poetry change python version to 3.x

arshbot picture arshbot · Mar 7, 2020 · Viewed 24.5k times · Source

According to poetry's docs, the proper way to setup a new project is with poetry new poetry-demo, however this creates a project based on the now deprecated python2.7 by creating the following toml file:

[tool.poetry]
name = "poetry-demo"
version = "0.1.0"
description = ""
authors = ["Harsha Goli <[email protected]>"]

[tool.poetry.dependencies]
python = "^2.7"

[tool.poetry.dev-dependencies]
pytest = "^4.6"

[build-system]
requires = ["poetry>=0.12"]
build-backend = "poetry.masonry.api"

How can I update this to 3.7? Simply changing python = "^2.7" to python = "^3.7" results in the following error when poetry install is run:

[SolverProblemError]
The current project's Python requirement (2.7.17) is not compatible with some of the required packages Python requirement:
  - zipp requires Python >=3.6

Because no versions of pytest match >=4.6,<4.6.9 || >4.6.9,<5.0
 and pytest (4.6.9) depends on importlib-metadata (>=0.12), pytest (>=4.6,<5.0) requires importlib-metadata (>=0.12).
And because no versions of importlib-metadata match >=0.12,<1.5.0 || >1.5.0
 and importlib-metadata (1.5.0) depends on zipp (>=0.5), pytest (>=4.6,<5.0) requires zipp (>=0.5).
Because zipp (3.1.0) requires Python >=3.6
 and no versions of zipp match >=0.5,<3.1.0 || >3.1.0, zipp is forbidden.
Thus, pytest is forbidden.
So, because poetry-demo depends on pytest (^4.6), version solving failed.

Answer

finswimmer picture finswimmer · Mar 9, 2020

Whenever you change dependencies by hand in your pyproject.toml you have to take care of these points:

  1. Run poetry lock afterwards or remove the poetry.lock file to force recreation of it. The reasons for this is, that poetry install takes the poetry.lock as input if can find one and not the pyproject.toml.

  2. If you change the python version and uses in-project virtualenv, remove the .venv before running poetry install. poetry doesn't change the python version of a venv once it is created, because it uses the python version itself to create the virtualenv.