Latest NumPy in Conda Env

1

I want to add compas v2.3.0 to an extant conda environment, but installation would downgrade numpy to v1. Is there a guess how soon compas will work with numpy v2?

micromamba install compas=2.3.0
...
Transaction
  
  Prefix: ...

  Updating specs:

   - compas=2.3.0

  Package     Version  Build            Channel           Size
────────────────────────────────────────────────────────────────
  Install:
────────────────────────────────────────────────────────────────

  + watchdog    4.0.1  py312h7900ff3_0  conda-forge      136kB
  + scipy      1.14.0  py312hc2bc53b_1  conda-forge       18MB
  + networkx      3.3  pyhd8ed1ab_1     conda-forge        1MB
  + compas      2.3.0  pyhd8ed1ab_0     conda-forge        1MB

  Downgrade:
────────────────────────────────────────────────────────────────

  - numpy       2.0.1  py312h1103770_0  conda-forge     Cached
  + numpy      1.26.4  py312heda63a1_0  conda-forge        7MB

  Summary:

  Install: 4 packages
  Downgrade: 1 packages

  Total download: 28MB

────────────────────────────────────────────────────────────────

Confirm changes: [Y/n] n
Aborted.
2

Hi,

you could install the other requirements explicitly, and then install compas without dependencies.

pip install numpy scipy networkx
pip install compas --no-deps

note that using numpy 2 might result in an error here and there, because we have not caought up yet with the breaking changes introduced there…

3

In my case, that’s

micromamba activate myenvironment
micromamba install watchdog scipy networkx
micromamba install --no-deps compas
micromamba deactivate

without pip. The cool kids now use pipx, but Micromamba beats both; 2.0.0 is at release candidate status. I avoid pip(x) nowadays.

I’m using Julia with PythonCall, not just Python, in Jupyter Lab. The related CondaPkg uses “Current” environment backend to install what it needs. If I start managing depedencies by hand, then I will set the backend to “Null” so CondaPkg leaves them to me.