Yep all great points - and I agree POLS should be prioritized. I think an obvious solution may be to use a null keyword? Then it's possible to explicitly provide it as an "alternative" while at the same time possibly even being more clear than the current syntax (which can throw off the first timers). If one can bolster intuition while preserving principle of least surprise then it's probably a win/win. Thoughts?
On Jun 19, 2014, at 12:10, Jeffrey Kegler <[email protected]> wrote: > Christopher is right that I could make null RHS's part of the syntax of other > rules.. I don't because: > > 1.) Null alternatives would be easy to create by accident. > > 2.) Null alternatives are very special in the adverbs they do and do not > allow. > > 3.) Null rules can make parent symbols nullable, which can have tricky or > unintended results. > > 4.) Allowing null alternatives does not eliminate the non-orthogonality. It > just moves the issue elsewhere -- null alternatives in a prioritized rule > would almost certanily mean a typo, or an attempt to do something that won't > really work, and probably should be banned. > > All of which means that forcing a user to specify a null RHS as a special > separate rule is in keeping with the Principle of Least Surprise. Note that > one of the people who might be victim of these surprises is me -- in my own > SLIF grammars, I don't regret having to make my null rules separate. > > Finally, I welcome experimentation. It's quite possible to create your own > front-end to the SLIF, if you think you can improve on the SLIF and handle > null rules in a way that combines safety and elegance. As I've said before, > the SLIF exists because of the experiments by Peter Stuifzand and Andrew > Rodland, experiments which took Marpa beyond where I'd gone and showed me > that I could and should improve on Marpa's original interface (the NAIF). > > -- jeffrey > > On 06/19/2014 11:42 AM, Christopher Layne wrote: >> If you ignore the null rule it'll blow your mind less if you think about it >> as: >> >> A ::= B | 'hello!' >> A ::= >> >> There's no conceptual reason that the null rule couldn't be the last >> "alternative" but I believe the DSL grammar won't allow it hence it's a >> separate statement. >> >> On Jun 18, 2014, at 9:37, Paul Bennett <[email protected]> wrote: >> >>>> A ::= B >>>> A ::= 'hello!' >>>> A ::= # empty >>> Point not taken. Point, in fact, blew my mind. More research is required. > > -- > You received this message because you are subscribed to the Google Groups > "marpa parser" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/d/optout. -- You received this message because you are subscribed to the Google Groups "marpa parser" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
