Cannot import mlflow

[thirtyseven]

🐞 bug report

When trying to use external pip dependency mlflow, errors are thrown when code contains:

import mlflow

Is this a regression?

Not sure, I encountered this error while trying to migrate a repository from WORKSPACE rules_python to bzlmod with a test that uses mlflow. The test passes under WORKSPACE but fails under bzlmod. However, when I tried to recreate the issue in a minimal repro, the test fails under both WORKSPACE and bzlmod.

Description

When a py_test or py_binary contains:

import mlflow

the following error is thrown when running the code with bazel run or bazel test:

ImportError: cannot import name 'Resource' from 'mlflow.models.resources' (/private/var/tmp/_bazel_tkaplan/5b13b869a536c448538040ea91145cae/sandbox/darwin-sandbox/1/execroot/_main/bazel-out/darwin_arm64-fastbuild/bin/test.runfiles/rules_python++pip+pypi_312_mlflow_skinny/site-packages/mlflow/models/resources/__init__.py)

🔬 Minimal Reproduction

https://github.com/thirtyseven/mlflow-repro

bazel test :test

🔥 Exception or Error

exec ${PAGER:-/usr/bin/less} "$0" || exit 1
Executing tests from //:test
-----------------------------------------------------------------------------
Traceback (most recent call last):
  File "/private/var/tmp/_bazel_tkaplan/5b13b869a536c448538040ea91145cae/sandbox/darwin-sandbox/1/execroot/_main/bazel-out/darwin_arm64-fastbuild/bin/test.runfiles/_main/test.py", line 1, in 
    import mlflow
  File "/private/var/tmp/_bazel_tkaplan/5b13b869a536c448538040ea91145cae/sandbox/darwin-sandbox/1/execroot/_main/bazel-out/darwin_arm64-fastbuild/bin/test.runfiles/rules_python++pip+pypi_312_mlflow_skinny/site-packages/mlflow/__init__.py", line 35, in 
    from mlflow import (
  File "/private/var/tmp/_bazel_tkaplan/5b13b869a536c448538040ea91145cae/sandbox/darwin-sandbox/1/execroot/_main/bazel-out/darwin_arm64-fastbuild/bin/test.runfiles/rules_python++pip+pypi_312_mlflow_skinny/site-packages/mlflow/models/__init__.py", line 46, in 
    from mlflow.models.model import Model, get_model_info, set_model, update_model_requirements
  File "/private/var/tmp/_bazel_tkaplan/5b13b869a536c448538040ea91145cae/sandbox/darwin-sandbox/1/execroot/_main/bazel-out/darwin_arm64-fastbuild/bin/test.runfiles/rules_python++pip+pypi_312_mlflow_skinny/site-packages/mlflow/models/model.py", line 20, in 
    from mlflow.models.resources import Resource, ResourceType, _ResourceBuilder
ImportError: cannot import name 'Resource' from 'mlflow.models.resources' (/private/var/tmp/_bazel_tkaplan/5b13b869a536c448538040ea91145cae/sandbox/darwin-sandbox/1/execroot/_main/bazel-out/darwin_arm64-fastbuild/bin/test.runfiles/rules_python++pip+pypi_312_mlflow_skinny/site-packages/mlflow/models/resources/__init__.py)

🌍 Your Environment

Operating System:

  
macOS 15.3.2
  

Output of bazel version:

  
Bazelisk version: 1.25.0
Build label: 8.1.1
Build target: @@//src/main/java/com/google/devtools/build/lib/bazel:BazelServer
Build time: Tue Feb 25 18:53:44 2025 (1740509624)
Build timestamp: 1740509624
Build timestamp as int: 1740509624
  

Rules_python version:

  
1.2.0
  

Anything else relevant?

[FrankPortman]

This is due to implicit packages, I'm pretty sure.

mlflow has a folder called resources and a file called resources.py in the same directory.

This might break other things for you, but try:

• enable_implicit_namespace_pkgs to True in pip parse.
• --incompatible_default_to_explicit_init_py

I forget if you need only one or both.

P.S. thanks for helping me with the PySpark stuff on the Slack yesterday.

[thirtyseven]

Whoops, I forgot to copy enable_implicit_namespace_pkgs=True from our WORKSPACE pip_parse. Thanks!