BUILD: conda unable to solve for a dev env using environment.yml in finite time

[rs2]

• I have read the installation guide.

Platform

win-64

Installation Method

conda install

pandas Version

latest master

Python Version

n/a

Installation Logs

$ conda env create -f environment.yml
Collecting package metadata (repodata.json): done
Solving environment: /

• Full disclosure: this is not a pandas issue per se, this is a conda issue. However, I see two ways to solve this issue for anyone trying to build pandas locally:

1. Use mamba instead of conda to create dev environment. mamba has a much faster SAT solver compared to conda.
2. (Probably, preferred) Pin package version numbers for each platform in https://github.com/pandas-dev/pandas/blob/c03cc2245538e3a2c182f6eb15cfb143897a9503/environment.yml.

[phofl]

I build this locally on a linux ubuntu machine yesterday and it worked as usual. Do you have acess to a linux machine to test there?

Pinning is not really an option since we want to develop against the newest versions. Same as the testing suite does test against newest versions

[rs2]

I build this locally on a linux ubuntu machine yesterday and it worked as usual. Do you have acess to a linux machine to test there?

Pinning is not really an option since we want to develop against the newest versions. Same as the testing suite does test against newest versions

I need to test specifically on win-64 as per the issue description. conda was unable to solve for the environment in finite time, whereas mamba took well under 10 seconds to do so.

[attack68]

note: #43804. My dev env didn't resolve with pip in 3hours becuase I detected at least 2 incompatabilities.

[rs2]

note: #43804. My dev env didn't resolve with pip in 3hours becuase I detected at least 2 incompatabilities.

As a workaround, do use mamba instead of conda. It invariably finds a solution in no time. See https://github.com/mamba-org/mamba#installation for details.

[phofl]

This was solved for me relatively quickly on win64 too

[rs2]

This was solved for me relatively quickly on win64 too

conda SAT solver can be sensitive to the set of available package versions. It may be able to find a solution in minutes at a particular time, and take a few hours just a short time later.

[mroeschke]

It appears this wasn't reproducible in another windows environment so closing for now. This issue should still remain searchable with the suggestion to use mamba

[asishm]

I've been running into this issue on wsl2 (ubuntu 20.04) as well. At first it was getting killed due to OOM with 4GB RAM allocated via wslconfig, but took 20 minutes to resolve the package dependencies.

Mamba was near instant.

[MarcoGorelli]

Use mamba instead of conda to create dev environment. mamba has a much faster SAT solver compared to conda.

@rs2 did you want to open a PR for this?