conda: init at miniconda3 4.3.31 #34872
Conversation
|
@GrahamcOfBorg build conda |
|
Success on aarch64-linux (full log) Partial log (click to expand)
|
|
Success on x86_64-linux (full log) Partial log (click to expand)
|
|
Success on aarch64-linux (full log) Partial log (click to expand)
|
|
Success on x86_64-linux (full log) Partial log (click to expand)
|
|
Thanks! I didn't test it but took a quick look and looks good to me. Great that this is being worked on! Just some general questions:
I'm relatively new to nix so I suppose these are quite easy to achieve, but just wondering what is the "correct" way. |
| @@ -0,0 +1,98 @@ | |||
| { lib | |||
| , pkgs | |||
| # $ conda-shell | ||
| # $ conda install spyder | ||
| let | ||
| minicondaInstaller = pkgs.stdenv.mkDerivation rec { |
There was a problem hiding this comment.
I think it makes more sense to use runCommand here
| name = "conda-shell"; | ||
| targetPkgs = pkgs: ( | ||
| with pkgs; [ | ||
| conda |
There was a problem hiding this comment.
There should likely be a way to extend the packages here.
| homepage = https://conda.io/; | ||
| platforms = lib.platforms.linux; | ||
| license = lib.licenses.bsd3; | ||
| maintainers = with lib.maintainers; [ jluttine fridh bhipple ]; |
There was a problem hiding this comment.
Please remove me from the maintainers. I don't intend to use it.
- Overlay-friendly extraPkgs list for FHS customization - Update maintainers - Simplify installer script download significantly - Explicitly mention packages used, instead of pkgs
| # Conda manages most pkgs itself, but expects a few to be on the system. | ||
| , condaDeps ? [ stdenv.cc xorg.libSM xorg.libICE xorg.libXrender libselinux ] | ||
| # Any extra nixpkgs you'd like available in the FHS env for Conda to use | ||
| , extraPkgs ? [ ] |
There was a problem hiding this comment.
This can be extended with an overlay as such:
$ cat ~/.config/nixpkgs/overlays/conda.nix ~/src/nixpkgs
self: super:
{
conda = super.conda.override {
extraPkgs = [ self.sssd self.firefox ];
};
}
$ conda-shell
$ ls -l /usr/bin/firefox
lrwxrwxrwx 1 nobody nogroup 78 Dec 31 1969 /usr/bin/firefox -> /nix/store/a258dljnrj0chhnb6dpir2106yimfkb9-conda-shell-usr-target/bin/firefox
$ ls -l /usr/lib/sssd/libsss_krb5.so
lrwxrwxrwx 1 nobody nogroup 90 Dec 31 1969 /usr/lib/sssd/libsss_krb5.so -> /nix/store/a258dljnrj0chhnb6dpir2106yimfkb9-conda-shell-usr-target/lib/sssd/libsss_krb5.so
We could condense condaDeps and extraPkgs into one for simplicity, though they're both overridable in an overlay. I think logically the first one is more for "you're almost certainly going to need this" and the 2nd one is more for user tools like git/emacs, but don't feel too strongly either way.
| src = fetchurl { | ||
| url = "https://repo.continuum.io/miniconda/Miniconda3-${version}-Linux-x86_64.sh"; | ||
| sha256 = "1rklq81s9v7xz1q0ha99w2sl6kyc5vhk6b21cza0jr3b8cgz0lam"; | ||
| }; |
There was a problem hiding this comment.
Thanks for the tip to get rid of mkDerivation @FRidh; this is WAY simpler!
|
Success on x86_64-linux (full log) Partial log (click to expand)
|
|
Success on aarch64-linux (full log) Partial log (click to expand)
|
|
@jluttine regarding your second question, conda has full support for managing its own environments in |
pkgs/top-level/all-packages.nix
Outdated
| @@ -6943,6 +6943,8 @@ with pkgs; | |||
| }; | |||
| purePackages = recurseIntoAttrs (callPackage ./pure-packages.nix {}); | |||
|
|
|||
| conda = callPackage ../development/interpreters/conda { }; | |||
There was a problem hiding this comment.
I think we should actually move the expression to pkgs/tools/package-management/
|
Sounds sensible to me. As a matter of general policy, would you prefer if I squashed and rebased my commits after code review updates to keep the history clean, or leave the shas intact so you can just review the updated diff? I guess github has a squash-merge option now so we can get the best of both worlds if I don't do so myself? I didn't see anything in the mmanual about this. |
|
In general it is preferred when commits are "ready to merge" and that would mean that unnecessary commits have been squashed. |
|
Looks like you used the github squash-merge option, so we get a nice clean history without the reviewer having to double check that the claimed "no-op" manual rebase and squash is actually so 👍 I see some other mergers don't use it, though -- it might be worth noting something in https://nixos.org/nixpkgs/manual/#idm140737316961072 to encourage doing so when the individual commits don't contain particularly useful information. |
|
Success on aarch64-linux (full log) Partial log (click to expand)
|
|
Success on x86_64-linux (full log) Partial log (click to expand)
|
|
I know not everyone likes squashing as a maintainer, as that means purging the contributors history (although of course the original commits are still existing on the original branch). |
Motivation for this change
This takes the work done in PR #26443, which itself was a derivative work from
PR #26245, and adds a few improvements, including:
On the last point, I don't think we really need to give a miniconda2 option, or
at least it isn't a high priority; conda itself can install and manage conda
environments and pythons, so AFAICT a miniconda2 install would just save a small
amount of download bandwidth when someone who wants exclusively python2 adds a
python2 environment.
CC @FRidh @jluttine @layus
Things done
build-use-sandboxinnix.confon non-NixOS)nix-shell -p nox --run "nox-review wip"./result/bin/)