersTo).toBeUndefined();
|
||
expect(colOf('parts.parameter', 'param_name')?.distinct).toBeUndefined();
|
||
expect(colOf('parts.parameter', 'param_value')?.required).toBe(true);
|
||
expect(ledgerOf('parts.parameter')?.createOnly).toBeUndefined();
|
||
});
|
||
|
||
it('案卷两张:part 一单两行(配方遍历的唯一来源)、case 一单一行', () => {
|
||
expect(ledgerOf('sub_eval.part')?.caseKey).toBe('eval_ref');
|
||
expect(ledgerOf('sub_eval.part')?.fields.filter((f) => f.primaryKey).map((f) => f.name)).toEqual(['eval_ref', 'internal_pn']);
|
||
/* 「两颗料是不是同一个类型」于是成了基座数得出来的数:不同取值恰好 1 种 */
|
||
expect(colOf('sub_eval.part', 'chip_type_id')).toMatchObject({ required: true, distinct: true });
|
||
expect(colOf('sub_eval.part', 'exists_status')?.tally).toEqual(['found', 'absent']);
|
||
/* 「在库」是一次 lookup 的结果,不是一句断言:判 found 就必须拿得出一个基座查得到的料号 */
|
||
expect(colOf('sub_eval.part', 'found_pn_ref')?.refersTo).toBe('parts.profile.internal_pn');
|
||
expect(ledgerOf('sub_eval.case')?.fields.filter((f) => f.primaryKey).map((f) => f.name)).toEqual(['eval_ref']);
|
||
});
|
||
|
||
it('P-11 明细带合成列 judgement_class —— 推荐等级要数的是「硬约束不符有几条」', () => {
|
||
expect(colOf('substitution.difference_item', 'judgement_class')).toMatchObject({
|
||
required: true,
|
||
tally: ['match', 'soft_mismatch', 'hard_mismatch', 'unknown', 'conditional'],
|
||
});
|
||
/* 「判断依据」不是文件引用 —— 不写 P-10 了,名字得说实话 */
|
||
expect(colOf('substitution.difference_item', 'evidence_note')?.required).toBe(true);
|
||
expect(colOf('substitution.difference_item', 'evidence_ref')).toBeUndefined();
|
||
expect(colOf('substitution.difference_item', 'unknown_disposition')?.requiredWhen).toEqual({ column: 'difference_status', value: 'unknown' });
|
||
});
|
||
|
||
it('P-11 主记录:退休三列,等级只剩三档', () => {
|
||
for (const gone of ['substitution_type', 'confidence_score', 'analysis_report_ref']) {
|
||
expect(colOf('substitution.relation', gone)).toBeUndefined();
|
||
}
|
||
expect(colOf('substitution.relation', 'recommendation_level')?.tally).toEqual(['recommended', 'conditional', 'not_recommended']);
|
||
});
|
||
|
||
it('P-06:关联的是本单案号(P-01 不在本活动里),类型加了 parametric 一档', () => {
|
||
expect(colOf('validation.record', 'eval_ref')?.required).toBe(true);
|
||
expect(colOf('validation.record', 'request_id')).toBeUndefined();
|
||
expect(colOf('validation.record', 'validation_type')?.tally).toContain('parametric');
|
||
});
|
||
|
||
it('这个活动不碰 P-01 / P-10 / 客户档案', () => {
|
||
for (const gone of ['demand.request', 'parts.document', 'customer.profile']) {
|
||
expect(ledgerOf(gone)).toBeUndefined();
|
||
}
|
||
});
|
||
});
|
||
|
||
describe('契约', () => {
|
||
it('启动两参,都选填 —— 缺了由第一步派工单补', () => {
|
||
const props = manifest.contract.input.data['properties'] as Record<string, Record<string, unknown>>;
|
||
expect(Object.keys(props)).toEqual(['originalInternalPn', 'candidateInternalPn']);
|
||
expect(manifest.contract.input.data['required']).toEqual([]);
|
||
/* 运营端启动表单:两格都是「可选可新写」的下拉 */
|
||
for (const key of ['originalInternalPn', 'candidateInternalPn']) {
|
||
expect(props[key]?.['x-control']).toBe('combo');
|
||
expect(props[key]?.['x-options']).toMatchObject({ ledger: 'parts.profile', value: 'internal_pn' });
|
||
}
|
||
});
|
||
|
||
it('零工具 —— 这是这次重写的立场,不是疏漏', () => {
|
||
expect(manifest.contract.resources.tools).toEqual([]);
|
||
expect(manifest.declarations.tools).toEqual([]);
|
||
});
|
||
|
||
it('单一出口,载荷是已落地的事实', () => {
|
||
expect(manifest.contract.output.mandatory.map((p) => p.name)).toEqual(['assessment-released']);
|
||
expect(manifest.contract.output.selective).toEqual([]);
|
||
});
|
||
|
||
it('判据全读基座数 —— 任何一边都不许是活动自己写的汇总列', () => {
|
||
const machine = manifest.contract.acceptanceCriteria.filter((c) => c.level === 'L1' || c.level === 'L2');
|
||
expect(machine.length).toBeGreaterThanOrEqual(5);
|
||
for (const c of machine) {
|
||
expect(c.predicate).toBeDefined();
|
||
/* 基座读数的形状:域_对象_rows / _missing / _distinct_列 / tally 计数 */
|
||
expect(c.predicate).toMatch(/_(rows|missing|distinct_[a-z_]+|[a-z_]+)\b/);
|
||
}
|
||
});
|
||
|
||
it('预算:调人三次(补数据 / 审核 / 放行),模型只出场一次', () => {
|
||
expect(manifest.contract.resources.budget.humanCalls).toBe(8);
|
||
expect(manifest.contract.authorization.humanCallBudget).toBe(8);
|
||
expect(manifest.contract.resources.models).toEqual(['deepseek-v4-flash']);
|
||
});
|
||
});
|
||
|
||
describe('案例模型与清单的对齐', () => {
|
||
it('四个计划项,剧本 item 与 CMMN humanTask id 一一对应', () => {
|
||
const cmmnIds = [...manifest.assets.businessModel.cmmnXml.matchAll(/<humanTask id="([^"]+)"/g)].map((m) => m[1] ?? '');
|
||
expect(cmmnIds).toEqual(ITEMS);
|
||
expect((manifest.assets.businessModel.steps as ReadonlyArray<{ item: string }>).map((s) => s.item)).toEqual(ITEMS);
|
||
});
|
||
|
||
it('两处不给 doneWhen(审核 / 放行),其余两处都给 —— 校验入参与逐项分析全 auto、入库/结论事实一落地就收步', () => {
|
||
for (const s of manifest.assets.businessModel.steps as ReadonlyArray<{ item: string; doneWhen?: string }>) {
|
||
if (NO_DONE_WHEN.has(s.item)) expect(s.doneWhen, s.item).toBeUndefined();
|
||
else expect(s.doneWhen, s.item).toBeDefined();
|
||
}
|
||
});
|
||
|
||
/*
|
||
* SKIP(2026-08-29,Task 3):`S0d_p08_master` / `S0d_p10_document` 两个计划项已随四步重写撤下 ——
|
||
* 「补芯片主数据」与「补原厂技术资料」不再是两张各自的数据工单,统一并入 `S1_inputs`
|
||
* (校验入参与补数据)一步之内。Task 3 的 brief 明写这一步不设 `view`(它是可选的,
|
||
* Task 7–9 各自补上自己那一张),所以这条「两张数据工单件各自绑自己的 view」眼下没有对应的
|
||
* 断言对象可钉 —— 不是「暂时还没做」,是这个二选一视图的设计前提本身已经不存在了。
|
||
*/
|
||
it.skip('两张数据工单件各自绑自己的 view(P-08 与 P-10 不共用同一份画面)', () => {
|
||
const steps = manifest.assets.businessModel.steps as ReadonlyArray<{ item: string; view?: string }>;
|
||
expect(steps.find((s) => s.item === 'S0d_p08_master')?.view).toBe('p08-master-create');
|
||
expect(steps.find((s) => s.item === 'S0d_p10_document')?.view).toBe('p10-document-create');
|
||
});
|
||
|
||
it('那两项在 CMMN 里都标了人工激活', () => {
|
||
for (const id of NO_DONE_WHEN) {
|
||
const block = new RegExp(`<humanTask id="${id}"[\\s\\S]*?</humanTask>`).exec(manifest.assets.businessModel.cmmnXml)?.[0] ?? '';
|
||
/* Task 3 起改用 `manualActivation="true"` 属性直接标在 humanTask 元素上(与 CMMN 解析器的两种认法之一同款),不再嵌套 <manualActivationRule> */
|
||
expect(block, id).toContain('manualActivation="true"');
|
||
}
|
||
});
|
||
|
||
it('除放行外每一项恰好属于一个阶段;3 段 + 放行 = x/4', () => {
|
||
const phases = manifest.assets.businessModel.phases ?? [];
|
||
expect(phases).toHaveLength(3);
|
||
const phaseItems = phases.flatMap((p) => p.items);
|
||
expect([...phaseItems].sort()).toEqual([...ITEMS.filter((i) => i !== 'T_release')].sort());
|
||
expect(new Set(phaseItems).size).toBe(phaseItems.length);
|
||
});
|
||
|
||
it('说明书点名的计划项都真实存在', () => {
|
||
const mentioned = [...manifest.assets.businessModel.narrative.matchAll(/\b(S\d[a-z]?_[a-zA-Z_]+|T_release)\b/g)].map((m) => m[1] ?? '');
|
||
for (const name of mentioned) expect(ITEMS, `说明书提到但不存在的计划项: ${name}`).toContain(name);
|
||
});
|
||
|
||
/*
|
||
* SKIP(2026-08-29,Task 3):这条测的是「决策卡骨架住在说明书的散文里,模型抄卡骨架时照读」——
|
||
* 那正是**零 `guide`**重写要撤掉的机制本身(卡骨架挪进配方 `auto.do[].kind:'card'` 的结构化
|
||
* `options` 字段,由引擎按声明发卡,不再靠模型从说明书里抄 JSON 片段)。新版 `NARRATIVE`
|
||
* 因此不再嵌任何 `"value":"..."` 这类卡面 JSON —— 不是「还没写全」,是这条断言检查的机制
|
||
* 在四步全 auto 的设计里已经不存在了(决策卡的选项由 Task 4–6 的配方声明,不由说明书文本决定)。
|
||
*/
|
||
it.skip('说明书里只剩审核那一种决策卡(approve/reject/custom:);候选放弃/创建那张已退役 —— 缺库一律自动派数据工单(2026-08-27 用户定)', () => {
|
||
const n = manifest.assets.businessModel.narrative;
|
||
expect(n).not.toContain('"value":"create"');
|
||
expect(n).not.toContain('"value":"abandon"');
|
||
expect(n).toContain('"value":"approve"');
|
||
expect(n).toContain('"value":"reject"');
|
||
expect(n).toContain('custom:');
|
||
});
|
||
|
||
/*
|
||
* SKIP(2026-08-29,Task 2 → Task 3 补回,复核后维持 skip):原计划这条断言的是
|
||
* 「T_release 剧本里 `judge criterion="..."` 的原文逐字对得上 acceptanceCriteria 的 label」——
|
||
* 而 Task 3 的四步重写是**全程零 guide**:`T_release` 的 `guide` 现在是 `''`,
|
||
* 「prose guide 里嵌 `judge criterion="..."` 给模型执行」这套机制随零 guide 重写整体撤掉了
|
||
* (裁量判据改由人工验收清单 `acceptanceChecklist` 让人亲自核,不再靠模型在 guide 里自判)。
|
||
* 这不是「Task 3 会把断言点的字符串换成新的两条」——是断言检查的机制本身不复存在,
|
||
* 无法在不改变检查对象的前提下让它变绿,继续保留跳过并记录清楚原因。
|
||
*/
|
||
it.skip('T_release 剧本里 judge criterion 的原文逐字对得上 acceptanceCriteria 的 label', () => {
|
||
const tRelease = (manifest.assets.businessModel.steps as ReadonlyArray<{ item: string; guide?: string }>).find((s) => s.item === 'T_release');
|
||
const quoted = [...(tRelease?.guide ?? '').matchAll(/judge criterion="([^"]+)"/g)].map((m) => m[1] ?? '');
|
||
expect(quoted).toHaveLength(3);
|
||
const labels = new Set(manifest.contract.acceptanceCriteria.map((c) => c.label));
|
||
for (const q of quoted) expect(labels.has(q), q).toBe(true);
|
||
});
|
||
});
|
||
|
||
describe('报告模版:编辑面与只读面各就各位', () => {
|
||
/*
|
||
* SKIP(2026-08-29,Task 2 → Task 7,复核后维持 skip):`p08-master-create` / `p10-document-create` /
|
||
* `eval-input-list` 三张局面已被 Task 7 的 `input-supplement` 一张顶掉,但这条测的形状
|
||
* 与新设计不同,不是简单换个 view 名字就能救活:①P-10(`parts.document`)整个不在新
|
||
* S1 的范围里(本活动不再碰它);②新设计里 `parts.profile` 两行由系统铺、人不加行
|
||
* (要建哪颗料是启动参数定死的事),`parts.parameter` 的行集由配方从
|
||
* `sub_eval.part × parts.type_param` 派生且 `prune:true`(手加的孤行下一拍就被删)——
|
||
* 这两张表都**不给** `data-edit-add`,只有 `parts.type_param`(新建类型的字段清单)才有。
|
||
* 新覆盖见下面「件① 评估输入补全」;继续 skip、不改断言内容,留给读的人对照差异。
|
||
*/
|
||
it.skip('P-08 数据工单件的表 1 + 表 2(标准化参数子表)都可加行可编辑:data-edit / data-edit-add 与主键 data-col 都在', () => {
|
||
const p08 = viewHtml('p08-master-create');
|
||
expect(p08).toContain('data-edit="parts.profile"');
|
||
expect(p08).toContain('data-edit-add="parts.profile"');
|
||
expect(p08).toContain('data-col="internal_pn"');
|
||
expect(p08).toContain('data-edit="parts.parameter"');
|
||
expect(p08).toContain('data-edit-add="parts.parameter"');
|
||
expect(p08).toContain('data-col="param_name"');
|
||
expect(p08).toContain('data-col="param_value"');
|
||
expect(p08).toContain('data-col="constraint_type"');
|
||
const p10 = viewHtml('p10-document-create');
|
||
expect(p10).toContain('data-edit="parts.document"');
|
||
expect(p10).toContain('data-col="doc_seq"');
|
||
});
|
||
|
||
/* SKIP(2026-08-29,Task 2 → Task 7):`eval-input-list` 局面已撤下,`sub_eval.input_gap` / `sub_eval.pn_check` 两张单据也已随重写退休(顶掉它们的是 `sub_eval.part` / `sub_eval.case`) */
|
||
it.skip('评估输入清单:缺口表可编辑(supplement),绑定区只读', () => {
|
||
const html = viewHtml('eval-input-list');
|
||
expect(html).toContain(`data-edit="${D}.input_gap"`);
|
||
expect(html).toContain('data-col="supplement"');
|
||
expect(html).toContain(`data-rows="mine:${D}.pn_check"`);
|
||
});
|
||
|
||
/*
|
||
* 2026-08-29 Task 8 改窄:这条原本把 `difference-review` 也算作「只读」,与 `landing-confirm`
|
||
* 并列断言零 `data-edit`/`data-col`。而 spec §4.4/§6.2(同一次重写的权威设计,同日写定)
|
||
* 明确把 S3 从纯只读改成「人就地改四态,配方重算」—— brief 给的 Step 1 测试(见下面
|
||
* 「件② 差异评估结论审核」组)直接要求 `data-edit="substitution.difference_item"` 与
|
||
* `data-col="difference_status"` 等必须存在,与这条断言原样矛盾。这条断言写在那处设计
|
||
* 定稿之前(或是与 landing-confirm 一起想当然地假设了两者都只读),现改窄成只管
|
||
* `landing-confirm`(它确实全只读,spec §6.3)。`difference-review` 的「只读面在哪」
|
||
* 改由「件②」组精确覆盖(哪几列可改、`substitution.relation` 那半张不许有 data-edit)。
|
||
*/
|
||
it('落账确认只读:没有任何 data-edit / data-col(差异审核已按 spec §4.4 变为可就地编辑,断言见「件②」组)', () => {
|
||
expect(viewHtml('landing-confirm')).not.toContain('data-edit=');
|
||
expect(viewHtml('landing-confirm')).not.toContain('data-col=');
|
||
});
|
||
|
||
it('每个可编辑格子的 data-col 都在对应台账对象的声明里(input-supplement 一屏三张表,三张台账对象的声明并起来看;difference-review 一张表)', () => {
|
||
const cases: Array<[string, Array<[string, string]>]> = [
|
||
['input-supplement', [['parts', 'profile'], ['parts', 'type_param'], ['parts', 'parameter']]],
|
||
['difference-review', [['substitution', 'difference_item']]],
|
||
];
|
||
for (const [view, objects] of cases) {
|
||
const cols = new Set([...viewHtml(view).matchAll(/data-col="([^"]+)"/g)].map((m) => m[1] ?? ''));
|
||
const declared = new Set(objects.flatMap(([domain, object]) => byAddress(domain, object)?.fields.map((f) => f.name) ?? []));
|
||
for (const c of cols) expect(declared.has(c), `${view}: ${c}`).toBe(true);
|
||
}
|
||
});
|
||
|
||
/* SKIP(2026-08-29,Task 2 → Task 8/9):`difference-review` / `landing-confirm` 两张局面已撤下(新的两张由 Task 8/9 重做);它们绑的字段(recommendation_level 已只剩三档、analysis_status 已不带 tally)到时也会一并对齐 */
|
||
it.skip('枚举列只用 data-class-of 投影,不摆 ASCII 原值', () => {
|
||
expect(viewHtml('difference-review')).toContain('data-class-of="difference_status"');
|
||
expect(viewHtml('difference-review')).toContain('data-class-of="recommendation_level"');
|
||
expect(viewHtml('difference-review')).not.toContain('data-slot="difference_status"');
|
||
const landing = viewHtml('landing-confirm');
|
||
for (const col of ['request_status', 'current_stage', 'analysis_status', 'recommendation_level']) {
|
||
expect(landing, col).toContain(`data-class-of="${col}"`);
|
||
expect(landing, col).not.toContain(`data-slot="${col}"`);
|
||
}
|
||
});
|
||
|
||
it('宿主拒推含脚本的模版', async () => {
|
||
const host = createMockHost({ contract: CONTRACT });
|
||
await expect(host.base.interaction.pushHtml('<div><script>x()</script></div>')).rejects.toThrow(/无脚本执行权/);
|
||
});
|
||
|
||
/*
|
||
* SKIP(2026-08-29,Task 2 → Task 7 部分交付):VIEWS 现在是三张新局面的名单,`input-supplement`
|
||
* 已随 Task 7 补回、这一条本身可以对它验(外壳确实是 page/card),但 `difference-review` /
|
||
* `landing-confirm` 两张还没有(viewHtml 返回空串,`toContain('class="card-head"')` 会假),
|
||
* 循环体没法只验一部分 —— 等 Task 8/9 补齐三张再一起解封。
|
||
*/
|
||
it.skip('三张模版的外壳都是搬自目标件的 page/card,不再含 E04/I01 那套视觉底(.proto-root)的痕迹', () => {
|
||
for (const view of VIEWS) {
|
||
const html = viewHtml(view);
|
||
expect(html, view).toContain('class="card-head"');
|
||
expect(html, view).toContain('class="card-foot"');
|
||
expect(html, view).not.toContain('proto-root');
|
||
}
|
||
});
|
||
});
|
||
|
||
/*
|
||
* 件① 评估输入补全(2026-08-29,Task 7):S1 的配方缺什么就发一张数据工单,`view: 'input-supplement'`
|
||
* 指向它。三块竖排——要评估什么(两颗料的存在性 + 主数据)/ 芯片类型(选型 + 新建类型时的核心参数
|
||
* 字段清单)/ 参数值(逐条实测/规格值);顶部一条红带摆配方算好的 `gate_reason`。
|
||
* 复用文件顶部已有的 `viewHtml` 短写(与 brief 草稿里的 `viewOf` 同一件事,不重复定义一个)。
|
||
*/
|
||
describe('件① 评估输入补全', () => {
|
||
const html = viewHtml('input-supplement');
|
||
|
||
it('在 viewTemplates 里,且 S1 指向它', () => {
|
||
expect(html).not.toBe('');
|
||
const steps = manifest.assets.businessModel.steps as ReadonlyArray<{ item: string; view?: string }>;
|
||
expect(steps.find((s) => s.item === 'S1_inputs')?.view).toBe('input-supplement');
|
||
});
|
||
|
||
it('三块都在,行都绑本单草稿(不是整表)', () => {
|
||
expect(html).toContain('data-rows="mine:sub_eval.part"');
|
||
expect(html).toContain('data-rows="mine:parts.parameter"');
|
||
expect(html).toContain('data-rows="mine:parts.type_param"');
|
||
});
|
||
|
||
it('芯片类型是从台账现读的下拉 —— 写死等于「只能选出包那天就有的那几种」', () => {
|
||
expect(html).toContain('data-options="parts.chip_type"');
|
||
expect(html).toContain('data-option-value="type_id"');
|
||
expect(html).toContain('data-option-label="type_name"');
|
||
/* data-options 与 data-slot 不能共存(渲染时 data-slot 会被剥掉)—— 当下值由 data-col 定 */
|
||
expect(/data-options="[^"]*"[^>]*data-slot=/.test(html)).toBe(false);
|
||
});
|
||
|
||
it('**主键列必挂 `data-col`** —— 它不是编辑面,是行的钥匙(漏了会把人改的行错配到别行上)', () => {
|
||
expect(html).toContain('data-col="internal_pn"');
|
||
expect(html).toContain('data-col="param_name"');
|
||
});
|
||
|
||
it('可改的三张表各有一道编辑闸', () => {
|
||
for (const address of ['parts.profile', 'parts.type_param', 'parts.parameter']) {
|
||
expect(html).toContain(`data-edit="${address}"`);
|
||
}
|
||
});
|
||
|
||
/*
|
||
* 2026-08-29 评审 Minor:brief 原话是「红带显隐用 data-class-of」,但 gate_reason 是
|
||
* 中文自由文本 —— classToken 只留 [a-z0-9-],中文字符全被剥空,不论空串还是有内容的
|
||
* 中文句子算出来的 token 都是空的,这条路本来就走不通(不是本任务的臆断,评审读了
|
||
* view-render.ts 的 classToken 实现确认)。真正管显隐的是 data-blank 出的空文本节点
|
||
* 配 CSS :has(:empty) ——源码里不留一个永远不起作用的 data-class-of 属性。
|
||
*/
|
||
it('模版不算不判:没有内联脚本,红带的显隐交给样式表(:has(:empty),不是 gate_reason 上算不出来的 data-class-of)', () => {
|
||
expect(html).not.toContain('<script');
|
||
expect(html).not.toContain('data-class-of="gate_reason"');
|
||
expect(html).toContain('class="jc-gate-text" data-slot="gate_reason" data-blank');
|
||
expect(html).toContain('.jc-gate-bar:has(.jc-gate-text:empty)');
|
||
});
|
||
|
||
it('表宽超出要能横着滑 —— 不许写成 overflow:hidden(E04/I01 抄错过一次)', () => {
|
||
expect(html).not.toMatch(/\.jc-tbl-wrap\s*\{[^}]*overflow\s*:\s*hidden/);
|
||
});
|
||
|
||
/*
|
||
* 2026-08-29 评审 Important:BIND_COMMON 的序号计数器规则原来只认 `.tbl-wrap`
|
||
* (继承自已删除的旧 P08_MASTER_CREATE,同一个 wrapper class,同一个坑),
|
||
* 而这张件的三张表全用平台的 `.jc-tbl-wrap` —— 两个 class 对不上,计数器永不触发,
|
||
* 只读态与编辑态下「#」列逐行都是空白。钉死 `.jc-tbl-wrap` 现在也接了这五条规则,
|
||
* 且这张件确实只用 `.jc-tbl-wrap`(没有表在用没打这条补丁就会数不出来的 `.tbl-wrap`)。
|
||
*/
|
||
it('三张表的序号列真的会数:.jc-tbl-wrap 也接了 CSS 计数器(不再只认 .tbl-wrap)', () => {
|
||
expect(html).toMatch(/\.jc-tbl-wrap\s+table\s*,\s*\.tbl-wrap\s+table\s*\{\s*counter-reset:\s*rn;?\s*\}/);
|
||
expect(html).toMatch(/\.jc-tbl-wrap\s+table\s+td\.rn\s*,\s*\.tbl-wrap\s+table\s+td\.rn\s*\{\s*counter-increment:\s*rn;?\s*\}/);
|
||
expect(html).toMatch(/\.jc-tbl-wrap\s+table\s+td\.rn\s+\.jc-rn-num::before\s*,\s*\.tbl-wrap\s+table\s+td\.rn\s+\.jc-rn-num::before\s*\{\s*content:\s*counter\(rn\);?\s*\}/);
|
||
// 这张件的三张表都走 .jc-tbl-wrap(没有 .tbl-wrap)—— 补丁真的接在会用到的类上
|
||
expect(html).not.toContain('class="tbl-wrap"');
|
||
const jcTblWrapCount = html.match(/class="jc-tbl-wrap"/g)?.length ?? 0;
|
||
expect(jcTblWrapCount).toBe(3);
|
||
});
|
||
});
|
||
|
||
/*
|
||
* 件② 差异评估结论审核(2026-08-29,Task 8):S3 的验收工单,`view: 'difference-review'`
|
||
* 指向它,`cardVariant: 'review'`。绑定契约(spec §6.2):对决头(原 PN → 候选 PN、类型徽)+
|
||
* 四态分布带 + 推荐等级大徽章只读;逐条明细可就地改(四态 / 影响说明 / 判断依据 / UNKNOWN 去向);
|
||
* 收口四格(推荐理由 / 差异摘要 / 风险 / 验证要求)只读 —— 它们是配方按规则算的,人要改就去改
|
||
* 上面那一条四态,系统重算。复用文件顶部已有的 `viewHtml` 短写。
|
||
*/
|
||
describe('件② 差异评估结论审核', () => {
|
||
const html = viewHtml('difference-review');
|
||
|
||
it('在 viewTemplates 里,S3 指向它,面孔是审核卡', () => {
|
||
expect(html).not.toBe('');
|
||
const s3 = (manifest.assets.businessModel.steps as ReadonlyArray<{ item: string; view?: string; cardVariant?: string }>)
|
||
.find((s) => s.item === 'S3_review');
|
||
expect(s3?.view).toBe('difference-review');
|
||
expect(s3?.cardVariant).toBe('review');
|
||
});
|
||
|
||
it('逐条明细可改,且**改的是四态**', () => {
|
||
expect(html).toContain('data-rows="mine:substitution.difference_item"');
|
||
expect(html).toContain('data-edit="substitution.difference_item"');
|
||
expect(html).toContain('data-col="difference_status"');
|
||
for (const col of ['impact_statement', 'evidence_note', 'unknown_disposition']) {
|
||
expect(html).toContain(`data-col="${col}"`);
|
||
}
|
||
expect(html).toContain('data-col="difference_item_id"');
|
||
});
|
||
|
||
it('**推荐理由 / 摘要 / 风险 / 验证要求只读** —— 让人直接改摘要,等于让他给自己没做的判断盖章', () => {
|
||
for (const col of ['recommendation_reason', 'difference_summary', 'verification_requirement']) {
|
||
expect(html).toContain(`data-slot="${col}"`);
|
||
expect(html).not.toContain(`data-col="${col}"`);
|
||
}
|
||
expect(html).not.toContain('data-edit="substitution.relation"');
|
||
});
|
||
|
||
it('等级与四态用 data-class-of 上色 —— 判断留在样式表,模版不判', () => {
|
||
expect(html).toContain('data-class-of="recommendation_level"');
|
||
expect(html).not.toContain('<script');
|
||
});
|
||
});
|
||
|
||
/*
|
||
* 2026-08-27 实测账,两条都真截图 + SQL 复核过:
|
||
* ① **带下划线的枚举值一律显示成「待定」**——平台有两套互相打架的消毒规则:
|
||
* `data-class-of` 投影出的 class 走 `view-render.ts` 的 `classToken`(只留 `[a-z0-9-]`,
|
||
* 吃掉下划线,`pin_to_pin` → `v-pintopin`);基座读数名走 `engines.ts` 的 `slug`
|
||
* (只留 `[a-z0-9_]`,吃掉连字符)。手写 `.v-pin_to_pin` 那种选择器永远匹配不上 ——
|
||
* 这条测试把 `manifest.declarations.ledger` 里每一个「模版真的摆了徽标」的 tally 列过一遍,
|
||
* 用与 `classToken` 一字不差的 `vtok` 算出选择器,断言对应的 CSS 规则确实存在。
|
||
* ② **缺口表补充内容每一行都摆红「必填」,而闸只认 3 行**——`requiredWhen` 是按行不按列,
|
||
* 静态 `class="req"` 把整列都标了必填。
|
||
*/
|
||
/*
|
||
* SKIP(2026-08-29,Task 2 → Task 7-9 补回):`BADGED` 列表里的每一项都指着已撤下的五张局面
|
||
* (p08-master-create 等)或已改动的字段(recommendation_level 掉了 alternate 档、
|
||
* analysis_status 不再带 tally、`sub_eval.pn_check` 整个改名 `sub_eval.part`)。
|
||
* 这整组「回归网」要等 Task 7–9 把新三张件补回来、字段清单也重新对齐后才谈得上继续钉。
|
||
*/
|
||
describe.skip('徽标别名与值域相关的坑(2026-08-27 实测账,回归网)', () => {
|
||
/** 必须与 `apps/api/src/modules/interaction/internal/domain/view-render.ts` 的 `classToken` 一字不差 */
|
||
const vtok = (v: string): string => v.toLowerCase().replace(/[^a-z0-9-]/g, '').slice(0, 24);
|
||
|
||
/** 每张模版真的摆了徽标的那几列(带下划线的 tally 值最容易漏,漏了就是「待定」那次事故) */
|
||
const BADGED: ReadonlyArray<{ view: string; domain: string; object: string; column: string }> = [
|
||
{ view: 'p08-master-create', domain: 'parts', object: 'profile', column: 'lifecycle_status' },
|
||
{ view: 'p08-master-create', domain: 'parts', object: 'parameter', column: 'constraint_type' },
|
||
{ view: 'p10-document-create', domain: 'parts', object: 'document', column: 'document_type' },
|
||
{ view: 'p10-document-create', domain: 'parts', object: 'document', column: 'document_status' },
|
||
{ view: 'eval-input-list', domain: D, object: 'pn_check', column: 'exists_status' },
|
||
{ view: 'eval-input-list', domain: D, object: 'pn_check', column: 'role' },
|
||
{ view: 'eval-input-list', domain: D, object: 'input_gap', column: 'status' },
|
||
{ view: 'difference-review', domain: 'substitution', object: 'relation', column: 'substitution_type' },
|
||
{ view: 'difference-review', domain: 'substitution', object: 'relation', column: 'recommendation_level' },
|
||
{ view: 'difference-review', domain: 'substitution', object: 'relation', column: 'risk_level' },
|
||
{ view: 'difference-review', domain: 'substitution', object: 'difference_item', column: 'constraint_type' },
|
||
{ view: 'difference-review', domain: 'substitution', object: 'difference_item', column: 'difference_status' },
|
||
{ view: 'difference-review', domain: 'substitution', object: 'difference_item', column: 'risk_level' },
|
||
{ view: 'landing-confirm', domain: 'substitution', object: 'relation', column: 'substitution_type' },
|
||
{ view: 'landing-confirm', domain: 'substitution', object: 'relation', column: 'recommendation_level' },
|
||
{ view: 'landing-confirm', domain: 'substitution', object: 'relation', column: 'analysis_status' },
|
||
{ view: 'landing-confirm', domain: 'substitution', object: 'relation', column: 'risk_level' },
|
||
{ view: 'landing-confirm', domain: 'demand', object: 'request', column: 'request_status' },
|
||
/* 运营侧看板把六张表的枚举列全摆成徽标 —— 每一列的每个值都得有文案,漏一个就是一枚「—」 */
|
||
{ view: BOARD_VIEW, domain: 'parts', object: 'profile', column: 'lifecycle_status' },
|
||
{ view: BOARD_VIEW, domain: 'parts', object: 'document', column: 'document_type' },
|
||
{ view: BOARD_VIEW, domain: 'parts', object: 'document', column: 'document_status' },
|
||
{ view: BOARD_VIEW, domain: 'demand', object: 'request', column: 'request_status' },
|
||
{ view: BOARD_VIEW, domain: 'demand', object: 'request', column: 'risk_level' },
|
||
{ view: BOARD_VIEW, domain: 'substitution', object: 'relation', column: 'substitution_type' },
|
||
{ view: BOARD_VIEW, domain: 'substitution', object: 'relation', column: 'recommendation_level' },
|
||
{ view: BOARD_VIEW, domain: 'substitution', object: 'relation', column: 'analysis_status' },
|
||
{ view: BOARD_VIEW, domain: 'substitution', object: 'relation', column: 'risk_level' },
|
||
{ view: BOARD_VIEW, domain: 'substitution', object: 'difference_item', column: 'constraint_type' },
|
||
{ view: BOARD_VIEW, domain: 'substitution', object: 'difference_item', column: 'difference_status' },
|
||
{ view: BOARD_VIEW, domain: 'substitution', object: 'difference_item', column: 'risk_level' },
|
||
{ view: BOARD_VIEW, domain: 'validation', object: 'record', column: 'validation_type' },
|
||
{ view: BOARD_VIEW, domain: 'validation', object: 'record', column: 'validation_status' },
|
||
{ view: BOARD_VIEW, domain: D, object: 'pn_check', column: 'role' },
|
||
{ view: BOARD_VIEW, domain: D, object: 'pn_check', column: 'exists_status' },
|
||
];
|
||
|
||
it('每张模版摆了徽标的每个 tally 列,声明里的每个值都有对应的 v-<token> 规则(选择器是从值算出来的,不是手写的)', () => {
|
||
for (const { view, domain, object, column } of BADGED) {
|
||
const tally = byAddress(domain, object)?.fields.find((f) => f.name === column)?.tally ?? [];
|
||
expect(tally.length, `${domain}.${object}.${column} 没有声明 tally`).toBeGreaterThan(0);
|
||
const html = viewHtml(view);
|
||
for (const value of tally) {
|
||
expect(html, `${view}: ${domain}.${object}.${column}=${value} → 应有 .v-${vtok(value)}::after`).toContain(`.v-${vtok(value)}::after`);
|
||
}
|
||
}
|
||
});
|
||
|
||
it('缺口表补充内容按行投影必填,不是整列静态 req(requiredWhen 是按行的,不是按列的)', () => {
|
||
const html = viewHtml('eval-input-list');
|
||
expect(html).not.toContain('class="req" data-h="补充内容"');
|
||
expect(html).toContain('data-class-of="status"');
|
||
expect(html).toContain('.sup.v-gap');
|
||
});
|
||
});
|
||
|
||
/*
|
||
* 2026-08-27 定向复审逮住的两条 Critical(都是这一轮版面重做新引入的回归,不是原有的坑):
|
||
* ① `sub_eval.input_gap` 的「键」(`item_key`)那一格丢了 `data-col` —— 一单 11 行,`item_key`
|
||
* 逐行不同,人保存**任何一行**时服务端 `resolveRowKey` 凑不出主键(它既不在件上、
|
||
* 又不是「全表一致」的那种案号型列),整格落不下去;
|
||
* ② `sub_eval.input_gap.status` 声明 `required: true`,件上又有「＋ 加一行」——
|
||
* 降成只读徽标后,人新加的行永远没有格子能把这一列填上,`_missing` 恒为正、
|
||
* 提交闸永远拦着,这是「墙不配路」的死路形态。
|
||
* 两条各配一张回归网,防的是同一类事故再出现在别的字段上。
|
||
*/
|
||
/*
|
||
* 解封(2026-08-29,Task 7):`EDITABLE` 列表已按新设计换成 `input-supplement` 一张件绑的
|
||
* 三张台账对象(`parts.profile` / `parts.type_param` / `parts.parameter`)—— `p10-document-create`
|
||
* 与 `sub_eval.input_gap` 两张随重写整个退休,不再有对应条目。这两条 Critical 回归网继续钉在
|
||
* 新件上:主键(internal_pn / type_id+param_name)都挂了 `data-col`,必填列都有可填的控件。
|
||
*/
|
||
describe('可编辑件的主键与必填列(2026-08-27 复审:主键丢钥匙 / 必填填不了,两条 Critical 回归)', () => {
|
||
/** input-supplement 一屏三张真能编辑的表与它们各自绑定的台账对象 */
|
||
const EDITABLE: ReadonlyArray<{ view: string; domain: string; object: string }> = [
|
||
{ view: 'input-supplement', domain: 'parts', object: 'profile' },
|
||
{ view: 'input-supplement', domain: 'parts', object: 'type_param' },
|
||
{ view: 'input-supplement', domain: 'parts', object: 'parameter' },
|
||
];
|
||
|
||
it('每一个主键列都有 data-col —— 它不是编辑面,是行的钥匙,丢了就整行保存不下去', () => {
|
||
for (const { view, domain, object } of EDITABLE) {
|
||
const table = byAddress(domain, object);
|
||
/*
|
||
* `caseKey` 声明的那一列(如 `eval_ref`)是例外:它是「一单一号,全表当然一致」的
|
||
* 案号型主键(与 I01 的 `round_ref` 同款),件上压根不摆,服务端 `resolveRowKey`
|
||
* 按「全表一致」自动补上(`apps/api/src/app/routes/desk.ts` 注释里点名的那条路)——
|
||
* 摆出来给人填反而是错的(那会让人以为能按行改案号)。真正逐行不同、必须挂
|
||
* `data-col` 的,是主键里除 caseKey 之外的那几列(如 `item_key`)。
|
||
*/
|
||
const pk = (table?.fields.filter((f) => f.primaryKey).map((f) => f.name) ?? []).filter((name) => name !== table?.caseKey);
|
||
expect(pk.length, `${domain}.${object} 没有非 caseKey 的主键列`).toBeGreaterThan(0);
|
||
const html = viewHtml(view);
|
||
for (const col of pk) {
|
||
expect(html, `${view}: 主键列 ${domain}.${object}.${col} 缺 data-col`).toContain(`data-col="${col}"`);
|
||
}
|
||
}
|
||
});
|
||
|
||
it('可编辑件上,每一个 required:true 的列都有真能填的控件(data-col)—— 必填但填不了是墙不配路', () => {
|
||
for (const { view, domain, object } of EDITABLE) {
|
||
const table = byAddress(domain, object);
|
||
const required = table?.fields.filter((f) => f.required === true).map((f) => f.name) ?? [];
|
||
expect(required.length, `${domain}.${object} 没有声明 required 列`).toBeGreaterThan(0);
|
||
const html = viewHtml(view);
|
||
for (const col of required) {
|
||
expect(html, `${view}: 必填列 ${domain}.${object}.${col} 没有可填的控件`).toContain(`data-col="${col}"`);
|
||
}
|
||
}
|
||
});
|
||
});
|
||
|
||
describe('活动入口', () => {
|
||
it('没有 run —— 行为全在清单里', () => {
|
||
expect(activity.run).toBeUndefined();
|
||
});
|
||
});
|
||
|
||
const CONTRACT = {
|
||
activityId: 'part-substitution-assessment',
|
||
version: '1.0.0',
|
||
data: {},
|
||
message: undefined,
|
||
outputPorts: { mandatory: ['assessment-released'], selective: [] },
|
||
prohibitions: [],
|
||
discretion: [],
|
||
} as const;
|
||
|
||
/* 验收件题头是验收绿(2026-08-27 用户定:与对话区验收卡同色,不再是目标件的红) */
|
||
describe('两张验收件的主题色', () => {
|
||
it('ACC1 / ACC2 的 --ledger 是验收绿 #0aa294,题头渐变不再带红;CONDITIONAL 徽标保留语义红', async () => {
|
||
const { ACC1_CSS, ACC2_CSS } = await import('../src/views/target-style.ts');
|
||
for (const css of [ACC1_CSS, ACC2_CSS]) {
|
||
expect(css).toContain('--ledger:#0aa294');
|
||
expect(css).not.toContain('--ledger:#b23a2e');
|
||
expect(css).not.toContain('#fdf3f1');
|
||
expect(css).toContain('rgba(10,162,148,.25)');
|
||
}
|
||
// ACC2 的 CONDITIONAL 徽标保留目标件的语义红;ACC1 目标件 2026-08-27 改版把它改成了品牌蓝,照抄
|
||
expect(ACC2_CSS).toContain('.st.cond{color:#b23a2e');
|
||
expect(ACC1_CSS).toContain('.st.cond{color:var(--brand-deep)');
|
||
});
|
||
});
|
||
|
||
/*
|
||
* 差异评估结论审核 · 目标件 2026-08-27 改版:「一个候选一张表」(用户:存在多个替代 PN 要对比差异时看不清楚)。
|
||
* 外层按 relation 逐个复制 <section class="cand">,里层明细只列该候选名下的行(data-by 父键过滤);
|
||
* 汇总表一个候选一条;品牌从 parts.profile 整表按料号挑(取值,不是算)。
|
||
*/
|
||
/*
|
||
* SKIP(2026-08-29,Task 2 → Task 8 补回):`difference-review` 局面已随重写撤下,
|
||
* 新版由 Task 8 重做(件② `difference-review` —— 差异评估结论审核)。
|
||
*/
|
||
describe.skip('difference-review:一个候选一张表', () => {
|
||
const html = viewHtml('difference-review');
|
||
it('候选块是嵌套行模版:外层 relation、里层明细按 substitution_relation_id 过滤;容器标签不在行模版里再出现', () => {
|
||
expect(html).toContain('<article class="cands" data-rows="mine:substitution.relation"><section class="cand">');
|
||
expect(html).toContain('data-rows="mine:substitution.difference_item" data-by="substitution_relation_id"');
|
||
const inner = html.slice(html.indexOf('<section class="cand">'), html.indexOf('</section></article>'));
|
||
expect(inner).not.toContain('<article');
|
||
expect(inner).toContain('class="cand-idx"');
|
||
expect(inner).toContain('data-slot="original_internal_pn"');
|
||
});
|
||
it('汇总表一个候选一条:推荐程度 / 风险级别按值上色,差异统计绑 difference_summary,品牌按料号从 parts.profile 挑', () => {
|
||
expect(html).toContain('<tbody data-rows="mine:substitution.relation"><tr>');
|
||
expect(html).toContain('<span class="rec" data-class-of="recommendation_level"></span>');
|
||
expect(html).toContain('class="stat" data-slot="difference_summary"');
|
||
expect(html).toContain('data-rows="parts.profile" data-by="internal_pn=candidate_internal_pn"');
|
||
expect(html).toContain('data-slot="substitution_relation_rows"');
|
||
expect(html).not.toContain('【AI 产出的差异明细】');
|
||
});
|
||
});
|
||
|
||
/* 评估输入清单的件上摆「候选参数现值」:一颗料一张、里层按料号过滤、只读(没有 data-edit) */
|
||
/*
|
||
* SKIP(2026-08-29,Task 2 → Task 7 补回):`eval-input-list` 局面已随重写撤下,
|
||
* 「候选参数现值」这块只读参考并入 Task 7 的 `input-supplement`。
|
||
*/
|
||
describe.skip('eval-input-list:候选参数现值只读参考', () => {
|
||
it('外层按核对表逐颗复制,里层 parts.parameter 按 internal_pn 过滤,不是可编辑表', () => {
|
||
const html = viewHtml('eval-input-list');
|
||
expect(html).toContain(`<article class="prefs" data-rows="mine:${D}.pn_check"><section class="pref">`);
|
||
expect(html).toContain('<tbody data-rows="parts.parameter" data-by="internal_pn"><tr>');
|
||
const block = html.slice(html.indexOf('<article class="prefs"'), html.indexOf('</section></article>'));
|
||
expect(block).not.toContain('data-edit=');
|
||
expect(block).toContain('data-slot="param_value"');
|
||
expect(html).toContain('【候选参数现值】');
|
||
});
|
||
});
|
||
|
||
/*
|
||
* 运营中心沙箱看板(2026-08-27 用户定:替换全流程分支图,只摆最近一次已放行实例的关键台账落账快照)。
|
||
* 模版归活动,「哪几行是这一单的」也归活动 —— `data` 声明:案号怎么铸、哪张表按哪一列圈;基座只照它查正式表。
|
||
*/
|
||
describe('activity-ledger:关键台账数据看板', () => {
|
||
const board = manifest.assets.viewTemplates.find((v) => v.name === BOARD_VIEW);
|
||
const html = board?.html ?? '';
|
||
const data = board?.data;
|
||
const declaredAddresses = new Set(manifest.declarations.ledger.map((l) => `${l.domain}.${l.object}`));
|
||
|
||
/*
|
||
* SKIP(2026-08-29,Task 2 → Task 9 补回):看板的 `data` 声明(`views/ledger-board.ts` 的
|
||
* `LEDGER_BOARD_DATA`)还圈着已退休的旧台账(`sub_eval.pn_check` / `sub_eval.verdict` /
|
||
* `demand.request` / `customer.profile` / `parts.document`)—— Task 9「运营中心看板重绑」
|
||
* 会把它换成新九张表;这条取数声明的自洽性检查到那时再验。html 自身的结构测试(下面三条)
|
||
* 不依赖 `declarations.ledger`,仍然有效、继续跑。
|
||
*/
|
||
it.skip('取数声明:案号台账是 record 且 key 是它的 caseKey、前缀 W05-;每条 rows 的表都已声明;from 只指 case 或前面的 as.列', () => {
|
||
expect(data).toBeDefined();
|
||
if (data === undefined) return;
|
||
const caseLedger = manifest.declarations.ledger.find((l) => `${l.domain}.${l.object}` === data.case.ledger);
|
||
expect(caseLedger?.kind).toBe('record');
|
||
expect(caseLedger?.caseKey).toBe(data.case.key);
|
||
expect(data.case.prefix).toBe('W05-');
|
||
const seen = new Set<string>();
|
||
for (const entry of data.rows) {
|
||
expect(declaredAddresses.has(entry.ledger), entry.ledger).toBe(true);
|
||
const columns = new Set(manifest.declarations.ledger.find((l) => `${l.domain}.${l.object}` === entry.ledger)?.fields.map((f) => f.name));
|
||
expect(columns.has(entry.by.column), `${entry.as} by ${entry.by.column}`).toBe(true);
|
||
for (const ref of [entry.by.from, entry.not?.from]) {
|
||
if (ref === undefined || ref === 'case') continue;
|
||
const at = ref.lastIndexOf('.');
|
||
const as = ref.slice(0, at);
|
||
const column = ref.slice(at + 1);
|
||
expect(seen.has(as), `${entry.as} 引用了后面 / 不存在的集合 ${as}`).toBe(true);
|
||
const source = data.rows.find((r) => r.as === as);
|
||
const sourceColumns = new Set(manifest.declarations.ledger.find((l) => `${l.domain}.${l.object}` === source?.ledger)?.fields.map((f) => f.name));
|
||
expect(sourceColumns.has(column), `${ref} 不是 ${source?.ledger} 的列`).toBe(true);
|
||
}
|
||
seen.add(entry.as);
|
||
}
|
||
// 六张主表 + 案号两张 + 同料号历史都圈了
|
||
expect([...seen]).toEqual(
|
||
expect.arrayContaining([`${D}.verdict`, `${D}.pn_check`, 'demand.request', 'customer.profile', 'parts.profile', 'parts.parameter', 'parts.document', 'substitution.relation', 'substitution.difference_item', 'validation.record', 'validation.history']),
|
||
);
|
||
});
|
||
|
||
it('html 里每个 data-rows 键都是声明里的一个 as(行由基座圈过,不是整表);嵌套只用 data-by 父键过滤', () => {
|
||
const as = new Set(data?.rows.map((r) => r.as) ?? []);
|
||
const keys = [...html.matchAll(/data-rows="([^"]+)"/g)].map((m) => m[1] ?? '');
|
||
expect(keys.length).toBeGreaterThan(8);
|
||
for (const key of keys) expect(as.has(key), key).toBe(true);
|
||
expect(html).toContain('<span class="pnrow" data-rows="sub_eval.pn_check" data-by="internal_pn">');
|
||
expect(html).toContain('<div class="pchips" data-rows="parts.parameter" data-by="internal_pn">');
|
||
expect(html).toContain('<tbody class="hist" data-rows="validation.history">');
|
||
expect(html).not.toContain('mine:');
|
||
});
|
||
|
||
it('无脚本、无 on*、无外链(过安装静态审);有 </body> 给出口注入报高脚本;三态由 board:state 投影;只读', () => {
|
||
expect(html.length).toBeGreaterThan(5000);
|
||
expect(html).not.toMatch(/<\s*script\b/i);
|
||
expect(html).not.toMatch(/\son[a-z]+\s*=/i);
|
||
expect(html).not.toMatch(/(src|href)\s*=\s*["']?\s*(https?:)?\/\//i);
|
||
expect(html.toLowerCase()).toContain('</body>');
|
||
expect(html).toContain('<div class="page" data-class-of="board:state">');
|
||
expect(html).toContain('.page.v-none .notice.none{display:flex}');
|
||
expect(html).toContain('.page.v-missing .notice.missing{display:flex}');
|
||
expect(html).not.toContain('data-edit=');
|
||
expect(html).not.toContain('data-col=');
|
||
});
|
||
|
||
it('目标件六张表的表头与页头要素都在;MATCH 行淡显、放弃的候选只在有 abandoned 时露面 —— 都是 class 交样式表,模版不判', () => {
|
||
for (const header of ['<th>内部料号</th>', '<th>替代关系 ID</th>', '<th>参数项</th>', '<th>关联料号</th>', '<th>验证 ID</th>', '<th>标准化参数集</th>']) expect(html).toContain(header);
|
||
expect(html).toContain('data-slot="case:instanceNo"');
|
||
expect(html).toContain('data-slot="case:ref"');
|
||
expect(html).toContain(`data-slot="${D}.verdict.original_internal_pn"`);
|
||
expect(html).toContain('data-slot="demand.request.request_id"');
|
||
expect(html).toContain('<tr data-class-of="difference_status">');
|
||
expect(html).toContain('tr.v-match td{opacity:.55}');
|
||
expect(html).toContain('.aband:has(i.v-abandoned){display:block}');
|
||
expect(html).toContain('data-slot="substitution_difference_item_difference_status_unknown"');
|
||
});
|
||
});
|