Conda installation issues

[d-chambers]

Description

As part of the DASCore manuscript, some authors mentioned they had issues installing DASCore with conda. We need to look into this more and see if there is anything we can do to speed this up, apart from recommending Mamba.

Unfortunately, I don't know anything about the OS/installation they were using but we should test the current version on the major ones.

Anyone who has a few minutes, if you could install a new DASCore conda env on your machine and report back in this issue on how long it takes and if you encounter any problems that would be helpful. Also, please list your OS, python version, and conda version.

edit: You can just copy and paste this on a posix system:

time conda create -n dascore_new -y -c conda-forge dascore

[danesnick]

Hey Derrick,

I did a couple test installations on machines I had access to.

• Ubuntu 22.04.3 LTS via Windows Subsystem for Linux (Kernel version 5.15.146.1-microsoft-standard-WSL2) using Anaconda 2024.02: installed without any issue and quite quickly actually. See command below:

  
  (base) ndanes@12600K-DESKTOP:~$ time conda create -n dascore_stable -c conda-forge dascore -y
  Channels:
  - conda-forge
  - defaults
  Platform: linux-64
  Collecting package metadata (repodata.json): done
  Solving environment: done

Package Plan

environment location: /home/ndanes/anaconda3/envs/dascore_stable

added / updated specs:

• dascore ...OMITTED... Downloading and Extracting Packages:

Preparing transaction: done Verifying transaction: done Executing transaction: done #

To activate this environment, use

#

$ conda activate dascore_stable

#

To deactivate an active environment, use

#

$ conda deactivate

real 0m40.574s user 0m32.389s sys 0m1.844s (dascore_stable) ndanes@12600K-DESKTOP:~$ python --version Python 3.12.2

- macOS Sonoma 14.3.1 (Apple M1 2020 Macbook Pro 16GB): Installed without issue using Anaconda 2024.02 arm64. I will say it took noticeably longer to install here, but didn't time it. Also Python 3.12.2

- Wendian (CentOS 7) Scratch file system using Anaconda 2023.07: 

real 6m0.050s user 1m57.693s sys 0m26.592s

Installed without issue, also Python 3.12.2.

[d-chambers]

Thanks @danesnick, so no issues on any of the installations but the Sonoma 14 installation was much slower than 6m right?

[ahmadtourei]

I just tested this on my Windows machine and could successfully install DASCore on a new environment using conda in ~119 seconds.

OS: Windows 10 (version 10.0.19045) Python: 3.12.2 Conda: 23.1.0 DASCore: 0.1.0

Note: Already tested it on my Ubuntu 20.04 and could install it without issues.

[sboltz]

I just tested it with Ubuntu 22 and it took ~90 seconds.

real 1m37.128s user 1m2.841s sys 0m19.311s

Regarding the "just use mamba" suggestion, libmamba is actually now the default package solver for new versions of conda as of about mid-2023: https://conda.org/blog/2023-07-05-conda-libmamba-solver-rollout/. Based on the style of the output from my install just now, I think that's what it used (I haven't used the python install on my personal machine in quite some time, though, so I'm not certain).

[danesnick]

Thanks @danesnick, so no issues on any of the installations but the Sonoma 14 installation was much slower than 6m right?

I didn't notice any issues, but yes it was a bit slower.

[SeunghooKim]

I did a couple of tests as well: 1) macOS Sonoma 14.3.1 (Apple M1 2021 Macbook Pro 16GB): worked without a problem, didn't give me time but it didn't seem to take too long.

2) GC compute engine (8 vCPU, 4 core, 30 GB memory): real 2m22.070s
user 1m56.196s
sys 0m13.264s

3) Windows 11 laptop: real 0m53.067s user 0m36.376s sys 0m7.788s

[SeisMatt]

Similarly to Seunghoo, I tested this on an Apple M1 Macbook Pro 32 GB, but on a slightly older OS – 12.6.6 No issue, and no time, but was setup in about under a minute.

[d-chambers]

Ok, thanks everyone. I think we have sufficient data to say this is not a wide-spread issue. I will close for now.