Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2014-07-29

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

All times shown according to UTC.

Time Nick Message
01:03 Aria Unicode woes eventually end. But the tunnel is the longest ever built.
01:26 jeffreykegler1 Yeah, well remember me on the other side. :-)
01:35 Aria Hehe.
01:38 jeffreykegler1 By the way, Roberto himself in PiL suggests slnunicode.
03:11 ronsavage jeffreykegler: We'll remember you /if/ you make to the other side......
05:06 ronsavage left #marpa
10:13 jdurand joined #marpa
10:14 jdurand I was reading http://en.wikipedia.org/wiki/VT​D-XML#Virtual_Token_Descriptor and have the fealing that Earley implementation in Marpa looks like what they call a Virtual Token Descriptor -;
12:22 jdurand_ joined #marpa
13:54 Aria Huh.
14:57 jeffreykegler joined #marpa
14:58 jeffreykegler http://vtd-xml.sourceforge.net/faq.html#​*_Can_you_explain_the_GPL_license_a_bit
14:58 jeffreykegler VTD-XML claims to be both GPL and patented.
14:59 Aria Ayuh.
14:59 Aria Damn software patents. I really hope we see them eliminated in the next decade.
15:02 jeffreykegler Glancing at it, there's no description of the algorithm, just of data structures.
15:03 jeffreykegler It's "advantage" over DOM (and Marpa) would be that it does not use pointers.
15:04 jeffreykegler I will note that Marpa could be implemented without pointers also, which can very much speeds up parsing, by moving the cost into the evaluation phase.
15:05 jeffreykegler As an aside, I know when I brought out Marpa years ago that it would be treated with skepticism.
15:06 jeffreykegler Because if you've been around a while, you've seen new parser after new parser comes out (one at least every two years) ...
15:06 jeffreykegler with revolutionary claims, but which then disappears.
15:07 jeffreykegler Marps's distinction is that there's theory behind it, and the theory is open.
15:08 jeffreykegler That is, it's not just a list of claims, and "Here try it!  It works fast on these examples."
15:08 Aria Heh, yeah.
15:09 jeffreykegler Few parsing algorithms are so bad that there are not some real-life exampples on which they perform remarkably well.
15:09 Aria "Why yes, you have a nice, easy LL grammar. Of course it performs well with your LL-shaped parser toolkit. Go away"
15:09 Aria Definitely.
15:11 jeffreykegler So I expect when I brought out Marpa, especially experienced folks, would say "Yeah.  Right, guy." ...
15:12 jeffreykegler until they say that it was theoretically strongly-based and ...
15:12 jeffreykegler also accomplished things that no other parser does at all,
15:13 jeffreykegler such as combining procedural parser with syntax-driven ...
15:13 jeffreykegler and parsing (just for a start) LR(k) in O(n) for all k.
15:13 jeffreykegler -> * "until they say" -> "until they saw"
16:54 jeffreykegler1 joined #marpa
17:14 jeffreykegler1 I just emailed a question about LuaJIT versions to the G+ group.  I expect most work on Kollos will take place here on the IRC channel ...
17:14 jeffreykegler1 but some of those who can contribute may not backlog this channel, so I like to "reach out" at crucial points.
17:16 jeffreykegler1 The TL;DR of the G+ post is that I'm thinking (at least at first) of requiring LuaJIT 2.0.2.
18:08 jdurand_ Re http://irclog.perlgeek.de/​marpa/2014-07-29#i_9099271 - in fact I say I was not impressed, and is glad to be able to generate thousands of lines of C Marpa grammar implementation, using my private EBNF parser -;
18:28 jdurand_ Re https://groups.google.com/d/msg/mar​pa-parser/HQfufECQ7Oo/PCBvxMkooC4J - I post here since I am not lua aware, yet. Nevertheless I want to point out the LuaJIT exposes the /only/ thing that I find really useful in all my object oriented programmer experience: exceptions
18:51 jeffreykegler1 By the way, revisiting that XML parser ...
18:52 jeffreykegler1 always interesting and often ignored are errors.  It's a lot harder to produce quality error messages than it is to parse quaranteed-100%-perfect files.
18:54 jeffreykegler1 The XML standard tries to get around the issue, IIRC, by stating the XML parsers do *not* have to deal with incorrect files, and that as far as the standard is concerned, there is not such thing as liberal XML.
18:55 jeffreykegler1 But in real life, broken and oddball files happen, and you want the parser to be as helpful as possible.
18:59 jdurand_ That's where a Marpa parser will be a big value added compared to others
18:59 jeffreykegler1 I am looking forward to it.
19:00 jdurand_ Since the very beginning I appreciated the pertinence of the error: where exactly at least, if not more important: the exact context - well both are important, and Marpa provides them with perfect accuracy
19:00 jdurand_ c.f. my latest google post then - I have a roadmap, it is just a question of time
19:01 jdurand_ very recently I have decided to drop forever ICONV in facour of ICU, because ICU natively handles streaming (keeping internally "partial" code points) - with ICONV it requires user-space logic that is just pain
19:02 jdurand_ "in favour"

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