pyTME circular import error

[DcShepherd]

Hello,

I have been following the tutorial for the installation of pyTME but I am running into this error when I run the pytest protocol.

============================================================================================= test session starts ==============================================================================================
platform linux -- Python 3.11.9, pytest-8.3.2, pluggy-1.5.0
rootdir: /home/doulin/clones/pyTME
configfile: pyproject.toml
plugins: napari-0.5.1, napari-plugin-engine-0.2.0, npe2-0.7.7
collected 0 items / 17 errors                                                                                                                                                                                  

==================================================================================================== ERRORS ====================================================================================================
___________________________________________________________________________ ERROR collecting tme/tests/preprocessing/test_compose.py ___________________________________________________________________________
ImportError while importing test module '/home/doulin/clones/pyTME/tme/tests/preprocessing/test_compose.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
tme/tests/preprocessing/test_compose.py:3: in <module>
    from tme.preprocessing import Compose
tme/__init__.py:1: in <module>
    from . import extensions
E   ImportError: cannot import name 'extensions' from partially initialized module 'tme' (most likely due to a circular import) (/home/doulin/clones/pyTME/tme/__init__.py)

Furthermore, check to see if the preprocess.py is working but then I get a similar error

preprocess.py --help
Traceback (most recent call last):
  File "/home/doulin/miniconda3/envs/pytme/bin/preprocess.py", line 12, in <module>
    from tme import Preprocessor, Density
  File "/home/doulin/clones/pyTME/tme/__init__.py", line 1, in <module>
    from . import extensions
ImportError: cannot import name 'extensions' from partially initialized module 'tme' (most likely due to a circular import) (/home/doulin/clones/pyTME/tme/__init__.py)

I tried deleting and reinstalling multiple times but I get the same error. Would you be able to point out where I might have made of a mistake during the installation.

[maurerv]

Hi @DcShepherd,

Thank you for reaching out. pytme comes with a set of c extensions that help performance but can be a bit tricky to build. I reproduce your error when installing like this

git clone https://github.com/KosinskiLab/pyTME.git
cd pyTME
python3 -mpip install .

Installing the library in editable mode fixed the issue for me python3 -mpip install -e .. The other installation methods should also yield a working installation though (https://kosinskilab.github.io/pyTME/quickstart/installation.html).

Which installation method did you use? :)

[DcShepherd]

Hi, Thanks for the quick reply, my installation process is shown below from this page: https://kosinskilab.github.io/pyTME/quickstart/installation.html

1. Conda environment set up

conda create \
   --name pytme \
   -c conda-forge \
   python=3.11 \
   pyfftw \
   napari \
   magicgui \
   pyqt

2. Then activate the environment and run the following.

conda activate pytme
pip install git+https://github.com/KosinskiLab/pyTME.git

3. Install Cupy

pip install "pytme[cupy]"

4. Install the GUI

pip install git+https://github.com/maurerv/napari-density-io.git

The good new is that the python3 -mpip install -e command seems to have fixed the issue

[maurerv]

Hi,

Thank you for sharing the installation process. I am glad to hear pytme is working now.

I was able to reproduce your error and have updated the installation instructions accordingly. Re-installing in editable mode did solve the issue, but it is a temporary solution at best. The editable mode can be avoided by modifying the file structure of the package. I plan to integrate this in the next 0.2.4 release.

Thanks again for reporting :)

[DcShepherd]

Hello,

Thanks for the info! I also had to install Pytorch rather than CuPy to get all the test to pass. I'll have to check if it is a config problem with this machine.