Question about poetry and dependency resolution. I've been running

poetry add meltano

on a clean project (basically, create a directory, cd into it,

poetry init

with no dependencies, then

poetry add meltano

). The dependency resolution took 5191 seconds (yes, 1.5 hrs!). And no, this is not on a 56k modem 😉 I went through poetry's issues, particularly https://github.com/python-poetry/poetry/issues/2094 where they say it's caused by the package not defining dependencies correctly. I'd like to understand what is going on here. As I understand it, meltano has been using poetry for dependency management since December, and I see a

poetry.lock

in the repo. So from what I understand, it's all clear here. That said, when I look at the logs, I see poetry trying pretty much every version of

boto3

under the sun, and I'm not sure why. Has anyone else hit this problem (namely, insanely slow dependency resolution)?

@laurent That's crazy! I just found your comment here: https://github.com/python-poetry/poetry/issues/2094#issuecomment-780790115 I haven't actually used Poetry to installed

meltano

before, only pip and pipx (see https://gitlab.com/meltano/meltano/-/merge_requests/1960), since Poetry is most useful when you have to manage tons of dependencies, while Meltano typically lives by itself in a venv and is used like an application rather than a library as part of a bigger project

good question, I haven't actually tried this, but will do and report back

Maybe 😄 Were you trying to just install Meltano locally, or contribute to Meltano?

just install it for my own use. I think I just went on autopilot, added it as a dependency for my project, like I'd add pandas or things like that. Is it fundamentally wrong?

It's not, but since you only care about the

meltano

executable and not about being able to reference the

meltano

module from your own source code, it doesn't need to be in your project like

pandas

would

ok, I see your point. So I guess I've been trying to shoehorn it into my usual approach, but it's not ideal, as you explain. I think my main concern is reproducibility, but the process you pointed to should be ok from the perspective, might just need to pin the git commit I'm using for meltano somehow. I'll figure it out.

I think the problem is that poetry doesn't use

poetry.lock

when you're adding

meltano

to an existing project, since it needs to check all of the dependencies against the packages already installed in the virtual environment, while

poetry install

can assume there's nothing in the virtual environment yet

yes, makes sense. Thanks for the quick reply!