Store the state and database files in ~/.conda-store by default

[nkaretnikov]

This work is part of #513. Original PR: #554.

[netlify]

✅ Deploy Preview for kaleidoscopic-dango-0cf31d canceled.

Name	Link
🔨 Latest commit	d412a2b
🔍 Latest deploy log	https://app.netlify.com/sites/kaleidoscopic-dango-0cf31d/deploys/653984ab861696000843d697

[nkaretnikov]

This should land only once Windows lands (#640), to make sure this is compatible.

[nkaretnikov]

Rebased on the Windows PR to be able to test on Windows, will remove redundant commits later.

[nkaretnikov]

Contents of CONDA_STORE_DIR:

% ls ~/.conda-store
celerybeat-schedule.bak  celerybeat-schedule.dat  celerybeat-schedule.dir  conda-store.sqlite  state  storage

% ls ~/.conda-store/state
default

% ls ~/.conda-store/state/default
a54148c1d1d435985989ea232bc982f0eb877991f6c70c4de1390dd76e7ef564-20231025-184722-649393-2-test-env  envs

ls ~/.conda-store/storage
archive  docker  lockfile  logs  yaml

[nkaretnikov]

Tested on Windows and Linux with this env. No files are created in current dir anymore, as expected.

channels:
- conda-forge
dependencies:
- python
- pip:
  - nothing
- ipykernel
- pytest
- requests
description: ''
name: test-env
prefix: null
variables: null

[nkaretnikov]

This change shouldn't affect Nebari since it uses a config file, which sets database_url, storage_class, and store_directory.

[costrouc]

LGTM. I don't think there is anything controvertial in this PR and I was able to run it locally and have it use ~/.conda-store/...

[jaimergp]

I wonder if we are on time to instead use well accepted standards like XDG on Linux, or %LOCALAPPDATA% on Windows.

[nkaretnikov]

@jaimergp I'm not quite sure that is better. Also, on Windows, I see this:

Conda and vscode (which is developed by MS) use /Users/<User>.

[jaimergp]

.conda needs to change too but that ship sailed too long ago and people's environments may live there, so it's tricky to change now. .vscode's issue is inherited from Electron (see lengthy thread). We don't have a good excuse I'm afraid :D

[nkaretnikov]

Yeah, I've changed my mind after reading relevant threads from other projects. We should use well-accepted standards.