CondaPkg removes and recreates environment often when using Windows.

[RefatIsmail96]

This issue occurs because CondaPkg detects a change in the load path. Specifically, it tries to compare Base.load_path and meta.load_path in the function _resolve_can_skip_1(conda_env, load_path, meta_file). The comparison fails because Base.load_path contains an uncapitalized directory (e.g.,"c:\Users...") for the active project, while meta.load_path has a capitalized path (e.g., "C:\Users...").

The problem seems to stem from how Julia initializes the Base.ACTIVE_PROJECT variable. Specifically, the issue arises from the line unsafe_string(Base.JLOptions().project) in the function Base.init_active_project.

While this appears to be more of a Julia-related issue rather than a problem within CondaPkg, I wanted to bring it to the attention of CondaPkg users in case others encounter the same issue.

As a simple workaround, you can manually activate the project with the correct path before calling CondaPkg. For example:
For example: Pkg.activate("C:\Users...")

[cjdoris]

Oh interesting. The thing is that CondaPkg is just calling Base.load_path() so why would the capitalisation be different from one session to the next?

Regardless, this is something we could fix by normalising all the paths before comparing.

[cjdoris]

Oh even better we can compare the load paths with samefile: https://docs.julialang.org/en/v1/base/file/#Base.Filesystem.samefile

[cjdoris]

Could you see if #167 fixes the issue for you? You can install it like pkg> add CondaPkg#compare-load-path.

[jaakkor2]

pkg> add CondaPkg#compare-load-path fixes the environment re-install whenever using PythonCall on Windows in CondaPkg 0.2.25. Julia launched with julia +lts --startup-file=no --project=debug. CondaPkg 0.2.24 did not have this problem.