[CT-1886] Include adapter_response in RunResultMsg

[jtcohen6]

adapter_response is being skipped during message serialization for execution results:

$ dbt --debug --log-format json run | jq .data.run_result.adapter_response
...
{}
...

dbt-core/core/dbt/contracts/results.py

Lines 132 to 142 in 7b464b8

	def to_msg(self):
	# TODO: add more fields
	msg = RunResultMsg()
	msg.status = str(self.status)
	msg.message = cast_to_str(self.message)
	msg.thread = self.thread_id
	msg.execution_time = self.execution_time
	msg.num_failures = cast_to_int(self.failures)
	msg.timing_info = [ti.to_msg() for ti in self.timing]
	# adapter_response
	return msg

Reproduction case

I drop a breakpoint in here:

dbt-core/core/dbt/task/runnable.py

Lines 232 to 237 in 7b464b8

	fire_event(
	NodeFinished(
	node_info=runner.node.node_info,
	run_result=result.to_msg(),
	)
	)

ipdb> result.adapter_response
{'_message': 'CREATE VIEW (0 processed)', 'code': 'CREATE VIEW', 'bytes_processed': 0, 'location': 'US', 'project_id': 'dbt-test-env', 'job_id': 'b70d9a54-38a6-4e21-ad07-d08e75ace428', 'slot_ms': 0}
ipdb> result.to_msg()
RunResultMsg(status='success', message='CREATE VIEW (0 processed)', timing_info=[TimingInfoMsg(name='compile', started_at=datetime.datetime(2023, 1, 24, 10, 38, 9, 423443), completed_at=datetime.datetime(2023, 1, 24, 10, 38, 9, 426940)), TimingInfoMsg(name='execute', started_at=datetime.datetime(2023, 1, 24, 10, 38, 9, 427426), completed_at=datetime.datetime(2023, 1, 24, 10, 38, 10, 391500))], thread='Thread-1 (worker)', execution_time=0.9696011543273926, adapter_response={}, num_failures=0)
ipdb> result.to_msg().adapter_response
{}

Previous behavior

This does represent a functional regression in v1.4, relative to v1.3, when we used different serialization (mashumaro) for the result:

$ dbt --debug --log-format json run | jq .data.run_result.adapter_response
...
{
  "_message": "CREATE VIEW (0 processed)",
  "bytes_processed": 0,
  "code": "CREATE VIEW",
  "job_id": "9180b859-944f-4f91-a559-d772e0697c6d",
  "location": "US",
  "project_id": "dbt-test-env",
  "slot_ms": 0
}
...

Same breakpoint as above:

ipdb> result.to_dict()['adapter_response']
{'_message': 'CREATE VIEW', 'code': 'CREATE VIEW', 'rows_affected': -1}

[jtcohen6]

I'm guessing this is because our proto representation of adapter_response can only be a dictionary with string keys + string values (given limited proto type system):

https://github.com/dbt-labs/dbt-core/blob/main/core/dbt/events/types.proto#L48

dbt-core/core/dbt/events/proto_types.py

Lines 64 to 66 in 7b464b8

	adapter_response: Dict[str, str] = betterproto.map_field(
	6, betterproto.TYPE_STRING, betterproto.TYPE_STRING
	)

Still, this would be better than not including the data at all:

     msg.timing_info = [ti.to_msg() for ti in self.timing] 
     msg.adapter_response = {k: str(v) for k, v in self.adapter_response.items()}
     return msg 

$ dbt --debug --log-format json run | jq .data.run_result.adapter_response
...
{
  "_message": "CREATE VIEW (0 processed)",
  "bytes_processed": "0",
  "code": "CREATE VIEW",
  "job_id": "10588bfb-7857-432a-950e-926e10c00a9e",
  "location": "US",
  "project_id": "dbt-test-env",
  "slot_ms": "0"
}
...