dojeda / poetry2conda Goto Github PK

Following is the error

sumit@Sumits-MacBook-Air backend % poetry2conda pyproject.toml environment.yaml
Traceback (most recent call last):
  File "/opt/homebrew/bin/poetry2conda", line 8, in <module>
    sys.exit(main())
  File "/opt/homebrew/lib/python3.9/site-packages/poetry2conda/convert.py", line 294, in main
    converted_obj = convert(args.pyproject, include_dev=args.dev, extras=args.extras)
  File "/opt/homebrew/lib/python3.9/site-packages/poetry2conda/convert.py", line 53, in convert
    dependencies, pip_dependencies = collect_dependencies(
  File "/opt/homebrew/lib/python3.9/site-packages/poetry2conda/convert.py", line 161, in collect_dependencies
    tag = constraint["tag"]
KeyError: 'tag'

I'm suspecting this dependency in my pyproject.yaml is the culprit.

pyppeteer = { git = "https://github.com/pyppeteer/pyppeteer.git", rev = "0fd15a7" }

It would be nice to generate a meta.yaml file to generate conda package with conda build (https://docs.conda.io/projects/conda-build/en/latest/user-guide/tutorials/build-pkgs.html)

It would be great to automatically update conda dependencies upon commit/push.

Looks like dependencies with a dot aren't handled well, see below

[tool.poetry.dependencies]
python = "^3.6.1"
ruamel.yaml = "^0.16"

# dev dependencies ...

% poetry2conda --dev pyproject.toml environment.yml
Traceback (most recent call last):
  File "/home/vvk/miniconda3/envs/birdfeeder/bin/poetry2conda", line 8, in <module>
    sys.exit(main())
  File "/home/vvk/miniconda3/envs/birdfeeder/lib/python3.8/site-packages/poetry2conda/convert.py", line 295, in main
    converted_obj = convert(args.pyproject, include_dev=args.dev, extras=args.extras)
  File "/home/vvk/miniconda3/envs/birdfeeder/lib/python3.8/site-packages/poetry2conda/convert.py", line 53, in convert
    dependencies, pip_dependencies = collect_dependencies(
  File "/home/vvk/miniconda3/envs/birdfeeder/lib/python3.8/site-packages/poetry2conda/convert.py", line 167, in collect_dependencies
    raise ValueError(
ValueError: This converter only supports normal dependencies and git dependencies. No path, url, python restricted, environment markers or multiple constraints. In your case, check the "ruamel" dependency. Sorry.

(Pdb) poetry_dependencies.items()
dict_items([('python', '^3.6.1'), ('ruamel', {'yaml': '^0.16'}), ('poetry2conda', '^0.3'), ('pre-commit', '^2.2.0'), ('pytest', '^5.4'), ('pytest-cov', '^2.7'), ('pytest-mock', '^3.1.0')])

I have a project that has a dependency for a package from an alternate source (not pypi). Is there a work around to to specify a package like that for pip?

Hello!

I think this is a great tool, but I think there is one package that stops me from using it. I'm talking about rdkit.

You see, it's impossible to install it as a git dependency and the simplest way to install it is via Conda.

Can I use your tool in this case? And how 😊?)

Hi, first of all many thanks for this project!

TL;DR: the generated pip section should include the --index-url and --extra-index-url using info from tool.poetry.source

I'm on a limbo of mixed environments and repos, but I need to switch back from a poetry-based to a conda-based build env.
I already have some projects built with poetry and saved on a private devpi index;
the directive to use the private repo is in pyproject.toml's tool.poetry.source:

[[tool.poetry.source]]
name = "private"
url = "https://my.private.repo/my/stable/+simple/"
default= true

I added the tool.poetry2conda.dependencies section to mark my packages with channel = "pip", and when I use poetry2conda pyproject.toml environment.yml, the yaml file contains the pip packages but no --index-url option.

Passing extra options in the environment.yml's pip section is possible from conda version 4.4.0 (see this pr).

It would be wonderful if the urls specified in the tool.poetry.source sections could be translated to a --index-url entry if default = true, or a --extra-index-url otherwhise.

Poetry by default will install your project’s package everytime you run install unless using the option --no-root. When exporting an environment.yaml and installing it to a conda virtual environment the root project package is not installed.

Also it is not possible to "sneak" this in as an optional, editable dependency or even a git dependency since Poetry is not happy about the self-loop.

I guess it would make sense to handle this bi-directional.

Hi,

Thanks for this tool that makes it really easy to keep my conda friends happy!

I don't really understand why I have to add an extra section to my pyproject.toml file, as in general I would personally want the environment to be named after the project, and to be warned in case I already have an environment with that name.
If I don't want it to have to same name, I could just change the environment name in the resulting environment.yml file, right?
What am I missing?

Thanks in advance for the response!

When the poetry2conda command fails, the target environment.yaml file is still created with empty contents. This is an undesirable behavior due to the usage of argparse.FileType. Ideally, the file should be written only when the conversion has succeeded.

It seems obvious, but when a constraing that does not follow semantic version is used on a dependency, poetry2conda will fail. Perhaps the most simple example is:

black = "^19.10b0"

which is present in our own pyproject.toml.

I think that there are two possible solutions here:

The pragmatic one, which is to catch semantic version parsing errors, show a warning and use pin what was declared. In the case of black = "^19.10b0", conda will not understand this caret version, but we could warn the use to not use a caret (black = "19.10b0") and we could pin it to black==19.10b0.

The more complex one: since poetry does understand this caret, we should probably re-use poetry caret-handling code. This is complex because we don't want to set poetry as a dependency (it would install poetry on the virtual environment, and this is discouraged on poetry's installation guide). However, I saw that poetry has recently been reorganized on a core package, which may respond to our need.

First of all great work. This looks amazing.

Here is the issue I found:

Tried installing opencv-python-headless = "^3.4.8.29"

Received ValueError: Invalid simple block '^3.4.8.29'

To anyone interested,
It has been a while since I updated this project and I currently don't have the bandwidth / time that this project requires.
I have seen that there are plenty of contributions and contributors so please let me know here if you are interested on becoming the maintainer of this project.

I am leaving this here to follow/remind the task of setting up a github workflow.