We should start caching strings, instead of copying them all around

[Yoric]

Many Strings will appear several times (e.g. tags). To avoid copying them, we should use a mechanism of atoms, that makes sure we only have one instance of the String in memory at any given time.

[Yoric]

@SimonSapin has developed a crate for this. We should start using it for:

• Id<T>;
• tags.

Given that the crate is pretty much not documented, this may mean documentation work, too.

[SimonSapin]

(I’m more or less the maintainer of that crate these days, but it was developed by @ glennw and @ kmcallister.)

Yes, docs need work. In the meantime, the main APIs are Atom::from(&str) -> Atom, Atom dereferences to str, and comparing or hashing atoms is fast (it compares/hashes u64’s).

[azasypkin]

Hey @SimonSapin,

I see that string_cache crate includes a huge list of predefined atoms that look specific to Servo [1].

Is it included in final binary for every consumer of this crate? If so, is it possible to override it somehow?

Thanks!

[1] https://github.com/servo/string-cache/blob/master/src/static_atom_list.rs

[SimonSapin]

@azasypkin Currently yes they’re included and no you can’t change them. There are plans to make Atom generic so that each user-defined "kind" of atom has its own list of static strings: servo/string-cache#22 and servo/string-cache#136

[azasypkin]

Okay, got it, thanks!

[Yoric]

Well, first part has landed.