RFC FS-1108 - Allow more undentations and remove inconsistencies

[Happypig375]

RFC Allow more undentations and remove inconsistencies

[Happypig375]

Linking suggestions:

• Allow undentation for constructors
• Relax indentation rules on mutually recursive functions when using with CE
• Relax some of the indentation rules
• Add new undentation rule for match with
• Relax indentation rules on anonymous records
• Allow undentation in method and constructor calls
• Alignment requirements for mutually recursive functions are non-intuitive
• Mutual Recursion with Async Workflow Possible Compiler Issue
• Composed function leads to unexpected indentation warning
• Warning only when function application is inside SynExpr.IfThenElse

[Happypig375]

3 more:

• Improve Indentation Rules for Generic Signatures
• Indentation of new lines should be dependent on the start of the previous line, not the end of it
• Relax indentation rules on Records

[auduchinok]

Here's another case, if it can be added. 🙂

[Happypig375]

@auduchinok I haven't included fsharp/fslang-suggestions#514 in here, which your case belongs to. Not sure if I can fix attribute indentation just as easily with the others.

[Happypig375]

I can't run FSharpQA locally because Perl is complaining about Win32::Process not being installed and there is no instructions on how to deal with this error, while trying to use StrawberryPerl's bundled cpanm to install Win32::Process tells me that ExtUtils::Manifest must be installed first, and it cannot be installed because I need... ExtUtils::Manifest? I can't resolve this circular reference.

[dsyme]

@Happypig375 I'll go through the suggestions and check the approvals. Best to keep to the approved suggestions to avoid having to factor out the unapproved ones?

[Happypig375]

@dsyme As written in the draft RFC, this tries to be a general fix to a lot of suggestions with the same underlying cause. Implementing the suggestions one by one by introducing special cases after special cases just introduce more inconsistency.

[dsyme]

The changes look plausible - please keep them minimal as we would want to check this very carefully

The RFC spec also looks plausible, I'll comment there seperately

[nojaf]

I've played around with this and was able to verify all my samples that used to produce warnings.
It all worked fine and I'm looking forward to this. Many thanks again @Happypig375 🎉!

[dsyme]

Marking as approved, as just two minor test cases need to be added (though I think they may actually already be under test)

[dsyme]

Some changes are needed, I've identified a failing scenario

[dsyme]

@Happypig375 I put this

<OtherFlags>$(OtherFlags) --langversion:preview</OtherFlags>

in

src/fsharp/FSharp.Compiler.Service/FSharp.Compiler.Service.fsproj 

and the compilation fails

C:\GitHub\dsyme\fsharp\src\fsharp\absil\ilwritepdb.fs(611,15): error FS0058: Possible incorrect indent
ation: this token is offside of context started at position (608:17). Try indenting this token further
or using standard formatting conventions. [C:\GitHub\dsyme\fsharp\src\fsharp\FSharp.Compiler.Service\
FSharp.Compiler.Service.fsproj]

The problem can be reduced to this:

beginmatch1with|1->()|2->
f()end

Thanks

(It's possible we'd think that this shouldn't be allowed, but it is)

[dsyme]

@Happypig375 Can you fix this one? #11772 (comment) thanks

[Happypig375]

Will come back a few days later, I'm busy now

[Happypig375]

I'm back

[Happypig375]

@dsyme Fixed

[dsyme]

Thank you for addressing the last remaining items

[goswinr]

Thanks for this work, will this produce a warning for this quirk?

Given this this operator for evaluation of side effects in pipelines

let inline(|>!)a f = f a ; a 

and this code snippet

if1=1then"same"else"different"|>! printfn "The numbers are %s."|> printfn "Yes, they are %s."

This never prints The numbers are same but it does print Yes, they are same. (Also asked on SO)

[Happypig375]

This PR does not add additional warnings.

[goswinr]

Thanks for clarification @Happypig375, I guess my desired warning is tracked in fsharp/fslang-suggestions#806