html5ever VS json

Would you consider using some libraries in your project? There are lots of good ones in the Rust ecosystem, and many of them are not part of any existing browsers.

For example:

- https://github.com/servo/html5ever (HTML parsing - note: this is used in Servo)

- https://github.com/parcel-bundler/lightningcss (CSS parsing)

- https://github.com/DioxusLabs/taffy (web layout)

- https://github.com/pop-os/cosmic-text (text layout and rendering)

Obviously you should be free to work on whatever you like, but just as a benchmark on the scope of your project: I spent ~6 months implementing just the CSS Grid algorithm in Taffy last year. An entire browser from literal scratch is probably a 10 year project for one person.

So I'm doing some web scraping in Rust, and so I will need to parse HTML. [scraper](https://docs.rs/scraper/latest/scraper/) (which uses [html5ever](https://github.com/servo/html5ever)) is doing fine except that it's the bottleneck of my application.

So I need a faster parser. I've tried [tl](https://docs.rs/tl/latest/tl/) which would've been perfect except that it doesn't actually work on the HTML I have. When I try to `query_selector` the elements I need, it returns nothing.

[Kuchiki](https://docs.rs/kuchiki/latest/kuchiki/) is abandonded.

I couldn't figure out how to get [lol-html](https://github.com/cloudflare/lol-html) to work for me (it's designed for re-writing HTML, whatever that means). It doesn't seem to have an API to extract the inner text of an element.

[html5gum](https://github.com/untitaker/html5gum) seems to be just an HTML tokenizer, or otherwise just too low-level. I have not yet tried [quick-xml](https://github.com/tafia/quick-xml/) but judging from the README, it's pretty low-level too. I mean, if these are the only options left then I will try them. Otherwise, I would love to use a parser that's faster but as ergonomic as `scraper` or `tl`.

At this point, I would be happy with an Lxml bridge/port of some sort. I don't need to mutate HTML, just parse and read data from it.

See https://github.com/servo/html5ever/tree/master/rcdom for an example implementation to imitate.

Oh wow. So serde_json doesn't roundtrip floats by default, it uses some imprecise faster algorithm https://github.com/serde-rs/json/issues/707

Good thing there's msgpack I guess.

tokio / hyper / toml / serde / serde_json / json5 / console

Serde JSON, an extension of the serde crate that enables the serialization and deserialization of Rust structs.

To serialize and deserialize data, we'll employ the popular serde crate along with serde_json. This will allow us to easily convert between Rust types and JSON when working with API requests and responses. For async operations we'll use the Rust futures crate.

From the Cargo.toml file above, we can see that the Perseus version at the time of publication is 0.4.2 and has the following dependencies that are common to both the engine side (server-side) and client side of a Perseus application: sycamore, serde, and serde_json.

serde_json

Number parsing uses a fairly naive but uses a lossless algorithm in musli-json. In serde_json they use a fork of lexical I haven't wrapped my head around. I wanted something simple to start with.

Your best best would be to use an enum. Either your own or something like the one from serde_json depending on what you are trying to do.

Don't do that if you care about memory usage. In your toy program, I wouldn't be surprised if memory usage was a lot better if you used Box instead. (even if it doesn't look like it, you can handle almost all the use cases of serde_json::Value with it, often not much less convenient)