0.11.0 small regression

[mackoj]

Hi,

Thanks for the fixes in the 0.11.0 version it helps a lot but after updating to 0.11.0 a few of my parsers face some small issues.
Where I use Parse(SomeCustomConvert()) the type of the input changed from.

A few examples:
Before: public typealias Input = ((String, Int), (String, String, String))
After: public typealias Input = (String, Int, (String, String, String))

Before: public typealias Input = ((String, Int, String), (String, Double, Double), ([Double], String, String), (Double?, Double, Double), (Double, Double))
After: public typealias Input = (String, Int, String, (String, Double, Double), ([Double], String, String), (Double?, Double, Double), (Double, Double))

Before: public typealias Input = ((String, Int?, Int?), String, String)
After: public typealias Input = (String, Int?, Int?, String, String)

For some reason, the first tuple always lost its tupling-ness.

To debug this when I saw No exact matches in call to initializer or Initializer 'init(_:with:)' requires the types X and Y be equivalent I just comment the (SomeCustomConvert()) to reveal the type when I add .map a the closing } of the Parse. Then I just redid the Input type to fix it.

// sorry for the multiple edits I did send it by mistake before finishing it

[stephencelis]

@mackoj There are some differences expected due to how buildPartialBlock accumulates, but we think it's probably a change that needs to happen in order to support many more parsers in a block. Apple's Regex builder similarly flattens things, but even more aggressively than we do (we may want to be as aggressive in the future).

The tupling changed slightly for our Date benchmark, which you can observe here:

645c688?w=1#diff-b9c31c5970042d2757410720d6dc581db52c8b951963f09113be33533a1042f0L71-R139

Can you post a full example of the issue, how you expect it to capture output, and then how it does capture output?

[mackoj]

This is an example of it.

I expect it to capture content like this ((String, Int), (String, String, String)) but it did it like this (String, Int, (String, String, String)).

parser:

conversion(left after, right before):

[mackoj]

I don't think it's a major issue some of the custom conversions were to go around the limitation of the number of parsers. With the 0.11.0 it's cleaner.

let failureTestDetails = Parse {
  Peek { "/".utf8 }
  PrefixThrough("\n".utf8).pipe {
    Parse(.case(XcodeBuildLog.testCaseFailureReason)) {
      PrefixUpTo(":".utf8).map(.string) // filename
      ":".utf8
      Int.parser() // line
      ":".utf8
      " error: -[".utf8
      PrefixUpTo(" ".utf8).map(.string) // suite
      " ".utf8
      PrefixUpTo("] : ".utf8).map(.string) // name
      "] : ".utf8
      PrefixUpTo("\n".utf8).map(.string) // failureMessage
      "\n".utf8
    }
  }
}