Feedback on initial version

[mfisher87]

Hey @mstudinger, just re-posting feedback I shared on the CryoCloud Slack as requested! Great idea to set up GitHub Discussions.

---

I'd suggest providing a bit of extra setup/background with conda to help users avoid some common pitfalls:

1. Avoid installing "everything" to base environment; that always results in disaster in my experience. I hear your point about the hassle of activation though. I can give more tips on that if you'd like 🙂 To achieve having "one true env" that is less brittle, I'd suggest avoiding base by creating an empty environment you can blow away at any time, e.g. named geo, and automatically activate that environment with .bashrc or the equivalent for your OS.

2. Remind users to check their prompt to be sure the right env is activated before installing packages; this is really the same tip as above but I think it's really important to avoid installing to base if you want to avoid the hassle of blowing away all of conda 🙂

3. Install opencv, pyproj, and segment-anything (thanks to the prolific Qiusheng Wu @giswqs ❤️ ) with conda (don't forget -c conda-forge, or you could even provide instructions for the user to set that up as the default). Mixing conda and pip can often get folks in trouble, and conda-forge has extremely good coverage of the PyPI ecosystem. I'd even explicitly advise folks to look first in conda-forge and use pip as a last resort. If you do need to use pip as a last resort, ensure it's installed to your environment with conda install pip first, or you could be installing pip packages to another conda environment. It's confusing, so the "last resort" rule of thumb helps avoid a lot of hassle.

[mstudinger]

This is great – thanks @mfisher87! As said on CryoCloud Slack this is exactly the kind of expert guidance that is incredibly helpful (and time saving!) for folks like me. Thanks for taking the time to put such a comprehensive response together and sharing it here. I feel package management has gotten a little less daunting already… 🙂

[giswqs]

If you are looking for a one-stop shop for commonly used geospatial packages, I would recommend geospatial.

conda create -n geo python=3.11
conda activate geo
conda install -c conda-forge mamba
mamba install -c conda-forge geospatial

[mfisher87]

Nice, TIL!

Mamba is one of the few things I'd suggest installing to the base environment over a non-base env :)

[mstudinger]

@giswqs: this is great and exactly the kind of guidance I was hoping to get. Thank you for sharing it!