Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2016-02-29

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

All times shown according to UTC.

Time Nick Message
00:44 idiosyncrat_ joined #marpa
00:45 idiosyncrat_ pczarn: and you are adding rules and symbols to the grammar dynamically?  During a parse?
00:50 koo7 joined #marpa
02:46 ronsavage joined #marpa
02:48 ilbot3 joined #marpa
02:48 Topic for #marpa is now Start here: http://savage.net.au/Marpa.html - Pastebin: http://scsys.co.uk:8002/marpa - Jeffrey's Marpa site: http://jeffreykegler.github.io/Marpa-web-site/ - IRC log: http://irclog.perlgeek.de/marpa/today
05:44 idiosyncrat_ Good night!
08:57 black_ant joined #marpa
10:06 koo7 joined #marpa
11:44 kaare_ joined #marpa
14:56 maybekoo5 joined #marpa
14:56 Idiosyncrat joined #marpa
16:27 pczarn joined #marpa
16:29 pczarn Idiosyncrat: no, I don't change the grammar during parsing
16:29 pczarn improved tracing: https://gist.github.com/pc​zarn/22df76657b11f8167fc8
16:33 Idiosyncrat Why both a bit with the symbol and a bit vector?
16:33 Idiosyncrat I end up using both.
16:33 Idiosyncrat s/Why both/Why *not* both/
16:33 Idiosyncrat AFK
16:50 pczarn Using both is a fine solution.
16:51 pczarn Another problem is that my grammar types are generic over symbol type. This means adding e.g. a bit makes the interface and thus code more complex.
16:54 pczarn Perhaps that's a mistake. Symbols are integers smaller than N so that a bit matrix NxN fits in the memory.
16:56 pczarn What's the largest grammar anyone would ever need to use or analyze?
19:38 maybekoo5 joined #marpa
21:31 ronsavage joined #marpa
22:09 idiosyncrat_ joined #marpa
22:10 idiosyncrat_ pczarn: IIRC some practical hand-generated grammars can have over 1000 symbols.
22:11 idiosyncrat_ Jean-Damien, what is your largest?
22:12 idiosyncrat_ Auio-generated grammars, much greater use of which I hope Marpa will enable, could be even larger.
22:13 idiosyncrat_ Never store the same data twice is a good guideline, but it's a common fault in programs to turn good guidelines into bad rules.
22:14 pczarn joined #marpa
22:14 idiosyncrat_ In the case is marking symbols as terminals, I ended up often having a pointer to the symbol in local context, and simply wanting the bit, but ..
22:14 idiosyncrat_ I also often needed to do bit manipulations involving all symbols or at least all terminals, ...
22:15 idiosyncrat_ So my experience is that you're likely to want *both* a bit in the symbol's "object", ...
22:15 idiosyncrat_ and a bit vector, so that you don't wind up having to built it every time you use it.
22:22 pczarn hmm, I checked, I'm currently building the bitset of terminals in 3 analyses, and 2 of them (based on RHS closure) need a bitset anyway. That's not counting FIRST and FOLLOW set computation.
22:25 pczarn I'm interested in the size of binarized grammars, I guess they easily reach thousands of symbols.
22:28 pczarn the important number is approximately sum(rule.rhs.length for rule in grammar.rules)
22:29 pczarn + number of symbols before binarization

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