Escape pipes and newlines in CSV to Markdown table cells

packages/markitdown/src/markitdown/converters/_csv_converter.py

@@ -12,6 +12,14 @@
 ACCEPTED_FILE_EXTENSIONS = [".csv"]
 
 
+def _escape_cell(value: str) -> str:
+    # A literal pipe or newline inside a cell would split it into extra
+    # columns or rows when rendered, so escape pipes and flatten newlines.
+    return value.replace("|", "\\|").replace("\r\n", " ").replace("\n", " ").replace(
+        "\r", " "
+    )
+
+
 class CsvConverter(DocumentConverter):
     """
     Converts CSV files to Markdown tables.
@@ -58,7 +66,9 @@ def convert(
         markdown_table = []
 
         # Add header row
-        markdown_table.append("| " + " | ".join(rows[0]) + " |")
+        markdown_table.append(
+            "| " + " | ".join(_escape_cell(cell) for cell in rows[0]) + " |"
+        )
 
         # Add separator row
         markdown_table.append("| " + " | ".join(["---"] * len(rows[0])) + " |")
@@ -70,7 +80,9 @@ def convert(
                 row.append("")
             # Truncate if row has more columns than header
             row = row[: len(rows[0])]
-            markdown_table.append("| " + " | ".join(row) + " |")
+            markdown_table.append(
+                "| " + " | ".join(_escape_cell(cell) for cell in row) + " |"
+            )
 
         result = "\n".join(markdown_table)
 

packages/markitdown/tests/test_csv_converter.py

@@ -0,0 +1,29 @@
+import io
+
+from markitdown.converters._csv_converter import CsvConverter
+from markitdown._stream_info import StreamInfo
+
+CSV_STREAM_INFO = StreamInfo(extension=".csv", mimetype="text/csv")
+
+
+def _convert(data: bytes) -> str:
+    return CsvConverter().convert(io.BytesIO(data), CSV_STREAM_INFO).markdown
+
+
+def test_pipe_in_cell_is_escaped():
+    markdown = _convert(b"name,note\nAlice,a|b\nBob,plain\n")
+    lines = markdown.splitlines()
+    # Header, separator and two data rows.
+    assert len(lines) == 4
+    # Every row must have the same number of unescaped column separators.
+    expected_pipes = lines[0].count("|")
+    for line in lines:
+        assert line.replace("\\|", "").count("|") == expected_pipes
+    assert "a\\|b" in markdown
+
+
+def test_newline_in_cell_does_not_break_row():
+    markdown = _convert(b'name,note\n"Bob","line1\nline2"\n')
+    lines = markdown.splitlines()
+    assert len(lines) == 3
+    assert "line1 line2" in lines[2]