Container types are inconsistent even within the same patch of Python protobuf

[lazarillo]

What version of protobuf and what language are you using?

Protobuf version: 4.25.3
Python version: 3.11

What operating system (Linux, Windows, ...) and version?

[local] Linux and [remote] Google Composer (Airflow) environment with image: composer-2.6.6-airflow-2.7.3

More details on this below.

What runtime / compiler are you using (e.g., python version or gcc version)

Python 3.11

What did you do?

When I create a protobuf message that contains a map type, for instance:

syntax = "proto3";

message OrderContract {
  map<string, string> custom = 1;
}

I am seeing that when working with it in Python, on my local machine (which is Ubuntu linux, also using Python 3.11, more details below) the custom element is registered as a google._upb._message.ScalarMapContainer, but on the remote Composer (Airflow) environment, the custom element is registered as a google.protobuf.internal.containers.ScalarMap.

Both of these environments are as exactly matched as possible: both are unix, both are using protobuf version 4.25.3, and both are using Python 3.11.

Problem

I need to perform introspection: I am creating some fake data, generating it using the faker library when I have hit a primitive element (str, int, etc), but introspecting further to get to a lower leaf when I have a protobuf Message, ScalarMapContainer, or some other protobuf (non-primitive) type.

It is difficult to debug this introspection since there is inconsistency even across the same Major.Minor.Patch.

I understand that these are supposed to be internals, but with no way to perform introspection, I am forced to work this way. I simply need to understand when and why google._upb._message... is used over google.protobuf.internal.containers.... I read that this happened with the change to protobuf 4.*, but this is clearly not the case, since I have the exact same version in both cases, and they are using the two different underlying libraries.

Anything else we should know about your project / environment

My linux machine details:

> uname -a
Linux <my-account> 6.5.0-26-generic #26~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Tue Mar 12 10:22:43 UTC 2 x86_64 x86_64 x86_64 GNU/Linux`

As mentioned above, the remote machine is a Google Cloud Composer image, version composer-2.6.6-airflow-2.7.3. It's details can be found here.

---

Here is an output of python -m pip show protobuf across the two environments:

My local:

Name: protobuf
Version: 4.25.3
Summary: 
Home-page: https://developers.google.com/protocol-buffers/
Author: protobuf@googlegroups.com
Author-email: protobuf@googlegroups.com
License: 3-Clause BSD License
Location: /home/<my-account>/.local/lib/python3.11/site-packages
Requires: 
.
.
.

Cloud Composer:

Name: protobuf
Version: 4.25.3
Summary: 
Home-page: https://developers.google.com/protocol-buffers/
Author: protobuf@googlegroups.com
Author-email: protobuf@googlegroups.com
License: 3-Clause BSD License
Location: /opt/python3.11/lib/python3.11/site-packages
Requires: 
.
.
.