String ending with character in parboiled2, when the string can contain that character

Question

I've come across a tricky problem writing a parboiled2 parser, which is that I need to match a portion of a line which is a string which has its end marked by a : character. This would be easy enough, except that the string can contain the : character.

At the moment I've got this which treats the string as a group of colon-terminated strings and concats them, but this consumes the trailing : which I don't want as the trailing : is not part of the string itself.

def address = rule { capture(oneOrMore(zeroOrMore(noneOf(":")) ~ ":")) }

I feel like I should be using &(":") somewhere in here but I'm struggling to work that in while matching the interstitial : characters.

Example successful matches (as part of a longer string):

• localhost: -> localhost
• 1::: -> 1::
• ::: -> ::

Mismatches:

• :

Any suggestions would be welcome, even if it's "you can't do this" so I can stop racking my brains.

---

The context for this is parsing the bind setting in an HAProxy configuration file. Some examples of valid strings given the following (simplified) case classes are:

case class Bind(endpoint: Endpoint, params: Seq[String])
case class Endpoint(address: Option[String], port: Option[Int])

• bind :80 -> Bind(Endpoint(None, Some(80)), Seq())
• bind localhost:80 -> Bind(Endpoint(Some("localhost"), Some(80)), Seq())
• bind localhost -> Bind(Endpoint(Some("localhost"), None), Seq())
• bind :80 param1 -> Bind(Endpoint(None, Some(80)), Seq("param1")))

In other words, if there is a string it needs to be terminated before the final : as that's the indicator that there is a port. The endpoint rule looks something like this:

def endpoint = rule { optional(address) ~ optional(':' ~ int) ~> Endpoint }

Ultimately the matchable string for the endpoint is terminated by either a space or the end of the line, so one option would be to just capture until the space and then parse the string separately, but I was hoping to do it within the main parser.

Answer 1

I think that the following should work for your problem description:

def noColons = rule { zeroOrMore(noneOf(":")) }
def colonWithNext = rule { ':' ~ &(noColons ~ ':') }
def address = rule { capture(oneOrMore(noColons).separatedBy(colonWithNext)) ~ ':' }

The problem with your code was the usage of the ~ combinator, since an expression like A ~ B only matches if at first A matches and then B matches, but it would mismatch at B if rule B is part of rule A. There's no backtracking involved here, the parboiled2 parser only backtracks for alternatives.

So, in this case you have to make sure to consume the ':' only if there's another one following it.