pin-env-yml

Pin conda/micromamba environment.yml packages to their currently installed versions — in-place, without bloat.

The Problem

Sharing a conda environment.yml with unpinned versions leads to non-reproducible environments:

# ❌ Unpinned — different installs on different machines
dependencies:
  - numpy
  - pandas
  - scikit-learn

Tools like conda-lock solve this by generating a separate lockfile. But sometimes you just want to pin the versions directly in your existing environment.yml, without changing your workflow.

That's exactly what pin-env-yml does.

Installation

pip install pin-env-yml

Requires Python 3.8+. No external dependencies — only Python stdlib.

Usage

CLI

# Reads env name from the 'name:' field in the YML
pin-env-yml environment.yml

# Or specify the env name explicitly
pin-env-yml environment.yml myenv

Python API

from pin_env_yml import pin_yml

# Pin in-place
pinned_count = pin_yml("environment.yml")
pinned_count = pin_yml("environment.yml", target_env_name="myenv")

Example

Before (environment.yml):

name: myproject
channels:
  - conda-forge
  - defaults
dependencies:
  - python
  - numpy
  - pandas
  - pip:
    - requests
    - httpx

Run:

pin-env-yml environment.yml
# 🔍 Analyzing YML file: environment.yml
# 📦 Matching against installed packages in env: myproject
#   📌 Pinned python → 3.11.9 (conda)
#   📌 Pinned numpy → 1.26.4 (conda)
#   📌 Pinned pandas → 2.2.2 (conda)
#   📌 Pinned requests → 2.32.3 (pip)
#   📌 Pinned httpx → 0.27.0 (pip)
#
# ✨ Done! Pinned 5 explicit dependencies in: environment.yml

After (environment.yml):

name: myproject
channels:
  - conda-forge
  - defaults
dependencies:
  - python=3.11.9
  - numpy=1.26.4
  - pandas=2.2.2
  - pip:
    - requests==2.32.3
    - httpx==0.27.0

How It Works

1. Reads the environment.yml file
2. Queries micromamba list -n <env> --json (falls back to conda list)
3. For each explicit dependency in the YML, substitutes the currently installed version
4. Writes the updated file back in-place
5. Uses = for conda packages and == for pip packages (correct convention for each)

Requirements

• Python 3.8+
• micromamba or conda must be in your PATH and the target environment must exist

License

MIT © Santt997