CoCoA » CoCoA-5

The involved lines are:

Lexer.C:391:            Keyword("Weights", TT_WEIGHTS),
Parser.C:178:    case TT_WEIGHTS: return "\"Weights\"";
Parser.C:1465:                case TT_WEIGHTS:

This is trickier than I originally thought.

The way the CoCoA-5 interpreter is designed a keyword is always seen as that keyword even it would make no sense in that context. Many computer languages work this way (FORTRAN is the only exception I know of), so it is probably a good idea.

In principle we could allow records to contain fields whose names are CoCoA-5 keywords; indeed PreprocessPoints does exactly that. Modifying the indexing function would not be hard: we could make rec["weights"] work as expected; it would be harder to make rec.weights work as expected since that would break the rule of a keyword always being seen as a keyword. However, if we were to do that then a printed record containing a field with name weights could not be cut-and-pasted as input -- for that to work we would have to modify parsing of a record which looks to be rather tricky (and would further break the rule about a keyword always being seen as a keyword).

Nevertheless I do find it unnatural/inconvenient that fields in a record are "arbitrarily" forbidden from having certain names.

I suppose the cleanest solution is to stick with the rule that keywords can never be used as identifiers in any context. It has the advantage that it is fairly easy to remember (assuming you have a list of keywords available), and very easy to implement (no need to change the existing code).

I note that the same limitation applies to tags in tagged values; this is probably unnecessary as a tag only ever appears as a string (right?).

I shall try to improve the error message when executing rec["weights"] which is unhelpful in its current form.

I should also modify RECORD::setField so that it gives an error if the field name is not a valid identifier for CoCoA-5. This ought to be easy.

Hint: Interpreter.C:2276--2288

Fixing the error messages, as suggested in my previous post, looks to be rather trickier than I had hoped :-(

In Interpreter.H:718 the mem fn setField is declared as virtual; why??

It works... more or less. A["weights"] gives the error I wanted, but A.weights still gives a less helpful error. To rectify this requires changing Parser.H:158; perhaps I'll add a new issue for this as it will probably take some time to do.