Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-05-21

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

All times shown according to UTC.

Time Nick Message
02:32 jeffreykegler joined #marpa
02:35 jeffreykegler ronsavage: re the FAQ -- maybe a special FAQ with lexing questions separated out might help folks, if you're interested in doing it.
02:36 jeffreykegler Most questions are about, not really Marpa itself, but lexing.
02:37 jeffreykegler It could be pieced together gradually, from various stray comments and responses to questions.
03:03 ronsavage joined #marpa
03:33 ronsavage jeffreykegler: OK. I'll start thinking about it.
04:22 jeffreykegler ronsavage: Thanks.
04:24 jeffreykegler The next step is perhaps that I should write up a short "tips" entry and maybe a couple of others, which can be used to start it off.
04:25 jeffreykegler Btw, one reason that Marpa's uptake has been slower than it might have been is that lexing has always been a roadblock, going back to when I simply did not provide one with Marpa
04:26 jeffreykegler Marpa at its core is a parser and the tradition is to assume that lexing is done by something else.  That's how, for example, yacc/bison work.
04:27 jeffreykegler And back when I started 6 years ago I saw yacc/bison as my main competition, which turned out in fact to be pretty much totally wrong.
04:27 jeffreykegler I wanted to kill yacc/bison off, when in fact they were already dead.
04:28 jeffreykegler And with them died most knowledge of the traditional lexing/parsing division, and how to write a lexer.
04:29 jeffreykegler A comeback I once got that I'll never forget --
04:30 jeffreykegler I tried to explain this lexer/parser distinction, and the two layer approach to someone, and they said something like
04:30 jeffreykegler "So in order to use Marpa to parse, I have to write a parser?"
04:32 jeffreykegler This person (whom I will not name) was particularly resistent to Marpa's charms, because he despised mathematical and Theory approaches ...
04:34 jeffreykegler A point of view which can be justified if you're looking at parsers, because the net result of 50 years of academic work on the topic is that most parsing is done using recursive descent,
04:34 jeffreykegler which dates to Ned Irons 1961 paper.
04:36 jeffreykegler So in terms of current practice we have all the academic work from 1961 to 2014 producing absolutely zero.
04:54 ronsavage joined #marpa
05:22 shadowpaste "jeffreykegler" at 108.70.210.81 pasted "Code for reply to Christopher Layne" (87 lines) at http://scsys.co.uk:8002/379375
06:59 ronsavage Is there any obvious stuff in your or my docs re lexing which I could copy into a lexer-oriented FAQ?
12:17 LLamaRider joined #marpa
15:25 jeffreykegler joined #marpa
15:27 jeffreykegler ronsavage: re http://irclog.perlgeek.de/​marpa/2014-05-21#i_8751986 -- let me think about that.
15:27 jeffreykegler I've certainly written a lot about lexing, but what is really suitable for a FAQ (and how much revison it would need) I'll need to mull over.
15:29 jeffreykegler Another factor is that with LATM, most of the answers about what techniques are best has changed.
15:29 jeffreykegler I frankly did not expect that LATM would have that much impact, so that's a pleasant surprise.
16:54 jeffreykegler Btw, one thing I am surprised that nobody has done yet is design a new language using the features new with Marpa --
16:55 jeffreykegler Feature 1: the ability to use a far wider syntax than before.
16:55 jeffreykegler Feature 2: the ability to use an ambiguous grammar.
16:56 jeffreykegler Feature 2 may seem strange, but it is how natural languages work -- they have very ambiguous grammars, but expect an utterance to have just one parse ...
16:57 jeffreykegler when it does not you clarify or report an error.
17:22 jeffreykegler joined #marpa

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