Question
Explicit File vs Environment.yml: Which Conda Restore Guarantees Exact Artifacts?
Neel Maple
0 reputation · 11 May 2025, 13:45 UTC
97K views1
Loading…
0 reputation · 11 May 2025, 13:45 UTC
Restore a broken conda environment after disk loss while ensuring reproducibility.
Need bit‑for‑bit exactness on the same platform, but also consider portability to a different OS/arch.
The conda documentation lists two restore paths: conda list --explicit followed by conda create --file, and conda env export/conda env create -f environment.yml. Each has trade‑offs around channel URLs, build strings, solver resolution, and platform dependency.
35275 reputation · 11 May 2025, 18:51 UTC
For bit-for-bit exactness on the same platform, the explicit file (generated via conda list --explicit) is the only method that guarantees the restoration of identical binary artifacts. An environment.yml file is a specification for the solver; it ensures version compatibility but does not guarantee that the same build strings or binaries are selected during restoration.
The fundamental difference lies in whether the Conda solver is invoked during the restoration process.
anacondalinux-64), using an explicit file on a different OS or architecture will typically fail as the referenced URLs will not exist for that platform.To restore a broken environment with maximum fidelity on the same platform, use the following scoped commands:
conda create --name restored_env --file spec-file.txtGenerate a new explicit list from the restored environment and compare it to the original spec file. If the lines match exactly, the binaries are identical.
conda list --explicit > verify-spec.txt
diff spec-file.txt verify-spec.txt
This guidance assumes you are using a standard Conda distribution. Note that explicit files rely on the availability of the original channel URLs; if a package was removed from a remote repository or hosted on a private server that is now offline, the restore will fail regardless of the method used.
Missing Diagnostic: Are you restoring to an identical OS version and hardware architecture (e.g., Linux x86_64 to Linux x86_64), or is this a cross-platform migration?
Use comments to ask for clarification. Post a solution as an answer.
35,275 reputation · 11 May 2025, 17:05 UTC
One practical gap with conda list --explicit is that it only records conda-installed packages. In mixed environments where pip was used inside conda, the explicit spec will be incomplete unless you also export pip requirements separately.
That means a conda create --file restore can be bit-for-bit for conda packages on the same platform, but the Python packages installed via pip will be missing or re-solved, breaking true exactness. For verification, compare conda list --explicit before and after restore and also diff pip freeze.