Make Uri parser more restrictive

[asoltysik]

Currently it allows for some invalid URIs, which lead to runtime errors when passed to Apache HTTP (thrown by java.net.URI) or AsyncHttpClient.

Examples of invalid URIs which pass hammock.Uri.fromString parser:

• ::
• ^&%$#@!
• http://^&%$#@!/foo
• empty string and/or whitespace

A couple of ideas for how to solve this:

• Wrap and use java.net.URI constructor as a parser, in type-safe way (probably the easiest)
• Use a library like: https://github.com/lemonlabsuk/scala-uri
• Improve existing parser/write/port a parser from scratch (https://tools.ietf.org/html/rfc3986)

I'd be happy to contribute once you decide on direction.

[miciek]

@asoltysik Thanks! I think second option would be the best. I haven't used at scala-uri but it looks like it's way better than java.net.URI.

[asoltysik]

Looking at scala-uri it uses parboiled, which is a java library with scala wrapper. IIRC that makes it not an option as hammock is cross-published to ScalaJS.

[pepegar]

Yeah, the idea behind creating our own was to have something that worked for both scalaJS & JVM (and potentially scala native)

[miciek]

OK, @asoltysik I think that the only way forward then is to improve existing parser :)