fix(md): keep inline HTML and autolinks inside GFM table cells

docling/backend/md_backend.py

@@ -132,6 +132,7 @@ class MarkdownDocumentBackend(DeclarativeDocumentBackend):
     _ENTITY_RE = re.compile(r"&(#\d+|#x[0-9a-fA-F]+|\w+);")
     _DELIMITER_CELL_RE = re.compile(r":?-+:?")
     _PIPE_ENTITY = "|"
+    _BR_TAG_RE = re.compile(r"<br[\s/>]", re.IGNORECASE)
 
     @staticmethod
     def _split_table_row(row: str) -> list[str]:
@@ -810,6 +811,18 @@ def _iterate_elements(  # noqa: C901
                     formatting=formatting,
                     hyperlink=hyperlink,
                 )
+
+        elif self.in_table and isinstance(
+            element, marko.inline.InlineHTML | marko.inline.AutoLink
+        ):
+            _log.debug(" - Inline HTML or autolink in a table: %s", element.children)
+            # The text inside a tag pair and the URL of an autolink arrive as
+            # RawText children, so only a <br> adds to the cell: a line break,
+            # as the HTML backend gives for <br> in a cell.
+            if isinstance(element, marko.inline.InlineHTML) and self._BR_TAG_RE.match(
+                str(element.children)
+            ):
+                self._append_table_text("\n")
         else:
             if not isinstance(element, str):
                 self._close_table(doc)

tests/test_backend_markdown.py

@@ -272,6 +272,35 @@ def test_convert_table_keeps_inline_code_spans():
     ]
 
 
+def test_convert_table_keeps_inline_html_and_autolinks():
+    """Inline HTML and an autolink inside a GFM table cell are part of that cell.
+
+    A <br> is a line break in the cell, the text inside other tags stays
+    without the tags, and an autolink keeps its URL.
+    """
+    markdown = """| Name<br>(unit) | Value |
+| --- | --- |
+| CO<sub>2</sub> | <kbd>Ctrl</kbd>+<kbd>C</kbd> |
+| a<br>b<br/>c<br />d | see <https://x.org> now |
+"""
+    conv_result = get_converter().convert_string(markdown, format=InputFormat.MD)
+    assert conv_result.status == ConversionStatus.SUCCESS
+
+    assert conv_result.document.texts == []
+    assert len(conv_result.document.tables) == 1
+    table_data = conv_result.document.tables[0].data
+    assert table_data.num_rows == 3
+    assert table_data.num_cols == 2
+    assert [cell.text for cell in table_data.table_cells] == [
+        "Name\n(unit)",
+        "Value",
+        "CO2",
+        "Ctrl+C",
+        "a\nb\nc\nd",
+        "see https://x.org now",
+    ]
+
+
 def test_convert_table_has_no_duplicate_cells():
     """
     Regression test: