http4s backend doesn't fail when there is additional path to consume after match

[hejfelix]

e.g. /my/path/is/good is matched by endpoint.get.in("my" / "path") even though there is more to consume.

[hejfelix]

Hmm, looks like this causes other problems. Please don't merge yet!

[hejfelix]

I added this test, and it causes the two backends to disagree:

  testServer(in_empty_path_out_string, () => pureResult(Either.right[Unit, String]("")), "empty path should match empty path") { baseUri =>
    sttp.get(uri"$baseUri").send().map(_.code shouldBe com.softwaremill.sttp.StatusCodes.Ok)
  }

Akka doesn't like this, http4s does (note: I'm not actually using http4s' matching mechanism in that interpeter)

[ghostbuster91]

Shouldn't emptyPath be defined as

  val in_empty_path_out_string: Endpoint[Unit, Unit, String] =
    endpoint.get.out(stringBody)

?

This change makes both http4s and akka servers to fail on empty path should not match nonempty path test case.

[hejfelix]

Honestly it's not really clear what the expected behaviour is here :/

[adamw]

I'd say that if there are no path constraints specified, then the path is unconstrained. Same as with the method .get matches only the GET method, while without it matches anything So:

endpoint // matches any path
endpoint.get.out(stringBody) // matches any path
endpoint.path("x" / "y") // matches *only* /x/y and /x/y/
endpoint.path("x") // matches *only* /x and /x/

What about "" in path then? Following the same logic:

endpoint.path("" / "y") // matches *only* //y and //y/
endpoint.path("") // matches *only* / and //

So I think it might be logical that specifying "" as the only path component would mean "the root". That would need special-casing, so that GET http://example.com matches such a path specification, as does GET http://example.com/. We could probably add a / to the path if it is empty before any matching.

[ghostbuster91]

Isn't more obvious for the user to pass / as a root? I think we should not allow empty string

[hejfelix]

We are hurting a bit by stringly types here. In the http4s dsl there's a singleton root path called Root IIRC

[adamw]

@ghostbuster91 well we'll always only be able to detect empty strings at runtime (not compile-time). But as I understand, we have two possibilities:

1. / as an "empty path" object & / as a method to combine paths:

endpoint.path(/) // matches only root (no path and /)
endpoint.path(/ "x" / "y" / "z") // matches only /x/y/z and /x/y/z/
endpoint.path(/ "x").path(/ "y" / "z")

2. "" as an empty path:

endpoint.path("") // matches only root (no path and /)
endpoint.path("x" / "y" / "z") // matches only /x/y/z and /x/y/z/
endpoint.path("x").path("y" / "z")

Which one is better, is of course the question? :)

[hejfelix]

I vote for 1

[adamw]

@hejfelix yeah, I'm for 1 as well. Created #41

[ghostbuster91]

I have a third proposition which is just a slightly modification of the 2nd.
Let's use endpoint without any path specification as a root. Path of endpoints which are built upon another will be treated as relative. And we should fail in runtime if the path is an empty string.

Cons:

• still stringly typed

Pros:

• concise syntax

Pro & Con:
There is no way to build an endpoint using another endpoint as a base and specify different absolute path.