Parsing Url is one of the most expensive things that async-h1 does about async-h1

[yoshuawuyts]

In most cases, we parse a url from host and path and then later on only use the path for routing.

When routing on the path we do need to parse the URL in order to filter out the query string, and normalize URL escaping. We can probably skip parsing the URL in async-h1, but I'm not sure to which degree it's feasible for e.g. tide.

If we think we'll have to parse URLs for tide anyway, is it still worth adding extra logic to async-h1 to not always have to parse them?

[jbr]

It might be worth checking if percent_encoding::percent_decode_str(path.split("?").next()) is faster than Url decode. However, this would have to happen at the same time as http-types v3, since there is currently no way to construct a Request without a full Url.

[jbr]

I also think it's worth asking why we need to parse urls for tide. Why does the host and path have to be represented as a Url specifically?

Alternatively, we could try to add a constructor for Url, or fork Url to add a constructor, since currently we are concatenating strings and then parsing them in order to construct a Url from scheme, host and path