fix(markdown): preserve wrapped table breaks and separate file links

Use the HTML-disabled Markdown pass to find tables exposed by wrapper normalization. Treat inline line separators as file-link boundaries so adjacent paths remain independently usable.

packages/app/src/components/markdown/html.browser.test.tsx

@@ -7,33 +7,36 @@ import { normalizeMarkdownHtml } from "./html";
 
 const FLAGS = { hardSoftBreaks: true, latexMath: false };
 
-it("renders HTML breaks on separate lines inside the same table cell", async () => {
-  const container = document.createElement("div");
-  document.body.appendChild(container);
-  const root = createRoot(container);
-  try {
-    act(() =>
-      root.render(
-        <EnrichedMarkdownText
-          markdown={normalizeMarkdownHtml(
-            "| A | B |\n|---|---|\n| 1 | first<br/>second |\n| 2 | first<br />second |\n| 3 | first<br>second |",
-          )}
-          md4cFlags={FLAGS}
-        />,
-      ),
-    );
-    await waitFor(() => expect(within(container).getAllByRole("row")).toHaveLength(4));
-    for (const row of within(container).getAllByRole("row").slice(1)) {
-      const cell = within(row).getAllByRole("cell")[1];
-      expect(cell.innerText).toBe("first\nsecond");
-      const first = document.createRange();
-      first.selectNodeContents(cell.firstChild!);
-      const second = document.createRange();
-      second.selectNodeContents(cell.lastChild!);
-      expect(second.getBoundingClientRect().top).toBeGreaterThan(first.getBoundingClientRect().top);
+it.each([false, true])(
+  "renders table-cell breaks on separate lines (details: %s)",
+  async (details) => {
+    const container = document.createElement("div");
+    document.body.appendChild(container);
+    const root = createRoot(container);
+    const table =
+      "| A | B |\n|---|---|\n| 1 | first<br/>second |\n| 2 | first<br />second |\n| 3 | first<br>second |";
+    const source = details ? `<details><summary>Results</summary>\n${table}\n</details>` : table;
+    try {
+      act(() =>
+        root.render(
+          <EnrichedMarkdownText markdown={normalizeMarkdownHtml(source)} md4cFlags={FLAGS} />,
+        ),
+      );
+      await waitFor(() => expect(within(container).getAllByRole("row")).toHaveLength(4));
+      for (const row of within(container).getAllByRole("row").slice(1)) {
+        const cell = within(row).getAllByRole("cell")[1];
+        expect(cell.innerText).toBe("first\nsecond");
+        const first = document.createRange();
+        first.selectNodeContents(cell.firstChild!);
+        const second = document.createRange();
+        second.selectNodeContents(cell.lastChild!);
+        expect(second.getBoundingClientRect().top).toBeGreaterThan(
+          first.getBoundingClientRect().top,
+        );
+      }
+    } finally {
+      act(() => root.unmount());
+      container.remove();
     }
-  } finally {
-    act(() => root.unmount());
-    container.remove();
-  }
-});
+  },
+);

packages/app/src/components/markdown/html.ts

@@ -38,9 +38,6 @@ export function normalizeMarkdownHtml(source: string): string {
   const htmlRanges: { start: number; end: number }[] = [];
   const tableRanges: { start: number; end: number }[] = [];
   function findHtml(node: Nodes) {
-    if (node.type === "table" && node.position) {
-      tableRanges.push({ start: node.position.start.offset!, end: node.position.end.offset! });
-    }
     if (node.type === "html" && node.position) {
       htmlRanges.push({ start: node.position.start.offset!, end: node.position.end.offset! });
     } else if ("children" in node) node.children.forEach(findHtml);
@@ -52,18 +49,22 @@ export function normalizeMarkdownHtml(source: string): string {
     }),
   );
   const codeRanges: { start: number; end: number }[] = [];
-  // Disable HTML parsing so code inside <details> remains visible to the Markdown parser.
+  // Disable HTML parsing so code and tables inside <details> remain visible.
   const tree = fromMarkdown(source, {
-    extensions: [{ disable: { null: ["htmlFlow", "htmlText"] } }],
+    extensions: [{ disable: { null: ["htmlFlow", "htmlText"] } }, gfmTable()],
+    mdastExtensions: [gfmTableFromMarkdown()],
   });
-  function protectCode(node: Nodes) {
+  function findMarkdownRanges(node: Nodes) {
+    if (node.type === "table" && node.position) {
+      tableRanges.push({ start: node.position.start.offset!, end: node.position.end.offset! });
+    }
     if (node.type === "code" || node.type === "inlineCode") {
       const start = node.position?.start.offset;
       const end = node.position?.end.offset;
       if (start !== undefined && end !== undefined) codeRanges.push({ start, end });
-    } else if ("children" in node) node.children.forEach(protectCode);
+    } else if ("children" in node) node.children.forEach(findMarkdownRanges);
   }
-  protectCode(tree);
+  findMarkdownRanges(tree);
   const links: (string | undefined)[] = [];
   let previousTag: { name: string; end: number } | null = null;
   const convert = (raw: string, offset: number): string => {

packages/app/src/components/markdown/renderer.test.ts

@@ -62,6 +62,43 @@ describe("shared Markdown source adaptation", () => {
       ["3", "first\u2028second"],
     ]);
   });
+  it("keeps table breaks inside a details wrapper without a blank line", async () => {
+    const source =
+      "<details><summary>Results</summary>\n| A | B |\n|---|---|\n| 1 | first<br>second |\n</details>";
+    const tree = await parseMarkdown(prepareMarkdownSource(source).render(), {
+      hardSoftBreaks: true,
+      latexMath: false,
+    });
+    const tables: ASTNode[] = [];
+    const collect = (node: ASTNode) => {
+      if (node.type === "Table") tables.push(node);
+      node.children?.forEach(collect);
+    };
+    collect(tree);
+    expect(tables).toHaveLength(1);
+    expect(tables[0].attributes?.bodyRowCount).toBe("1");
+    expect(tables[0].children?.[1]?.children?.[0]?.children?.[1]?.children).toEqual([
+      { type: "Text", content: "first\u2028second" },
+    ]);
+  });
+  it("recognizes file links separately on either side of a table break", async () => {
+    const prepared = prepareMarkdownSource(
+      "| A | B |\n|---|---|\n| 1 | src/first.ts<br>src/second.ts |",
+      "/repo",
+    );
+    const tree = recognizeTextLinks(
+      await parseMarkdown(prepared.render(), { hardSoftBreaks: true, latexMath: false }),
+      prepared.linkRecognition.text,
+      prepared.linkRecognition.inlineCode,
+    );
+    const links: string[] = [];
+    const collect = (node: ASTNode) => {
+      if (node.type === "Link") links.push(String(node.attributes?.url));
+      node.children?.forEach(collect);
+    };
+    collect(tree);
+    expect(links).toEqual(["src/first.ts", "src/second.ts"]);
+  });
   it.each([
     ["| A<br>B | C |\n|---|---|\n| x | y |", "| A\u2028B | C |\n|---|---|\n| x | y |"],
     [

packages/app/src/components/markdown/source.ts

@@ -10,9 +10,10 @@ import type { Link, Nodes, Paragraph, PhrasingContent, Root } from "mdast";
 
 // Keep this regex in the syntax shared by JavaScript, ICU and java.util.regex.
 // The Unicode flag makes Metro expand it to escapes rejected by native engines.
-const PATH_CHARACTER = String.raw`[a-zA-Z0-9\u0080-\uFFFF_@.%+-]`;
+// U+2028 is an inline line break, including between table-cell file links.
+const PATH_CHARACTER = String.raw`[a-zA-Z0-9\u0080-\u2027\u2029-\uFFFF_@.%+-]`;
 const PATH_LETTER = String.raw`[a-zA-Z]`;
-const PATH_SEGMENT = String.raw`${PATH_CHARACTER}*[a-zA-Z0-9\u0080-\uFFFF_@%+-]`;
+const PATH_SEGMENT = String.raw`${PATH_CHARACTER}*[a-zA-Z0-9\u0080-\u2027\u2029-\uFFFF_@%+-]`;
 const LINE_SUFFIX = String.raw`(?::\d+(?::\d+)?(?:-\d+(?::\d+)?)?)?`;
 const FILE_EXTENSIONS = [...ASSISTANT_FILE_EXTENSIONS]
   .map((extension) => extension.replace(/[a-z]/g, (letter) => `[${letter}${letter.toUpperCase()}]`))
@@ -29,7 +30,7 @@ function createTextPattern(path: string, forgeReferencePattern?: string): RegExp
   const forgeReference = forgeReferencePattern ? `|${forgeReferencePattern}` : "";
   // The native adapter mistakes a literal + in lookbehind for a quantifier; use its hex escape.
   return new RegExp(
-    String.raw`(?:${BARE_URL}${forgeReference}|(?<![a-zA-Z0-9\u0080-\uFFFF_@.%\x2b~:/\\-])(?:${path}|${FILE_NAME})${LINE_SUFFIX}(?![a-zA-Z0-9\u0080-\uFFFF_@%+:/\\-]|\.${PATH_CHARACTER}))`,
+    String.raw`(?:${BARE_URL}${forgeReference}|(?<![a-zA-Z0-9\u0080-\u2027\u2029-\uFFFF_@.%\x2b~:/\\-])(?:${path}|${FILE_NAME})${LINE_SUFFIX}(?![a-zA-Z0-9\u0080-\u2027\u2029-\uFFFF_@%+:/\\-]|\.${PATH_CHARACTER}))`,
     "g",
   );
 }