to_dict returns wrong enum fields when numbering is not consecutive

[michael-sayapin]

Protobuf spec file like this:

syntax = "proto3";
package message;
message TeraMessage {
    int64 timestamp = 1;
    enum MessageType {
        MESSAGE_TYPE_UNKNOWN = 0;
        MESSAGE_TYPE_ACTION_MESSAGE = 1;
        // @exclude MESSAGE_TYPE_COMMAND_MESSAGE = 2; // DEPRECATED
        MESSAGE_TYPE_CONFIG_MESSAGE = 3;
        MESSAGE_TYPE_HEARTBEAT_MESSAGE = 4;
    }
    MessageType message_type = 5;
    bytes message = 6;
}

Generates the following Python bindings:

from dataclasses import dataclass
import betterproto

class TeraMessageMessageType(betterproto.Enum):
    MESSAGE_TYPE_UNKNOWN = 0
    MESSAGE_TYPE_ACTION_MESSAGE = 1
    # NB: notice that 2 is missing
    MESSAGE_TYPE_CONFIG_MESSAGE = 3
    MESSAGE_TYPE_HEARTBEAT_MESSAGE = 4

@dataclass
class TeraMessage(betterproto.Message):
    timestamp: int = betterproto.int64_field(1)
    message_type: "TeraMessageMessageType" = betterproto.enum_field(5)
    message: bytes = betterproto.bytes_field(6)

To reproduce the bug:

>>> from my.path.message import TeraMessage, TeraMessageMessageType
>>> message = TeraMessage(message_type=TeraMessageMessageType.MESSAGE_TYPE_CONFIG_MESSAGE)
>>> message.to_dict()
{'messageType': 'MESSAGE_TYPE_HEARTBEAT_MESSAGE'}
>>> message.to_json()
'{"messageType": "MESSAGE_TYPE_HEARTBEAT_MESSAGE"}'
>>> TeraMessage().parse(bytes(message)).message_type == TeraMessageMessageType.MESSAGE_TYPE_CONFIG_MESSAGE
True

[boukeversteegh]

Awesome bugreport! Thanks!

Would you be willing to add this test-case to our standard tests?

You basically only need to:

• add a folder with that proto file in it (and call the message Test)
• plus a json with the same name containing the faulty value
• add the name of the folder to betterproto/tests/inputs/config.py in xfail

Details are described here: Standard Tests Development Guide

[michael-sayapin]

Sure @boukeversteegh ! Created a PR https://github.com/danielgtaylor/python-betterproto/pull/94