AttributeError: 'bool' object has no attribute 'lower'

[ZhymabekRoman]

Hello @s3rius! Thank you for creating taskiq.

I encountered an exception (something like AttributeError: 'bool' object has no attribute 'lower') with taskiq when attempting to use my custom SimpleRetryMiddleware. The problem occurred with the following code:

@taskiq_broker.task(retry_on_error=True)
async def _process_not_reached_deal(deal_details: BitrixDealResultData) -> Optional[bool]:

Unfortunately, I don't have the full traceback of the exception to provide for troubleshooting.

I've implemented a fix, and the code is now working correctly:

_LABEL_PARSERS: Dict[LabelType, Callable[[str], Any]] = {
    LabelType.INT: int,
    LabelType.STR: str,
    LabelType.FLOAT: float,
    LabelType.BOOL: lambda x: str(x).lower() == "true",
    LabelType.BYTES: base64.b64decode,
    LabelType.ANY: lambda x: x,
}

However, I'm not entirely sure if the boolean value conversion logic (lambda x: str(x).lower() == "true") is the most appropriate approach. I'd appreciate your thoughts on this implementation.

[s3rius]

Thanks for finding it out and making a fix. Actually that case is quite interesting, because labels should always be strings during transmission and serialization. Can you please provide with more details of your setup?

Also, I guess this setup will work, because for strings the str will do nothing, but for actual booleans it will be actually doing everything correctly. Can you publish a PR? Or I can do it myself.

[ZhymabekRoman]

I have been unable to replicate the problem when using a minimal code example. Could the issue be related to either the taskiq-litestar or taskiq-pipelines integration library?

[s3rius]

It actually should not.

[ZhymabekRoman]

I suppose problem might be somewhere here:

Custom retry middleware with task scheduling logic with delay execution

from datetime import datetime, timedelta
from logging import getLogger
from typing import Any

from taskiq.abc.middleware import TaskiqMiddleware
from taskiq.abc.schedule_source import ScheduleSource
from taskiq.exceptions import NoResultError
from taskiq.kicker import AsyncKicker
from taskiq.message import TaskiqMessage
from taskiq.result import TaskiqResult

logger = getLogger("taskiq.retry_middleware")


class SimpleRetryMiddleware(TaskiqMiddleware):
  """Middleware to add retries."""

  def __init__(
      self,
      default_retry_count: int = 3,
      default_retry_label: bool = False,
      no_result_on_retry: bool = True,
      retry_delay: int = 60,
  ) -> None:
      self.schedule_source = None
      self.default_retry_count = default_retry_count
      self.default_retry_label = default_retry_label
      self.no_result_on_retry = no_result_on_retry
      self.retry_delay = retry_delay

  def set_schedule_source(self, source: ScheduleSource) -> None:
      self.schedule_source = source

  async def on_error(
      self,
      message: "TaskiqMessage",
      result: "TaskiqResult[Any]",
      exception: BaseException,
  ) -> None:
      """Retry on error.

      This middleware is used to retry
      tasks on errors.

      If error is found during the execution
      this function is invoked.

      :param message: Message that caused the error.
      :param result: execution result.
      :param exception: found exception.
      """
      # Valid exception
      if isinstance(exception, NoResultError):
          return

      retry_on_error = message.labels.get("retry_on_error")
      if isinstance(retry_on_error, str):
          retry_on_error = retry_on_error.lower() == "true"

      retry_delay: int = int(message.labels.get("retry_delay", self.retry_delay))

      if retry_on_error is None:
          retry_on_error = self.default_retry_label
      # Check if retrying is enabled for the task.
      if not retry_on_error:
          return

      # Didn't workd out
      # schedule_label = datetime.datetime.utcnow() + datetime.timedelta(
      #     seconds=self.retry_timeout
      # )
      # message.labels["time"] = schedule_label

      kicker: AsyncKicker[Any, Any] = AsyncKicker(
          task_name=message.task_name,
          broker=self.broker,
          labels=message.labels,
      ).with_task_id(message.task_id)

      # Getting number of previous retries.
      retries = int(message.labels.get("_retries", 0)) + 1
      kicker.with_labels(_retries=retries)
      max_retries = int(message.labels.get("max_retries", self.default_retry_count))

      if retries < max_retries:
          schedule_time = datetime.now() + timedelta(seconds=retry_delay)
          logger.info(
              f"Task '{message.task_name}' invocation failed. Retrying in {retry_delay} seconds. Retry {retries}/{max_retries}, schedule time: {schedule_time}",  # noqa: E501
          )

          logger.info(f"Schedule source: {self.schedule_source}")
          if self.schedule_source is None:
              logger.error("Schedule source or retry delay is not set")
              await kicker.kiq(
                  *message.args,
                  **message.kwargs,
              )
              return

          await kicker.schedule_by_time(
              self.schedule_source,
              schedule_time,
              *message.args,
              **message.kwargs,
          )

          if self.no_result_on_retry:
              result.error = NoResultError()

      else:
          logger.warning(
              "Task '%s' invocation failed. Maximum retries count is reached.",
              message.task_name,
          )

I suppose the problem that I didn't use prepare_label somewhere here :/

Additional: taskiq broker initialize logic

from typing import Any

from taskiq import TaskiqScheduler
from taskiq.schedule_sources import LabelScheduleSource

# from taskiq_nats import NatsBroker
from taskiq_pipelines.middleware import PipelineMiddleware
from taskiq_redis import ListQueueBroker, RedisAsyncResultBackend, RedisScheduleSource

from branched_call_bot.config import cnf
from branched_call_bot.services.taskiq import SimpleRetryMiddleware

# from .redis import async_redis_pool

redis_async_result: RedisAsyncResultBackend[Any] = RedisAsyncResultBackend(
    redis_url=cnf.REDIS.url,
)

retry_middleware = SimpleRetryMiddleware(
    default_retry_count=cnf.TASKIQ.RETRY_COUNT,
    retry_delay=cnf.TASKIQ.RETRY_DELAY,
)

pipeline_middleware = PipelineMiddleware()
# nats_broker = NatsBroker(
#         servers=cnf.NATS.SERVERS,
#         queue=cnf.NATS.QUEUE,
#     )

redis_broker = ListQueueBroker(cnf.REDIS.url)

taskiq_broker = redis_broker.with_middlewares(
    pipeline_middleware,
    retry_middleware,
).with_result_backend(redis_async_result)


label_scheduler_source = LabelScheduleSource(taskiq_broker)
redis_scheduler_source = RedisScheduleSource(url=cnf.REDIS.url)

retry_middleware.set_schedule_source(redis_scheduler_source)  # type: ignore

taskiq_scheduler = TaskiqScheduler(
    broker=taskiq_broker,
    sources=[redis_scheduler_source, label_scheduler_source],
)

# taskiq_litestar.init(
#     broker=taskiq_broker,
#     app="branched_call_bot.clients.litestar:app",
# )