[superseded] simplify code that handles hints/warnings

[timotheecour]

• WarningsToStr, HintsToStr hardcoded tables wasn't DRY; I removed it in favor of proc msgToHumanStr*(a: TMsgKind): string. This simplifies a bunch of code related to hints/warnings
• added hintLinkingVerbose to always show link command; happy to make that the new meaning for hintLink if we don't want to create a new hint for that (the existing hintLink isn't very useful as it doesn't show anything except "Link")

note

I could also use an inverted index table, eg:

const humanStrToMsg* = block: ## Inverted index of `proc humanStrToMsg`
  var ret = initTable[string, TMsgKind]()
  for a in TNoteKind:
    let str = a.msgToHumanStr.normalizeStyle
    ret[str] = a
  ret

proc normalizeStyle(a: string): string=
  # normalizes: _Foo_Bar => Foobar so that `cmpIgnoreStyle` doesn't need to be used

however, humanStrToMsg is only used in non performance critical places so I went for simplicity instead (with same time complexity as before PR)

[krux02]

Hmm, I am not convinced that this is any better or worse. It is just different.

[timotheecour]

I don't think it's good style to have to write stuff like this:

const	
   WarningsToStr* = ["CannotOpenFile", "OctalEscape",	
     "XIsNeverRead", "XmightNotBeenInit",	
     "Deprecated", "ConfigDeprecated",	
     "SmallLshouldNotBeUsed", "UnknownMagic",	
     "RedefinitionOfLabel", "UnknownSubstitutionX",	
     "LanguageXNotSupported", "FieldXNotSupported",	
     "CommentXIgnored",	
     "TypelessParam", "UseBase", "WriteToForeignHeap",	
     "UnsafeCode", "EachIdentIsTuple", "ShadowIdent",	
     "ProveInit", "ProveField", "ProveIndex", "GcUnsafe", "GcUnsafe2", "Uninit",	
     "GcMem", "Destructor", "LockLevel", "ResultShadowed",	
     "Spacing", "User"]	
 
    HintsToStr* = [	
     "Success", "SuccessX", "CC", "LineTooLong",	
     "XDeclaredButNotUsed", "ConvToBaseNotNeeded", "ConvFromXtoItselfNotNeeded",	
     "ExprAlwaysX", "QuitCalled", "Processing", "CodeBegin", "CodeEnd", "Conf",	
     "Path", "CondTrue", "CondFalse", "Name", "Pattern", "Exec", "Link", "Dependency",	
     "Source", "Performance", "StackTrace", "GCStats", "GlobalVar",	
     "User", "UserRaw", "ExtendedContext",	
   ]

which is completely redundant / DRY violation of the corresponding enum members; my PR gets rid of it and computes these automatically (and simplifies code that gets human string from enum)

[Araq]

which is completely redundant / DRY violation of the corresponding enum members; my PR gets rid of it and computes these automatically (and simplifies code that gets human string from enum)

Regardless of whether your chances are good or bad, this is just not true. These names are exposed to the command line so any internal compiler renaming would influence Nim's command line syntax.

[timotheecour]

These names are exposed to the command line so any internal compiler renaming would influence Nim's command line syntax.

but there are only 4 exceptions out of 60 names (hints/warnings), which I cleanly handle in msgToHumanStr :

proc msgToHumanStr*(a: TMsgKind): string =
   case a # handle exceptions here
   of warnInconsistentSpacing: result = "Spacing"
   of hintConditionAlwaysTrue: result = "CondTrue"
   of hintConditionAlwaysFalse: result = "CondFalse"
   of hintExecuting: result = "Exec"
   of hintLinking: result = "Link"
   else: # general case
     case a
     of warnMin..warnMax: result = ($a)["warn".len .. ^1]
     of hintMin..hintMax: result = ($a)["hint".len .. ^1]
     else: doAssert false

IMO it's cleaner to handle the general case (allowing for the occasional, rare exception) rather than duplicating the enum names in WarningsToStr, HintsToStr.
The rest of code is also simplified thanks to that.

[timotheecour]

/cc @Araq tests are green

[timotheecour]

ping @Araq

[Araq]

I like the old version better and objectively your code adds new imports / dependencies that the old variant didn't have.

[timotheecour]

removing the TODO; this has since been more or less implemented via #15775 and hint:link + --listcmd for the 2 aspects of this PR
pr_simplify_hint_warn