Use cucumber/messages where possible

[mpkorstanje]

🤔 What's the problem you're trying to solve?

We currently have cucumber/messages
as a message protocol. This protocol allows tools to consume messages in a standardized format.

However not all implementations of the Gherkin parser use the message protocol when there is an implementation available. This issue represents the current status.

• C++
• .Net
  • (dotnet) Migrate gherkin messages to cucumber messages #426
• Elixir
• Go
• Java
• Perl
• PHP
• Python
• Ruby

✨ What's your proposed solution?

Migrate the missing implementations to messages.

⛏ Have you considered any alternatives or workarounds?

No response

📚 Any additional context?

No response

[clrudolphi]

Following up on the discussion in Query PR#81

.Net: I've made a comparison of the type structures currently present in the Gherkin.CucumberMessages.Types folder against those in Cucumber/Messages. There are two differences that I'd like clarification on.

1. PickleStep has a property called Type. In Gherkin.C.M that property is of type StepKeywordType while in Cucumber/Messages the corresponding type is called PickleStepType. The former has 6 values while the latter has a subset of those (4 values). Complicating things a bit more, Cucumber/Messages has a type called StepKeywordType but it has 5 values.

Gherkin StepKeywordType	Messages PickleStepType	Messages StepKeywordType
Unknown	Unknown	Unknown
Context	Context	Context
Action	Action	Action
Outcome	Outcome	Outcome
Conjunction		Conjunction
Unspecified

Is this a reflection of enum values that are not used and can be dropped, or an error in specifying the schema for Messages?
What course of action do you recommend?

2. In Gherkin.CucumberMessages the Source type has a property called MediaType that is a string, while in Cucumber/Messages this property is of the enumeration called SouceMediaType. This might cause a problem if anyone has stuck a value in there that doesn't map cleanly to the enumeration. I haven't yet checked the Ast parser and AstMessageConverter to confirm what values they are using to populate this property. I'll come back if there is a confirmed problem.

[mpkorstanje]

Note: I've edited your table for clarity.

1a.

In the messages domain, there is a difference between pickle step type and a keyword type. The latter describes.. well grammar and isn't too useful if you want to verify that context, action and outcome are always specified in that order. The former does that by working out for each conjunction what its operation is.

The conjunctions are assumed to be the preceding pickle step type.

I.e. A conjunction after a context step is a context step, a action step after an action step and an out come step after an outcome step. So

Step	Stepkeyword type	Picklestep type
Given something	Context	Context
And something else	Conjunction	Context
When something	Action	Action
And something else	Conjunction	Action
Then something	Outcome	Outcome
But not something else	Conjunction	Outcome

1b.

Looking at the code it seems that StepKeywordType.Unspecified can only occur if a step did not have keyword type defined in the gherkin-languages.json which seems like it sould be an error. So you could replace all instances of Unspecified with Unknown.

2. I don't think this is a problem. We only support feature files and mostly in theory markdown files. So I wouldn't expect any custom values there.