Change hash attr to algo agnostic one (sha256 -> hash, cargoSha256 -> cargoHash)

[Kranzes]

As requested previously in #177, this change can help with minimizing the nitpicks on new nixpkgs PRs because of the new fetchers' hash attribute name conventions.

This works for:

• sha256 -> hash
• cargoSha256 -> cargoHash
• vendorSha256 -> vendorHash
• and more

Note: It currently only supports sha256 hashes as I saw that in a couple of lines below we only get the hash for sha256 anyway, also, the other algorithms aren't that popular anyway.

[Kranzes]

Do I need to change anything in the tests?

[Mic92]

I don't think I want nix-update to also refactor the code. Replacing hashes is somewhat reliable because they are practically unique but replacing random sha256 = variables could replace instances it shouldn't replace. I think for that it's better to have a tool that understand the language and picks up what a fetcher call is rather just doing search and replace.

[Kranzes]

Fair. I guess while at it also close the issue about this? Btw, isn't --format refactoring too in some sense?

replacing random sha256 = variables could replace instances it shouldn't replace

So far it has been pretty reliable to me... 🤷‍♂️

[Mic92]

Yes but --format is using a reliable nix formatter that has support for nix syntax on AST level rather than just doing substitution. Here an example:

  some-package = some-generic-builder {
    sha256 = "foo";
  };

We sometimes have builder functions i.e. postgres where we maintain multiple versions of the same program. Those function may not accept hash but only sha256 as an input and would be broken as a result of this. Even if we fix this in every instance in nixpkgs there are still third-party repos that also can use nix-update.

[Mic92]

You can just make yourself a wrapper that uses sed on all git-modified files to do the replacement.