D2L -- multiple grammars.

[jeffreykegler]

Unlike in LPeg, we'll have to deal with grammar sets. PEG has unified lexing/parsing, whereas typically Marpa will have at least a G1/L0 pair. (This is the same matter that I did not get right in my first pass at the LUIF.)

[rns]

D2L currently has no special syntax to define lexical rules, my thought (in the passage about luif.grammar_new() and luif.grammar_loadstring()) was that they can be inferred and the lexical grammar auto-built.

Now that D2L must produce KIR, is it ok to infer lexical rules to produce the linked lexical grammar for the structural?

On the second thought, '~' can mark lexical rules by being the first field in a rule table, just as '|' and '||' that can make the code simpler which can be a good thing to start with.

[jeffreykegler]

D2L needs to allow the grammar to be specified in some way. Even if we stick intially to G1/L0, in practical applications some L0 rules will need to be defined directly as L0 rules, rather than indirectly as charclasses and strings in L0 rules.

[rns]

Fixed in b9b2554.

[jeffreykegler]

Hmmm. Interesting idea.

On Sat, May 2, 2015 at 11:41 PM, rns notifications@github.com wrote:

Fixed in b9b2554
b9b2554
.

—
Reply to this email directly or view it on GitHub
#13 (comment).

[rns]

Another option can be a special key, e.g. lexical in the structural grammar table, which must contain all lexical rules. I think lexical rules can in theory be precedenced (can they?) and, in that case, '~' as the first field followed by '|' or '||' doesn't seem to be a good idea.