Poetry: Enable multiple subset of dependencies

[nico-unity]

• [ X] I have searched the issues of this repo and believe that this is not a duplicate.
• [ X] I have searched the documentation and believe that my question is not covered.

Feature Request

Currently using the latest version of poetry on python 3.7.0.

Using poetry when developing on a local machine with cpu and then building a Docker to launch the code on a GPU machine is painful with some module and current poetry segmentation of dev and core dependencies. A bit like the command poetry install --no-dev it would be very useful if we could define X set of dependencies ex: poetry install --dependencies test. The dependencies argument by default could be the current poetry behaviour and then we could had what ever subset of dependencies in the pyproject.toml, ex:

[tool.poetry.dependencies]
python = "^3.6"
kfp = "^0.1.31"
tensorflow_gpu = "^2.0.0"

[tool.poetry.dev-dependencies]
matplotlib = "^3.1"

[tool.poetry.test-dependencies]
python = "^3.6"
kfp = "^0.1.31"
tensorflow = "^2.0.0"

If multi dependencies is not possible another solution could be enable installation of strictly the dev dependencies ex: poetry install --no-prd

Thank you.

[vbersier]

Isn't it possible to achieve something similar with optional dependencies and extras?

[nico-unity]

Extra would still be install when poetry install is called no ? So there no way to install either only dev dependencies .

[vbersier]

You specify which extras get installed as per the documentation, and you can make as many extra categories as you want.

[nico-unity]

And what's the behaviour of those extras ?

[vbersier]

I'm no expert as I started using poetry only a few days ago, but they are optional sets of packages that you can opt-in to install. So for your use-case, you'd make something like:

[tool.poetry.dependencies]
python = "^3.6"
tensorflow = {version="^2.0.0", optional=true}
tensorflow-gpu = {version="^2.0.0", optional=true}

[tool.poetry.dev-dependencies]
matplotlib = "^3.1"

[tool.poetry.extras]
cpu = ["tensorflow"]
gpu = ["tensorflow-gpu"]

In your dev environment you'd do poetry install --extras "cpu", and in your Dockerfile you'd do poetry install --no-dev --extras "gpu"

[nico-unity]

Ok sorry I did a mistake previously, seems like this solves it ! Thanks. closing the issue.