CODEWITHSUNDEEP
pythoncondapython-poetry
edd313
edd313

Reputation: 1469

Conda to poetry environment

I have a conda environment that I would like to convert to a poetry environment.

What I have tried is to translate the environment.yaml of the conda environment into a pyproject.toml file that poetry can read. Here you have the steps:

  1. Generate the yaml file

    conda env export --from-history > environment.yaml

    The --from-history flag includes only the packages that I explicitly asked for. Here it is how the file looks like after installing numpy.

    # environment.yaml

name: C:\Users\EDOCIC\Screepts\My_projects\Tests\conda2poetry\condaenv
channels:
  - defaults
dependencies:
  - numpy

  2. Manually create the pyproject.toml file out of environment.yaml. I added the numpy version, which I got from conda env export. Here it is the result:

    # pyproject.toml

[tool.poetry]
name = "conda2poetry"
version = "0.1.0"
description = ""
authors = [""]

[tool.poetry.dependencies]
python = "~3.7"
numpy = "^1.21.5"

[tool.poetry.dev-dependencies]

[build-system]
requires = ["poetry-core>=1.0.0"]
build-backend = "poetry.core.masonry.api"

  3. Create the environment with poetry init, which will automatically read the toml file.

The process seems to work but it's quite manual and prone to mistakes. Is there a better way?

Upvotes: 11

Views: 10606

Answers (1)

merv
merv

Reputation: 76950

No, there is not a better way. Conda is a generic package manager and does not discern Python versus non-Python packages, therefore this has to be done with manual curation.

Additionally, package names might also differ. For example py-opencv(conda-forge) vs opencv-python (PyPi).

Tips

In addition to pulling down the --from-history YAML, it may also help to dump out a pip list --format=freeze. This could help with resolving any tricky packages that use different names in Conda versus PyPI.

If the environment uses any PyPI packages directly, this won't be seen from a conda env export --from-history. However, these will appear when using conda list (entries with channel pypi) or plain conda env export, which would have a dependencies.pip: section if there are any.

Upvotes: 5

Related Questions