Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2015-09-12

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:13 ronsavage joined #marpa
01:08 koo7 joined #marpa
01:19 idiosyncrat1 joined #marpa
02:05 ronsavage I'm part-way thru splitting Chapter 3 (the FAQ) of the Marpa Papers out into a separate github repo, but, even more importantly, it's lunchtime!
03:59 koo7 joined #marpa
04:00 koo7 joined #marpa
04:44 idiosyncrat joined #marpa
05:41 idiosyncrat Good night!
05:53 idiosyncrat1 left #marpa
06:12 pczarn joined #marpa
10:53 idiosyncrat joined #marpa
15:10 MerrittClarksdal joined #marpa
15:10 djns hi
15:45 koo7 joined #marpa
16:30 pczarn hi
16:31 pczarn I have tracing, events and actions working in terms of the original grammar.
17:47 pczarn My precomputed grammar type has a function from internal dotted rules to sets of external dotted rules.
17:49 idiosyncrat sets?
17:50 idiosyncrat as in not singletons?
17:51 pczarn Singletons? Nulling rule elimination can add more than one dotted rule.
17:53 idiosyncrat An external rule can create many internal rules, but in my scheme not vice versa
17:56 pczarn Yes, a binarized nullable rule can create at most 3 internal rules (plus a "nulling" rule)
18:01 pczarn If there's an internal dotted rule where a nullable symbol was eliminated, debugging should show nulling rules.
18:03 pczarn Nulling events can be implemented the same way.
18:15 idiosyncrat Marpa does not allow nulling rules, just nulling symbols.
18:15 idiosyncrat Any nulling tree is pruned back to its topmost symbol.
18:16 idiosyncrat Not I am saying nulling not nullable -- nullable rules are rewritten to either nulling symbols or non-nullable rules
18:17 idiosyncrat The semantics and structure of the missing nulling rules could easily be replaced -- they don't depend on the input, but are constants and in fact could be precomputed from the grammar.
18:18 idiosyncrat But Marpa does not do this, because nobody seems to miss nulling rules -- very few Marpa users seem to even notice they're gone.
18:19 idiosyncrat This is after years of experience with actual users.
18:19 idiosyncrat pczarn: again, you can implement as you wish, and I'd look forward to benefiting from the result of any experimentation
18:20 idiosyncrat ... but I am sharing the experience with Marpa
18:22 pczarn What about nulling events and nulling forests?
18:23 idiosyncrat A nulling forest is pruned back to its topmost symbol.
18:23 idiosyncrat Nulling events are propagated up the forest to that topmost symbol.
18:25 pczarn Ah. Thank you. A partial function to external dotted rules, then.
18:26 idiosyncrat Again, you can implement as you like, but this system of null handling has worked very smoothly in Marpa ...
18:26 idiosyncrat despite my early fears otherwise.
18:28 idiosyncrat I was surprised -- I expected that *somebody* was going to miss his nulling parse forest, but no.
18:28 idiosyncrat pczarn: Maybe one of your users will be the first to complain. :-)
18:30 pczarn The bocage doesn't have to be aware of nulling parse forests after construction, though.
18:31 idiosyncrat Right.
18:31 idiosyncrat Traces of the forest remain in that any events it contained were propagated up to the symbol that replaced it.
18:32 idiosyncrat But there is just a nulling symbol representing what was potentially a huge, complex structure.
18:32 idiosyncrat One way to think about it ...
18:32 idiosyncrat If the problem is about the structure of nulling forests, it no longer has anything to do with the input ...
18:33 idiosyncrat and therefore is not really a parsing problem anymore -- it's combinatorics.
19:03 pczarn joined #marpa
19:30 idiosyncrat joined #marpa
19:43 MerittClarksdale Greetings
22:05 jdurand joined #marpa
23:26 idiosyncrat joined #marpa

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary