Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2017-01-01

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

All times shown according to UTC.

Time Nick Message
01:12 jpgainsborough Idiosyncrat, what do you think of allowing regex type patterns to be embedded in a BNF
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
03:39 idiosyncrat_ jpgainsborough: it's an interesting idea (re https://irclog.perlgeek.de/​marpa/2017-01-01#i_13830213)
03:39 idiosyncrat_ Q's that arise are what kind of regexes?
03:40 idiosyncrat_ Perl's?  Lua's?  A custom type best suited to Marpa's features?
03:40 idiosyncrat_ Each of these has its own advantages and problems.
03:46 jpgainsborough I am most familiar with Perl's, so when I mention Regex, thats what I mean
03:46 jpgainsborough I have not looked at Lua's so I cannot have an opinion on that
05:04 idiosyncrat_ The problems to adopting Perl regexes are several
05:04 idiosyncrat_ They would tie Marpa to Perl, making it hard to create Marpa's for other languages.
05:06 idiosyncrat_ If you look at the POD, you'll see Perl regexes are vastly complex, with many, many operations which make no sense in the Marpa context.  It's hard to see that Perl regexes in full could or should be incorporated into Lua.
05:07 idiosyncrat_ Finally, even by XS standards, the interface to Perl regexes is gruesome, and IIRC very very poorly documented.
05:08 idiosyncrat_ I have changed my mind about this kind of thing.  In the past, where I've seen obstacles, Marpa users helped me see past them by creating prototypes.  This helped me by pointing out the way in dealing with the implementation gotchas ...
05:09 idiosyncrat_ and giving me a feeling for how nifty the feature would be.
05:09 idiosyncrat_ The SLIF comes out of a prototype by Peter Stuifzand.
05:10 idiosyncrat_ The idea of using 2 Marpa grammars to parse, one feeding into the other, comes from Andrew Rodland (hobbs).
06:11 hobbs and that wasn't even what I thought was the cool part of what I was doing :-P
06:11 hobbs but I'll admit the result of it was nice
06:14 hobbs what I built at the same time was a way (at a substantial cost in speed and memory both) to use arbitrary perl regexes as tokens, combined with marpa's speculative parsing, in an attempt to let me write grammars as free-form as, say, Parse::RecDescent only without the recursive descent :)
06:16 hobbs the two-grammar thing came later as a hack to deal with a really sort of unusual feature of the language I was parsing (TAP).
06:19 hobbs because TAP parsers (which grew up in real life long before anyone had the idea of formalizing the language) decide on a line-by-line basis whether a line of input is actually something they understand to be TAP, or "junk" that the program under test happened to output to the same stream
06:20 hobbs actual TAP lines participate in the grammar of a TAP stream, while junk lines are like ignored whitespace in other languages -- maybe kept around for curiosity's sake, but best swept under the rug as fast as possible.
14:56 sirdancealot joined #marpa
21:27 ronsavage joined #marpa
22:10 ronsavage joined #marpa
22:40 choroba joined #marpa
23:17 idiosyncrat_ joined #marpa

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