Unable to use request body with lazy listener when socket mode is enabled (v1.11.0)

[JordanGibson]

Reproducible in:

slack-bolt==1.11.0
slack-sdk==3.13.0
Python 3.8.8
ProductName: macOS
ProductVersion: 12.0
BuildVersion: 21A5506j
Darwin Kernel Version 21.1.0: Thu Aug 19 02:54:44 PDT 2021; root:xnu-8019.40.29~26/RELEASE_ARM64_T8101

The slack_bolt version

slack-bolt==1.11.0
slack-sdk==3.13.0

Python runtime version

Python 3.8.8

OS info

ProductName: macOS
ProductVersion: 12.0
BuildVersion: 21A5506j
Darwin Kernel Version 21.1.0: Thu Aug 19 02:54:44 PDT 2021; root:xnu-8019.40.29~26/RELEASE_ARM64_T8101

Steps to reproduce:

(Share the commands to run, source code, and project settings (e.g., setup.py))

1. Begin with the starter project for bolt-python in socket mode
2. Enabling socket mode on the app
3. Add functionality for FaaS (or at least, mock it) following the documentation
4. Execute command /start-process sampletext

Expected result:

For the functions respond_to_slack_within_3_seconds and run_long_process to be called same with the body parameter for both calls.

Actual result:

The run_long_process body is empty ({}), and the respond_to_slack_within_3_seconds body is populated. This is evident as when I log the body on first execution of each call to both functions, the following is the output of the application (sensitive information redacted).

⚡️ Bolt app is running!
{'token': '***REDACTED***', 'team_id': '***REDACTED***', 'team_domain': '***REDACTED***', 'channel_id': '***REDACTED***', 'channel_name': '***REDACTED***', 'user_id': '***REDACTED***', 'user_name': '***REDACTED***', 'command': '/start-process', 'text': 'test', 'api_app_id': '***REDACTED***', 'is_enterprise_install': '***REDACTED***', 'response_url': ***REDACTED***', 'trigger_id': '***REDACTED***'}
{}
Failed to run an internal function ('text')

I do receive an Accepted response from respond_to_slack_within_3_seconds as it completes successfully, but I do not receive any messages from the run_long_process call as it fails to complete due to the body being empty.

For futher information, the respond parameter in run_long_process is assigned. For example, I removed the {body['text']} interpolation and it worked, sending the message Completed! (task: ) (task empty as I removed the replacement).

When in HTTP mode, it works fine (when changing the SocketModeHandler(app, ...).start() to app.start(.....))

app.py to reproduce

import os
import time

from slack_bolt import App

from slack_bolt.adapter.socket_mode import SocketModeHandler

app = App(token=os.environ.get("SLACK_BOT_TOKEN"), process_before_response=True)


def respond_to_slack_within_3_seconds(body, ack):
    text = body.get("text")
    if text is None or len(text) == 0:
        ack(":x: Usage: /start-process (description here)")
    else:
        ack(f"Accepted! (task: {body['text']})")


def run_long_process(respond, body):
    time.sleep(5)  # longer than 3 seconds
    respond(f"Completed! (task: {body['text']})")


app.command("/start-process")(
    ack=respond_to_slack_within_3_seconds,  # responsible for calling `ack()`
    lazy=[run_long_process]  # unable to call `ack()` / can have multiple functions
)

if __name__ == "__main__":
    SocketModeHandler(app, os.environ["SLACK_APP_TOKEN"]).start()

[JordanGibson]

Update

After some investigation, I have been able to determine the root cause of this.

The culprit of this is how we create request objects when in socket mode (see slack_bolt/request/request.py:50). Currently, if the body is created as a dict instance, then we set the body to empty because we are unable to guarantee that order of the body will be perserved if we were to convert it to a str. As of Python 3.6+, dicts are ordered by insertion order.

[seratch]

Hi @JordanGibson, thanks for taking the time to report this issue!

I've checked this in detail and figured out the issue is a regression bug caused by #546 since v1.11.0 (the latest version). I will come up with a pull request to fix this and will release a patch version shortly.