Difficulties when mandating NumPy availability

[knoepfel]

Phlex v0.1.0 includes an optional dependency on NumPy. If the intention is to use NumPy to translate between (e.g.) C++ std::vector<T> types and Python types, then NumPy should be a required dependency.

Unfortunately, our CI system does not currently build (or test) with NumPy. This is evidenced by the lack of py:vec test in the PR that incorporated Python support.

An attempt to mandate NumPy using CMake is shown in PR #219. The failure to import NumPy is explicit when running the tests. Specifically:

 52/61 Test #54: py:vec ..................................***Failed    0.16 sec
  Using configuration file: /__w/phlex/phlex/phlex-src/test/python/pyvec.jsonnet
  [2026-01-07 19:03:44.324] [info] Number of worker threads: 4
  ModuleNotFoundError: No module named 'numpy'
  ImportError: numpy._core.multiarray failed to import
  [2026-01-07 19:03:44.409] [debug] Routed 0 messages in 0 microseconds (-nan microseconds per message)
  [2026-01-07 19:03:44.409] [info] 
  [2026-01-07 19:03:44.409] [info] CPU time: 0.06370s  Real time: 0.08574s  CPU efficiency:  74.30%
  [2026-01-07 19:03:44.409] [info] Max. RSS: 30.628 MB
  build with numpy support, but numpy not importable

We should mandate the use of NumPy and solve this problem.

[knoepfel]

I've learned more. If I activate the CI container's environment, I see:

root@2e866cf18b4f:~/build-phlex# spack find py-numpy
==> In environment /opt/spack-environments/phlex-ci
==> 9 root specs
-- no arch / no compilers ---------------------------------------
[+] catch2  [+] cmake  [+] lcov  [+] llvm@21.1.4:+link_llvm_dylib+llvm_dylib+zstd targets=x86  [+] ninja   -  phlex  [+] py-cmake-format  [+] py-gcovr  [+] py-pip

-- linux-debian13-x86_64_v3 / %c,cxx=gcc@15.2.0 -----------------
py-numpy@2.3.5
==> 1 installed package
==> 0 concretized packages to be installed (show with `spack find -c`)

So py-numpy is installed in the CI container...it's just not available:

root@2e866cf18b4f:~/build-phlex# spack python -c 'import numpy'
Traceback (most recent call last):
  File "<input>", line 1, in <module>
ModuleNotFoundError: No module named 'numpy'

If I then explicitly load it, I see the following:

root@2e866cf18b4f:~/build-phlex# spack load py-numpy
root@2e866cf18b4f:~/build-phlex# spack python -c 'import numpy'
Traceback (most recent call last):
  File "/spack/opt/spack/linux-x86_64_v3/py-numpy-2.3.5-qczi2aqli7yp2w27zwczzidj5rhvbe43/lib/python3.14/site-packages/numpy/_core/__init__.py", line 22, in <module>
    from . import multiarray
  File "/spack/opt/spack/linux-x86_64_v3/py-numpy-2.3.5-qczi2aqli7yp2w27zwczzidj5rhvbe43/lib/python3.14/site-packages/numpy/_core/multiarray.py", line 11, in <module>
    from . import _multiarray_umath, overrides
  File "/spack/opt/spack/linux-x86_64_v3/py-numpy-2.3.5-qczi2aqli7yp2w27zwczzidj5rhvbe43/lib/python3.14/site-packages/numpy/_core/overrides.py", line 5, in <module>
    from numpy._core._multiarray_umath import (
    ...<3 lines>...
    )
ModuleNotFoundError: No module named 'numpy._core._multiarray_umath'

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/spack/opt/spack/linux-x86_64_v3/py-numpy-2.3.5-qczi2aqli7yp2w27zwczzidj5rhvbe43/lib/python3.14/site-packages/numpy/__init__.py", line 112, in <module>
    from numpy.__config__ import show_config
  File "/spack/opt/spack/linux-x86_64_v3/py-numpy-2.3.5-qczi2aqli7yp2w27zwczzidj5rhvbe43/lib/python3.14/site-packages/numpy/__config__.py", line 4, in <module>
    from numpy._core._multiarray_umath import (
    ...<3 lines>...
    )
  File "/spack/opt/spack/linux-x86_64_v3/py-numpy-2.3.5-qczi2aqli7yp2w27zwczzidj5rhvbe43/lib/python3.14/site-packages/numpy/_core/__init__.py", line 48, in <module>
    raise ImportError(msg) from exc
ImportError: 

IMPORTANT: PLEASE READ THIS FOR ADVICE ON HOW TO SOLVE THIS ISSUE!

Importing the numpy C-extensions failed. This error can happen for
many reasons, often due to issues with your setup or how NumPy was
installed.

We have compiled some common reasons and troubleshooting tips at:

    https://numpy.org/devdocs/user/troubleshooting-importerror.html

Please note and check the following:

  * The Python version is: Python3.13 from "/usr/bin/python3"
  * The NumPy version is: "2.3.5"

and make sure that they are the versions you expect.
Please carefully study the documentation linked above for further help.

Original error was: No module named 'numpy._core._multiarray_umath'


The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "<input>", line 1, in <module>
  File "/spack/opt/spack/linux-x86_64_v3/py-numpy-2.3.5-qczi2aqli7yp2w27zwczzidj5rhvbe43/lib/python3.14/site-packages/numpy/__init__.py", line 117, in <module>
    raise ImportError(msg) from e
ImportError: Error importing numpy: you should not try to import numpy from
        its source directory; please exit the numpy source tree, and relaunch
        your python interpreter from there.

This should be sorted out.

[knoepfel]

The error reported above is a red herring and due to the fact that spack python points to the /usr/bin/python3 and not what is installed in the environment. If I type simply python -c 'import numpy', the command finishes successfully.

I am recreating the CI container with python and py-numpy included as roots of the Spack environment. I have verified that doing this solves the NumPy problem described here. PR to follow.

[knoepfel]

I chased this down a little more, and it's more nuanced than I thought. I posted the following to the Spack Slack channel:

I have python and py-numpy as roots of an environment (with a view). I activate the environment and then build C++ code within that activated environment. Some of that C++ code uses NumPy. I then run the relevant unit test and see a failure that numpy cannot be imported by the test:

root@57d5324fea63:~/build-phlex# ctest -R py:vec -V
...
54: ModuleNotFoundError: No module named 'numpy'

However, if I instead:

1. Explicitly spack load py-numpy within the activated environment
2. Build the code
3. Explicitly spack unload py-numpy (this part is optional)

I can then successfully run the test:

root@57d5324fea63:~/build-phlex# ctest -R py:vec
Test project /root/build-phlex
    Start 54: py:vec
1/1 Test #54: py:vec ...........................   Passed    0.13 sec

Anyone know what might be going on?

I would have expected that the equivalent of an explicit spack load py-numpy was happening whenever the environment was first activated

Harmen Stoppels then responded:

not exactly, PYTHONPATH is not set in the env in recent versions of spack, instead if python from the view is invoked it locates numpy like any other virtual environment
check if python3 -c 'import numpy' works, if yes, then maybe the wrong python executable is run by ctest

[knoepfel]

The reason for the behavior observed above is that where python -c 'import numpy' used the executable location to identify potential site-packages candidates, phlex -c ... does not. One solution to this is to make sure that the relevant site-packages paths get added to the sys.path variable. This is the approach chosen by #219.