Lazy static completions

[lavafroth]

Changes

• Replace HX_TAGS, HX_ATTRIBUTE_VALUES and TEXT_STORE from being OnceLocks to being lazy statics
• init_text_store and init_hx_tags are now implicit
• Lazy statics no longer require .get() for accessing attributes

[ThePrimeagen]

could we break these up into 2 prs?

please do the lazy statics. i like that change a lot :)

then we can avoid the clones

[lavafroth]

The point about lesser clones is a by-product of using lazy statics. The lazy static forces the use of static string slices in the cases I mentioned unless one goes out of their way to clone things. 😅

[lavafroth]

Also, it would be better in the long run to have the binary representation of the completion structs embedded in the final executable during comptime so we don't have to call .leak() on the completions vector. We could use the include! macro to do something like that.

[ThePrimeagen]

ok, very interesting. i am going to try to review this once more and make sure i get what is going on. :)

i may just need to download it and look at it in my editor

[ThePrimeagen]

i am also going to get #22 merged in first.

may cause conflict, sorry

[lavafroth]

Sure thing! I've synced this branch with upstream.

[ThePrimeagen]

psst, sorry about not merging this

if you remove the conflicts, i can help get this in

[lavafroth]

Sure, I'll resolve the conflicts by tomorrow.

[lavafroth]

I'm only considering the inclusion of lazy static for now, if I try to add more changes this PR becomes really massive.

[lavafroth]

Hey just a heads up, it looks like OnceCell / OnceLock are actually more mature for our use case. There's also talk about deprecating lazy_static. Sorry for troubling you, I'm closing this.