mnote/wolai-frontend/convex/_utils/attachmentExtract.ts

import JSZip from "jszip";

const MAX_TEXT_CHARS = 120_000;

function clampText(input: string, maxChars = MAX_TEXT_CHARS): string {
  const trimmed = String(input || "").replace(/\s+\n/g, "\n").trim();
  if (!trimmed) return "";
  if (trimmed.length <= maxChars) return trimmed;
  return `${trimmed.slice(0, maxChars)}…`;
}

function decodeXmlEntities(input: string): string {
  return input
    .replace(/&lt;/g, "<")
    .replace(/&gt;/g, ">")
    .replace(/&amp;/g, "&")
    .replace(/&quot;/g, '"')
    .replace(/&apos;/g, "'")
    .replace(/&#x([0-9a-fA-F]+);/g, (_, hex) => {
      const code = Number.parseInt(hex, 16);
      if (!Number.isFinite(code)) return "";
      return String.fromCodePoint(code);
    })
    .replace(/&#(\d+);/g, (_, dec) => {
      const code = Number.parseInt(dec, 10);
      if (!Number.isFinite(code)) return "";
      return String.fromCodePoint(code);
    });
}

function getExtension(fileName: string | null | undefined): string {
  const raw = String(fileName ?? "").trim().toLowerCase();
  const idx = raw.lastIndexOf(".");
  if (idx === -1) return "";
  return raw.slice(idx + 1).replace(/[^a-z0-9]+/g, "");
}

function detectKind(args: { mimeType?: string | null; fileName?: string | null }): "pdf" | "docx" | "pptx" | "xlsx" | null {
  const mime = String(args.mimeType ?? "").toLowerCase().trim();
  if (mime === "application/pdf") return "pdf";
  if (mime === "application/vnd.openxmlformats-officedocument.wordprocessingml.document") return "docx";
  if (mime === "application/vnd.openxmlformats-officedocument.presentationml.presentation") return "pptx";
  if (mime === "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet") return "xlsx";

  const ext = getExtension(args.fileName ?? null);
  if (ext === "pdf") return "pdf";
  if (ext === "docx") return "docx";
  if (ext === "pptx") return "pptx";
  if (ext === "xlsx") return "xlsx";
  return null;
}

function extractTextFromXmlByTag(xml: string, tagName: string): string {
  const out: string[] = [];
  const re = new RegExp(`<${tagName}[^>]*>([\\s\\S]*?)<\\/${tagName}>`, "g");
  let m: RegExpExecArray | null;
  while ((m = re.exec(xml))) {
    const raw = m[1] ?? "";
    const clean = decodeXmlEntities(raw).replace(/\s+/g, " ").trim();
    if (clean) out.push(clean);
  }
  return out.join(" ").trim();
}

function extractDocxText(documentXml: string): string {
  const paras = documentXml.split(/<w:p[\s>]/g);
  const out: string[] = [];
  for (const p of paras) {
    const line: string[] = [];
    const re = /<w:t[^>]*>([\s\S]*?)<\/w:t>/g;
    let m: RegExpExecArray | null;
    while ((m = re.exec(p))) {
      const raw = m[1] ?? "";
      const clean = decodeXmlEntities(raw).replace(/\s+/g, " ").trim();
      if (clean) line.push(clean);
    }
    const joined = line.join("").trim();
    if (joined) out.push(joined);
  }
  return out.join("\n").trim();
}

function extractPptxText(slideXmls: string[]): string {
  const out: string[] = [];
  for (const xml of slideXmls) {
    const line = extractTextFromXmlByTag(xml, "a:t");
    if (line) out.push(line);
  }
  return out.join("\n\n").trim();
}

function extractXlsxText(args: { sharedStringsXml: string | null; sheetXmls: string[] }): string {
  const sharedStrings: string[] = [];
  if (args.sharedStringsXml) {
    const re = /<t[^>]*>([\s\S]*?)<\/t>/g;
    let m: RegExpExecArray | null;
    while ((m = re.exec(args.sharedStringsXml))) {
      const raw = m[1] ?? "";
      const clean = decodeXmlEntities(raw).replace(/\s+/g, " ").trim();
      if (clean) sharedStrings.push(clean);
    }
  }

  const out: string[] = [];
  for (const xml of args.sheetXmls) {
    const rows = xml.split(/<row[\s>]/g);
    for (const r of rows) {
      const cells: string[] = [];

      // t="s" => sharedStrings index；t="inlineStr" => <is><t>...；默认 => <v>number
      const cellRe = /<c\b[^>]*?(?:t="([^"]+)")?[^>]*>([\s\S]*?)<\/c>/g;
      let cm: RegExpExecArray | null;
      while ((cm = cellRe.exec(r))) {
        const t = (cm[1] ?? "").trim();
        const body = cm[2] ?? "";
        if (t === "inlineStr") {
          const inline = extractTextFromXmlByTag(body, "t");
          if (inline) cells.push(inline);
          continue;
        }
        const vMatch = /<v>([\s\S]*?)<\/v>/.exec(body);
        if (!vMatch) continue;
        const rawV = decodeXmlEntities(String(vMatch[1] ?? "")).trim();
        if (!rawV) continue;
        if (t === "s") {
          const idx = Number.parseInt(rawV, 10);
          const s = Number.isFinite(idx) ? (sharedStrings[idx] ?? "") : "";
          if (s) cells.push(s);
        } else {
          cells.push(rawV);
        }
      }

      const line = cells.join(" ").replace(/\s+/g, " ").trim();
      if (line) out.push(line);
    }
  }
  return out.join("\n").trim();
}

export type AttachmentExtractOk = {
  ok: true;
  strategy: string;
  text: string;
  meta?: Record<string, unknown>;
};
export type AttachmentExtractResult =
  | AttachmentExtractOk
  | { ok: false; strategy: string; reason: string; meta?: Record<string, unknown> };

export async function extractTextFromAttachment(args: {
  mimeType: string | null;
  fileName: string | null;
  bytes: ArrayBuffer;
}): Promise<AttachmentExtractResult> {
  const kind = detectKind({ mimeType: args.mimeType, fileName: args.fileName });
  if (!kind) {
    return { ok: false, strategy: "unsupported", reason: "暂不支持该附件类型" };
  }

  if (kind === "pdf") {
    try {
      const mod = await import("pdfjs-dist/legacy/build/pdf.mjs");
      const data = new Uint8Array(args.bytes);
      // eslint-disable-next-line @typescript-eslint/no-explicit-any
      const loadingTask = (mod as any).getDocument({ data });
      // eslint-disable-next-line @typescript-eslint/no-explicit-any
      const pdf = await (loadingTask as any).promise;
      const out: string[] = [];
      const pages = Number(pdf?.numPages ?? 0) || 0;
      for (let i = 1; i <= pages; i += 1) {
        // eslint-disable-next-line @typescript-eslint/no-explicit-any
        const page = await (pdf as any).getPage(i);
        // eslint-disable-next-line @typescript-eslint/no-explicit-any
        const content = await (page as any).getTextContent();
        // eslint-disable-next-line @typescript-eslint/no-explicit-any
        const items = Array.isArray((content as any)?.items) ? (content as any).items : [];
        const line = items
          .map((it: any) => (typeof it?.str === "string" ? it.str : ""))
          .join(" ")
          .replace(/\s+/g, " ")
          .trim();
        if (line) out.push(line);
      }
      const text = clampText(out.join("\n\n"));
      if (!text) {
        return { ok: false, strategy: "pdfjs", reason: "未提取到可用文本", meta: { pages } };
      }
      return { ok: true, strategy: "pdfjs", text, meta: { pages } };
    } catch (err) {
      const message = err instanceof Error ? err.message : String(err);
      return { ok: false, strategy: "pdfjs", reason: message };
    }
  }

  try {
    const zip = await JSZip.loadAsync(args.bytes);
    if (kind === "docx") {
      const file = zip.file("word/document.xml");
      const xml = file ? await file.async("string") : "";
      const text = clampText(extractDocxText(xml));
      if (!text) return { ok: false, strategy: "docx.xml", reason: "未提取到可用文本" };
      return { ok: true, strategy: "docx.xml", text };
    }

    if (kind === "pptx") {
      const slideFiles = Object.keys(zip.files)
        .filter((p) => /^ppt\/slides\/slide\d+\.xml$/i.test(p))
        .sort((a, b) => a.localeCompare(b, undefined, { numeric: true }));
      const slideXmls: string[] = [];
      for (const p of slideFiles) {
        const f = zip.file(p);
        if (!f) continue;
        // eslint-disable-next-line no-await-in-loop
        slideXmls.push(await f.async("string"));
      }
      const text = clampText(extractPptxText(slideXmls));
      if (!text) return { ok: false, strategy: "pptx.xml", reason: "未提取到可用文本" };
      return { ok: true, strategy: "pptx.xml", text, meta: { slides: slideXmls.length } };
    }

    if (kind === "xlsx") {
      const sharedStringsXml = await (async () => {
        const f = zip.file("xl/sharedStrings.xml");
        if (!f) return null;
        return await f.async("string");
      })();

      const sheetFiles = Object.keys(zip.files)
        .filter((p) => /^xl\/worksheets\/sheet\d+\.xml$/i.test(p))
        .sort((a, b) => a.localeCompare(b, undefined, { numeric: true }));
      const sheetXmls: string[] = [];
      for (const p of sheetFiles) {
        const f = zip.file(p);
        if (!f) continue;
        // eslint-disable-next-line no-await-in-loop
        sheetXmls.push(await f.async("string"));
      }
      const text = clampText(extractXlsxText({ sharedStringsXml, sheetXmls }));
      if (!text) return { ok: false, strategy: "xlsx.xml", reason: "未提取到可用文本" };
      return { ok: true, strategy: "xlsx.xml", text, meta: { sheets: sheetXmls.length } };
    }

    return { ok: false, strategy: "unsupported", reason: "暂不支持该附件类型" };
  } catch (err) {
    const message = err instanceof Error ? err.message : String(err);
    return { ok: false, strategy: "zip", reason: message };
  }
}