fix(uri_utils): unquote percent-encoded base64 payload in data URIs (#2481)

Signed-off-by: reginaldalfret <reginaldalfret@gmail.com>

Author: reginaldalfret

packages/markitdown/src/markitdown/_uri_utils.py

@@ -70,6 +70,6 @@ def parse_data_uri(uri: str) -> Tuple[str | None, Dict[str, str], bytes]:
         elif len(part) > 0:
             attributes[part.lower()] = ""
 
-    content = base64.b64decode(data) if is_base64 else unquote_to_bytes(data)
+    content = base64.b64decode(unquote_to_bytes(data)) if is_base64 else unquote_to_bytes(data)
 
     return mime_type, attributes, content

packages/markitdown/tests/test_module_misc.py

@@ -191,6 +191,19 @@ def test_data_uris() -> None:
     assert attributes["charset"] == "utf-8"
     assert data == b"Hello, World!"
 
+    # Test percent-encoded base64 payload (RFC 2397 compliance)
+    data_uri = "data:text/plain;base64,SGVsbG8%3D"
+    mime_type, attributes, data = parse_data_uri(data_uri)
+    assert mime_type == "text/plain"
+    assert len(attributes) == 0
+    assert data == b"Hello"
+
+    data_uri = "data:application/octet-stream;base64,%2B/8="
+    mime_type, attributes, data = parse_data_uri(data_uri)
+    assert mime_type == "application/octet-stream"
+    assert len(attributes) == 0
+    assert data == b"\xfb\xff"
+
 
 def test_file_uris() -> None:
     expected_path = os.path.abspath("/path/to/file.txt")