Different behaviour with or without the pre-commit hook (reportAttributeAccessIssue)

[ReHoss]

Hi there,

When running pre-commit run with the environment given below, I get this error

error: "run_data_generator" is not a known attribute of module "examples.python_scripts.generating_trajectory_data.trajectory_generator"(reportAttributeAccessIssue)

while running pyright does not return any error.

Could you explain why this behaviour is observed?

Working tree:

control_dde/
├── examples/
│   ├── __init__.py
│   └── python_scripts/
│       ├── __init__.py
│       └── generating_trajectory_data/
│           ├── __init__.py
│           └── trajectory_generator.py
└── test/
    └── test_generator_and_training.py

test_generator_and_training.py

import pathlib
from typing import Any, Generator

import pytest
from control_dde.utils import PATH_PROJECT_ROOT
from examples.python_scripts.generating_trajectory_data import trajectory_generator


@pytest.fixture(params=[94])
def seed(request) -> Generator[int, None, None]:
    yield request.param


@pytest.fixture(params=["debug_pytest"])
def mlflow_experiment_name(request) -> Generator[str, None, None]:
    yield request.param


@pytest.fixture(params=["unique_trajectory_divided", "multiple_trajectories"])
def data_type(request) -> Generator[str, None, None]:
    yield request.param


@pytest.fixture(params=[20])
def num_steps(request) -> Generator[int, None, None]:
    yield request.param


@pytest.fixture(params=["zero", "uniform"])
def control_type(request) -> Generator[str, None, None]:
    yield request.param


@pytest.fixture(params=[1, 4])
def num_trajectories(request) -> Generator[int, None, None]:
    yield request.param


@pytest.fixture(
    params=["pendulum", "mackey_glass", "vanderpol", "linear_time_invariant"]
)
def env_name(request) -> Generator[str, None, None]:
    yield request.param


@pytest.fixture
def dict_env_params() -> dict[str, Any]:
    return {
        "pendulum": {},
        "mackey_glass": {},
        "vanderpol": {},
        "linear_time_invariant": {
            "dt": 0.01,
            "matrix_a": [[1, 2], [3, 4]],
            "matrix_b": [[-1], [-2]],
            "matrix_c": [[0, 1]],
            "control_penalty": 1,
            "dtype": "float32",
        },
    }


@pytest.fixture
def dict_configuration(
    seed: int,
    mlflow_experiment_name: str,
    data_type: str,
    num_steps: int,
    control_type: str,
    num_trajectories: int,
    env_name: str,
    dict_env_params: dict[str, Any],
) -> dict[str, Any]:
    return {
        "seed": seed,
        "mlflow_experiment_name": mlflow_experiment_name,
        "data": {
            "type": data_type,
            "num_steps": num_steps,
            "control_type": control_type,
            "num_trajectories": num_trajectories,
        },
        "env": {"name": env_name, "params": dict_env_params[env_name]},
    }


def test_generator_from_list_configurations(dict_configuration: dict[str, Any]):
    trajectory_generator.run_data_generator(
        dict_config=dict_configuration,
        path_config_absolute=None,
        path_project_root=pathlib.Path(PATH_PROJECT_ROOT),
        plot_trajectory_data=True,
    )

pyrightconfig.json

{
  "venvPath": "venv",
  "venv": "venv_control_dde",
}

.pre-commit-config.yaml

  - repo: https://github.com/RobertCraigie/pyright-python
    rev: v1.1.373
    hooks:
      - id: pyright
        exclude: 'oderl/.*'
        files: "."
        args: ["--pythonpath", "."]
        additional_dependencies: [ torch, pytest, typing_extensions, jaxtyping, matplotlib, gymnasium, seaborn, torchvision, scipy ]

[RobertCraigie]

should your import be this?

from control_dde.examples.python_scripts.generating_trajectory_data import trajectory_generator

---

if you could share a repository to reproduce this issue, that'd be a great help :)

[chrismv48]

This is happening to me too. Driving me nuts

[RobertCraigie]

I believe this will be due to dependency mismatches or other pre-commit environment differences, so I'm closing in favour of #203.