Revert "python3Packages.dataset: use sqlalchemy_1_4"

[dotlambda]

This partially reverts #380276 because only unversioned attributes may be included in dependencies.¹

Things done

• Built on platform:
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• Tested, as applicable:
  • NixOS tests in nixos/tests.
  • Package tests at passthru.tests.
  • Tests in lib/tests or pkgs/test for functions and "core" functionality.
• Ran nixpkgs-review on this PR. See nixpkgs-review usage.
• Tested basic functionality of all binary files, usually in ./result/bin/.
• Nixpkgs Release Notes
  • Package update: when the change is major or breaking.
• NixOS Release Notes
  • Module addition: when adding a new NixOS module.
  • Module update: when the change is significant.
• Fits CONTRIBUTING.md, pkgs/README.md, maintainers/README.md and other READMEs.

---

Add a 👍 reaction to pull requests you find important.

Footnotes

1. https://nixos.org/manual/nixpkgs/stable/#contributing-guidelines ↩

[philiptaron]

@wolfgangwalther, I thought you argued against this approach.

[wolfgangwalther]

I thought you argued against this approach.

Right, that was in #404946 (comment), which was implemented in #444420. This technically only applies to by-name packages, not python-modules - but that was the case for the earlier guideline as well.

I'd say outside of by-name both is fine, but I don't think "because only unversioned attributes may be included in dependencies." is a sufficient reason / motivation for this kind of change - there is no such hard requirement.

Edit: The above was written purely from the by-name perspective. By looking at the original PR, I found the comment referring to the python manual at https://nixos.org/manual/nixpkgs/stable/#contributing-guidelines, which clearly states that this is not allowed for python. It would help to link such information in the PR body directly.

[dotlambda]

The above was written purely from the by-name perspective.

Even in pkgs/by-name, overrideScope should be used instead of versioned Python packages.

[wolfgangwalther]

The above was written purely from the by-name perspective.

Even in pkgs/by-name, overrideScope should be used instead of versioned Python packages.

I was not talking about python at all, but about the interface for by-name packages and whether they put stuff in all-packages.nix or not. This is independent of which language/ecosystem the package is built with.

[wolfgangwalther]

To be clear: No objection from my side.

[xfnw]

diff lgtm

[dotlambda]

nixpkgs-review result

Generated using nixpkgs-review.

Command: nixpkgs-review pr 477562
Commit: 99a1d8d30942fc679856035eff56bd2f78c6da22

---

x86_64-linux

✅ 5 packages built:

• apkleaks
• apkleaks.dist
• jadx
• quark-engine
• quark-engine.dist

[philiptaron]

Yeah, I guess that's fair. That is a disheartening thread.

[philiptaron]

The package removals listed are all due to marking dataset broken.

[linsui]

Is this something "wrong"? Nix thought these packages are also broken so won't build them? Or a bug or the CI?

[dotlambda]

Is this something "wrong"? Nix thought these packages are also broken so won't build them? Or a bug or the CI?

That's how it's supposed to be.

[linsui]

Then why do you break androguard and fdroidserver intentionally? 😕 Shouldn't the CI check them?

[dotlambda]

Then why do you break androguard and fdroidserver intentionally?

I didn't break them. They are simply not rebuilt by this PR.

[linsui]

Because there is no rebuild? Got it, thanks!