rst.whichMsgClass is buggy: relies on $enum instead of $ast(enum)

[timotheecour]

/cc @a-mr
root cause for #17257 (comment)

Currently we have a more basic problem that I haven't learned how to handle rstMessage invocation in test, it just throws AssertionDeffect by some reason, which is ridiculous :-(

the problem is this code:

  MsgKind* = enum          ## the possible messages
    meCannotOpenFile = "cannot open '$1'",
    meExpected = "'$1' expected",
    ...
    meGeneralParseError = "general parse error",
    meInvalidDirective = "invalid directive: '$1'",
    mwRstStyle = "RST style: $1"

proc whichMsgClass*(k: MsgKind): MsgClass =
  ## returns which message class `k` belongs to.
  case ($k)[1]
  of 'e', 'E': result = mcError
  of 'w', 'W': result = mcWarning
  of 'h', 'H': result = mcHint
  else: assert false, "msgkind does not fit naming scheme"

relies on $k which doesn't reflect the ast symbol of the enum

Additional Information

1.5.1 07df4fe

proposed solution

I'd like all 3 things below:

fix 1

use range:

const
  MsgKindErrors = meCannotOpenFile..mwRedefinitionOfLabel
  MsgKindWarnings = mwRedefinitionOfLabel.next..mwRstStyle
  MsgKindHints = mwRstStyle.next..MsgKind.high

then:

proc whichMsgClass*(k: MsgKind): MsgClass =
  case k
  of MsgKindErrors: mcError
  of MsgKindWarnings: mcWarning
  of MsgKindHints: mcHint

fix 2

replace

    # TODO: find out hot to catch warning here instead of throwing a defect
    echo input8
    expect(AssertionDefect):
      let output8 = input8.toHtml

(eg from #17257)
accordingly now that the bug gets fixed, ditto for all (or most) of the other expect(AssertionDefect):

fix 3 + proposal: enumutils.symbolName

(EDIT: => #17281)
not needed to fix this issue but still useful in general (and would provide an alternative way to fix this): add proc nativeString(a: enum): string which returns the native string for an enum, ignoring the string definition. This should address remaining concerns for using this kind of enums:

type A = enum
  a0 = "human name"

(refs #15775 (comment))
since it gives both stringified form (most common case) as well as access to native name ("a0"):

let x = A.low
assert $x == "human name" # common case
assert nativeString(x) == "human name" # when needed

[a-mr]

To have symbolName (fix 3) is generally useful. And it's closest to the original idea of code $k[1]; to change it to k.symbolName[1]

[timotheecour]

@a-mr fix 3 has been merged so you can go ahead and use it to close this issue