Problem: Cannot install mfix 22.1.1 from source/pip.
Platform: Windows 10
Python version: 3.9.13 [MSC v.1929 64 bit (AMD64)] on win32
Pip version: 22.1
I had trouble with Anaconda (couldn’t connect to repository) so I tried the build using pip. The build seems to fail at flask/vtk. This is the result:
python3 -m pip install mfix-22.1.1.tar.gz
Processing mfix-22.1.1.tar.gz
Installing build dependencies … done
Getting requirements to build wheel … done
Preparing metadata (pyproject.toml) … done
Collecting flask<2,>=1
Using cached Flask-1.1.4-py2.py3-none-any.whl (94 kB)
ERROR: Could not find a version that satisfies the requirement vtk>=9 (from mfix) (from versions: none)
ERROR: No matching distribution found for vtk>=9
I also tried to manually install vtk, which worked (vtk 9.1.0) but then running the pip install mfix… command returns the same “no matching distribution” error.
I’m wondering if anyone has had this error or could help me find the root cause. I have had similar trouble with nodeworks.
I am trying again with Anaconda in the meantime, trying to reset the repository sites. My company utilizes internal repos, and I suspect the customization is interfering with communication to retrieve mfix from Anaconda.
This is an update on installation from Anaconda. Considering the headline of my question, this is hiding the solution for Conda. I would like to share a solution though. I was getting this error:
CondaHTTPError: HTTP 000 CONNECTION FAILED for url https://conda.anaconda.org/conda-forge/win-64/current_repodata.json
I was able to solve the issue connecting to conda-forge by adding the Anaconda root directory, .\Scripts and .\Library\bin to PATH and by copying libcrypto-1_1-x64.* libssl-1_1-x64.* from .\Library in Anaconda folder to .\DLL in Anaconda folder.
This seems so obvious in retrospect, but I was able to build from source/pip by setting up a virtual environment (I use venv) using Python 3.9.13 (but not yet in 3.10) and installing from the tarball there. So, it’s very similar to the conda approach (i.e., create an environment and build within that sandbox). For some reason I continue to have trouble with conda including
ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header,
got 80 from PyObject
I just hope these comments could help someone in the future. I would love to use the mainstream Anaconda approach, but getting it to work still eludes me right now.
I assume your getting the vtk build from pip?
I believe the install is taking from pip (pyPI) as a dependency of the other package(s).
I haven’t looked at the mfix source to trace it back.I do find that I have fewer problems when building in WSL Linux. I don’t have the luxury of using a “real” linux OS at this time though.