feat: 账本 ledger 模块（任务书 03）
一张只追加的表（序号、种类、内容、哈希），禁改禁删两条触发器。 有自己编号、只发生一次的 8 种记录加部分唯一索引，重复就报错：阶段开、阶段关、运行登记、 确认启动（确认没启动的结果也算这一条，防止同一次运行既扣次又退还）、运行结果、模型调用开始、结束、消融结算。 写入时只查四种拒收：只收开着的阶段的提交；替换必须接本人同一阶段、同一对象、同一种提交的最新一版，实现与核验类不能替换； 结果先于确认启动（没启动除外）；本轮消融已有格启动后不能关解释阶段。 写提案时作者自评拆成封存。提供打开（核对表结构版本，非账本库不改动）、追加、同步事务、读取、未完成的动作。 第一版的可见性、截止序号、通用唯一键、jev 记录和多余的前提检查已按新版任务书清掉。 三轮全新上下文审阅；flow 传错参数一类的意见按 CLAUDE.md"罕见情况怎么办"第 2 条只记录不加代码。 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
+6
-1
@@ -13,6 +13,7 @@
|
||||
- `docs/结构.md`：13 个模块加 vocab、步骤清单、账本记录种类、可替换点、检查规则、粗估规模（6cf9f0c）。
|
||||
- 任务书 01（工具链 + vocab）：npm 包、TypeScript、`node --test`、dependency-cruiser 规则全写，临时违规实测 14 条都报错；vocab 1030 行，`npm run check` 通过，`npm test` 24 条通过。结构.md 第八部分补了 vocab 对外内容。
|
||||
- 任务书 02（rules）：src/rules 13 个文件、1761 行，`npm test` 192 条通过。预期值来源有三个：方法.md 的手算例；`tests/fixtures/rules/` 下不看实现写的 Python 脚本（重跑后 JSON 逐字节相同）；参考项目 calc15、calc14b 里与现行方法一致的部分。审阅做了两轮，每轮都换一个全新上下文的子 Agent。结构.md 第八部分补了 rules 对外内容。
|
||||
- 任务书 03（ledger）：src/ledger 5 个文件、584 行，`npm test` 226 条通过（ledger 34 条）。一张只追加的表，两条禁改禁删触发器，8 处唯一约束（阶段开、阶段关、运行登记、确认启动或确认没启动、运行结果、模型调用开始、结束、消融结算），写入时只查四种拒收，提案的作者自评拆成封存另存。按第一版写过可见性和通用唯一键，按新版全部清掉了。全新上下文的审阅做了三轮。结构.md 第八部分补了 ledger 对外内容。
|
||||
|
||||
## 关键决定（用户原话见各提交说明）
|
||||
|
||||
@@ -26,6 +27,7 @@
|
||||
- 每份预测只对一个候选、每份挑刺只针对一个目标；入口、因子、实现条款的编号由程序按位置分配；实现提交不设代码字段（程序从工作目录冻结）；运行结果用"没启动 / 正常退出 / 超时 / 被取消"一个字段表示启动与结束；四种 token 数都可记测不到。用户原话："3、4、5、6：同意保留你的写法。方法.md 第五部分的表（入口编号、实现的代码字段）由监督者改。"
|
||||
- 账本改为一张只追加的表，全部记录都不能改（ca2cd52）。用户原话："同意全部只追加"。
|
||||
- 谁能看什么归 materials（按任务定材料，第一条消息和查询工具是同一份材料的两种给法），ledger 只存；只在"只能一次"处加唯一约束；Jev 不单独成记录，记在它判断的那件事里（0b0cbb0）。用户原话："给到派的材料有哪些"和"一开始灌进去还是主动等派去查"是分开的事情；Jev"本质只是一种可以代替程序的判别器"。
|
||||
- 有自己编号、按定义只发生一次的记录加唯一约束，重复就报错；记录之间的前后顺序检查一律不加，flow 调 ledger 是模块之间的调用，不算边界（493ea6d）。用户原话："前后顺序检查一律不加。flow 调用 ledger 是模块之间的调用，不算系统边界；顺序由 flow 的唯一入口和它的测试保证。"
|
||||
- 计算时间不再是预算，只记录；预算只有运行次数（42347e4）。用户原话："CPU 秒不当预算但是要记录"。
|
||||
- 解释的每格三个数改为可选，关掉消融时不交。用户原话："同意 A：解释的 cells 去掉 minItems，改为可选"。
|
||||
- rules 的约定（方法.md 没写死的地方）：入池 P < 0.5 严格；采纳和消融中发现都要严格大于采纳线；成本用 max(0, ĝ₀)；没测的格附"含它的最好组合"的净值；战绩在分数原值上判；编号一律按自然序比较。用户原话："1–8 都同意"。消融成分权重跟随 λ、分歧量不掺开放预留，由用户改进方法.md（a050771）。
|
||||
@@ -55,6 +57,9 @@
|
||||
- 核验提交 `body` 的字段说明要补一句：同一行同时出现在因子和保障里时，在说明里写一句提醒（方法.md 4.3 ⑨）。写 submissions 时一并改。
|
||||
- 给 config：每批上限必须 ≥ 1（为 0 时 R̂ 是无穷大，选格会卡住）。
|
||||
- 给 flow：同一轮里选批和选格都用随机规则时，传不同的"版本"算种子，否则两次抽取用同一串随机数。
|
||||
- 给 materials：被退回的提案原样存在退回记录里，作者自评没有拆成封存，所以不能把别人的退回记录给研究者看。方法.md 第三部分"作者自评从交上来那一刻起就封存"和"被退回的提交原样记"之间怎么处理，待用户定。
|
||||
- 给 flow：确认启动前要先登记运行；模型调用要先开始再结束；阶段要先开再关；消融编号要由采纳确定地算出（否则换编号能重复结算）；ledger 都不查。
|
||||
- 待用户定：结构.md 第 10 种记录的"整个环境的用量"（实验员会话的 CPU、内存等）在 ledger 里还没有对应的种类；私人笔记的版本号不检查。
|
||||
- 写 plugin 时实测：源码导入路径带 `.ts` 扩展名，Pi 加载插件能否接受。
|
||||
|
||||
## 工作方式（用户 2026-10-01 定）
|
||||
@@ -68,4 +73,4 @@
|
||||
|
||||
## 下一步
|
||||
|
||||
任务书 02（rules）已核对通过（60c21ae）。`任务书/03-ledger.md` 已写好，等用户另开 Agent 执行，监督者核对。
|
||||
任务书 03（ledger）已完成并提交，等监督者核对。
|
||||
|
||||
+29
@@ -383,3 +383,32 @@ Jev 是一个判别器，和程序检查地位相同，不单独成一种记录
|
||||
- `CELL_RULES`、`CellRuleName`：选格规则登记表，四条：信息价值减成本、全测、随机抽同样多的格、都不测。
|
||||
- `runBatchRule(name, input)`：按名字跑一条选批规则，结果带上规则名和版本，flow 写决定记录时从这里取。
|
||||
- `runCellRule(name, input)`：按名字跑一条选格规则，结果带上规则名和版本。
|
||||
|
||||
### ledger
|
||||
|
||||
下面是 `src/ledger/index.ts` 导出的全部内容。账本是一个 SQLite 文件里的一张只追加的表；外面拿不到数据库连接，只能经下面这些函数读写。ledger 不判断谁能看到哪些记录，那归 materials。
|
||||
|
||||
**打开与写入**
|
||||
|
||||
- `Ledger.open(path)`：打开一份研究的账本，新文件就建表，旧文件核对表结构版本号，对不上就报错。
|
||||
- `ledger.close()`：关闭数据库。
|
||||
- `ledger.append(kind, content)`：追加一条记录并返回它。先按种类检查内容，再做第四部分说的四种拒收检查（第 1 条是只收开着的阶段的提交，即有开阶段记录、没有关阶段记录），都在同一个事务里。有自己编号、按定义只发生一次的记录按编号加了唯一约束，重复写就报错：阶段开、阶段关、运行登记、确认启动（"结果是没启动"也算在这一条里，一次运行两样只能有一样）、运行结果、模型调用开始、模型调用结束、消融结算；其余记录照写。记录之间的先后（例如结果前要先有登记）不查。写提案时，作者自评从提案里拆出来，在同一个事务里另存成一条封存。
|
||||
- `ledger.transaction(fn)`：里面的几次追加要么全部写进去，要么一条都不写；fn 必须是同步函数，返回值类型排除了 Promise，异步函数在类型检查时就被挡住；不能嵌套。
|
||||
|
||||
**读取**
|
||||
|
||||
- `ledger.records()`：全部记录，按序号排，读出时按种类重新检查内容。
|
||||
- `ledger.stageClosed(stageId)`：这个阶段关了没有。
|
||||
- `ledger.latestSubmission(slot)`：某份提交（阶段、作者、对象、编号）的最新一版。
|
||||
- `ledger.pending()`：未完成的动作，即登记了还没有结果的运行（带着有没有确认启动）和开始了还没有结束的模型调用，恢复时由 flow 逐个核对。
|
||||
- `Pending`（类型）：`pending()` 的返回值。
|
||||
|
||||
**类型**
|
||||
|
||||
- `Kind`：记录的种类，共 19 种：`config`（研究配置）、`stageOpen` / `stageClose`（阶段开、关）、`submission`（提交的每一版）、`rejection`（退回，带原样内容和理由，理由来自 Jev 的带原始概率和结论）、`note`（私人笔记的每一版）、`protocol`（冻结协议）、`seal` / `unseal`（作者自评的封存、解封，都用记录序号指明是哪一条）、`code`（冻结代码）、`runPlanned` / `runStarted` / `runResult`（运行的登记要跑、确认启动、结果）、`decision`（决定）、`settlement`（一次消融的结算）、`callStarted` / `callUsage` / `callEnded`（模型调用的开始，附这次任务的全部材料；每次回复的用量；结束）、`event`（事件）。
|
||||
- `AppendKind`：调用方能追加的种类，除了 `seal`（封存只由 ledger 在写提案时一并写）。
|
||||
- `Content<K>`：某种记录存下的内容。
|
||||
- `Input<K>`：追加时交的内容；只有提交和存下的不同：提案带着作者自评。
|
||||
- `LedgerRecord<K>`：一条记录：序号、种类、内容的哈希、内容。
|
||||
- `SubmissionSlot`（类型）：一份提交不论第几版：阶段、作者、提交种类（版本号只在同一种提交里往下接）、针对谁（候选或格，没有就空）、这一份的编号（只有提案、挑刺这两种一人能交好几份的才有，其余必须为空）。
|
||||
- `StoredSubmission`、`SubmissionInput`（类型）：存下的提交和追加时交的提交，按提交种类区分内容类型；版本号从 1 起，替换时加 1，被替换的就是上一版。
|
||||
|
||||
@@ -0,0 +1,13 @@
|
||||
// ledger：研究的全部记录怎么存、怎么保证不能改、写入时拒收什么、花钱的动作怎么先登记
|
||||
// （docs/结构.md 第二部分、第四部分）。
|
||||
export { Ledger, type Pending } from "./ledger.ts";
|
||||
export {
|
||||
type AppendKind,
|
||||
type Content,
|
||||
type Input,
|
||||
type Kind,
|
||||
type LedgerRecord,
|
||||
type StoredSubmission,
|
||||
type SubmissionInput,
|
||||
type SubmissionSlot,
|
||||
} from "./kinds.ts";
|
||||
@@ -0,0 +1,232 @@
|
||||
import { Type, type Static, type TSchema } from "typebox";
|
||||
import { Value } from "typebox/value";
|
||||
import {
|
||||
Id,
|
||||
JudgedResult,
|
||||
ModelUsage,
|
||||
ProcessResult,
|
||||
Range,
|
||||
RunDefinition,
|
||||
STAGES,
|
||||
SUBMISSION_KINDS,
|
||||
SUBMISSION_SCHEMAS,
|
||||
type Json,
|
||||
type Submission,
|
||||
type SubmissionKind,
|
||||
} from "../vocab/index.ts";
|
||||
|
||||
// 账本记录的种类（docs/结构.md 第四部分的 11 种记录，细分成下面 19 种），每种一份内容定义（TypeBox）。
|
||||
// 下面写"只有一条""只一次"的，由 schema.ts 的唯一约束保证。
|
||||
// 配置、冻结协议、决定、结算的内容由别的模块产生，这里只检查最少的标识字段，其余只检查是 Json。
|
||||
|
||||
const CLOSED = { additionalProperties: false } as const;
|
||||
const Text = Type.String({ minLength: 1 });
|
||||
const AnyJson = Type.Unsafe<Json>({});
|
||||
const Round = Type.Integer({ minimum: 0 });
|
||||
const Version = Type.Integer({ minimum: 1 });
|
||||
const Seq = Type.Integer({ minimum: 1 });
|
||||
const OptionalId = Type.Union([Id, Type.Null()]);
|
||||
|
||||
// ---------------------------------------------------------------------------
|
||||
// 提交（记录 3）
|
||||
// ---------------------------------------------------------------------------
|
||||
|
||||
/**
|
||||
* 一人在同一阶段对同一对象能交好几份的提交种类：提案（每人 1 至 M 条，方法.md 4.3 ②）、
|
||||
* 挑刺（回应时按挑刺编号逐条答复，方法.md 第五部分）。这两种的每一份带一个由程序分配的编号 item，
|
||||
* 替换时按 SubmissionSlot 找同一份的最新一版（docs/结构.md 第四部分"写入时由 ledger 拒绝的几种情况"）。
|
||||
* 其余种类一人对一个对象只交一份，item 在字段定义里就只能为空，所以不能另起一个编号绕开"不能替换"。
|
||||
*/
|
||||
const MULTI_ITEM_KINDS: ReadonlySet<SubmissionKind> = new Set(["proposal", "critique"]);
|
||||
|
||||
/**
|
||||
* 一份提交不论第几版：哪个阶段、谁交的、哪种提交、针对谁（候选或格，没有就空）、这一份的编号（见 MULTI_ITEM_KINDS）。
|
||||
* 提交种类在位置里：版本号只在同一种提交里往下接，别的种类顶替不了不能替换的实现，提案和挑刺用了同一个编号也接不上。
|
||||
*/
|
||||
export interface SubmissionSlot {
|
||||
stageId: string;
|
||||
author: string;
|
||||
kind: SubmissionKind;
|
||||
target: string | null;
|
||||
item: string | null;
|
||||
}
|
||||
|
||||
/**
|
||||
* 账本里存的提案不带作者自评：写入时 ledger 把它拆出来另存成一条封存（方法.md 第三部分"作者自评从交上来那一刻起就封存"）。
|
||||
* 这样给别人看提案时不用去掉什么。
|
||||
*/
|
||||
const STORED_BODY: Record<SubmissionKind, TSchema> = {
|
||||
...SUBMISSION_SCHEMAS,
|
||||
proposal: Type.Omit(SUBMISSION_SCHEMAS.proposal, ["authorEstimate"]),
|
||||
};
|
||||
|
||||
type StoredBody<K extends SubmissionKind> = K extends "proposal"
|
||||
? Omit<Submission<"proposal">, "authorEstimate">
|
||||
: Submission<K>;
|
||||
|
||||
/** 一份提交的一版。version 从 1 起，替换时加 1，被替换的就是上一版。 */
|
||||
type SubmissionOf<Body extends Record<SubmissionKind, unknown>> = {
|
||||
[K in SubmissionKind]: Omit<SubmissionSlot, "kind"> & { version: number; kind: K; content: Body[K] };
|
||||
}[SubmissionKind];
|
||||
|
||||
export type StoredSubmission = SubmissionOf<{ [K in SubmissionKind]: StoredBody<K> }>;
|
||||
/** 写入时交给 ledger 的提交：提案带着作者自评。 */
|
||||
export type SubmissionInput = SubmissionOf<{ [K in SubmissionKind]: Submission<K> }>;
|
||||
|
||||
/** 某种提交一版的完整字段定义：外层的位置和版本号，加上按提交种类检查的 content。 */
|
||||
function submissionSchema(kind: SubmissionKind, body: TSchema): TSchema {
|
||||
return Type.Object(
|
||||
{
|
||||
stageId: Id,
|
||||
author: Id,
|
||||
target: OptionalId,
|
||||
item: MULTI_ITEM_KINDS.has(kind) ? Id : Type.Null(),
|
||||
version: Version,
|
||||
kind: Type.Literal(kind),
|
||||
content: body,
|
||||
},
|
||||
CLOSED,
|
||||
);
|
||||
}
|
||||
|
||||
const SUBMISSION_KIND_NAMES = Object.keys(SUBMISSION_KINDS) as SubmissionKind[];
|
||||
const SUBMISSION_FORMS = {
|
||||
input: Object.fromEntries(SUBMISSION_KIND_NAMES.map((k) => [k, submissionSchema(k, SUBMISSION_SCHEMAS[k])])),
|
||||
stored: Object.fromEntries(SUBMISSION_KIND_NAMES.map((k) => [k, submissionSchema(k, STORED_BODY[k])])),
|
||||
} as Record<"input" | "stored", Record<SubmissionKind, TSchema>>;
|
||||
|
||||
// ---------------------------------------------------------------------------
|
||||
// 种类表
|
||||
// ---------------------------------------------------------------------------
|
||||
|
||||
export const KINDS = {
|
||||
/** 1 研究配置：冻结的配置、步骤清单、规则、方法参数、提示词哈希。内容由配置模块产生，这里只检查是 Json。 */
|
||||
config: AnyJson,
|
||||
|
||||
/** 2 阶段开：阶段名、第几轮（第零轮是 0）、这个阶段的应交名单。阶段编号由 flow 起，一个编号只开一次；同一个阶段名可以开好几次（如讨论）。 */
|
||||
stageOpen: Type.Object(
|
||||
{ stageId: Id, stage: Type.Enum(STAGES, { type: "string" }), round: Round, roster: Type.Array(Id, { minItems: 1 }) },
|
||||
CLOSED,
|
||||
),
|
||||
/** 2 阶段关：关闭即公开，此后这个阶段不再收提交。一个阶段只关一次。 */
|
||||
stageClose: Type.Object({ stageId: Id }, CLOSED),
|
||||
|
||||
/** 3 提交的每一版（经 submissions 收下后由 flow 写入）。字段定义按提交种类选，见 checkContent。 */
|
||||
submission: Type.Unsafe<StoredSubmission>({}),
|
||||
/**
|
||||
* 3 退回：被退回的那份内容原样存（它可能正是因为不合格式才被退回），连同退回理由。
|
||||
* 理由来自 Jev 的，jev 里带原始概率和结论；程序检查退回的，jev 为空。
|
||||
*/
|
||||
rejection: Type.Object(
|
||||
{
|
||||
stageId: Id,
|
||||
author: Id,
|
||||
kind: Type.Enum(SUBMISSION_KIND_NAMES, { type: "string" }),
|
||||
submitted: AnyJson,
|
||||
reason: Text,
|
||||
jev: Type.Union([Type.Object({ probabilities: AnyJson, conclusion: AnyJson }, CLOSED), Type.Null()]),
|
||||
},
|
||||
CLOSED,
|
||||
),
|
||||
/** 3 私人笔记的每一版：每次修改追加一版，以版本号最大的为准。 */
|
||||
note: Type.Object({ author: Id, version: Version, body: Text }, CLOSED),
|
||||
|
||||
/** 4 冻结协议：每个候选一份。内容由 flow 合成，先存 Json。 */
|
||||
protocol: Type.Object({ candidate: Id, protocol: AnyJson }, CLOSED),
|
||||
|
||||
/** 5 作者自评的封存：proposal 是那一版提案记录的序号。只由 ledger 在写提案时一并写入。 */
|
||||
seal: Type.Object({ proposal: Seq, range: Range }, CLOSED),
|
||||
/** 5 解封：放出哪几条封存（封存记录的序号）。合并而成的候选有几位作者就有几条。 */
|
||||
unseal: Type.Object({ candidate: Id, seals: Type.Array(Seq, { minItems: 1 }) }, CLOSED),
|
||||
|
||||
/** 6 冻结代码：候选或消融格的代码版本号（由 workspace 给出）和因子位置表。位置表先存 Json。 */
|
||||
code: Type.Object({ object: Id, version: Text, locations: AnyJson }, CLOSED),
|
||||
|
||||
/** 7 运行·登记要跑（占名额）：运行编号由 flow 起，一个编号只登记一次；用途、轮次、对象（候选或格；基线没有）、运行定义。 */
|
||||
runPlanned: Type.Object(
|
||||
{
|
||||
run: Id,
|
||||
purpose: Type.Enum(["baseline", "forward", "ablation"], { type: "string" }),
|
||||
round: Round,
|
||||
object: OptionalId,
|
||||
definition: RunDefinition,
|
||||
},
|
||||
CLOSED,
|
||||
),
|
||||
/** 7 运行·确认启动（扣次）。一次运行只有一条，而且和"结果是没启动"只能有一样。 */
|
||||
runStarted: Type.Object({ run: Id }, CLOSED),
|
||||
/** 7 运行·结果：进程层面的结果，以及拿到产物时的独立判分（没有就空）。一次运行只有一条。 */
|
||||
runResult: Type.Object({ run: Id, process: ProcessResult, judged: Type.Union([JudgedResult, Type.Null()]) }, CLOSED),
|
||||
|
||||
/**
|
||||
* 8 决定：选批、采纳、选格、押的因子、入口分配、判重合并。decision 是决定种类。
|
||||
* 输入输出存 Json：ledger 不认识 rules。判重合并的输入里带 Jev 的原始概率。
|
||||
*/
|
||||
decision: Type.Object(
|
||||
{
|
||||
decision: Id,
|
||||
round: Round,
|
||||
rule: Text,
|
||||
ruleVersion: Type.Integer({ minimum: 1 }),
|
||||
params: AnyJson,
|
||||
seed: Type.Union([Text, Type.Null()]),
|
||||
input: AnyJson,
|
||||
output: AnyJson,
|
||||
},
|
||||
CLOSED,
|
||||
),
|
||||
|
||||
/** 9 结算：一次消融一条，各格结局、配合关系、解释的确认或否定等先存 Json。 */
|
||||
settlement: Type.Object({ ablation: Id, result: AnyJson }, CLOSED),
|
||||
|
||||
/** 10 模型调用开始：先登记再调用（方法.md 第九部分）。附这次任务的全部材料，用来重现模型当时看到了什么。一次调用只有一条。 */
|
||||
callStarted: Type.Object({ call: Id, actor: Id, stageId: OptionalId, model: Text, materials: AnyJson }, CLOSED),
|
||||
/** 10 模型调用每次回复的用量，reply 从 1 起。 */
|
||||
callUsage: Type.Object({ call: Id, reply: Type.Integer({ minimum: 1 }), usage: ModelUsage }, CLOSED),
|
||||
/** 10 模型调用结束：正常结束，或失败并写明原因。一次调用只有一条。 */
|
||||
callEnded: Type.Union([
|
||||
Type.Object({ call: Id, ending: Type.Literal("completed") }, CLOSED),
|
||||
Type.Object({ call: Id, ending: Type.Literal("failed"), reason: Text }, CLOSED),
|
||||
]),
|
||||
|
||||
/** 11 事件：违规、"按配置跳过"、"没测"及原因、停下的原因、运行状态未知等。event 是事件种类。 */
|
||||
event: Type.Object({ event: Id, reason: Text, objects: Type.Array(Id) }, CLOSED),
|
||||
} as const;
|
||||
|
||||
export type Kind = keyof typeof KINDS;
|
||||
export type Content<K extends Kind> = Static<(typeof KINDS)[K]>;
|
||||
/** 调用方能追加的种类：封存只由 ledger 在写提案时一并写。 */
|
||||
export type AppendKind = Exclude<Kind, "seal">;
|
||||
/** 写入时交的内容：除了提交（提案带作者自评），都和存下的内容相同。 */
|
||||
export type Input<K extends AppendKind> = K extends "submission" ? SubmissionInput : Content<K>;
|
||||
|
||||
/** 一条记录。序号就是写入的先后。 */
|
||||
export type LedgerRecord<K extends Kind = Kind> = {
|
||||
[P in K]: { seq: number; kind: P; hash: string; content: Content<P> };
|
||||
}[K];
|
||||
|
||||
function errorsOf(schema: TSchema, value: unknown): string {
|
||||
return Value.Errors(schema, value)
|
||||
.slice(0, 3)
|
||||
.map((e) => `${e.instancePath || "/"} ${e.message}`)
|
||||
.join("；");
|
||||
}
|
||||
|
||||
/**
|
||||
* 按种类检查内容。提交按提交种类选字段定义：写入时用带作者自评的那份（input），存下和读出时用不带的（stored）。
|
||||
*/
|
||||
export function checkContent(kind: Kind, content: unknown, form: "input" | "stored"): void {
|
||||
let schema: TSchema = KINDS[kind];
|
||||
let label = kind as string;
|
||||
if (kind === "submission") {
|
||||
const sub = (content as { kind?: unknown } | null)?.kind;
|
||||
if (typeof sub !== "string" || !Object.hasOwn(SUBMISSION_KINDS, sub)) {
|
||||
throw new Error(`账本记录"submission"的提交种类不认识：${String(sub)}`);
|
||||
}
|
||||
schema = SUBMISSION_FORMS[form][sub as SubmissionKind];
|
||||
label = `submission（提交种类 ${sub}）`;
|
||||
}
|
||||
if (!Value.Check(schema, content)) {
|
||||
throw new Error(`账本记录"${label}"的内容不符合定义：${errorsOf(schema, content)}`);
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,209 @@
|
||||
import { DatabaseSync, type SQLInputValue } from "node:sqlite";
|
||||
import { canonical, hash } from "../vocab/index.ts";
|
||||
import {
|
||||
checkContent,
|
||||
KINDS,
|
||||
type AppendKind,
|
||||
type Content,
|
||||
type Input,
|
||||
type Kind,
|
||||
type LedgerRecord,
|
||||
type SubmissionInput,
|
||||
type SubmissionSlot,
|
||||
} from "./kinds.ts";
|
||||
import { checkWrite, type WriteState } from "./rejections.ts";
|
||||
import { CREATE_SCHEMA, schemaState } from "./schema.ts";
|
||||
|
||||
// 一份研究的账本：一个 SQLite 文件、一张只追加的表（docs/结构.md 第四部分）。
|
||||
// 同一时刻只有一个控制进程打开它，由 plugin 的文件锁保证，这里不另做判断。
|
||||
// 谁能看到哪些记录不归这里，由 materials 按任务定；ledger 把记录原样交给程序。
|
||||
|
||||
interface Row {
|
||||
seq: number;
|
||||
kind: string;
|
||||
content: string;
|
||||
hash: string;
|
||||
}
|
||||
|
||||
/** 还没有结果的花钱动作（方法.md 第九部分）。恢复时 flow 逐个核对，怎么核对不归 ledger。 */
|
||||
export interface Pending {
|
||||
/** 登记了、还没有结果的运行，以及它有没有"确认启动"记录。按登记先后排。 */
|
||||
runs: Array<{ plan: LedgerRecord<"runPlanned">; started: boolean }>;
|
||||
/** 开始了、还没有结束的模型调用。按开始先后排。 */
|
||||
calls: Array<LedgerRecord<"callStarted">>;
|
||||
}
|
||||
|
||||
export class Ledger {
|
||||
readonly path: string;
|
||||
readonly #db: DatabaseSync;
|
||||
#inTransaction = false;
|
||||
|
||||
private constructor(path: string) {
|
||||
this.path = path;
|
||||
this.#db = new DatabaseSync(path);
|
||||
}
|
||||
|
||||
/** 打开一份研究的账本：新文件就建表，旧文件核对表结构版本号，对不上就报错。 */
|
||||
static open(path: string): Ledger {
|
||||
const ledger = new Ledger(path);
|
||||
if (schemaState(ledger.#db, path) === "new") {
|
||||
ledger.#atomic(() => {
|
||||
for (const sql of CREATE_SCHEMA) ledger.#db.exec(sql);
|
||||
});
|
||||
}
|
||||
return ledger;
|
||||
}
|
||||
|
||||
close(): void {
|
||||
this.#db.close();
|
||||
}
|
||||
|
||||
/**
|
||||
* 事务：里面的几次追加要么全部写进去，要么一条都不写（方法.md 4.3 ⑱）。
|
||||
* fn 必须是同步函数：await 之后的写入发生在事务提交之后，回滚管不到它。
|
||||
* 所以返回值类型排除了 Promise，`npm run check` 会挡住 async 函数。不支持嵌套，嵌套时 SQLite 直接报错。
|
||||
*/
|
||||
transaction<T>(fn: () => T extends PromiseLike<unknown> ? never : T): T {
|
||||
return this.#atomic(fn as () => T);
|
||||
}
|
||||
|
||||
#atomic<T>(fn: () => T): T {
|
||||
this.#db.exec("BEGIN IMMEDIATE");
|
||||
this.#inTransaction = true;
|
||||
try {
|
||||
const result = fn();
|
||||
this.#db.exec("COMMIT");
|
||||
return result;
|
||||
} catch (error) {
|
||||
this.#db.exec("ROLLBACK");
|
||||
throw error;
|
||||
} finally {
|
||||
this.#inTransaction = false;
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 追加一条记录，返回存下的那一条。先按种类检查内容，再做写入时的四种拒收检查（rejections.ts）。
|
||||
* 写提案时，作者自评从提案里拆出来，在同一个事务里另存成一条封存，返回的是提案那一条。
|
||||
*/
|
||||
append<K extends AppendKind>(kind: K, content: Input<K>): LedgerRecord<K> {
|
||||
const write = (): LedgerRecord<K> => {
|
||||
if (kind !== "submission") return this.#put(kind, content as Content<K>);
|
||||
const input = content as SubmissionInput;
|
||||
checkContent("submission", input, "input");
|
||||
if (input.kind !== "proposal") return this.#put(kind, content as Content<K>);
|
||||
const { authorEstimate, ...proposal } = input.content;
|
||||
const record = this.#put("submission", { ...input, content: proposal });
|
||||
this.#put("seal", { proposal: record.seq, range: authorEstimate });
|
||||
return record as LedgerRecord<K>;
|
||||
};
|
||||
return this.#inTransaction ? write() : this.#atomic(write);
|
||||
}
|
||||
|
||||
#put<K extends Kind>(kind: K, content: Content<K>): LedgerRecord<K> {
|
||||
checkContent(kind, content, "stored");
|
||||
checkWrite(kind, content, this.#writeState);
|
||||
const digest = hash(content);
|
||||
const result = this.#db
|
||||
.prepare("INSERT INTO records (kind, content, hash) VALUES (?, ?, ?)")
|
||||
.run(kind, canonical(content), digest);
|
||||
return { seq: Number(result.lastInsertRowid), kind, hash: digest, content } as LedgerRecord<K>;
|
||||
}
|
||||
|
||||
/** 写入时拒收检查要读的状态，和写入在同一个事务里读。 */
|
||||
readonly #writeState: WriteState = {
|
||||
stageClosed: (stageId) => this.stageClosed(stageId),
|
||||
stageOpen: (stageId) =>
|
||||
this.#first("stageOpen", "json_extract(content, '$.stageId') = ?", stageId) as LedgerRecord<"stageOpen"> | undefined,
|
||||
latest: (slot) => this.latestSubmission(slot),
|
||||
runStarted: (run) => this.#first("runStarted", "json_extract(content, '$.run') = ?", run) !== undefined,
|
||||
ablationStarted: (round) =>
|
||||
this.#one(
|
||||
`SELECT s.seq FROM records s JOIN records p
|
||||
ON p.kind = 'runPlanned' AND json_extract(p.content, '$.run') = json_extract(s.content, '$.run')
|
||||
WHERE s.kind = 'runStarted'
|
||||
AND json_extract(p.content, '$.purpose') = 'ablation'
|
||||
AND json_extract(p.content, '$.round') = ?`,
|
||||
round,
|
||||
) !== undefined,
|
||||
};
|
||||
|
||||
// -------------------------------------------------------------------------
|
||||
// 读取
|
||||
// -------------------------------------------------------------------------
|
||||
|
||||
/** 全部记录，按序号排；读出的每条都按种类重新检查内容。 */
|
||||
records(): LedgerRecord[] {
|
||||
return this.#all("SELECT * FROM records ORDER BY seq").map((row) => this.#record(row));
|
||||
}
|
||||
|
||||
/** 这个阶段关了没有。 */
|
||||
stageClosed(stageId: string): boolean {
|
||||
return this.#first("stageClose", "json_extract(content, '$.stageId') = ?", stageId) !== undefined;
|
||||
}
|
||||
|
||||
/** 某份提交的最新一版（本人在这个阶段对这个对象、这个编号交的这种提交里版本号最大的那一版），一版都没有就是 undefined。 */
|
||||
latestSubmission(slot: SubmissionSlot): LedgerRecord<"submission"> | undefined {
|
||||
const row = this.#one(
|
||||
`SELECT * FROM records
|
||||
WHERE kind = 'submission'
|
||||
AND json_extract(content, '$.stageId') = ?
|
||||
AND json_extract(content, '$.author') = ?
|
||||
AND json_extract(content, '$.kind') = ?
|
||||
AND json_extract(content, '$.target') IS ?
|
||||
AND json_extract(content, '$.item') IS ?
|
||||
ORDER BY json_extract(content, '$.version') DESC LIMIT 1`,
|
||||
slot.stageId,
|
||||
slot.author,
|
||||
slot.kind,
|
||||
slot.target,
|
||||
slot.item,
|
||||
);
|
||||
return row === undefined ? undefined : (this.#record(row) as LedgerRecord<"submission">);
|
||||
}
|
||||
|
||||
/** 未完成的动作：登记了但没有结果的运行，开始了但没有结束的模型调用。按运行编号、调用编号对上。 */
|
||||
pending(): Pending {
|
||||
const runs = this.#all(
|
||||
`SELECT * FROM records p WHERE p.kind = 'runPlanned'
|
||||
AND NOT EXISTS (SELECT 1 FROM records r WHERE r.kind = 'runResult'
|
||||
AND json_extract(r.content, '$.run') = json_extract(p.content, '$.run'))
|
||||
ORDER BY p.seq`,
|
||||
).map((row) => {
|
||||
const plan = this.#record(row) as LedgerRecord<"runPlanned">;
|
||||
return { plan, started: this.#writeState.runStarted(plan.content.run) };
|
||||
});
|
||||
const calls = this.#all(
|
||||
`SELECT * FROM records c WHERE c.kind = 'callStarted'
|
||||
AND NOT EXISTS (SELECT 1 FROM records e WHERE e.kind = 'callEnded'
|
||||
AND json_extract(e.content, '$.call') = json_extract(c.content, '$.call'))
|
||||
ORDER BY c.seq`,
|
||||
).map((row) => this.#record(row) as LedgerRecord<"callStarted">);
|
||||
return { runs, calls };
|
||||
}
|
||||
|
||||
/** 某种记录里满足条件的第一条（按序号），没有是 undefined。 */
|
||||
#first(kind: Kind, where: string, ...params: SQLInputValue[]): LedgerRecord | undefined {
|
||||
const row = this.#one(`SELECT * FROM records WHERE kind = ? AND ${where} ORDER BY seq LIMIT 1`, kind, ...params);
|
||||
return row === undefined ? undefined : this.#record(row);
|
||||
}
|
||||
|
||||
#one(sql: string, ...params: SQLInputValue[]): Row | undefined {
|
||||
return this.#db.prepare(sql).get(...params) as Row | undefined;
|
||||
}
|
||||
|
||||
#all(sql: string, ...params: SQLInputValue[]): Row[] {
|
||||
return this.#db.prepare(sql).all(...params) as unknown as Row[];
|
||||
}
|
||||
|
||||
/** 把一行还原成记录，并按种类检查内容；种类不认识或内容不符合定义都直接报错。 */
|
||||
#record(row: Row): LedgerRecord {
|
||||
if (!Object.hasOwn(KINDS, row.kind)) {
|
||||
throw new Error(`账本第 ${row.seq} 条记录的种类"${row.kind}"不认识`);
|
||||
}
|
||||
const kind = row.kind as Kind;
|
||||
const content: unknown = JSON.parse(row.content);
|
||||
checkContent(kind, content, "stored");
|
||||
return { seq: Number(row.seq), kind, hash: row.hash, content } as LedgerRecord;
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,74 @@
|
||||
import { SUBMISSION_KINDS, type Stage } from "../vocab/index.ts";
|
||||
import type { Content, Kind, LedgerRecord, SubmissionSlot } from "./kinds.ts";
|
||||
|
||||
// 写入时由 ledger 拒绝的四种情况（docs/结构.md 第四部分"写入时由 ledger 拒绝的几种情况"），在写入的同一个事务里查。
|
||||
// 只有这四种。记录之间别的先后（例如运行结果前要先有登记）和提交种类与阶段对不对得上都不查：
|
||||
// flow 调 ledger 是模块之间的调用，不是系统边界。"只发生一次"由 schema.ts 的唯一约束管。
|
||||
// 提交内容的字段归 submissions，该不该开下一个阶段归 flow。
|
||||
|
||||
/** 检查要读的几样当前状态，由 Ledger 在同一个事务里提供。 */
|
||||
export interface WriteState {
|
||||
stageClosed(stageId: string): boolean;
|
||||
/** 开这个阶段的那条记录，没有是 undefined。 */
|
||||
stageOpen(stageId: string): LedgerRecord<"stageOpen"> | undefined;
|
||||
/** 某份提交现在的最新一版，一版都没有是 undefined。 */
|
||||
latest(slot: SubmissionSlot): LedgerRecord<"submission"> | undefined;
|
||||
runStarted(run: string): boolean;
|
||||
/** 这一轮有没有消融格的运行已经确认启动。 */
|
||||
ablationStarted(round: number): boolean;
|
||||
}
|
||||
|
||||
/** 实现和核验阶段的提交不能替换（方法.md 第五部分总原则）：实现、修复、核验、消融实现、消融核验。 */
|
||||
const NOT_REPLACEABLE: ReadonlySet<Stage> = new Set([
|
||||
"implementation",
|
||||
"verification",
|
||||
"ablationImplementation",
|
||||
"ablationVerification",
|
||||
]);
|
||||
|
||||
export function checkWrite(kind: Kind, content: unknown, state: WriteState): void {
|
||||
if (kind === "submission") checkSubmission(content as Content<"submission">, state);
|
||||
if (kind === "runResult") checkRunResult(content as Content<"runResult">, state);
|
||||
if (kind === "stageClose") checkStageClose(content as Content<"stageClose">, state);
|
||||
}
|
||||
|
||||
/**
|
||||
* 第 1 条：只收开着的阶段的提交，即有开阶段记录、没有关阶段记录。
|
||||
* 第 2 条：新的一版必须正好接在本人在这个阶段、对同一个对象的最新一版后面（第一版是 1）；
|
||||
* 实现、修复、核验、消融实现、消融核验只能交第一版。
|
||||
*/
|
||||
function checkSubmission(s: Content<"submission">, state: WriteState): void {
|
||||
const where = `提交 ${s.kind}（阶段 ${s.stageId}，作者 ${s.author}，对象 ${s.target ?? "无"}，编号 ${s.item ?? "无"}，第 ${s.version} 版）`;
|
||||
if (state.stageOpen(s.stageId) === undefined || state.stageClosed(s.stageId)) {
|
||||
throw new Error(`拒收${where}：阶段 ${s.stageId} 没有开着，只收开着的阶段的提交`);
|
||||
}
|
||||
const stage = SUBMISSION_KINDS[s.kind].stage;
|
||||
if (s.version > 1 && NOT_REPLACEABLE.has(stage)) {
|
||||
throw new Error(`拒收${where}：${stage} 阶段的提交不能替换（方法.md 第五部分）`);
|
||||
}
|
||||
const latest = state.latest(s)?.content.version ?? 0;
|
||||
if (s.version !== latest + 1) {
|
||||
throw new Error(
|
||||
latest === 0
|
||||
? `拒收${where}：本人在这个阶段对这个对象还没有交过，第一份应是第 1 版`
|
||||
: `拒收${where}：要替换的第 ${s.version - 1} 版不是本人在这个阶段对这个对象的最新一版（最新是第 ${latest} 版）`,
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
/** 第 3 条：运行结果要写在"确认启动"之后。例外是"没启动"：它本来就没有启动记录（方法.md 第七部分）。 */
|
||||
function checkRunResult(r: Content<"runResult">, state: WriteState): void {
|
||||
if (r.process.ending !== "notStarted" && !state.runStarted(r.run)) {
|
||||
throw new Error(`拒收运行 ${r.run} 的结果（${r.process.ending}）：这次运行还没有"确认启动"记录`);
|
||||
}
|
||||
}
|
||||
|
||||
/** 第 4 条：这一轮已经有消融格确认启动了，就不能再关解释阶段（方法.md 4.3 ⑭：解释冻结必须早于任何一格启动）。 */
|
||||
function checkStageClose(c: Content<"stageClose">, state: WriteState): void {
|
||||
const open = state.stageOpen(c.stageId);
|
||||
if (open?.content.stage === "explanation" && state.ablationStarted(open.content.round)) {
|
||||
throw new Error(
|
||||
`不能关解释阶段 ${c.stageId}：第 ${open.content.round} 轮已经有消融格确认启动，解释冻结必须早于任何一格启动（方法.md 4.3 ⑭）`,
|
||||
);
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,56 @@
|
||||
import type { DatabaseSync } from "node:sqlite";
|
||||
|
||||
// 账本的表结构（docs/结构.md 第四部分"全部只追加"）：只有一张表，每一行是序号、种类、内容、内容的哈希。
|
||||
// 表结构一改就把版本号加 1；旧文件版本号对不上时打开报错，不迁移，也不逐字核对建表语句。
|
||||
|
||||
export const SCHEMA_VERSION = 1;
|
||||
|
||||
/** 新文件要执行的建表语句，由 Ledger 在一个事务里执行，最后写入版本号。 */
|
||||
export const CREATE_SCHEMA = [
|
||||
// 序号用 AUTOINCREMENT：只增不复用，就是写入的先后。STRICT 让列的类型真的生效。
|
||||
`CREATE TABLE records (
|
||||
seq INTEGER PRIMARY KEY AUTOINCREMENT,
|
||||
kind TEXT NOT NULL,
|
||||
content TEXT NOT NULL,
|
||||
hash TEXT NOT NULL
|
||||
) STRICT`,
|
||||
// 整张表禁止修改、禁止删除。要更正就追加一条新记录。
|
||||
`CREATE TRIGGER records_no_update BEFORE UPDATE ON records
|
||||
BEGIN SELECT RAISE(ABORT, '账本只能追加：禁止修改记录'); END`,
|
||||
`CREATE TRIGGER records_no_delete BEFORE DELETE ON records
|
||||
BEGIN SELECT RAISE(ABORT, '账本只能追加：禁止删除记录'); END`,
|
||||
// 唯一约束：有自己编号、按定义只发生一次的记录，同一个编号只能有一条，重复就报错。其余记录照写，不设唯一键。
|
||||
// 记录之间的先后（例如结束前要先有开始）不在这里查。
|
||||
`CREATE UNIQUE INDEX stage_open_once ON records (json_extract(content, '$.stageId')) WHERE kind = 'stageOpen'`,
|
||||
`CREATE UNIQUE INDEX stage_close_once ON records (json_extract(content, '$.stageId')) WHERE kind = 'stageClose'`,
|
||||
`CREATE UNIQUE INDEX run_planned_once ON records (json_extract(content, '$.run')) WHERE kind = 'runPlanned'`,
|
||||
// "确认启动"只能有一条，而且"确认没启动"（结果是没启动）也算在这一条里：
|
||||
// 两样都有，就是同一次运行既扣次又退还（方法.md 第七部分"什么时候扣次数"）。
|
||||
`CREATE UNIQUE INDEX run_started_or_not_started_once ON records (json_extract(content, '$.run'))
|
||||
WHERE kind = 'runStarted' OR (kind = 'runResult' AND json_extract(content, '$.process.ending') = 'notStarted')`,
|
||||
`CREATE UNIQUE INDEX run_result_once ON records (json_extract(content, '$.run')) WHERE kind = 'runResult'`,
|
||||
`CREATE UNIQUE INDEX call_started_once ON records (json_extract(content, '$.call')) WHERE kind = 'callStarted'`,
|
||||
`CREATE UNIQUE INDEX call_ended_once ON records (json_extract(content, '$.call')) WHERE kind = 'callEnded'`,
|
||||
`CREATE UNIQUE INDEX settlement_once ON records (json_extract(content, '$.ablation')) WHERE kind = 'settlement'`,
|
||||
`PRAGMA user_version=${SCHEMA_VERSION}`,
|
||||
];
|
||||
|
||||
/**
|
||||
* 看这个文件是新文件还是当前版本的账本，别的情况报错；是账本才设连接参数，不改别人的库。
|
||||
* synchronous=FULL：提交即落盘。账本存的是"预测冻结在结果之前"的证据，掉电丢掉最后几条等于丢证据。
|
||||
* 版本号存在 SQLite 自带的 user_version 里；新文件的 user_version 是 0，而且没有任何表。
|
||||
*/
|
||||
export function schemaState(db: DatabaseSync, path: string): "new" | "current" {
|
||||
const version = Number((db.prepare("PRAGMA user_version").get() as { user_version: number }).user_version);
|
||||
const objects = Number((db.prepare("SELECT count(*) AS n FROM sqlite_master").get() as { n: number }).n);
|
||||
const state = version === SCHEMA_VERSION ? "current" : version === 0 && objects === 0 ? "new" : undefined;
|
||||
if (state !== undefined) {
|
||||
db.exec("PRAGMA journal_mode=WAL");
|
||||
db.exec("PRAGMA synchronous=FULL");
|
||||
return state;
|
||||
}
|
||||
throw new Error(
|
||||
`账本 ${path} 的表结构版本是 ${version}（库里已有 ${objects} 个表、索引或触发器），本程序只认版本 ${SCHEMA_VERSION}。` +
|
||||
"旧研究不迁移，请另开一份新研究",
|
||||
);
|
||||
}
|
||||
@@ -0,0 +1,103 @@
|
||||
import { mkdtempSync, rmSync } from "node:fs";
|
||||
import { tmpdir } from "node:os";
|
||||
import { join } from "node:path";
|
||||
import type { TestContext } from "node:test";
|
||||
import { Ledger } from "../src/ledger/index.ts";
|
||||
import type { ModelUsage, ProcessResult, Range, RunDefinition, Stage, Submission } from "../src/vocab/index.ts";
|
||||
|
||||
// ledger 测试共用：每个测试在自己的临时目录里建一个真实的数据库文件，测试结束时连目录一起删掉。
|
||||
|
||||
export function tempDir(t: TestContext): string {
|
||||
const dir = mkdtempSync(join(tmpdir(), "inforsi-ledger-"));
|
||||
t.after(() => rmSync(dir, { recursive: true, force: true }));
|
||||
return dir;
|
||||
}
|
||||
|
||||
export function freshLedger(t: TestContext): { ledger: Ledger; path: string } {
|
||||
const path = join(tempDir(t), "ledger.db");
|
||||
const ledger = Ledger.open(path);
|
||||
t.after(() => ledger.close());
|
||||
return { ledger, path };
|
||||
}
|
||||
|
||||
export const range = (q1: number, q5: number, q9: number): Range => ({ q1, q5, q9 });
|
||||
|
||||
export function proposal(estimate: Range, body = "提案正文"): Submission<"proposal"> {
|
||||
return {
|
||||
factors: [{ name: "换优化器", method: "把 SGD 换成 Adam", keySettings: "学习率 0.01", continues: null }],
|
||||
safeguards: [],
|
||||
metrics: { defined: [], reused: [] },
|
||||
authorEstimate: estimate,
|
||||
body,
|
||||
};
|
||||
}
|
||||
|
||||
export const prediction = (r: Range): Submission<"firstPrediction"> => ({ range: r, body: "理由" });
|
||||
export const critique = (body = "挑刺内容"): Submission<"critique"> => ({ target: "whole", body });
|
||||
export const wrapUp = (body = "收尾"): Submission<"wrapUp"> => ({ body });
|
||||
|
||||
export const implementation = (body = "无"): Submission<"implementation"> => ({
|
||||
factorLocations: [{ factor: "p1", locations: ["solve.py:10-20"] }],
|
||||
safeguardLocations: [],
|
||||
unassigned: [],
|
||||
concerns: "无",
|
||||
body,
|
||||
});
|
||||
|
||||
export const definition: RunDefinition = {
|
||||
command: ["python", "solve.py"],
|
||||
workdir: "/work",
|
||||
image: { digest: `sha256:${"0".repeat(64)}`, runtime: "runc" },
|
||||
wallSeconds: 120,
|
||||
cores: 1,
|
||||
inheritEnv: false,
|
||||
env: {},
|
||||
};
|
||||
|
||||
export const exited: ProcessResult = {
|
||||
ending: "exited",
|
||||
exitCode: 0,
|
||||
cpuSeconds: { status: "measured", value: 100 },
|
||||
wallSeconds: { status: "measured", value: 101 },
|
||||
peakMemoryBytes: { status: "measured", value: 1e8 },
|
||||
};
|
||||
|
||||
export const notStarted: ProcessResult = { ending: "notStarted", reason: "启动前取消" };
|
||||
|
||||
/** 开一个阶段，返回开阶段那条记录的序号。 */
|
||||
export function openStage(
|
||||
ledger: Ledger,
|
||||
stageId: string,
|
||||
stage: Stage,
|
||||
round: number,
|
||||
roster: string[] = ["r1", "r2"],
|
||||
): number {
|
||||
return ledger.append("stageOpen", { stageId, stage, round, roster }).seq;
|
||||
}
|
||||
|
||||
export function closeStage(ledger: Ledger, stageId: string): number {
|
||||
return ledger.append("stageClose", { stageId }).seq;
|
||||
}
|
||||
|
||||
/** 登记一次运行。 */
|
||||
export function planRun(ledger: Ledger, run: string, purpose: "baseline" | "forward" | "ablation", round: number): number {
|
||||
return ledger.append("runPlanned", { run, purpose, round, object: purpose === "baseline" ? null : `obj-${run}`, definition }).seq;
|
||||
}
|
||||
|
||||
/** 一次模型调用的开始记录，materials 是这次任务的全部材料。 */
|
||||
export const callStart = (call: string, actor = "r1", stageId: string | null = null) => ({
|
||||
call,
|
||||
actor,
|
||||
stageId,
|
||||
model: "glm-5.3",
|
||||
materials: { firstMessage: "任务说明……", tools: ["read", "submit"] },
|
||||
});
|
||||
|
||||
const unknownCount = { status: "unmeasured" as const, reason: "服务没报" };
|
||||
export const usage: ModelUsage = {
|
||||
inputTokens: unknownCount,
|
||||
cacheReadTokens: unknownCount,
|
||||
cacheWriteTokens: unknownCount,
|
||||
outputTokens: unknownCount,
|
||||
cost: unknownCount,
|
||||
};
|
||||
@@ -0,0 +1,57 @@
|
||||
import assert from "node:assert/strict";
|
||||
import { join } from "node:path";
|
||||
import { test } from "node:test";
|
||||
import { Ledger } from "../src/ledger/index.ts";
|
||||
import { callStart, exited, freshLedger, notStarted, planRun, tempDir, usage } from "./ledger-helpers.ts";
|
||||
|
||||
// 未完成的动作（方法.md 第九部分"中断与恢复时的记账含义"：先登记、再执行、再登记结果）：
|
||||
// 登记了但没有结果的运行、开始了但没有结束的模型调用能被找出来；写了结果之后就不再出现。
|
||||
|
||||
test("运行：登记了没结果的被找出来，带着有没有确认启动；写了结果就不再出现", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
planRun(ledger, "base", "baseline", 0);
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
planRun(ledger, "run2", "forward", 1);
|
||||
ledger.append("runStarted", { run: "base" });
|
||||
ledger.append("runStarted", { run: "run1" });
|
||||
ledger.append("runResult", { run: "base", process: exited, judged: null });
|
||||
assert.deepEqual(
|
||||
ledger.pending().runs.map((r) => [r.plan.content.run, r.started]),
|
||||
[
|
||||
["run1", true],
|
||||
["run2", false],
|
||||
],
|
||||
);
|
||||
ledger.append("runResult", { run: "run2", process: notStarted, judged: null });
|
||||
assert.deepEqual(ledger.pending().runs.map((r) => r.plan.content.run), ["run1"]);
|
||||
ledger.append("runResult", { run: "run1", process: exited, judged: null });
|
||||
assert.deepEqual(ledger.pending().runs, []);
|
||||
});
|
||||
|
||||
test("模型调用：开始了没结束的被找出来，带着这次任务的全部材料；有用量也还算没结束；结束后不再出现", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
ledger.append("callStarted", callStart("k1", "r1", "s1"));
|
||||
ledger.append("callStarted", callStart("k2", "e1"));
|
||||
ledger.append("callUsage", { call: "k1", reply: 1, usage });
|
||||
ledger.append("callEnded", { call: "k2", ending: "failed", reason: "接口超时" });
|
||||
const calls = ledger.pending().calls;
|
||||
assert.deepEqual(calls.map((c) => c.content.call), ["k1"]);
|
||||
assert.deepEqual(calls[0]!.content.materials, callStart("k1").materials);
|
||||
ledger.append("callEnded", { call: "k1", ending: "completed" });
|
||||
assert.deepEqual(ledger.pending(), { runs: [], calls: [] });
|
||||
});
|
||||
|
||||
test("关掉再打开：未完成的动作照样找得出来", (t) => {
|
||||
const path = join(tempDir(t), "ledger.db");
|
||||
const ledger = Ledger.open(path);
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
ledger.append("runStarted", { run: "run1" });
|
||||
ledger.append("callStarted", callStart("k1"));
|
||||
const before = ledger.pending();
|
||||
ledger.close();
|
||||
const reopened = Ledger.open(path);
|
||||
t.after(() => reopened.close());
|
||||
assert.deepEqual(reopened.pending(), before);
|
||||
assert.equal(before.runs.length, 1);
|
||||
assert.equal(before.calls.length, 1);
|
||||
});
|
||||
@@ -0,0 +1,158 @@
|
||||
import assert from "node:assert/strict";
|
||||
import { test } from "node:test";
|
||||
import type { Ledger } from "../src/ledger/index.ts";
|
||||
import {
|
||||
closeStage,
|
||||
critique,
|
||||
exited,
|
||||
freshLedger,
|
||||
implementation,
|
||||
notStarted,
|
||||
openStage,
|
||||
planRun,
|
||||
prediction,
|
||||
proposal,
|
||||
range,
|
||||
wrapUp,
|
||||
} from "./ledger-helpers.ts";
|
||||
|
||||
// 写入时由 ledger 拒绝的四种情况（docs/结构.md 第四部分"写入时由 ledger 拒绝的几种情况"），每种一条拒收、一条对应的正常情况。
|
||||
// 预期值来自文字：
|
||||
// 1. 只收开着的阶段（有开阶段记录、没有关阶段记录）的提交，别的拒收；
|
||||
// 2. 要替换的那一份不是本人在这个阶段、对同一个对象的最新一版 → 拒收；实现、修复、核验、消融实现、消融核验不能替换（方法.md 第五部分）；
|
||||
// 3. 运行结果写在"确认启动"之前 → 拒收；"没启动"的结果不需要启动记录（方法.md 第七部分）；
|
||||
// 4. 这一轮消融已经有格确认启动了，才来关这一轮的解释阶段 → 拒收（方法.md 4.3 ⑭）。
|
||||
|
||||
function submitWrapUp(ledger: Ledger, stageId: string, author: string, version: number, body = "收尾") {
|
||||
return ledger.append("submission", { stageId, author, target: null, item: null, version, kind: "wrapUp", content: wrapUp(body) });
|
||||
}
|
||||
|
||||
test("拒收 1：只收开着的阶段的提交：没开过的不收，关了以后不收、替换也不收；开着时收", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
assert.throws(() => submitWrapUp(ledger, "s1", "r1", 1), /没有开着/);
|
||||
openStage(ledger, "s1", "wrapUp", 1);
|
||||
submitWrapUp(ledger, "s1", "r1", 1);
|
||||
closeStage(ledger, "s1");
|
||||
assert.throws(() => submitWrapUp(ledger, "s1", "r2", 1), /没有开着/);
|
||||
assert.throws(() => submitWrapUp(ledger, "s1", "r1", 2, "关闭后想替换"), /没有开着/);
|
||||
assert.equal(ledger.records().filter((r) => r.kind === "submission").length, 1);
|
||||
});
|
||||
|
||||
test("拒收 2 的正常情况：替换本人在这个阶段、对同一个对象的最新一版，版本号加 1", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
openStage(ledger, "s1", "firstPrediction", 1);
|
||||
const slot = { stageId: "s1", author: "r1", target: "c1", item: null, kind: "firstPrediction" as const };
|
||||
ledger.append("submission", { ...slot, version: 1, content: prediction(range(1, 2, 3)) });
|
||||
ledger.append("submission", { ...slot, version: 2, content: prediction(range(1, 2, 4)) });
|
||||
ledger.append("submission", { ...slot, version: 3, content: prediction(range(1, 2, 5)) });
|
||||
assert.equal(ledger.latestSubmission(slot)?.content.version, 3);
|
||||
});
|
||||
|
||||
test("拒收 2：要替换的那一份不是最新一版、不是本人的、不是同一个对象的、不在这个阶段，都拒收", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
openStage(ledger, "s1", "firstPrediction", 1);
|
||||
openStage(ledger, "s2", "firstPrediction", 1);
|
||||
const slot = { stageId: "s1", author: "r1", target: "c1", item: null, kind: "firstPrediction" as const };
|
||||
ledger.append("submission", { ...slot, version: 1, content: prediction(range(1, 2, 3)) });
|
||||
ledger.append("submission", { ...slot, version: 2, content: prediction(range(1, 2, 4)) });
|
||||
const next = { ...slot, version: 3, content: prediction(range(1, 2, 6)) };
|
||||
// 想替换第 1 版，但最新是第 2 版。
|
||||
assert.throws(() => ledger.append("submission", { ...slot, version: 2, content: prediction(range(0, 2, 4)) }), /最新是第 2 版/);
|
||||
// 要替换的第 3 版不存在。
|
||||
assert.throws(() => ledger.append("submission", { ...slot, version: 4, content: prediction(range(1, 2, 6)) }), /最新是第 2 版/);
|
||||
// 不是本人的、不是同一个对象的、不在这个阶段的：本人在那里还没有交过。
|
||||
assert.throws(() => ledger.append("submission", { ...next, author: "r2" }), /还没有交过/);
|
||||
assert.throws(() => ledger.append("submission", { ...next, target: "c2" }), /还没有交过/);
|
||||
assert.throws(() => ledger.append("submission", { ...next, stageId: "s2" }), /还没有交过/);
|
||||
assert.equal(ledger.latestSubmission(slot)?.content.version, 2);
|
||||
});
|
||||
|
||||
test("拒收 2：同一个人对同一个候选的几份挑刺按各自的编号分开算版本", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
openStage(ledger, "s1", "critique", 1);
|
||||
const slot = { stageId: "s1", author: "r1", target: "c1", kind: "critique" as const };
|
||||
ledger.append("submission", { ...slot, item: "k1", version: 1, content: critique("第一份") });
|
||||
ledger.append("submission", { ...slot, item: "k2", version: 1, content: critique("第二份") });
|
||||
ledger.append("submission", { ...slot, item: "k1", version: 2, content: critique("第一份改") });
|
||||
assert.throws(() => ledger.append("submission", { ...slot, item: "k3", version: 2, content: critique("x") }), /还没有交过/);
|
||||
});
|
||||
|
||||
test("拒收 2：实现、修复、核验、消融实现、消融核验只收第一版，不能替换", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
openStage(ledger, "impl", "implementation", 1, ["e1"]);
|
||||
openStage(ledger, "fix", "implementation", 1, ["e1"]);
|
||||
openStage(ledger, "ver", "verification", 1, ["v1"]);
|
||||
openStage(ledger, "aimpl", "ablationImplementation", 1, ["e1"]);
|
||||
openStage(ledger, "aver", "ablationVerification", 1, ["v1"]);
|
||||
const base = { author: "e1", target: "c1", item: null };
|
||||
const verification = { clauses: [], factors: [], safeguards: [], unassigned: [], body: "无" };
|
||||
const ablImpl = { alsoOff: [], cannotRun: null, concerns: "无" };
|
||||
const cases = [
|
||||
{ ...base, stageId: "impl", kind: "implementation" as const, content: implementation() },
|
||||
{ ...base, stageId: "fix", kind: "repair" as const, content: implementation("修复") },
|
||||
{ ...base, author: "v1", stageId: "ver", kind: "verification" as const, content: verification },
|
||||
{ ...base, stageId: "aimpl", target: "cell1", kind: "ablationImplementation" as const, content: ablImpl },
|
||||
{ ...base, author: "v1", stageId: "aver", target: "cell1", kind: "ablationVerification" as const, content: { factors: [] } },
|
||||
];
|
||||
for (const c of cases) {
|
||||
ledger.append("submission", { ...c, version: 1 });
|
||||
assert.throws(() => ledger.append("submission", { ...c, version: 2 }), /不能替换/, c.kind);
|
||||
}
|
||||
});
|
||||
|
||||
test("拒收 2：版本号只在同一种提交里接：不能替换的实现不能被别的种类顶替，提案和挑刺用同一个编号也接不上", (t) => {
|
||||
// 同一位置上，实现 v1 之后交收尾 v2 拒收，最新一版仍是实现；提案和挑刺用同一个编号，版本号各算各的。
|
||||
const { ledger } = freshLedger(t);
|
||||
openStage(ledger, "impl", "implementation", 1, ["e1"]);
|
||||
const implSlot = { stageId: "impl", author: "e1", target: "c1", item: null };
|
||||
ledger.append("submission", { ...implSlot, version: 1, kind: "implementation", content: implementation() });
|
||||
assert.throws(() => ledger.append("submission", { ...implSlot, version: 2, kind: "wrapUp", content: wrapUp() }), /还没有交过/);
|
||||
assert.equal(ledger.latestSubmission({ ...implSlot, kind: "implementation" })?.content.kind, "implementation");
|
||||
|
||||
openStage(ledger, "mix", "proposal", 1);
|
||||
const mixSlot = { stageId: "mix", author: "r1", target: null, item: "x1" };
|
||||
ledger.append("submission", { ...mixSlot, version: 1, kind: "proposal", content: proposal(range(1, 2, 3)) });
|
||||
assert.throws(() => ledger.append("submission", { ...mixSlot, version: 2, kind: "critique", content: critique() }), /还没有交过/);
|
||||
assert.equal(ledger.latestSubmission({ ...mixSlot, kind: "critique" }), undefined);
|
||||
});
|
||||
|
||||
test("拒收 3：没有确认启动就写运行结果，拒收；确认启动后可以写", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
assert.throws(() => ledger.append("runResult", { run: "run1", process: exited, judged: null }), /还没有"确认启动"/);
|
||||
ledger.append("runStarted", { run: "run1" });
|
||||
ledger.append("runResult", { run: "run1", process: exited, judged: null });
|
||||
});
|
||||
|
||||
test("拒收 3 的例外：确认没启动的结果不需要启动记录", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
ledger.append("runResult", { run: "run1", process: notStarted, judged: null });
|
||||
assert.deepEqual(ledger.pending().runs, []);
|
||||
});
|
||||
|
||||
test("拒收 4：这一轮已经有消融格确认启动，就不能关这一轮的解释阶段", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
openStage(ledger, "expl", "explanation", 2);
|
||||
planRun(ledger, "cellrun", "ablation", 2);
|
||||
ledger.append("runStarted", { run: "cellrun" });
|
||||
assert.throws(() => closeStage(ledger, "expl"), /解释冻结必须早于任何一格启动/);
|
||||
assert.equal(ledger.stageClosed("expl"), false);
|
||||
});
|
||||
|
||||
test("拒收 4 的正常情况：只登记没启动、别的轮的消融启动、本轮正向启动，都不挡关解释阶段；关别的阶段不受限", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
openStage(ledger, "expl", "explanation", 2);
|
||||
planRun(ledger, "cell-planned", "ablation", 2); // 只登记，还没确认启动
|
||||
planRun(ledger, "cell-round1", "ablation", 1);
|
||||
ledger.append("runStarted", { run: "cell-round1" });
|
||||
planRun(ledger, "fwd-round2", "forward", 2);
|
||||
ledger.append("runStarted", { run: "fwd-round2" });
|
||||
closeStage(ledger, "expl");
|
||||
assert.equal(ledger.stageClosed("expl"), true);
|
||||
openStage(ledger, "wrap", "wrapUp", 2);
|
||||
planRun(ledger, "cell2", "ablation", 2);
|
||||
ledger.append("runStarted", { run: "cell2" });
|
||||
closeStage(ledger, "wrap");
|
||||
assert.equal(ledger.stageClosed("wrap"), true);
|
||||
});
|
||||
@@ -0,0 +1,299 @@
|
||||
import assert from "node:assert/strict";
|
||||
import { join } from "node:path";
|
||||
import { DatabaseSync } from "node:sqlite";
|
||||
import { test } from "node:test";
|
||||
import { Ledger } from "../src/ledger/index.ts";
|
||||
import { hash } from "../src/vocab/index.ts";
|
||||
import {
|
||||
callStart,
|
||||
closeStage,
|
||||
exited,
|
||||
freshLedger,
|
||||
implementation,
|
||||
notStarted,
|
||||
openStage,
|
||||
planRun,
|
||||
proposal,
|
||||
range,
|
||||
tempDir,
|
||||
} from "./ledger-helpers.ts";
|
||||
|
||||
// 预期值照 docs/结构.md 第四部分"账本记录"、方法.md 第七部分"什么时候扣次数"和第九部分的文字写：
|
||||
// 整张表禁改禁删；按定义只发生一次的记录重复就报错，其余照写；事务要么全写要么全不写；
|
||||
// 关掉再打开读出不变；表结构版本不对打开报错。
|
||||
|
||||
function countRows(path: string): number {
|
||||
const db = new DatabaseSync(path);
|
||||
const n = Number((db.prepare("SELECT count(*) AS n FROM records").get() as { n: number }).n);
|
||||
db.close();
|
||||
return n;
|
||||
}
|
||||
|
||||
test("触发器：直接用 SQL 执行 UPDATE、DELETE 都被拒绝，记录不变", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
ledger.append("config", { n: 2 });
|
||||
const before = ledger.records();
|
||||
const db = new DatabaseSync(path);
|
||||
t.after(() => db.close());
|
||||
assert.throws(() => db.exec("UPDATE records SET content = '{\"n\":3}'"), /禁止修改/);
|
||||
assert.throws(() => db.exec("UPDATE records SET hash = 'x' WHERE seq = 1"), /禁止修改/);
|
||||
assert.throws(() => db.exec("DELETE FROM records"), /禁止删除/);
|
||||
assert.throws(() => db.exec("DELETE FROM records WHERE seq = 1"), /禁止删除/);
|
||||
assert.deepEqual(ledger.records(), before);
|
||||
});
|
||||
|
||||
// 唯一约束（schema.ts）：有自己编号、按定义只发生一次的记录，同一个编号写第二条就报错，内容完全相同也报错，
|
||||
// 不多写一条；别的编号照常写。
|
||||
|
||||
test("唯一约束：同一个阶段开两次报错", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
openStage(ledger, "s1", "proposal", 1);
|
||||
assert.throws(() => openStage(ledger, "s1", "proposal", 1), /stage_open_once/);
|
||||
openStage(ledger, "s2", "proposal", 1);
|
||||
assert.equal(countRows(path), 2);
|
||||
});
|
||||
|
||||
test("唯一约束：同一个阶段关两次报错", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
openStage(ledger, "s1", "proposal", 1);
|
||||
closeStage(ledger, "s1");
|
||||
assert.throws(() => closeStage(ledger, "s1"), /stage_close_once/);
|
||||
assert.equal(countRows(path), 2);
|
||||
});
|
||||
|
||||
test("唯一约束：同一次运行登记两次报错", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
assert.throws(() => planRun(ledger, "run1", "forward", 1), /run_planned_once/);
|
||||
planRun(ledger, "run2", "forward", 1);
|
||||
assert.equal(countRows(path), 2);
|
||||
});
|
||||
|
||||
test("唯一约束：同一次运行确认启动两次报错", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
ledger.append("runStarted", { run: "run1" });
|
||||
assert.throws(() => ledger.append("runStarted", { run: "run1" }), /run_started_or_not_started_once/);
|
||||
assert.equal(countRows(path), 2);
|
||||
});
|
||||
|
||||
test("唯一约束：确认启动和\"结果是没启动\"算同一条，一次运行只能有一样，两种先后都报错", (t) => {
|
||||
// 登记→确认启动→结果没启动、登记→结果没启动→确认启动，第三步都报错：一次运行不能既扣次又退还（方法.md 第七部分）。
|
||||
const { ledger, path } = freshLedger(t);
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
ledger.append("runStarted", { run: "run1" });
|
||||
assert.throws(() => ledger.append("runResult", { run: "run1", process: notStarted, judged: null }), /run_started_or_not_started_once/);
|
||||
planRun(ledger, "run2", "forward", 1);
|
||||
ledger.append("runResult", { run: "run2", process: notStarted, judged: null });
|
||||
assert.throws(() => ledger.append("runStarted", { run: "run2" }), /run_started_or_not_started_once/);
|
||||
// 正常情况：确认启动后写启动过的结果。
|
||||
ledger.append("runResult", { run: "run1", process: exited, judged: null });
|
||||
assert.equal(countRows(path), 5); // 两条登记、一条确认启动、两条结果
|
||||
});
|
||||
|
||||
test("唯一约束：同一次运行写两个结果报错", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
ledger.append("runStarted", { run: "run1" });
|
||||
ledger.append("runResult", { run: "run1", process: exited, judged: null });
|
||||
assert.throws(() => ledger.append("runResult", { run: "run1", process: exited, judged: null }), /run_result_once/);
|
||||
assert.equal(countRows(path), 3);
|
||||
});
|
||||
|
||||
test("唯一约束：同一次模型调用开始两次报错", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
ledger.append("callStarted", callStart("k1"));
|
||||
assert.throws(() => ledger.append("callStarted", callStart("k1")), /call_started_once/);
|
||||
ledger.append("callStarted", callStart("k2"));
|
||||
assert.equal(countRows(path), 2);
|
||||
});
|
||||
|
||||
test("唯一约束：同一次模型调用结束两次报错", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
ledger.append("callStarted", callStart("k1"));
|
||||
ledger.append("callEnded", { call: "k1", ending: "completed" });
|
||||
assert.throws(() => ledger.append("callEnded", { call: "k1", ending: "failed", reason: "又报了一次" }), /call_ended_once/);
|
||||
assert.equal(countRows(path), 2);
|
||||
});
|
||||
|
||||
test("唯一约束：同一次消融结算两次报错", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
ledger.append("settlement", { ablation: "a1", result: { cells: { p1: "valid" } } });
|
||||
assert.throws(() => ledger.append("settlement", { ablation: "a1", result: { cells: { p1: "valid" } } }), /settlement_once/);
|
||||
ledger.append("settlement", { ablation: "a2", result: { cells: { p1: "insufficient" } } });
|
||||
assert.equal(countRows(path), 2);
|
||||
});
|
||||
|
||||
test("没有唯一约束的记录照写：同样的事件、决定写两次就是两条", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
const e = { event: "untested", reason: "剩余次数不够", objects: ["cell3"] };
|
||||
const first = ledger.append("event", e);
|
||||
const second = ledger.append("event", e);
|
||||
assert.equal(second.seq, first.seq + 1);
|
||||
const d = { decision: "batch", round: 1, rule: "bestExpected", ruleVersion: 1, params: {}, seed: null, input: { x: 1 }, output: ["c1"] };
|
||||
ledger.append("decision", d);
|
||||
ledger.append("decision", d);
|
||||
assert.equal(countRows(path), 4);
|
||||
});
|
||||
|
||||
test("内容不符合定义时不写入：多出的字段、提交内容不合格、提交编号 item 和提交种类对不上", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
openStage(ledger, "s1", "wrapUp", 1);
|
||||
assert.throws(() => ledger.append("runStarted", { run: "r1", extra: 1 } as never), /runStarted/);
|
||||
assert.throws(
|
||||
() => ledger.append("submission", { stageId: "s1", author: "r1", target: null, item: null, version: 1, kind: "wrapUp", content: { body: "" } }),
|
||||
/wrapUp/,
|
||||
);
|
||||
// 提案、挑刺一人能交好几份，必须带编号；实现一人对一个候选只交一份，编号必须为空，
|
||||
// 否则另起一个编号就能绕开"实现不能替换"。
|
||||
openStage(ledger, "prop", "proposal", 1);
|
||||
assert.throws(
|
||||
() => ledger.append("submission", { stageId: "prop", author: "r1", target: null, item: null as never, version: 1, kind: "proposal", content: proposal(range(1, 2, 3)) }),
|
||||
/proposal/,
|
||||
);
|
||||
openStage(ledger, "impl", "implementation", 1, ["e1"]);
|
||||
assert.throws(
|
||||
() => ledger.append("submission", { stageId: "impl", author: "e1", target: "c1", item: "x" as never, version: 1, kind: "implementation", content: implementation() }),
|
||||
/implementation/,
|
||||
);
|
||||
assert.equal(countRows(path), 3); // 三条开阶段
|
||||
});
|
||||
|
||||
test("写提案：作者自评拆成一条封存，提案记录里不带它", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
openStage(ledger, "s1", "proposal", 1);
|
||||
const record = ledger.append("submission", {
|
||||
stageId: "s1", author: "r1", target: null, item: "r1-1", version: 1, kind: "proposal", content: proposal(range(1, 2, 3)),
|
||||
});
|
||||
const all = ledger.records();
|
||||
const stored = all.find((r) => r.kind === "submission")!;
|
||||
assert.equal(JSON.stringify(stored.content).includes("authorEstimate"), false);
|
||||
assert.equal(JSON.stringify(record.content).includes("authorEstimate"), false);
|
||||
const { authorEstimate: _, ...withoutEstimate } = proposal(range(1, 2, 3));
|
||||
assert.deepEqual(stored.content.content, withoutEstimate);
|
||||
const seal = all.find((r) => r.kind === "seal")!;
|
||||
assert.deepEqual(seal.content, { proposal: record.seq, range: range(1, 2, 3) });
|
||||
assert.equal(seal.seq, record.seq + 1);
|
||||
});
|
||||
|
||||
test("退回：被退回的内容原样存，带退回理由；来自 Jev 的带原始概率和结论", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
const malformed = { factors: "不是数组" };
|
||||
ledger.append("rejection", { stageId: "s1", author: "r1", kind: "proposal", submitted: malformed, reason: "因子清单格式不对", jev: null });
|
||||
ledger.append("rejection", {
|
||||
stageId: "s6", author: "r2", kind: "response", submitted: { body: "改了评测" }, reason: "冻结条款改动了评测",
|
||||
jev: { probabilities: { yes: 0.97, no: 0.03 }, conclusion: "changesEvaluation" },
|
||||
});
|
||||
const rejections = ledger.records().filter((r) => r.kind === "rejection");
|
||||
assert.deepEqual(rejections.map((r) => r.content.submitted), [malformed, { body: "改了评测" }]);
|
||||
assert.deepEqual(rejections[1]!.content.jev, { probabilities: { yes: 0.97, no: 0.03 }, conclusion: "changesEvaluation" });
|
||||
});
|
||||
|
||||
test("事务：中途出错，一条都没写进去；中途被拒收，前面写的也撤销；正常结束则全部写进去", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
assert.throws(
|
||||
() =>
|
||||
ledger.transaction(() => {
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
ledger.append("runStarted", { run: "run1" });
|
||||
throw new Error("算到一半出错");
|
||||
}),
|
||||
/算到一半出错/,
|
||||
);
|
||||
assert.equal(countRows(path), 0);
|
||||
assert.throws(
|
||||
() =>
|
||||
ledger.transaction(() => {
|
||||
planRun(ledger, "run9", "forward", 1);
|
||||
ledger.append("runResult", { run: "run9", process: exited, judged: null });
|
||||
}),
|
||||
/确认启动/,
|
||||
);
|
||||
assert.equal(countRows(path), 0);
|
||||
ledger.transaction(() => {
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
ledger.append("runStarted", { run: "run1" });
|
||||
});
|
||||
assert.equal(countRows(path), 2);
|
||||
});
|
||||
|
||||
test("事务：异步函数被类型检查挡住", () => {
|
||||
// 只给 npm run check 看，不执行：await 之后的写入发生在事务提交之后，回滚管不到。
|
||||
const typeOnly = (ledger: Ledger) =>
|
||||
// @ts-expect-error 事务函数的返回值类型排除了 Promise。
|
||||
ledger.transaction(async () => {
|
||||
ledger.append("config", { n: 1 });
|
||||
});
|
||||
assert.equal(typeof typeOnly, "function");
|
||||
});
|
||||
|
||||
test("关掉再打开：读出的记录不变，序号接着往下排", (t) => {
|
||||
const path = join(tempDir(t), "ledger.db");
|
||||
const ledger = Ledger.open(path);
|
||||
ledger.append("config", { researchers: ["r1", "r2"] });
|
||||
openStage(ledger, "s1", "proposal", 1);
|
||||
ledger.append("submission", { stageId: "s1", author: "r1", target: null, item: "r1-1", version: 1, kind: "proposal", content: proposal(range(1, 2, 3)) });
|
||||
ledger.append("note", { author: "r2", version: 1, body: "笔记" });
|
||||
closeStage(ledger, "s1");
|
||||
planRun(ledger, "run1", "forward", 1);
|
||||
ledger.append("callStarted", callStart("k1"));
|
||||
const before = ledger.records();
|
||||
ledger.close();
|
||||
const reopened = Ledger.open(path);
|
||||
t.after(() => reopened.close());
|
||||
assert.deepEqual(reopened.records(), before);
|
||||
assert.equal(before.length, 8); // 配置、开阶段、提案、封存、笔记、关阶段、登记运行、调用开始
|
||||
assert.deepEqual(before.map((r) => r.hash), before.map((r) => hash(r.content)));
|
||||
assert.equal(reopened.append("runStarted", { run: "run1" }).seq, 9);
|
||||
});
|
||||
|
||||
test("打开：表结构版本不对报错；不是空文件也不是账本的库报错", (t) => {
|
||||
const dir = tempDir(t);
|
||||
const wrongVersion = join(dir, "v99.db");
|
||||
const db = new DatabaseSync(wrongVersion);
|
||||
db.exec("PRAGMA user_version = 99");
|
||||
db.close();
|
||||
assert.throws(() => Ledger.open(wrongVersion), /版本是 99/);
|
||||
|
||||
const foreign = join(dir, "foreign.db");
|
||||
const other = new DatabaseSync(foreign);
|
||||
other.exec("CREATE TABLE t (x INTEGER)");
|
||||
other.close();
|
||||
assert.throws(() => Ledger.open(foreign), /版本是 0/);
|
||||
});
|
||||
|
||||
test("打开：不是账本的库报错后，那个库的 journal_mode 不变", (t) => {
|
||||
const foreign = join(tempDir(t), "foreign.db");
|
||||
const other = new DatabaseSync(foreign);
|
||||
other.exec("CREATE TABLE t (x INTEGER)");
|
||||
const modeOf = (db: DatabaseSync) => (db.prepare("PRAGMA journal_mode").get() as { journal_mode: string }).journal_mode;
|
||||
const before = modeOf(other);
|
||||
other.close();
|
||||
assert.notEqual(before, "wal");
|
||||
assert.throws(() => Ledger.open(foreign), /版本是 0/);
|
||||
const reopened = new DatabaseSync(foreign);
|
||||
t.after(() => reopened.close());
|
||||
assert.equal(modeOf(reopened), before);
|
||||
});
|
||||
|
||||
test("读出时按种类检查内容：绕过 ledger 插进去的坏记录，读取时报错", (t) => {
|
||||
const { ledger, path } = freshLedger(t);
|
||||
const db = new DatabaseSync(path);
|
||||
t.after(() => db.close());
|
||||
db.exec(`INSERT INTO records (kind, content, hash) VALUES ('runStarted', '{"run":"run1","x":1}', 'h')`);
|
||||
assert.throws(() => ledger.records(), /runStarted/);
|
||||
});
|
||||
|
||||
test("查询：阶段关了没有、某份提交的最新一版（同一人对同一候选的另一份挑刺不算这一份的版本）", (t) => {
|
||||
const { ledger } = freshLedger(t);
|
||||
openStage(ledger, "s1", "critique", 1);
|
||||
assert.equal(ledger.stageClosed("s1"), false);
|
||||
const slot = { stageId: "s1", author: "r1", kind: "critique" as const, target: "c1", item: "k1" };
|
||||
assert.equal(ledger.latestSubmission(slot), undefined);
|
||||
ledger.append("submission", { ...slot, version: 1, content: { target: "whole", body: "第一版" } });
|
||||
const v2 = ledger.append("submission", { ...slot, version: 2, content: { target: "whole", body: "第二版" } });
|
||||
ledger.append("submission", { ...slot, item: "k2", version: 1, content: { target: "p1", body: "另一份" } });
|
||||
assert.deepEqual(ledger.latestSubmission(slot), v2);
|
||||
closeStage(ledger, "s1");
|
||||
assert.equal(ledger.stageClosed("s1"), true);
|
||||
});
|
||||
Reference in New Issue
Block a user