Poetry: Python dependency management and packaging made easy

github.com

17 points·sdispater··5 comments

Why not use conda for dependency management & packaging? The GH page makes no mention of conda. Not sure if it's all that different from Poetry, as I haven't used it yet. I do wonder, however, if this is yet another new tool when there exist more mature solutions.

From the documentation, the move to a single file for build & dependency specification is great. conda requires a rather Byzantine set of files with specific names at specific relative locations, which is troublesome.

From first-glance, Poetry seems the best of the new generation of package managers. I do not like pipenv.

When the overhead for miniconda distributed is ~50mb, the python packages I use are ~500-750mb. I don't understand why this is such a point of contention in this community.

Is there something that this handles that pip/requirements doesn't? Python has always been a breeze for this situation.

Poetry comes with an exhaustive and efficient dependency resolver which pip does not have.

It's also a packaging tool, which means you can build and publish packages with (on PyPI for instance).

And finally, it always creates a virtualenv for each project so that you always work isolated from the global Python installation.

What shell/terminal setup is the demo gif in the github using? Looks neat.

Reply on news.ycombinator.com