Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2015-09-21

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

All times shown according to UTC.

Time Nick Message
00:01 ronsavage For those using Pg, the problem with utf8 was that when DBD::Pg V 3.0.0 came out I had set the pg_enable_utf8 option (in the DBI connection attributes stored in config files) to 0 for one module, and left it at the default -1 for the other 2 modules (of the 3 causing problems). Now I'm on V 3.5.1 and have re-written the utf8 handling code in those modules, it took me a long time to track the issue down.
00:46 idiosyncrat_ MerittClarksdale: I guess it's recursive descent?
00:46 MerittClarksdale thats what i heard
00:47 idiosyncrat_ I do wonder if they might not have a smaller and more flexible parser if they used Marpa.
00:47 idiosyncrat_ And if they might no be able to produce vastly better error messages.
00:47 idiosyncrat_ ronsavage: Thanks!
00:55 idiosyncrat_ By "smaller" I mean the amount of code -- to be sure the Libmarpa library is not tiny.
00:57 jdurand_ ronsavage: thx &  well done -;
00:58 MerittClarksdale ive often wondered about that, if a c/c++ compiler with marpa would outdo all others
01:20 idiosyncrat_ Parsing technologies are amazing "sticky".
01:21 idiosyncrat_ LALR, despite its difficulties, managed to be considered *the* parsing technology for 3 decades.
01:22 idiosyncrat_ And using recursive descent to parse C means rolling the clock back to 1961.
01:23 idiosyncrat_ Most of you will have read it, but I tell this story in http://jeffreykegler.github.​io/Ocean-of-Awareness-blog/
01:23 idiosyncrat_ which remains my most popular blog post, by a long shot.
02:31 ronsavage jk: In a sense, it's your saddest post too.
03:07 idiosyncrat_ ronsavage: in that progress in parsing is so slow?
05:11 rns left #marpa
06:20 ronsavage joined #marpa
06:33 ronsavage jk: Basically yes, in that the glacier-on-Valium-like progress in parsing has been so slow (until recently :-).
09:53 ronsavage joined #marpa
10:37 lwa joined #marpa
14:38 idiosyncrat joined #marpa
14:40 idiosyncrat ronsavage: It seems that academic decided the parsing problem was solved -- by LALR/yacc
14:40 idiosyncrat Even today, you'll find LALR is taught as *the* solution to parsing.
14:41 idiosyncrat And you don't get a lot of funding to pursue solutions to solved problems. :-)
14:42 idiosyncrat I recognized this problem when I started Marpa, and that is why the dual strategy -- papers *and* implementation.
14:43 idiosyncrat Because I thought my writings would get little attention until practitioners, actually using the new parser, forced attention to the Earley/Leo/Marpa algorithm.
14:44 idiosyncrat Pessimistic as I was, I am surprised at how slow the process is.
14:45 idiosyncrat Because a 2nd factor had kicked in -- with academics having decided parsing was solved and/or just not a rewarding area to pursue, ...
14:45 idiosyncrat they stopped teaching it.
14:47 idiosyncrat BNF was a universal skill among programmers when I graduated but, after decades when it was close to useless, programmers stopped learning it ...
14:47 idiosyncrat and I mean even some of the top people in field.
14:48 idiosyncrat And without BNF, you don't really know what the parsing problem is.
14:48 idiosyncrat So the world divided between those who thought the parsing problem was solved, and those who didn't realize there was a problem.
14:52 idiosyncrat There seems to be a new interest in BNF and Earley's.
14:53 idiosyncrat Perhaps it's just conincidence, but it might be a bow wave of the Marpa meme.
17:02 MerittClarksdale BNFs are a good high level feature that makes writing a compiler accessible to anyone
17:03 MerittClarksdale It is the SQL of parsers, as with no need to know the detail of hash functions to be able to write database code
17:04 MerittClarksdale Even experts can benefit from high level features as it improves productivity
22:08 ronsavage joined #marpa

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