dev_harness/dev_scripts/check_harness.py

"""检查 DevHarness 必需文件、核心文档和任务归档的基本结构。"""

from __future__ import annotations

import argparse
import re
from pathlib import Path

from wiki_docs import WikiDocsError, load_config, parse_mirror


ROOT = Path(__file__).resolve().parents[1]
CORE_PAGE_PATHS = {
    "Home": "docs/README.md",
    "Project-Profile": "docs/00-project-profile.md",
    "Development-Workflow": "docs/01-workflow.md",
    "Architecture-and-Code-Map": "docs/02-architecture-and-code-map.md",
    "Business-Rules-and-Glossary": "docs/03-business-rules-and-glossary.md",
    "Local-Development-and-Verification": (
        "docs/04-local-development-and-verification.md"
    ),
    "Common-Changes": "docs/05-common-changes.md",
    "Troubleshooting": "docs/06-troubleshooting.md",
    "New-Project-Documentation-Setup": (
        "docs/07-new-project-documentation-setup.md"
    ),
    "Task-Archive-Template": "docs/templates/task-archive.md",
}
CORE_DOCUMENT_REQUIREMENTS = {
    "docs/README.md": (
        "## 第一次阅读",
        "## 五分钟开始",
        "## 简单修改从哪里开始",
        "## 事实来源",
    ),
    "docs/00-project-profile.md": (
        "## 基本信息",
        "## 技术栈与运行环境",
        "## 阅读入口",
        "## 常用命令",
        "## 环境、配置与凭据",
    ),
    "docs/01-workflow.md": (
        "## 面向初级维护者的修改边界",
        "## 每个任务的文档影响",
        "## 稳定文档与任务归档",
        "## 效率与范围控制",
        "### 严格控制范围",
        "### 渐进执行和修复",
        "### 复用已验证事实",
        "### 明确停止条件",
    ),
    "docs/02-architecture-and-code-map.md": (
        "## 项目定位",
        "## 代码地图",
        "## 两条主要执行路径",
        "## 不可破坏的边界",
    ),
    "docs/03-business-rules-and-glossary.md": (
        "## 核心术语",
        "## 工单状态",
        "## 稳定业务规则",
        "## 新项目需要补充什么",
    ),
    "docs/04-local-development-and-verification.md": (
        "## 环境要求",
        "## 第一次运行",
        "## 常用调试方式",
        "## 完成修改前",
    ),
    "docs/05-common-changes.md": (
        "## 风险分级",
        "## 修改 Wiki 文案",
        "## 调整 Harness 检查",
        "## 看懂 Agent 的修改",
    ),
    "docs/06-troubleshooting.md": (
        "## 排查顺序",
        "## 必须停止的情况",
    ),
    "docs/07-new-project-documentation-setup.md": (
        "## 初始化顺序",
        "## 完成标准",
    ),
}
REQUIRED_FILES = (
    "AGENTS.md",
    "CLAUDE.md",
    "README.md",
    "docs/00-project-profile.md",
    "docs/01-workflow.md",
    "docs/templates/task-archive.md",
    *CORE_DOCUMENT_REQUIREMENTS,
    "wiki-docs.json",
    "dev_scripts/wiki_docs.py",
    "dev_scripts/sync_wiki_docs.py",
    ".gitea/issue_template/epic.md",
    ".gitea/issue_template/mvp.md",
    ".gitea/issue_template/task.md",
)
ARCHIVE_HEADINGS = (
    "## 背景与目标",
    "## 最终方案",
    "## 修改文件",
    "## 验收结果",
    "## 测试",
    "## 相关提交",
)


def check_required_files(errors: list[str]) -> None:
    for relative_path in REQUIRED_FILES:
        if not (ROOT / relative_path).is_file():
            errors.append(f"缺少必需文件：{relative_path}")


def check_project_profile(errors: list[str], warnings: list[str], strict: bool) -> None:
    profile = ROOT / "docs" / "00-project-profile.md"
    if not profile.is_file():
        return
    if "<填写" in profile.read_text(encoding="utf-8"):
        message = "项目档案仍有未填写内容"
        (errors if strict else warnings).append(message)


def check_archives(errors: list[str]) -> None:
    task_dir = ROOT / "docs" / "task"
    for path in task_dir.glob("*.md"):
        if not re.match(r"^\d+-.+\.md$", path.name):
            errors.append(f"归档文件名不符合 <编号>-<标题>.md：{path.name}")
        content = path.read_text(encoding="utf-8")
        for heading in ARCHIVE_HEADINGS:
            if heading not in content:
                errors.append(f"{path.name} 缺少章节：{heading}")
        if "**未验证部分**：" not in content:
            errors.append(f"{path.name} 没有记录未验证部分")


def missing_sections(content: str, required: tuple[str, ...]) -> list[str]:
    return [section for section in required if section not in content]


def check_core_documents(errors: list[str], root: Path = ROOT) -> None:
    """检查初级维护者所需主题页的固定结构。"""

    for relative_path, required in CORE_DOCUMENT_REQUIREMENTS.items():
        path = root / relative_path
        if not path.is_file():
            continue
        try:
            _, body = parse_mirror(path.read_text(encoding="utf-8"))
        except (OSError, UnicodeDecodeError, WikiDocsError):
            continue
        for section in missing_sections(body, required):
            errors.append(f"{relative_path} 缺少核心章节：{section}")


def check_task_template(errors: list[str], root: Path = ROOT) -> None:
    path = root / ".gitea" / "issue_template" / "task.md"
    if not path.is_file():
        return
    content = path.read_text(encoding="utf-8")
    required = (
        "## 依赖与并行",
        "- 前置工单：无 / #编号",
        "- 是否允许与前置工单并行：是 / 否",
        "- 原因：",
        "## 文档影响",
        "- [ ] 不影响长期文档，原因：",
        "- [ ] 更新架构与代码地图",
        "- [ ] 更新业务规则与术语",
        "- [ ] 更新常见修改或故障排查",
    )
    for section in missing_sections(content, required):
        errors.append(f"单元任务模板缺少：{section}")


def check_agent_efficiency_rules(errors: list[str], root: Path = ROOT) -> None:
    path = root / "AGENTS.md"
    if not path.is_file():
        return
    content = path.read_text(encoding="utf-8")
    required = (
        "### 效率与范围控制",
        "#### 严格控制范围",
        "#### 渐进执行和修复",
        "#### 复用已验证事实",
        "#### 明确停止条件",
        "单元任务是唯一正式实施单位",
        "高风险修改必须停止",
        "用户没有明确验收通过前不得关闭",
        "长期文档必须先修改 Wiki",
        "提交只包含当前工单相关文件",
    )
    for section in missing_sections(content, required):
        errors.append(f"AGENTS.md 缺少：{section}")


def check_claude_code_entry(errors: list[str], root: Path = ROOT) -> None:
    """检查 Claude Code 入口直接复用共同 Agent 规则。"""

    path = root / "CLAUDE.md"
    if not path.is_file():
        return
    content = path.read_text(encoding="utf-8")
    lines = {line.strip() for line in content.splitlines()}
    if "@AGENTS.md" not in lines:
        errors.append("CLAUDE.md 缺少独立的 @AGENTS.md 导入")
    required = (
        "共同规则事实来源",
        "只记录 Claude Code 特有",
        "只修改 `AGENTS.md`",
        "## 模型路由",
        "## Agent 交接",
        "## Haiku 只读约束",
        "当前模型足以完成任务时不升级模型",
        "Opus 输出方案后必须等待用户确认",
        "不让 Haiku 决定最终根因",
        "只读必须通过子 Agent 工具权限实现",
    )
    for section in missing_sections(content, required):
        errors.append(f"CLAUDE.md 缺少：{section}")


def core_mapping_errors(configured_mappings: dict[str, str]) -> list[str]:
    errors: list[str] = []
    for page, expected_path in CORE_PAGE_PATHS.items():
        if configured_mappings.get(page) != expected_path:
            errors.append(
                f"核心 Wiki 页面映射缺失或路径错误：{page} -> {expected_path}"
            )
    return errors


def check_wiki_mirrors(errors: list[str]) -> None:
    """检查每份本地文档都有显式映射和可追踪的镜像头。"""

    try:
        config = load_config()
    except WikiDocsError as exc:
        errors.append(str(exc))
        return

    configured_mappings = {mapping.page: mapping.path for mapping in config.mappings}
    errors.extend(core_mapping_errors(configured_mappings))

    mapped_paths = {mapping.path for mapping in config.mappings}
    actual_paths = {
        path.relative_to(ROOT).as_posix() for path in (ROOT / "docs").rglob("*.md")
    }
    for path in sorted(actual_paths - mapped_paths):
        errors.append(f"docs 中存在未登记的 Wiki 镜像：{path}")

    for mapping in config.mappings:
        path = ROOT / mapping.path
        if not path.is_file():
            errors.append(f"缺少 Wiki 镜像：{mapping.path}")
            continue
        try:
            metadata, _ = parse_mirror(path.read_text(encoding="utf-8"))
        except (OSError, UnicodeDecodeError, WikiDocsError) as exc:
            errors.append(f"Wiki 镜像无效 {mapping.path}：{exc}")
            continue
        if metadata.get("generated") != "true (请先修改 Gitea Wiki，禁止直接编辑本文件)":
            errors.append(f"{mapping.path} 没有只读镜像标记")
        if metadata.get("wiki_page") != mapping.page:
            errors.append(f"{mapping.path} 的 wiki_page 与映射不一致")
        revision = metadata.get("wiki_revision", "")
        if re.fullmatch(r"[0-9a-f]{40,64}", revision) is None:
            errors.append(f"{mapping.path} 的 wiki_revision 无效")
        if not metadata.get("synchronized_at"):
            errors.append(f"{mapping.path} 缺少 synchronized_at")


def main() -> int:
    parser = argparse.ArgumentParser(description="检查 DevHarness 项目结构")
    parser.add_argument(
        "--strict",
        action="store_true",
        help="项目档案有占位内容时返回失败",
    )
    args = parser.parse_args()

    errors: list[str] = []
    warnings: list[str] = []
    check_required_files(errors)
    check_project_profile(errors, warnings, args.strict)
    check_wiki_mirrors(errors)
    check_core_documents(errors)
    check_task_template(errors)
    check_agent_efficiency_rules(errors)
    check_claude_code_entry(errors)
    check_archives(errors)

    for warning in warnings:
        print(f"警告：{warning}")
    for error in errors:
        print(f"错误：{error}")

    if errors:
        print(f"检查失败：{len(errors)} 个问题")
        return 1
    print("DevHarness 检查通过")
    return 0


if __name__ == "__main__":
    raise SystemExit(main())