feat(init)!: allow turning off aliases for libs and plugins

[mcornella]

Standards checklist:

• The PR title is descriptive.
• The PR doesn't replicate another PR which is already open.
• I have read the contribution guide and followed all the instructions.
• The code follows the code style guide detailed in the wiki.
• The code is mine or it's from somewhere with an MIT-compatible license.
• The code is efficient, to the best of my ability, and does not waste computer resources.
• The code is stable and I have tested it myself, to the best of my abilities.
• If the code introduces new aliases, I provide a valid use case for all plugin users down below.

---

Proof of concept work for backing up and restoring aliases on libs and plugins that have them turned off by the user via ztyle:

zstyle ':omz:*' aliases no
zstyle ':omz:lib:*' aliases no
zstyle ':omz:plugins:*' aliases no
zstyle ':omz:plugins:git' aliases no

Aliases could also be selectively enabled for specific components, such as:

zstyle ':omz:plugins:*' aliases no
zstyle ':omz:plugins:git' aliases yes

Work in collaboration with @carlosala.

Fixes #9414

[carlosala]

We can see in #10510 what does this PR closes also 👌🏻

[carlosala]

Just a quick reminder here, when merging this we should soft-deprecate

ohmyzsh/lib/directories.zsh

Line 12 in f7d903f

zstyle -T ':omz:directories' aliases || return 0

as is not following the ':omz:lib:<file> structure and document properly also that in README (removing also

ohmyzsh/README.md

Line 279 in f7d903f

### Remove directories aliases

)

[carlosala]

LGTM! 😉

[carlosala]

After testing with https://github.com/romkatv/zsh-bench, I cannot see a difference while removing all aliases and leaving all of them. It seems that the way we are keeping and restoring aliases has no performance issues.
I think we can merge this, and see how the community receives it.
I'm quite happy we finally sorted it :)

[mcornella]

I'm currently debugging why setting galiases does not remove the global aliases, don't merge it yet.

[carlosala]

At least from the documentation perspective, it should work exactly the same. Even though, if it's giving issues, I only see global aliases defined here and here. We could have some workaround there 👍🏻

[mcornella]

OK it looks like the same happens with aliases and the rest of special parameters when we're doing array=() (empty array).

When that happens, the resulting hash table does not actually get modified, I don't know exactly why by looking at the code: https://github.com/zsh-users/zsh/blob/c006d760/Src/Modules/parameter.c#L1767-L1807

I have pushed a version that uses unalias directly in a "smarter" way to avoid many calls, but I don't know if it is actually slower. Could you try it out with zsh-bench again?

---

EDIT: it looks like the culprit is that the hashtable is actually empty, so this check matches and nothing is done: https://github.com/zsh-users/zsh/blob/c006d7609703afcfb2162c36d4f745125df45879/Src/Modules/parameter.c#L1773-L1774

[carlosala]

I think the code is a bit more difficult to read. What about adding always a garbage alias, maybe alias __omz_tmp=<whatever>. Then we are sure the hash table is never empty (and we could use unalias only for that).
What do you think?

[mcornella]

That looks like a hack, I have instead pushed a change that doesn't resort to weird tricks, a simple unalias was enough. It should not be less fast, or even faster.

[carlosala]

Now it's more readable I'd say, LGTM!
Incredible job Marc😉

[mcornella]

Shipped! 🚀 Thank you for the review!