Make SemanticModel's Analyze*Flow methods public. Remove corresponding extension methods.

[JoshVarty]

Both AnalyzeControlFlow and AnalyzeDataFlow are marked internal within SemanticModel.

They're exposed via extension methods in ModelExtensions.AnalyzeControlFlow and ModelExtension.AnalyzeDataFlow

I don't think accessing these methods through extension methods provides any value. I propose the extension methods be removed (for the above methods, and their other overload) and we simply mark the corresponding methods within SemanticModel as public.

If the team agrees, I can go ahead and submit a PR for this.

[jmarolf]

This looks like a mistake from when we unified the interface ISemanticModel to the abstract class SemanticModel. Go ahead and submit a PR.

[gafter]

If I recall, the language-specific extension methods are type safe (e.g. Taking language-specific StatementSyntax)

[JoshVarty]

There are language-specific extension methods and they should be left alone.

But I don't think that the language-agnostic extension methods serve a purpose that wouldn't be achieved by making the following methods public instead of internal.

SemanticModel.AnalyzeControlFlow(SyntaxNode)
SemanticModel.AnalyzeControlFlow(SyntaxNode, SyntaxNode)
SemanticModel.AnalyzeDataFlow(SyntaxNode)
SemanticModel.AnalyzeDataFlow(SyntaxNode, SyntaxNode)

[DustinCampbell]

That seems fair to me.

[JoshVarty]

As discussed in #190 this is by design.