Define the relationship between qualified identifiers and path literals

[RCHowell]

For qualified identifiers, are we allowing the [ ] notation in paths?

Today, we have several places in the grammar which use such paths, but I’m unsure if these are supposed to be that way. The existing grammar and defines a special "simple path"

pathSimple
    : symbolPrimitive pathSimpleSteps*;

pathSimpleSteps
    : BRACKET_LEFT key=literal BRACKET_RIGHT             # PathSimpleLiteral
    | BRACKET_LEFT key=symbolPrimitive BRACKET_RIGHT     # PathSimpleSymbol
    | PERIOD key=symbolPrimitive                         # PathSimpleDotSymbol
    ;

And the AST has used the general expr in place of where a qualified identifier would be expected. In the past, this has allowed us to use an expr.path type where a qualified identifier is expected.

This rule is used across many grammar rules where one would expected qualified identifiers. For example,

removeCommand
    : REMOVE pathSimple;

Which means we can do REMOVE hello[0][goodbye] which looks wrong?

Our spec has qualified identifier as simply

<qualified identifier> ::= <identifier> ('.' <identifier>)+

<identifier> ::= ($ | _ | letter) ($ | _ | letter | digit)*
               | " <quoted-body> "

Which would suggest hello[0][goodbye] is not a qualified identifier, implying we have a bug in our grammar. But do we want to expand qualified identifiers to include the bracket path notation? It seems useful!