fix: definition-list直前のテキストをparagraphに包まない

Wikidotの挙動に合わせ、definition-list直前のテキスト行を
paragraphコンテナに包まず裸テキスト+line-breakとして出力。
normalizeHtmlのタグ前後改行処理も改善。

Author: r74tech

packages/parser/src/parser/rules/block/paragraph.ts

@@ -100,6 +100,18 @@ export const paragraphRule: BlockRule = {
       return { success: false };
     }
 
+    // Wikidot: text lines immediately before a definition list are not
+    // wrapped in <p>. Check if next token starts a definition list.
+    const nextPos = ctx.pos + result.consumed;
+    const nextToken = ctx.tokens[nextPos];
+    if (nextToken?.type === "COLON" && nextToken.lineStart) {
+      return {
+        success: true,
+        elements: [...elements, { element: "line-break" }],
+        consumed: result.consumed,
+      };
+    }
+
     return {
       success: true,
       elements: [

tests/fixtures/definition-list/basic/expected.json

@@ -1,17 +1,11 @@
 {
   "elements": [
     {
-      "element": "container",
-      "data": {
-        "type": "paragraph",
-        "attributes": {},
-        "elements": [
-          {
-            "element": "text",
-            "data": "ALPHA"
-          }
-        ]
-      }
+      "element": "text",
+      "data": "ALPHA"
+    },
+    {
+      "element": "line-break"
     },
     {
       "element": "definition-list",
@@ -75,17 +69,11 @@
       ]
     },
     {
-      "element": "container",
-      "data": {
-        "type": "paragraph",
-        "attributes": {},
-        "elements": [
-          {
-            "element": "text",
-            "data": "BETA"
-          }
-        ]
-      }
+      "element": "text",
+      "data": "BETA"
+    },
+    {
+      "element": "line-break"
     },
     {
       "element": "definition-list",

tests/integration/fixture-render.test.ts

@@ -74,8 +74,11 @@ function normalizeHtml(html: string): string {
       .replace(/\r\n/g, "\n")
       // onclick属性を削除（Wikidot固有のJS）
       .replace(/ onclick="[^"]*"/g, "")
-      // 連続する空白・改行を単一スペースに（HTML的に等価）
-      .replace(/\s+/g, " ")
+      // タグ前後の改行を削除（ブロック要素の前後の改行はHTML的に無意味）
+      .replace(/\n\s*</g, "<")
+      .replace(/>\s*\n/g, ">")
+      // 残りの連続空白を単一スペースに（HTML的に等価）
+      .replace(/[ \t]+/g, " ")
       // 属性順序を正規化（HTML的に等価）
       .replace(
         /<(\w+)((?:\s+[a-zA-Z_][\w-]*(?:="[^"]*")?)+)\s*(\/?)>/g,