Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2016-05-07

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

All times shown according to UTC.

Time Nick Message
00:33 idiosyncrat_ Thinking aloud about future directions:
00:34 idiosyncrat_ I plan to extend the SLIF to have it's own Lua-based semantics, which will make it much more efficient for custom processing -- Perl callbacks are, and will remain, expensive.
00:35 idiosyncrat_ But at the moment I don't think I'll do that in Marpa::R3 ...
00:36 idiosyncrat_ the reason being that forcing the user to be "bi-lingual" is an obstacle to new users, and some may not consider enhanced efficiency worth it.
00:37 idiosyncrat_ So rather than make Lua's presence inside Marpa::R3 visible, I may, when the time comes to expose the Lua semantics, declare Marpa::R3 "stable, bug fix only" ...
00:38 idiosyncrat_ and do the Lua stuff in a new, alpha, experimental, Marpa::R4.
01:11 idiosyncrat_ My thinking is that if I start to add the Lua semantics to Marpa::R3, that in the minds of some users I will be wrecking the interface:
01:13 idiosyncrat_ it will probably have to be an all-or-nothing thing -- Lua and Perl's data conventions are dramatically different.  An interface can try to hide the big differences, but that's just leaving the user to be pecked to death by a host of trick minor differences.
01:14 idiosyncrat_ The reason I am doing this thinking-out-loud is that I am interested in comments.
01:15 idiosyncrat_ I usually over-estimate people's readiness to adapt to new ideas, but I have also under-estimated it in the past -- the biggest example as LTM vs. LATM --
01:16 idiosyncrat_ LTM (longest token match) is traditional and easier theoretically while
01:17 idiosyncrat_ LATM (longest acceptable token match), that is, the longest token that the grammar will actually accept, is completely new, and requires a bit of parsing theory to explain.
01:17 idiosyncrat_ So I went with LTM, and only later introduced LATM as an option.
01:19 idiosyncrat_ Turns out for most people LATM is DWIM, so its conceptual difficulties are irrelevant.  And despite the fact that everything else is LTM, users have developed no attachment at all to it.
01:22 idiosyncrat_ So perhaps a bi-lingual Marpa::R3 is not as big an obstacle as I think it is.
01:49 ronsavage Appreciate the thinking-out-aloud. Big question: If Lua is visible in R3, is it mandatory, or could uses continue with SLIF? If mandatory, I'm opposed. Clean up everything else first.
01:50 ronsavage Reasoning: You don't want problems with Lua and problems with non-Lua changes all at the same time. As for R3 and R4, I think that's a clear-cut split for people to understand.
01:50 ronsavage 'uses' => 'users'!
04:23 idiosyncrat_ joined #marpa
04:24 idiosyncrat_ ronsavage: most of the potential ways of making Lua visible, make it mandatory.
04:25 idiosyncrat_ Because of the inconsistency in their ideas of data, the semantics have essentially to be one or the other.
04:26 idiosyncrat_ There are CPAN modules which translate back and forth, but they leave lots of corner cases unresolved.  IMHO trying to mix the two causes more problems than Lua will fix -- better to stick with Perl and its overhead.
04:28 idiosyncrat_ There may be certain internal uses of Lua, uses where, since they are internal the data involved is not open-ended and the problems of data conversion are controlled.
04:28 idiosyncrat_ Those I feel free to go with anytime.
04:42 kaare_ joined #marpa
04:57 idiosyncrat_ good night
07:09 maybekoo2 joined #marpa
14:06 rns joined #marpa
14:08 rns idiosyncrat: re http://irclog.perlgeek.de/marpa/2016-05-06#i_12445911 -- the results under BSDs are the same or worse -- you were right: I had to understand the reason better -- doing it now.
14:15 rns current resoning: 'dynamic' in test_dynamic comes from linktype and is required by test target, so instead of skipping test section altogether we need to keep it, but ensure that it does nothing.
14:26 rns Ok, a PR filed along the above lines, tested (and passed :) under cygwin, windows xp and windows 7.
14:30 rns idiosyncrat: re doing semantics via Lua callbacks in Marpa::R4 and keeping Marpa::R3 as a refactoring/bugfix/new feature/incompatible change development line of Marpa::R2 -- this looks good to me: proper separation of concerns.
15:15 rns AFK, will backlog.
15:15 rns left #marpa
16:30 idiosyncrat_ joined #marpa
16:30 idiosyncrat_ Just uploaded Marpa-R3-4.001_006 -- another build fix aimed at FreeBSD
16:52 idiosyncrat_ [2016-05-07T16:47:45Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.14.4] [6c3f58aa-1473-11e6-bc05-f7fa6ed20d30] [2016-05-07T16:47:45Z]
16:52 idiosyncrat_ [2016-05-07T16:46:58Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.22.0] [5069fafe-1473-11e6-97ef-cc0159dcf47f] [2016-05-07T16:46:58Z]
16:53 idiosyncrat_ [2016-05-07T16:43:51Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.20.3] [e11bc1a0-1472-11e6-a739-c611687016d1] [2016-05-07T16:43:51Z]
16:54 idiosyncrat_ [2016-05-07T16:41:50Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.24.0 RC3] [9926280e-1472-11e6-8c55-9c217b4f5a8e] [2016-05-07T16:41:50Z]
16:54 idiosyncrat_ [2016-05-07T16:40:48Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd-thread-multi] [perl-v5.16.3] [73abb2b0-1472-11e6-b82f-a3499540e6ea] [2016-05-07T16:40:48Z]
16:55 idiosyncrat_ [2016-05-07T16:38:46Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.22.1] [2b70c026-1472-11e6-a721-9301b9e05817] [2016-05-07T16:38:46Z]
16:56 idiosyncrat_ [2016-05-07T16:32:56Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.18.2] [5a94ca2e-1471-11e6-9512-a4c5bc07ce75] [2016-05-07T16:32:56Z]
16:57 idiosyncrat_ [2016-05-07T16:49:48Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.18.4] [b5e371ee-1473-11e6-9cf7-a41e1d2cf337] [2016-05-07T16:49:48Z]
16:57 idiosyncrat_ Looks like we have this one beat.
16:57 idiosyncrat_ rns: Thanks!
17:07 idiosyncrat_ [2016-05-07T16:57:48Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.20.0] [d39acf74-1474-11e6-af32-cdb5b08f7ff6] [2016-05-07T16:57:48Z]
17:07 idiosyncrat_ There's also a bunch of linux passes.
17:08 idiosyncrat_ [2016-05-07T16:54:48Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.20.1] [68bd16f8-1474-11e6-b110-91650c7706a4] [2016-05-07T16:54:48Z]
17:08 idiosyncrat_ [2016-05-07T16:52:45Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.20.2] [1f6b7af8-1474-11e6-a4a4-87c9e7ccd979] [2016-05-07T16:52:45Z]
17:09 idiosyncrat_ [2016-05-07T16:50:45Z] [Slaven Rezić (SREZIC)] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [amd64-freebsd] [perl-v5.10.1] [d7febda6-1473-11e6-84f8-c5a0cc0196fa] [2016-05-07T16:50:45Z]
18:13 idiosyncrat_ [ruslan shvedov] [pass] [JKEGL/Marpa-R3-4.001_006.tar.gz] [MSWin32-x86-multi-thread] [perl-v5.22.1] [2b4f11f5-6e07-1014-9f9e-58fa49d7832b] [2016-05-07T17:48:15Z]
18:13 idiosyncrat_ rns: Thanks!
18:19 maybekoo2 left #marpa
20:54 idiosyncrat_ joined #marpa
21:25 kaare_ joined #marpa
22:25 Pursuit joined #marpa

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