LexRaku confuses an object attribute named y with the y/// operator

[pryrt]

the Raku lexer is confused by an object attribute of y or tr. In certain syntax, LexRaku interprets it as the y/// or tr/// operator, not the attribute of the object instance obj.

verified using: SciTE 5.4.3 Scintilla:5.4.3 Lexilla:5.3.1:

code to lex(from https://uzluisf.gitlab.io/raku-by-example/classes-and-objects):

class Point2D {
    has Str $!id = 'P2'; # private attr. with default value
    has Numeric $.x;     # attr. with read-only accesor
    has Numeric $.y;     # same as above
}

# Initializing an object instance. Private attributes
# cannot be initialized during object construction.
my $p1 = Point2D.new(x => 3, y => 4);

say "x: ", $p1.x; # invoking the read-only accessor
say "y: ", $p1.y; # same as above

# An attribute with a read-only accesor cannot be modified
# from outside the class. To do so, add the `is rw` trait 
# to make it read-write. For example: `has Numeric $.x is rw;`

# this semicolon ends the y/// operation;

say "ok";

(to maintain compatiblity with Perl, Raku still has y/// as an alias of the tr/// operator . I confirmed that .tr behaves identically to .y in the examples above)

[nyamatongwe]

For say "y: ", $p1.y;, it is checking the next character ';' with IsValidRegOrQAdjacent which allows most punctuation including ; to be the delimiter and then switching to style SCE_RAKU_REGEX. So y;dol;wne; is a valid translation to the lexer.

Even space is valid to IsValidRegOrQAdjacent so y is treated as the start of a translation but is then stopped early inside ProcessValidRegQlangStart.

It is possible that the set of valid delimiters for y and tr operators is more constrained than for rx and IsValidRegOrQAdjacent or its caller should check that.