9e6f1c4481
- src/muse 新版全模块（装配/共享/上下文/任务运行/作品规划/故事世界/正文写作/审校修订/知识方法/作者经验/效果评测/交付连载/资料研究/正式变更/元数据/接入/基础设施/编排）+ 测试树（单元/契约/集成/架构/迁移/端到端/夹具） - 129 项功能全部实现与自动验证（功能覆盖.json/矩阵），含 W31 补齐的规则与代价/节奏安排/伏笔与承诺 - 旧实现按处置清单退出（702 条中 324 删，保护合同与未迁移条目留存有据）；web/app.py 旧工作台退役，新工作台为唯一写入口 - 数据库/旧库迁移：真实旧库内容批次迁移链（端点守卫/PG作品正文映射/质量资产缺省投影） - 运行手册 docs/运行手册.md；W30 本机服务阶段一已运行（infra PG 为正式内容权威） - R2 执行证据与私有运行材料在 .agents.local/改造/R2-20260909/（不入库）
287 lines
12 KiB
Python
287 lines
12 KiB
Python
"""PG作品依赖与逐块共享候选：只经B01/B03/B05公开入口提交、精确读回。"""
|
||
|
||
from dataclasses import asdict
|
||
|
||
from pydantic import TypeAdapter
|
||
|
||
from muse.作品规划.接口 import 档案保存
|
||
from muse.正式变更.接口 import 固定哈希
|
||
from muse.正文写作.接口 import 可见文本, 可见文本哈希, 正文结构哈希, 正文草稿
|
||
from muse.资料研究.接口 import 导入请求
|
||
from PG作品映射 import PG目标ID, 核对PG组, 指针
|
||
from 建立映射 import 稳定JSON, 迁移错误
|
||
from 转换PG正文 import 转换PG正文
|
||
|
||
|
||
def PG分流(源, 映射, 已知源):
|
||
# 分流身份只取固定清单；依赖是否到齐不改变原路由，分批期间留reserved。
|
||
from PG作品映射 import PG源身份, PG绑定
|
||
|
||
组, 章 = PG绑定(映射, 源)
|
||
作者 = 映射.值.get("pg_author_id")
|
||
if not isinstance(作者, str) or not 作者:
|
||
raise 迁移错误("pg_mapping_invalid", "PG映射必须绑定应用配置作者，不从旧owner/tenant推导")
|
||
作品身份 = PG源身份(组["work"])
|
||
if 组["use"] == "reference":
|
||
owner, kind, scope, target = "B03", "reference", 作品身份.对象键, 作品身份.对象键
|
||
else:
|
||
scope = PG目标ID(作品身份, 作者, "work")
|
||
if 章 is None:
|
||
owner, kind, target = "B01", "profile", scope
|
||
elif 指针(源) == 章["chapter"]:
|
||
owner, kind, target = "B01", "directory", PG目标ID(源, 作者, "chapter")
|
||
else:
|
||
owner, kind, target = "B05", "body", PG目标ID(PG源身份(章["chapter"]), 作者, "chapter")
|
||
return {
|
||
"owner": owner,
|
||
"type_id": "pg_" + kind,
|
||
"scope": scope,
|
||
"target_ref": target,
|
||
"state": "pending",
|
||
"content": {"pg_kind": kind, "use": 组["use"], "work_source": 组["work"]},
|
||
}
|
||
|
||
|
||
def 准备PG请求(装配, 身份, 源, 判定, 映射, 已知源):
|
||
if 映射.值.get("pg_author_id") != 身份.作者 or not 身份.允许写正式内容:
|
||
raise 迁移错误("target_denied", "PG目标作者必须来自当前应用身份")
|
||
当前分流 = PG分流(源, 映射, 已知源)
|
||
if any(当前分流[k] != getattr(判定, k) for k in ("owner", "scope", "target_ref", "type_id")):
|
||
raise 迁移错误("pg_mapping_invalid", "PG分流身份发生变化")
|
||
组, 作品, 章节 = 核对PG组(映射, 源, 已知源)
|
||
kind = 当前分流["content"]["pg_kind"]
|
||
command_source = 指针(源)
|
||
dependencies = [指针(作品)]
|
||
if kind == "profile":
|
||
work_id = 判定.scope
|
||
结构 = 装配.要求作品().新档案结构(身份, work_id, "work_core", 1)
|
||
payload = asdict(档案保存(work_id, 0, {"名称": 作品.原行["title"]}, 结构))
|
||
elif kind == "directory":
|
||
i = next(i for i, (c, _) in enumerate(章节) if 指针(c) == 指针(源))
|
||
目录 = 装配.要求作品().读取目录(身份, 判定.scope)
|
||
前章 = [PG目标ID(c, 身份.作者, "chapter") for c, _ in 章节[:i]]
|
||
if 目录["revision"] != i or [c["chapter_id"] for c in 目录["chapters"]] != 前章:
|
||
raise 迁移错误("pg_directory_stale", "目录必须由本清单的真实前序章节构成")
|
||
dependencies.extend(指针(c) for c, _ in 章节[: i + 1])
|
||
payload = {
|
||
"work_id": 判定.scope,
|
||
"chapter_id": 判定.target_ref,
|
||
"title": 源.原行["title"],
|
||
"expected_revision": i,
|
||
"chapter_ids": [*前章, 判定.target_ref],
|
||
}
|
||
elif kind == "body":
|
||
章, 块组 = next((c, b) for c, b in 章节 if any(指针(源) == 指针(x) for x in b))
|
||
当前 = 装配.要求正文().读取正文(身份, 判定.target_ref, 分支="main")
|
||
if 当前["work_id"] != 判定.scope or 当前["revision"] != 0:
|
||
raise 迁移错误("target_base_stale", "PG候选只面向真实本作品章的空正式基线")
|
||
dependencies.extend([指针(章), *(指针(b) for b in 块组)])
|
||
command_source = 指针(块组[0])
|
||
payload = {
|
||
"chapter_id": 判定.target_ref,
|
||
"branch_id": "main",
|
||
"base_revision": 0,
|
||
"base_document_hash": 当前["document_hash"],
|
||
"draft": 转换PG正文(块组),
|
||
}
|
||
else:
|
||
for 章, 块组 in 章节:
|
||
dependencies.extend([指针(章), *(指针(b) for b in 块组)])
|
||
文本, 对照 = 组装PG参考(章节)
|
||
command_source = 指针(作品)
|
||
请求 = 导入请求(
|
||
kind="reference",
|
||
title=作品.原行["title"],
|
||
origin="legacy-pg:" + 稳定JSON([身份.作者, 作品.source.对象键]),
|
||
content=文本,
|
||
authorized_uses=(),
|
||
)
|
||
payload = {"request": asdict(请求), "chapter_map": 对照}
|
||
return {
|
||
"owner": 判定.owner,
|
||
"scope": 判定.scope,
|
||
"subject_ref": 判定.target_ref,
|
||
"pg_kind": kind,
|
||
"legacy_source": 指针(源),
|
||
"command_source": command_source,
|
||
"dependencies": dependencies,
|
||
"mapping_hash": 固定映射哈希(映射),
|
||
"payload": payload,
|
||
}
|
||
|
||
|
||
def 组装PG参考(章节):
|
||
文本, 对照, 起 = [], [], 0
|
||
for 章, 块组 in 章节:
|
||
draft = 转换PG正文(块组) if 块组 else None
|
||
text = 可见文本(TypeAdapter(正文草稿).validate_python(draft["document"])) if draft else ""
|
||
对照.append(
|
||
{
|
||
**指针(章),
|
||
"order_no": 章.原行["order_no"],
|
||
"source_start": 起,
|
||
"source_end": 起 + len(text),
|
||
"draft": draft,
|
||
}
|
||
)
|
||
起 += len(text) + 1
|
||
文本.append(text)
|
||
return "\n".join(文本), 对照
|
||
|
||
|
||
def 固定映射哈希(映射):
|
||
import hashlib
|
||
|
||
return hashlib.sha256(稳定JSON(映射.值).encode()).hexdigest()
|
||
|
||
|
||
def PG命令ID(身份, 冻结):
|
||
from 映射台账 import 迁移命令ID
|
||
|
||
return 迁移命令ID(身份.作者, 冻结["command_source"]["source_key"])
|
||
|
||
|
||
def 提交PG请求(装配, 身份, 冻结):
|
||
p, kind = 冻结["payload"], 冻结["pg_kind"]
|
||
cmd = PG命令ID(身份, 冻结)
|
||
if kind == "profile":
|
||
return 装配.要求作品().保存档案(身份, cmd, TypeAdapter(档案保存).validate_python(p))
|
||
if kind == "directory":
|
||
return 装配.要求作品().添加章节(
|
||
身份,
|
||
cmd,
|
||
p["work_id"],
|
||
p["chapter_id"],
|
||
p["title"],
|
||
预期目录版本=p["expected_revision"],
|
||
)
|
||
if kind == "body":
|
||
draft = TypeAdapter(正文草稿).validate_python(p["draft"]["document"])
|
||
return 装配.要求正文().创建人工候选(
|
||
身份, cmd, p["chapter_id"], p["base_revision"], draft, 分支=p["branch_id"]
|
||
)
|
||
if kind == "reference":
|
||
with 装配.要求数据库().连接() as 连, 连.transaction():
|
||
结果 = 装配.要求资料().导入(
|
||
连, 身份.作者, TypeAdapter(导入请求).validate_python(p["request"])
|
||
)
|
||
# 重复位只是本次调用情况，不冒充另一个不可变导入回执。
|
||
版本 = 装配.要求资料().读取版本(连, 结果.source_id, 结果.revision)
|
||
return {
|
||
"source_id": 结果.source_id,
|
||
"revision": 结果.revision,
|
||
"content_hash": 结果.content_hash,
|
||
"import_result": 版本["import_result"],
|
||
}
|
||
raise 迁移错误("target_request_invalid", "未知PG目标种类")
|
||
|
||
|
||
def 核对PG产物(装配, 身份, 冻结, 回执):
|
||
p, kind = 冻结["payload"], 冻结["pg_kind"]
|
||
if not 身份.允许写正式内容 or not 身份.作者:
|
||
raise 迁移错误("target_denied", "PG核对需要应用作者身份")
|
||
if kind == "reference":
|
||
with 装配.要求数据库().连接(只读=True) as 连:
|
||
v = 装配.要求资料().读取版本(连, 回执["source_id"], 回执["revision"])
|
||
# B03公开版本带原文；与冻结请求逐码点核对，不仅比较自报哈希。
|
||
import hashlib
|
||
|
||
h = hashlib.sha256(p["request"]["content"].encode()).hexdigest()
|
||
if (
|
||
v["content_hash"] != h
|
||
or 回执["content_hash"] != h
|
||
or v["content"] != p["request"]["content"]
|
||
or v["import_result"] != 回执["import_result"]
|
||
or v["import_result"].get("导入者") != 身份.作者
|
||
):
|
||
raise 迁移错误("target_mismatch", "参考原文版本与冻结来源或作者不同")
|
||
来源 = [
|
||
s for s in 装配.要求资料().列出来源(连) if str(s["source_id"]) == 回执["source_id"]
|
||
]
|
||
if (
|
||
len(来源) != 1
|
||
or 来源[0]["origin"] != p["request"]["origin"]
|
||
or 来源[0]["kind"] != "reference"
|
||
):
|
||
raise 迁移错误("target_mismatch", "参考源身份不同")
|
||
return [
|
||
{
|
||
"id": 回执["source_id"],
|
||
"revision": 回执["revision"],
|
||
"kind": "source",
|
||
"owner": "B03",
|
||
"scope": 冻结["scope"],
|
||
"subject_ref": 冻结["subject_ref"],
|
||
"content_hash": h,
|
||
}
|
||
]
|
||
真实 = 装配.要求作品().正式.读取回执(身份, PG命令ID(身份, 冻结))
|
||
if 稳定JSON(真实) != 稳定JSON(回执) or 回执.get("action") != "manual_save":
|
||
raise 迁移错误("target_receipt_invalid", "PG依赖/候选必须来自真实人工保存回执，不接受确认")
|
||
if kind == "profile":
|
||
rows = [
|
||
r
|
||
for r in 回执["results"]
|
||
if r.get("work_id") == p["work_id"] and r.get("revision") == 1
|
||
]
|
||
v = 装配.要求作品().读取档案版本依据(身份, p["work_id"], 1)
|
||
if (
|
||
len(rows) != 1
|
||
or v["content_hash"] != 固定哈希(p["content"])
|
||
or v["schema_binding"] != rows[0]["schema_binding"]
|
||
):
|
||
raise 迁移错误("target_mismatch", "PG作品档案版本与真实回执/冻结不同")
|
||
id_, revision = p["work_id"], 1
|
||
elif kind == "directory":
|
||
revision = p["expected_revision"] + 1
|
||
rows = [
|
||
r
|
||
for r in 回执["results"]
|
||
if r.get("work_id") == p["work_id"]
|
||
and r.get("revision") == revision
|
||
and r.get("chapter_ids") == p["chapter_ids"]
|
||
]
|
||
v = 装配.要求作品().读取节点目录(身份, p["work_id"], 版本=revision)
|
||
chapters = [c for c in v["nodes"] if c["kind"] == "chapter"]
|
||
if (
|
||
len(rows) != 1
|
||
or [c["node_id"] for c in chapters] != p["chapter_ids"]
|
||
or chapters[-1]["title"] != p["title"]
|
||
):
|
||
raise 迁移错误("target_mismatch", "PG目录版本与真实回执/原章不同")
|
||
id_ = p["chapter_id"]
|
||
elif kind == "body":
|
||
id_ = 回执["target_ref"]
|
||
rows = [
|
||
r for r in 回执["results"] if r.get("candidate_id") == id_ and r.get("revision") == 1
|
||
]
|
||
v = 装配.要求正文().读取候选(身份, id_, 版本=1)
|
||
doc = TypeAdapter(正文草稿).validate_python(v["document"])
|
||
base = 装配.要求正文().读取正文(
|
||
身份, p["chapter_id"], 分支=p["branch_id"], 版本=p["base_revision"]
|
||
)
|
||
if (
|
||
len(rows) != 1
|
||
or v["origin"] != "author"
|
||
or any(v[k] != p[k] for k in ("chapter_id", "branch_id", "base_revision"))
|
||
or v["candidate_hash"] != rows[0]["candidate_hash"]
|
||
or 正文结构哈希(doc) != p["draft"]["document_hash"]
|
||
or 可见文本哈希(doc) != p["draft"]["visible_text_hash"]
|
||
or base["work_id"] != 冻结["scope"]
|
||
or base["document_hash"] != p["base_document_hash"]
|
||
):
|
||
raise 迁移错误("target_mismatch", "PG正文候选版本、原文或基线不同")
|
||
revision = 1
|
||
else:
|
||
raise 迁移错误("target_request_invalid", "未知PG目标种类")
|
||
return [
|
||
{
|
||
"id": id_,
|
||
"revision": revision,
|
||
"kind": {"profile": "work", "directory": "chapter", "body": "candidate"}[kind],
|
||
"owner": 冻结["owner"],
|
||
"scope": 冻结["scope"],
|
||
"subject_ref": 冻结["subject_ref"],
|
||
}
|
||
]
|