@yanhaidao/wecom 2.3.190 → 2.3.270

package/src/wecom_msg_adapter/markdown_adapter.ts

@@ -0,0 +1,331 @@
+/**
+ * 将较完整的 Markdown 降级转换为更适合企业微信 markdown_v2 的子集。
+ *
+ * 保守策略：
+ * - 保留：标题、粗体、斜体、引用、链接、行内代码、普通列表、表格
+ * - 降级：代码块、图片、任务列表、分隔线、HTML、脚注、复杂语法
+ * - 清理：多余空行、非法控制字符、过深嵌套
+ */
+export function toWeComMarkdownV2(markdown: unknown, maxLength = 4096): string {
+  if (!markdown) return "";
+
+  let text = String(markdown).replace(/\r\n/g, "\n").replace(/\r/g, "\n");
+
+  const extracted = extractInlineCodeSpans(text);
+  text = extracted.text;
+  const inlineCodeStore = extracted.store;
+
+  text = convertFencedCodeBlocks(text);
+  text = convertIndentedCodeBlocks(text);
+  text = convertImages(text);
+  text = convertTaskLists(text);
+  text = convertThematicBreaks(text);
+  text = stripHtml(text);
+  text = removeFootnotes(text);
+  text = removeUnfriendlyExtensions(text);
+  text = flattenDeepNesting(text);
+  text = normalizeTables(text);
+  text = restoreInlineCodeSpans(text, inlineCodeStore);
+  text = cleanupWhitespace(text);
+
+  if (maxLength != null && text.length > maxLength) {
+    text = truncateSafely(text, maxLength);
+  }
+
+  return text;
+}
+
+const INLINE_CODE_PREFIX = "\uFFF0INLINECODE";
+const INLINE_CODE_SUFFIX = "\uFFF1";
+
+function extractInlineCodeSpans(text: string): { text: string; store: string[] } {
+  const store: string[] = [];
+  const replaced = text.replace(/`([^`\n]+?)`/g, (_, content: string) => {
+    const idx = store.length;
+    store.push(content);
+    return `${INLINE_CODE_PREFIX}${idx}${INLINE_CODE_SUFFIX}`;
+  });
+  return { text: replaced, store };
+}
+
+function restoreInlineCodeSpans(text: string, store: string[]): string {
+  const re = new RegExp(`${INLINE_CODE_PREFIX}(\\d+)${INLINE_CODE_SUFFIX}`, "g");
+  return text.replace(re, (_, idxStr: string) => {
+    const idx = Number(idxStr);
+    const content = idx >= 0 && idx < store.length ? store[idx] : "";
+    return `\`${content}\``;
+  });
+}
+
+function convertFencedCodeBlocks(text: string): string {
+  return text.replace(/```([a-zA-Z0-9_+\-]*)\n([\s\S]*?)```/g, (_, lang: string, code: string) => {
+    const safeLang = (lang || "").trim();
+    const safeCode = String(code || "").replace(/^\n+|\n+$/g, "");
+    if (!safeCode.trim()) return "";
+
+    const title = safeLang ? `代码（${safeLang}）：` : "代码：";
+    return `\n${title}\n${safeCode}\n`;
+  });
+}
+
+function convertIndentedCodeBlocks(text: string): string {
+  const lines = text.split("\n");
+  const out: string[] = [];
+  let buffer: string[] = [];
+
+  const flushBuffer = () => {
+    if (!buffer.length) return;
+    const block = buffer
+      .map(line => (line.startsWith("    ") ? line.slice(4) : line))
+      .join("\n")
+      .replace(/\s+$/g, "");
+
+    if (block) {
+      out.push("代码：");
+      out.push(...block.split("\n"));
+    }
+    buffer = [];
+  };
+
+  for (const line of lines) {
+    if (/^    \S/.test(line)) {
+      buffer.push(line);
+    } else {
+      flushBuffer();
+      out.push(line);
+    }
+  }
+
+  flushBuffer();
+  return out.join("\n");
+}
+
+function convertImages(text: string): string {
+  text = text.replace(/!\[([^\]]*)\]\(([^)]+)\)/g, (_, alt: string, url: string) => {
+    const safeAlt = (alt || "").trim() || "图片";
+    const safeUrl = (url || "").trim();
+    return `[图片：${safeAlt}](${safeUrl})`;
+  });
+
+  text = text.replace(/!\[([^\]]*)\]\[[^\]]*\]/g, (_, alt: string) => {
+    const safeAlt = (alt || "").trim() || "图片";
+    return `图片：${safeAlt}`;
+  });
+
+  return text;
+}
+
+function convertTaskLists(text: string): string {
+  text = text.replace(/^(\s*[-*+]\s+)\[x\]\s+/gim, "✅ ");
+  text = text.replace(/^(\s*[-*+]\s+)\[\s\]\s+/gm, "⬜ ");
+  return text;
+}
+
+function convertThematicBreaks(text: string): string {
+  return text.replace(/^\s*([-*_])(\s*\1){2,}\s*$/gm, "────────");
+}
+
+function stripHtml(text: string): string {
+  text = text.replace(/<script\b[^>]*>[\s\S]*?<\/script>/gi, "");
+  text = text.replace(/<style\b[^>]*>[\s\S]*?<\/style>/gi, "");
+
+  text = text.replace(/<br\s*\/?>/gi, "\n");
+  text = text.replace(/<\/p\s*>/gi, "\n");
+  text = text.replace(/<p\b[^>]*>/gi, "");
+
+  const simpleTags = [
+    "div", "span", "b", "strong", "i", "em", "u",
+    "font", "small", "big", "section", "article",
+    "header", "footer", "main",
+  ];
+
+  for (const tag of simpleTags) {
+    const re = new RegExp(`</?${tag}\\b[^>]*>`, "gi");
+    text = text.replace(re, "");
+  }
+
+  text = text.replace(/<[^>]+>/g, "");
+  text = decodeHtmlEntities(text);
+
+  return text;
+}
+
+function decodeHtmlEntities(text: string): string {
+  const map: Record<string, string> = {
+    "&amp;": "&",
+    "&lt;": "<",
+    "&gt;": ">",
+    "&quot;": "\"",
+    "&#39;": "'",
+    "&nbsp;": " ",
+  };
+
+  return text.replace(/&amp;|&lt;|&gt;|&quot;|&#39;|&nbsp;/g, m => map[m] ?? m);
+}
+
+function removeFootnotes(text: string): string {
+  text = text.replace(/^\[\^[^\]]+\]:\s+.*(?:\n(?: {2,}|\t).*)*/gm, "");
+  text = text.replace(/\[\^[^\]]+\]/g, "[注]");
+  return text;
+}
+
+function removeUnfriendlyExtensions(text: string): string {
+  text = text.replace(/~~(.*?)~~/g, "$1");
+  text = text.replace(/==(.*?)==/g, "$1");
+  text = text.replace(/(?<!~)~([^~\n]+)~(?!~)/g, "$1");
+  text = text.replace(/\^([^^\n]+)\^/g, "$1");
+
+  text = text.replace(
+    /```(?:mermaid|math|latex|tex|graphviz|plantuml)\n([\s\S]*?)```/gi,
+    "\n内容略（不支持的扩展块）\n",
+  );
+
+  return text;
+}
+
+function flattenDeepNesting(text: string): string {
+  const lines = text.split("\n");
+  const out: string[] = [];
+
+  for (let line of lines) {
+    if (/^\s{4,}[-*+]\s+/.test(line)) {
+      line = line.replace(/^\s+/, "  ");
+    }
+
+    if (/^\s*(>\s*){2,}/.test(line)) {
+      const content = line.replace(/^\s*(>\s*)+/, "");
+      line = `> ${content}`;
+    }
+
+    out.push(line);
+  }
+
+  return out.join("\n");
+}
+
+function normalizeTables(text: string): string {
+  // Pass 1: stitch broken table rows back together.
+  // Models may split a single row across multiple lines in several ways:
+  //   a) first part ends without |, continuation does NOT start with |
+  //   b) first part ends without |, blank line(s), continuation starts with |
+  //   c) first part ends without |, blank line(s), lone | on its own line
+  // We absorb any blank lines that follow an incomplete row and keep merging
+  // until the accumulated row ends with |.
+  const rawLines = text.split("\n");
+  const stitched: string[] = [];
+
+  for (let idx = 0; idx < rawLines.length; idx++) {
+    const line = rawLines[idx]!;
+    const trimmed = line.trim();
+    const prev = stitched[stitched.length - 1];
+    const prevTrim = prev !== undefined ? prev.trim() : "";
+    const prevIsIncomplete = prevTrim.startsWith("|") && !prevTrim.endsWith("|");
+
+    if (prevIsIncomplete) {
+      if (trimmed === "") {
+        // Blank line inside a broken row — absorb it and keep waiting for the rest
+        continue;
+      }
+      if (trimmed.includes("|")) {
+        // Continuation (starting with | or not) — stitch into the pending row
+        stitched[stitched.length - 1] =
+          prev! + (trimmed.startsWith("|") ? trimmed : line);
+        continue;
+      }
+    }
+
+    stitched.push(line);
+  }
+
+  // Pass 2: convert each table block to plain text lines.
+  const out: string[] = [];
+  let i = 0;
+
+  while (i < stitched.length) {
+    if (looksLikeTableRow(stitched[i]!)) {
+      const tableBlock: string[] = [stitched[i]!];
+      let j = i + 1;
+
+      while (j < stitched.length && looksLikeTableRow(stitched[j]!)) {
+        tableBlock.push(stitched[j]!);
+        j += 1;
+      }
+
+      if (out.length > 0 && out[out.length - 1]?.trim() !== "") {
+        out.push("");
+      }
+
+      out.push(...tableToPlainText(tableBlock));
+
+      if (j < stitched.length && stitched[j]?.trim() !== "") {
+        out.push("");
+      }
+
+      i = j;
+    } else {
+      out.push(stitched[i]!);
+      i += 1;
+    }
+  }
+
+  return out.join("\n");
+}
+
+function looksLikeTableRow(line: string): boolean {
+  const stripped = String(line).trim();
+  if (!stripped.startsWith("|")) return false;
+  return (stripped.match(/\|/g) || []).length >= 2;
+}
+
+function isTableSeparatorRow(row: string): boolean {
+  const inner = row.replace(/^\|/, "").replace(/\|$/, "");
+  return inner.split("|").every(c => /^[\s\-:]+$/.test(c) && c.includes("-"));
+}
+
+function extractTableCells(line: string): string[] {
+  const raw = line.trim();
+  const parts = raw.split("|");
+  const inner = raw.startsWith("|") ? parts.slice(1) : parts;
+  const cells = (raw.endsWith("|") ? inner.slice(0, -1) : inner).map(p => p.trim());
+  return cells.filter(c => c.length > 0);
+}
+
+function tableToPlainText(lines: string[]): string[] {
+  const result: string[] = [];
+  for (const line of lines) {
+    if (!line.trim()) continue;
+    if (isTableSeparatorRow(line.trim())) continue;
+    const cells = extractTableCells(line);
+    if (cells.length === 0) continue;
+    result.push(cells.join(" | "));
+  }
+  return result;
+}
+
+function cleanupWhitespace(text: string): string {
+  const lines = text.split("\n").map(line => {
+    let s = line.replace(/[ \t]+$/g, "");
+    s = s.replace(/[\x00-\x08\x0b\x0c\x0e-\x1f]/g, "");
+    return s;
+  });
+
+  text = lines.join("\n");
+  text = text.replace(/\n{3,}/g, "\n\n");
+  return text.trim();
+}
+
+function truncateSafely(text: string, maxLength: number): string {
+  if (text.length <= maxLength) return text;
+
+  const suffix = "\n\n（内容过长，已截断）";
+  const allowed = maxLength - suffix.length;
+  if (allowed <= 0) return text.slice(0, maxLength);
+
+  let truncated = text.slice(0, allowed);
+  const cut = truncated.lastIndexOf("\n");
+  if (cut > maxLength * 0.7) {
+    truncated = truncated.slice(0, cut);
+  }
+
+  return truncated.replace(/\s+$/g, "") + suffix;
+}