Fsharp: VFPT-like Unused Declaration and Symbol graying

[saul]

The Visual F# Power Tools featured automatic 'graying' of unused open declarations and unused symbols. (Although unused opens wasn't 100% reliable).

I think this would be a great feature to port over to the main extension.

[dhwed]

FYI @vasily-kirichenko, there are two issues I've found with the VFPT unused open feature:

• Type aliases were not considered, so if you had an open just for a type alias, the open would be marked as unused.
• If SomeNamespace cannot be opened due to a missing reference (e.g. a C# type extension exists in SomeNamespace for a type that is defined in an assembly not referenced by the current F# project), then open SomeNamespace.SubNamespace would always be marked as unused. I'm not too worried about this issue as I've already worked around it by putting the C# type extensions from the referenced project in the same namespace as the types they are extending.

[vasily-kirichenko]

I'm against porting the old VFPT code. It's too inefficient and, in case of graying out opens, is inherently unreliable. We should reimplement it from scratch, on lower level.

[cartermp]

@vasily-kirichenko any inkling as to where it would need to be reimplemented? Perhaps it could surface as a lightbulb codefix? I'm not sure how tied to ISymbol the Roslyn code is, but perhaps there's some stuff that could be ported over as well.

[vasily-kirichenko]

I'm not talking about the UI part, it's obviously need to be rewritten, but it's the easy part. I meant the actual algorithm. In VFPT I found all private and internal declarations in a file, then called GetUsesOfSymbolInProject _for each one_, which is not super efficient as we have to iterate all resolution in a project as many times as we have private declarations in a file, see https://github.com/Microsoft/visualfsharp/blob/master/src/fsharp/NameResolution.fs#L1372. Maybe that's not a big deal, but quadratic complexity is not a very good thing in general.

What about graying out unused opens, the implementation is hacky and unreliable and I was fixing it many times and I don't like to go this track one more time. Maybe later, when I know the compiler and the service better, I will re-implement it the right way, but not now.

[cartermp]

Sounds perfectly reasonable 👍

[saul]

@vasily-kirichenko just to be clear, are you saying that you don't think greying unused opens is worth it? And that greying unused symbols is worth the effort/is reliable (I've found it to be reliable in 2015 at least)

[vasily-kirichenko]

@saul Yes, graying unused private and internal symbols is reliable and simple, but not very efficient, I'm looking at it now. But unused opens... No, sometime later.

[vasily-kirichenko]

https://github.com/Microsoft/visualfsharp/pull/2358

[saul]

@vasily-kirichenko you're a Roslyn god

[dsyme]

Implemented!