lang: parser: Grab entire variable names during lexing

This is inelegant, but I stronly believe that allowing dollar signs and their
variable name suffixes to both be first class citizens, is a bad idea.
The dollar sign should initiate a token that is exempt from any other
processing.

Note that it would be possible to construct a regex similar to the one for
IDENTIFIER, that only matches valid variable names. But it will be complex.
It seems more maintainable to just consume all allowed characters, and then
check validity with some simple golang in the parser actions.

Fixes purpleidea#728

lang/interpret_test/TestAstFunc2/lexer-parser0.txtar

@@ -1,16 +1,16 @@
 -- main.mcl --
 import "fmt"
-# unfortunately, for now `in` is a reserved keyword, see:
-# https://github.com/purpleidea/mgmt/issues/728
 $map = 55
-$fn = func($in) { # in is a special keyword
+# in is a special keyword, but accepted in a variable name
+$fn = func($in) {
 	13
 }
 test fmt.printf("%d", $fn(0)) {}
-func fn($in) { # in is a special keyword
+func fn($in) {
 	42 + $map
 }
 test fmt.printf("%d", $fn(0)) {}
 test fmt.printf("%d", fn(0)) {}
 -- OUTPUT --
-# err: errLexParse: parser: `syntax error: unexpected IN, expecting MAP_IDENTIFIER or IDENTIFIER` @5:2
+Vertex: test[13]
+Vertex: test[97]

lang/parser/lexer.nex

@@ -159,6 +159,15 @@
 			lval.str = yylex.Text()
 			return ARROW
 		}
+/\$[a-z0-9_\.]+/	{
+			yylex.pos(lval) // our pos
+			// drop the dollar sign
+			lval.str = strings.TrimLeft(yylex.Text(), "$")
+			if strings.Contains(lval.str, ".") {
+				return DOTTED_VARNAME
+			}
+			return UNDOTTED_VARNAME
+		}
 /\./		{
 			yylex.pos(lval) // our pos
 			lval.str = yylex.Text()
@@ -168,11 +177,7 @@
 			}
 			return DOT
 		}
-/\$/		{
-			yylex.pos(lval) // our pos
-			lval.str = yylex.Text()
-			return DOLLAR
-		}
+
 /bool/		{
 			yylex.pos(lval) // our pos
 			lval.str = yylex.Text()

lang/parser/lexparse.go

@@ -54,6 +54,7 @@ const (
 	ErrParseError             = interfaces.Error("parser")
 	ErrParseSetType           = interfaces.Error("can't set return type in parser")
 	ErrParseResFieldInvalid   = interfaces.Error("can't use unknown resource field")
+	ErrInvalidVariableName    = interfaces.Error("invalid variable name")
 	ErrParseAdditionalEquals  = interfaces.Error(errstrParseAdditionalEquals)
 	ErrParseExpectingComma    = interfaces.Error(errstrParseExpectingComma)
 )

lang/parser/parser.y

@@ -93,7 +93,8 @@ func init() {
 %token OPEN_BRACK CLOSE_BRACK
 %token IF ELSE
 %token BOOL STRING INTEGER FLOAT
-%token EQUALS DOLLAR
+%token EQUALS DOTTED_VARNAME
+%token UNDOTTED_VARNAME
 %token COMMA COLON SEMICOLON
 %token ELVIS DEFAULT ROCKET ARROW DOT
 %token BOOL_IDENTIFIER STR_IDENTIFIER INT_IDENTIFIER FLOAT_IDENTIFIER
@@ -1406,11 +1407,13 @@ undotted_identifier:
 	}
 ;
 var_identifier:
-	// eg: $ foo (dollar prefix + identifier)
-	DOLLAR undotted_identifier
+	UNDOTTED_VARNAME
 	{
 		posLast(yylex, yyDollar) // our pos
-		$$.str = $2.str // don't include the leading $
+		if $1.str == "" || strings.HasPrefix($1.str, "_") || strings.HasSuffix($1.str, "_") {
+			yylex.Error(fmt.Sprintf("%s: %s", ErrInvalidVariableName, $1.str))
+		}
+		$$.str = $1.str
 	}
 ;
 colon_identifier:
@@ -1439,13 +1442,24 @@ dotted_identifier:
 		$$.str = $1.str + interfaces.ModuleSep + $3.str
 	}
 ;
-// there are different ways the lexer/parser might choose to represent this...
 dotted_var_identifier:
-	// eg: $ foo.bar.baz (dollar prefix + dotted identifier)
-	DOLLAR dotted_identifier
+	DOTTED_VARNAME
+	{
+		posLast(yylex, yyDollar) // our pos
+		for _, ident := range strings.Split($1.str, ".") {
+			if ident == "" || strings.HasPrefix(ident, "_") || strings.HasSuffix(ident, "_") {
+				yylex.Error(fmt.Sprintf("%s: %s", ErrInvalidVariableName, $1.str))
+			}
+		}
+		$$.str = $1.str
+	}
+|	UNDOTTED_VARNAME
 	{
 		posLast(yylex, yyDollar) // our pos
-		$$.str = $2.str // don't include the leading $
+		if $1.str == "" || strings.HasPrefix($1.str, "_") || strings.HasSuffix($1.str, "_") {
+			yylex.Error(fmt.Sprintf("%s: %s", ErrInvalidVariableName, $1.str))
+		}
+		$$.str = $1.str
 	}
 ;
 capitalized_res_identifier: