Python type hint for Reader/Writer should probably be IO[bytes]

[naegelejd]

Currently, the Python BinaryProtocolReader accepts a binary input stream or a string filename:

class BinaryProtocolReader(ABC):
    def __init__(
        self,
        stream: Union[BufferedReader, BytesIO, BinaryIO, str],
        expected_schema: Optional[str],
    ) -> None:
    ....

However, Pyright 1.1.405 (2025-09-03) complains when the stream is of type IO[bytes], e.g.:

# Run mrd_phantom as subprocess and capture stdout
proc = subprocess.Popen(["mrd_phantom"], stdout=subprocess.PIPE)

if proc.stdout is None:
    raise RuntimeError("Failed to capture stdout")

with mrd.BinaryMrdReader(proc.stdout) as reader:    # <-- Pyright error
    ...

Pyright complains that

Argument of type "IO[bytes] | None" cannot be assigned to parameter "stream" of type "BufferedReader[_BufferedReaderStream] | BytesIO | BinaryIO | str" in function "__init__"
  Type "IO[bytes] | None" is not assignable to type "BufferedReader[_BufferedReaderStream] | BytesIO | BinaryIO | str"
    Type "IO[bytes]" is not assignable to type "BufferedReader[_BufferedReaderStream] | BytesIO | BinaryIO | str"
      "IO[bytes]" is not assignable to "BufferedReader[_BufferedReaderStream]"
      "IO[bytes]" is not assignable to "BytesIO"
      "IO[bytes]" is not assignable to "BinaryIO"
      "IO[bytes]" is not assignable to "str"Pylance[reportArgumentType](https://github.com/microsoft/pylance-release/blob/main/docs/diagnostics/reportArgumentType.md)

As I understand it, IO[bytes] is the generic ABC for all binary IO streams (source):

class typing.IO
class typing.TextIO
class typing.BinaryIO
Generic type IO[AnyStr] and its subclasses TextIO(IO[str]) and BinaryIO(IO[bytes]) represent the types of I/O streams such as returned by [open()](https://docs.python.org/3.8/library/functions.html#open).

This change appears to be recommended by type checkers like Ruff: astral-sh/ruff#15532

Similar changes may be needed for BinaryProtocolWriter, NDJsonProtocolReader, and NDJsonProtocolReader and they can also be tested using subprocess.PIPE.