dwarf-fortress-annals/dfannals/factcheck.py

# pyright: reportMissingImports=false, reportAttributeAccessIssue=false
# 说明：LSP 的 Python 环境看不到本项目包（未安装进它的 site-packages），
# 会把「包内互相 import」误报为缺失模块。运行时导入已由执行验证通过。

"""专名事实校验：正文里出现的专有名词必须能在史料中找到。

策略（按访谈决策：自动直推、不阻断，但标注可疑处）：
- 抽出正文中的拉丁字母专名序列（本项目的约定是专名保留英文原文）
- 与史料已知专名集合比对
- 未知专名既不阻断推送，也不静默放过，而是连同上下文一起标注出来

注意：这一步只能拦住"凭空出现的人名地名"，拦不住编造的对白与心理描写——
后者按用户选择属于允许的自由演绎范围。
"""
from __future__ import annotations

import re
from dataclasses import dataclass

from dfannals.legends import World, pretty

# 正文里允许出现、但不算专名的常见英文词（游戏术语/通用词）
COMMON_WORDS = {
    "Dwarf", "Dwarves", "Dwarven", "Fortress", "Fort", "Mountain", "Mountains", "Hill",
    "River", "Lake", "Forest", "Desert", "Ocean", "Cave", "Caves", "Town", "City",
    "King", "Queen", "Lord", "Lady", "Captain", "General", "Soldier", "Miner", "Smith",
    "Hammer", "Axe", "Sword", "Shield", "Gold", "Silver", "Iron", "Steel", "Copper",
    "Bronze", "Wood", "Stone", "Water", "Beer", "Ale", "Wine", "Goblin", "Goblins",
    "Elf", "Elves", "Human", "Humans", "Kobold", "Kobolds", "Troll", "Trolls",
    "Beast", "Beasts", "Titan", "Titans", "Dragon", "Dragons", "Demon", "Demons",
    "God", "Gods", "Temple", "Shrine", "Tomb", "Market", "Tavern", "Library",
    "Legend", "Legends", "Annals", "Chronicle", "Chapter", "Year", "Years",
    "North", "South", "East", "West", "Spring", "Summer", "Autumn", "Winter",
    "The", "A", "An", "And", "Or", "But", "In", "On", "At", "By", "Of", "To", "For",
}

# 拉丁专名序列：首字母大写的词，可含 ' - 与数字
# 英文专名的边界不能用 \b：Unicode 下汉字也算「词字符」，于是 `在Mistbreath的`
# 这种紧贴汉字的英文名根本不会被匹配——实测正是这个盲点让一个编造的地名混过了校验。
# 改成前后看「是不是英文字母」，与汉字相邻也能抓到。
NAME_RE = re.compile(r"(?<![A-Za-z])[A-Z][A-Za-z'\-]*(?:\s+[A-Z][A-Za-z'\-]*)*(?![A-Za-z])")
_CJK = re.compile(r"[\u4e00-\u9fff]")


@dataclass
class Suspicion:
    name: str
    count: int
    context: str


def known_names(world: World) -> set[str]:
    """史料中出现过的全部专名，含单词与整名两种粒度。"""
    names: set[str] = set()

    def add(value: str) -> None:
        value = (value or "").strip()
        if not value:
            return
        names.add(value)
        # 同时收录展示形式：正文里专名会做首字母大写，两边必须一致
        shown = pretty(value)
        names.add(shown)
        for token in re.split(r"[\s,]+", shown):
            token = token.strip("'\"()[]")
            if token and token[0].isupper():
                names.add(token)
            # 原样小写形式也保留，避免漏报
            names.add(value.lower())

    # 世界名必须拆词：否则 "The World of Prophecy" 里的 Prophecy 会被当成凭空编造
    add(world.name)
    add(world.altname)

    for fig in world.figures.values():
        add(fig.name)
        add(fig.race)
        add(fig.profession)
    for ent in world.entities.values():
        add(ent.name)
        add(ent.type)
        add(ent.race)
    for site in world.sites.values():
        add(site.name)
        add(site.type)
    return names


def _is_candidate(name: str, known: set[str]) -> bool:
    if name in known:
        return False
    tokens = name.split()
    # 全部由常见词组成的串不是专名
    if all(t in COMMON_WORDS for t in tokens):
        return False
    # 全大写的是史料里的枚举值（GOBLIN、MINER、DARK_FORTRESS），不是人名地名
    return not all(len(t) >= 2 and t.isupper() for t in tokens)


def check(text: str, world: World) -> list[Suspicion]:
    """返回可疑专名列表（按出现次数降序）。"""
    known = known_names(world)
    # 大小写不该决定是不是幻觉：史料里是 YETI/EAGLE_MAN，正文写成 Yeti/Eagle_man
    # 是同一件事（模型从种族字段学来的），不能报成凭空编造。
    folded = {k.casefold() for k in known}
    found: dict[str, list[str]] = {}

    for match in NAME_RE.finditer(text):
        raw = match.group(0)
        if raw.casefold() in folded:
            continue
        # 逐级回退：整串不认，就试着拆成更短的已知名字，减少误报
        if not _is_candidate(raw, known):
            continue
        if any(part.casefold() in folded for part in raw.split()):
            # 名字里有一部分是史料已知的（例如 "Urist the Bold"），不当作凭空编造
            continue
        start = max(0, match.start() - 20)
        end = min(len(text), match.end() + 20)
        found.setdefault(raw, []).append(text[start:end].replace("\n", " "))

    return [
        Suspicion(name=n, count=len(ctxs), context=ctxs[0])
        for n, ctxs in sorted(found.items(), key=lambda kv: (-len(kv[1]), kv[0]))
    ]


def report(suspicions: list[Suspicion], top: int = 12) -> str:
    if not suspicions:
        return "专名校验：未发现史料之外的专有名词。"
    lines = [f"专名校验：发现 {len(suspicions)} 个可疑专名（不阻断推送，仅标注）"]
    for s in suspicions[:top]:
        lines.append(f"  · {s.name} ×{s.count} —— 上下文：…{s.context}…")
    if len(suspicions) > top:
        lines.append(f"  （其余 {len(suspicions) - top} 个见校验报告文件）")
    return "\n".join(lines)


def annotate(chapter_md: str, suspicions: list[Suspicion]) -> str:
    """在章节末尾追加可见的校验提示。"""
    if not suspicions:
        return chapter_md
    names = "、".join(f"{s.name}×{s.count}" for s in suspicions[:10])
    block = (
        "\n\n---\n\n"
        f"> ⚠️ **史料核验提示**：以下专名未在 legends 史料中找到，可能是演绎或错误：{names}。\n"
        "> 其余对白与心理描写属于允许的文学演绎，不在校验范围内。\n"
    )
    return chapter_md.rstrip() + block