import { CAPABILITY_IDS } from "@umt/contracts";

// 汇报格式的机器核对(AG-5 评测用):fake 模型评测与真实模型评测共用。
// 只查能机械判断的项;"结论在第一句"按开头的套话与是否含状态词近似判断。

export interface ReportLintIssue {
  rule:
    | "too_long"
    | "uuid"
    | "hash"
    | "raw_bytes"
    | "iso_time"
    | "internal_field"
    | "capability_id"
    | "heading"
    | "no_conclusion_first"
    | "asks_user_for_id"
    | "lecture";
  detail: string;
}

export interface ReportLintOptions {
  /** 最多几行(非空行);缺省 8 */
  maxLines?: number;
}

const UUID = /\b[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}\b/i;
const HASH = /\b[0-9a-f]{40}(?:[0-9a-f]{24})?\b/i;
/** 7 位以上的裸数字(原始字节数、原始计数);年份、日期、带千分位的不算 */
const RAW_NUMBER = /(?<![\d.:\-/])\d{7,}(?![\d.:\-/])/;
const ISO_TIME = /\b\d{4}-\d{2}-\d{2}T\d{2}:\d{2}/;
/** 内部字段名:snake_case、camelCase 与已知的内部词 */
const INTERNAL_WORDS =
  /\b(?:adapter|healthStatus|health_status|mediaEntityId|slotId|planId|runId|taskId|downloadTaskId|sourceReleaseId|payloadHash|idempotency|capabilityId|invocationId)\b/;
const SNAKE = /\b[a-z]+_[a-z_]+\b/;
const CAMEL = /\b[a-z]+[A-Z][A-Za-z]+\b/;
/** 开头套话(结论没放第一句) */
const WEAK_OPENERS = /^(?:#|\*\*?检查结果|检查结果|结论[:：]|以下|下面|好的|当然|首先|根据|经过|我(?:已|先|查)|让我)/;
const ASK_ID = /(?:请|麻烦)?(?:提供|给(?:我)?(?:一个)?)[^。\n]{0,12}(?:ID|hash|编号|哈希)/i;
const LECTURE = /(?:不等于|并不代表|不能证明|无证据|建议定期|值得注意的是|需要注意的是)/;

/** 允许出现在汇报里的英文词(技术词保留原文) */
const ALLOWED_TOKENS = new Set(["qBittorrent", "iPhone", "macOS"]);

/** 核对一段汇报;返回空数组 = 通过 */
export function lintReport(text: string, opts: ReportLintOptions = {}): ReportLintIssue[] {
  const maxLines = opts.maxLines ?? 8;
  const issues: ReportLintIssue[] = [];
  const lines = text
    .split("\n")
    .map((l) => l.trim())
    .filter((l) => l !== "" && !/^\|?\s*:?-{3,}/.test(l));
  if (lines.length > maxLines) issues.push({ rule: "too_long", detail: `${lines.length} 行，超过 ${maxLines} 行` });
  const find = (re: RegExp) => text.match(re)?.[0];
  const hit = (rule: ReportLintIssue["rule"], re: RegExp) => {
    const m = find(re);
    if (m) issues.push({ rule, detail: m });
  };
  hit("uuid", UUID);
  hit("hash", HASH);
  hit("raw_bytes", RAW_NUMBER);
  hit("iso_time", ISO_TIME);
  hit("internal_field", INTERNAL_WORDS);
  const snake = find(SNAKE);
  if (snake) issues.push({ rule: "internal_field", detail: snake });
  const camel = text.match(new RegExp(CAMEL, "g"))?.find((w) => !ALLOWED_TOKENS.has(w));
  if (camel) issues.push({ rule: "internal_field", detail: camel });
  const cap = CAPABILITY_IDS.find((id) => text.includes(id));
  if (cap) issues.push({ rule: "capability_id", detail: cap });
  if (/^#{1,6} /m.test(text)) issues.push({ rule: "heading", detail: "用了 Markdown 标题" });
  const first = lines[0] ?? "";
  if (!first || WEAK_OPENERS.test(first)) issues.push({ rule: "no_conclusion_first", detail: first.slice(0, 40) });
  hit("asks_user_for_id", ASK_ID);
  hit("lecture", LECTURE);
  return issues;
}