@zxh126/pi-qqbot 2.4.1 → 2.6.2

package/util/chunk.ts

@@ -1,42 +1,107 @@
-export function splitText(text: string, maxBytes = 1900): string[] {
-  if (Buffer.byteLength(text, "utf8") <= maxBytes) return text ? [text] : [];
-  const parts: string[] = [];
-  let current = "";
-  let currentBytes = 0;
-  for (const seg of text.split(/(?<=\n)/)) {
-    const segBytes = Buffer.byteLength(seg, "utf8");
-    if (segBytes > maxBytes) {
-      if (current) {
-        parts.push(current);
-        current = "";
-        currentBytes = 0;
-      }
-      let buf = "";
-      let bufBytes = 0;
-      for (const ch of seg) {
-        const chBytes = Buffer.byteLength(ch, "utf8");
-        if (bufBytes + chBytes > maxBytes) {
-          parts.push(buf);
-          buf = ch;
-          bufBytes = chBytes;
-        } else {
-          buf += ch;
-          bufBytes += chBytes;
-        }
-      }
-      current = buf;
-      currentBytes = bufBytes;
-      continue;
-    }
-    if (currentBytes + segBytes > maxBytes) {
-      parts.push(current);
-      current = seg;
-      currentBytes = segBytes;
-    } else {
-      current += seg;
-      currentBytes += segBytes;
-    }
-  }
-  if (current) parts.push(current);
-  return parts;
-}
+/**
+ * Markdown 文本切分器：QQ 单条消息有字节上限（默认 1900），需要按行切分，
+ * 并保持 GFM 表格、代码块的完整性（长表格不拦腰截断，否则 md 渲染出破表）。
+ */
+const CODE_FENCE = /^```/;
+const GFM_TABLE_LINE = /^\|.+\|$/;
+
+export function splitText(text: string, maxBytes = 1900): string[] {
+  if (Buffer.byteLength(text, "utf8") <= maxBytes) return text ? [text] : [];
+  const parts: string[] = [];
+  let current = "";
+  let currentBytes = 0;
+  let inCodeBlock = false;
+  let tableLines: string[] = [];
+
+  const flushCurrent = () => {
+    if (!current) return;
+    parts.push(current);
+    current = "";
+    currentBytes = 0;
+  };
+  /** 单行超过上限：按字符硬切（罕见，保持旧行为兜底） */
+  const pushCharSplit = (line: string) => {
+    let buf = "";
+    let bufBytes = 0;
+    for (const ch of line) {
+      const chBytes = Buffer.byteLength(ch, "utf8");
+      if (bufBytes + chBytes > maxBytes) {
+        parts.push(buf);
+        buf = ch;
+        bufBytes = chBytes;
+      } else {
+        buf += ch;
+        bufBytes += chBytes;
+      }
+    }
+    if (buf) parts.push(buf);
+  };
+  const appendLine = (line: string) => {
+    const lineBytes = Buffer.byteLength(line, "utf8");
+    if (!current) {
+      if (lineBytes > maxBytes) {
+        pushCharSplit(line);
+        return;
+      }
+      current = line;
+      currentBytes = lineBytes;
+      return;
+    }
+    const joinBytes = 1; // 拼接的换行符
+    if (currentBytes + joinBytes + lineBytes > maxBytes) {
+      flushCurrent();
+      if (lineBytes > maxBytes) {
+        pushCharSplit(line);
+        return;
+      }
+      current = line;
+      currentBytes = lineBytes;
+    } else {
+      current += "\n" + line;
+      currentBytes += joinBytes + lineBytes;
+    }
+  };
+  /** 表格整块提交：先尽力并入 current，放不下就整块另起一条消息 */
+  const flushTable = () => {
+    if (tableLines.length === 0) return;
+    const block = tableLines.join("\n");
+    tableLines = [];
+    const blockBytes = Buffer.byteLength(block, "utf8");
+    if (!current) {
+      current = block;
+      currentBytes = blockBytes;
+      return;
+    }
+    if (currentBytes + 1 + blockBytes > maxBytes) {
+      flushCurrent();
+      current = block;
+      currentBytes = blockBytes;
+    } else {
+      current += "\n" + block;
+      currentBytes += 1 + blockBytes;
+    }
+  };
+
+  for (const rawLine of text.split("\n")) {
+    const line = rawLine.endsWith("\r") ? rawLine.slice(0, -1) : rawLine;
+    if (CODE_FENCE.test(line)) {
+      flushTable();
+      inCodeBlock = !inCodeBlock;
+      appendLine(line);
+      continue;
+    }
+    if (inCodeBlock) {
+      appendLine(line);
+      continue;
+    }
+    if (GFM_TABLE_LINE.test(line)) {
+      tableLines.push(line);
+      continue;
+    }
+    flushTable();
+    appendLine(line);
+  }
+  flushTable();
+  if (current) parts.push(current);
+  return parts.length > 0 ? parts : [text];
+}