[antlr-interest] Disregarding subrule ordering

[Bryan Ewbank]

To make a perhaps subtle distinction, this is a /semantic/ error, so
is more easily done in a separate pass (or are them fighting words?). 
In other words, think about adding this to a treeparser, rather than
the initial parse.

The parser would simply consume parameters passively:

    object :
        KW_OBJECT^ ID params
        ;

    params :
        LEFT_CURLY^
         ( param0 | param1 | param2 ... )*
        RIGHT_CURLY!
        ;

Now, the tree parser would look at the params object in the context of
the ID'd object and know (from a lookup table?) which elements are
available, as well as which have been seen...

    object : #( OBJECT id:ID params[#id->getText()] );

    params
    [ const string objName ]
    {
        SetOfNames legalParameters = getLegalParameters(objName);
        SetOfNames reqdParameters = getRequiredParams(objName)
        SetOfNames seen;
        string paramName;
    }
    :   #( LEFT_CURLY
            (
                paramName=parameter[legalParameters]
                {
                    // limited to one instance
                    if seenParameters contains paramName exists then
                        error ...
                    add paramName to seenParameters
                }
            )*
        )
        {
            // force required to exist
            for every p in reqdParameters
                if seenParameters does_not_contain p then
                    error ...
        }
    ;

    parameter
    [ const SetOfNames legalParameters ]
    returns [ string paramName ]
    :
        #( ... ??? ... )
        {
            if legalParameters does_not_contain paramName then
                error ...
        }
    ;

On 12/1/05, Royne Borrud <royne.borrud at gmail.com> wrote:
> Hi.
> My goal is to parse a statement similar to this:
>
> object name {
>     optionalParam1
>     param1
>     param3
>     param2
>     optionalParam3
> }
>
> My problem is that while I do not want to enforce a specific ordering
> of the parameters, I do want to make sure that all three parameters
> occur once and only once, and that the optional parameters occur once
> or not at all. Is this possible without adding specific code to handle
> this? I'm beginning to think not... :/
> If there isn't, has anyone got any hints for how to go about this?
>