weekly Rakudo status meetings with phase transitions: Tue 19:00 UTC | IR clogs at irclog.perlgeek.de/phasers/today
Set by moderator on 8 March 2011.
sorear PRE REPORT 05:41
DID
* got a yapsi commit bit
* starting to lay groundwork for multi dispatch in niecza 05:42
- figured out how to unify multi token with multi method
- also thought about "my multi token"
- multi tokens will do real MMD, then the system will be extended to type-based MMD 05:43
* gave a commit bit to mberends++, who added term:<now>
* implemented term:<xxx> overloading
* solved the "build requires too new .net" problem 05:44
* implemented parameter type constraints
WILL DO
* finish MMD-y stuff
* try again to get .WHO-based qualification working (jnthn++) 05:45
* after nominal MMD works, branch yapsi and fill it with niecza workarounds
* then slowly fix stuff
BLOCKERS 05:46
* $external only
EOR
07:32 cognominal left 07:37 cognominal joined 14:47 TimToady left 14:48 TimToady joined 16:14 colomon joined 16:43 benabik joined
diakopter ./ 17:59
jnthn o/ 18:00
moritz_ \\o
jnthn Is it time already? :)
moritz_ no, in 1h
jnthn oh!
Was gonna say...
diakopter oh, US summer time
jnthn :P
wtf, you already have summer?! In March?
:P
moritz_ today *really* felt like spring 18:01
18:02 colomon left
moritz_ could pre-report 18:02
* reviewed some rakudo pull requests
* translated a bit of PIR in nqp to Perl 6 code
* started to get rid of P6Regex.pir, and then let jnthn++ do all the hard work
that's about it 18:03
pmichaud_ good afternoon, #phasers 18:10
diakopter ur 50 min early
pmichaud_ I probably won't be here in 50 min
diakopter neither
jnthn o/ pmichaud_ 18:12
moritz_ pmichaud_: jdhore (same github nick) will be our release manager this month. Could you please add him as a rakudo committer? 18:14
jnthn Pre-report (partly so pmichaud_++ gets to see it... :-)) 18:16
Missed last week's #phasers because I was on the flight back from an excellent hackathon (mberends++ for organizing and hosting it), so here's two weeks worth of report. :-)
* Finished up the lexical setting work, tossed the confusingly named non-lexical setting we had before
* Got some initial static lexpad infrastructure in place, though not really using it just yet
* Moved NQP into a HLL of its own and called it just NQP, not NQP-rx. This will not break things if people then want to have both loaded, and it was needed for using HLL mapping on lexpad/lexinfo
* Started working on the work that will let us create meta-objects at compile time and then either use them at runtime right off or "deserialize" them; we'll eventually be able to migrate to a more real serialization format, but for now this will do OK 18:17
* Broke meta-objects into their own PBC
* Made it so that compiled NQP output loads the things it needs, to compilation of programs to PIR results in something usable
* Moved NQPMu and native types into the setting, though they ain't lexical yet (will be soon, just didn't get to that yet)
* Some work towards lexically scoped types
* Started to prepare for .WHO, stashes and the like
* Implemented very basic use statement in NQP; will flesh out with time
* Implemented basic support for MAIN
* It's now possible to write compilers entirely in NQP, no PIR needed
* Did that with the NQP compiler, pretty much
* Eliminated built-ins cheats by re-writing them into NQP and putting them in the setting
* Worked with moritz++ to also get P6Regex PBC's compilation managed entirely by NQP, eliminating P6Regex.pir
* Planned a bit on GLOBAL merging, after discussions with TimToady++ and others
PLAN
* Write my talk for OSDC.TW
* Get HLL.pbc fully NQP-compiled (no HLL.pir)
* Work towards sane package handling, including lexical packaging and nested packages
* Note that much of what I'm figuring out in NQP will be directly applicable to Rakudo, and should make us far more spec compliant there too
BLOCKERS
* Teaching Thu and Fri, which is always tiring. Then flying to Taiwan, where I'll be jetlagged and have lots of interesting things to go see. :-) I'll try and commit bits now and then. :-)
EOR
18:18 colomon joined
pmichaud_ my report: nothing perl6-ish to report. :( 18:20
moritz_++ # jdhore++ added as committer to rakudo repo 18:22
...was it something I wrote? 18:45
:-)
afk
18:45 pmichaud_ is now known as pmichaud 18:46 mberends joined 18:50 [particle] left 18:53 [particle] joined, mberends left
colomon \\o 19:10
moritz_ it's #phasers time
anybody got a report?
colomon pmichaud: you were 40 minutes early. :)
tadzik I has
colomon I've got a short one. 19:11
moritz_ tadzik goes first, I'd say
tadzik not much, but something because I missed last week's one
so, neutro is now considered dead & buried, I worked out a modular, testable solution which came out to be similar to masak's Pls. It's called Pies, and has quite a complete implementation called panda, which is to be blag about soon, with the complete deprecation of neutro 19:12
also, there is an implementation of Pies being a client side of mberends++ Popolnik, so testing and benchmarking modules 19:13
tadzik peeks at git log
fixed some weird Test.pm behaviour 19:14
Rakudo now does not load the .pir files if it's older than .pm file (when loading modules)
colomon tadzik++
tadzik fixed a number of things about the new IO capabilities 19:15
I think that's it, unless I forgot something
probably fixed a few bugs, but don't remember now
moritz_ tadzik++
colomon is panda usable now? 19:16
tadzik yes
colomon url?
tadzik oh, and it has a special, R*-friendly version
github/tadzik/Pies
colomon tadzik++
Util arrives; thrown by time change. 19:17
tadzik ++colomon # I sense testing
Util I attended the VM Summit at PyCon. Blog post pending.
Also, I poked at making `use lib` work in Rakudo; thwarted so far by lack of import list handling.
EOR
19:17 masak joined, mberends joined
moritz_ colomon: you next? 19:18
colomon sure
tadzik Util++
BREAKING NEWS: There is no Karma Bot at the channel, we can as well say "thank you" :) 19:19
s, cnr
colomon I did a bunch of benchmarking for masak's p5, with a few blog posts. And got a few more things working in the ABC module. .EOR
diakopter colomon: see the log before you joined 19:20
lots of other
masak since last I attended phasers, I've been implementing -n and -p in Rakudo. I've also posted about the p5 task, and plan to post about it again tonight. .EOR
jnthn pre-ported
colomon oh, hey, anyone here have a framework to use bit bisect with Rakudo? 19:22
I'd like to go hunting for some of the slowdowns in the last six months. (From the rakudo benchmarks.) 19:23
*git bisest
*git bisect
moritz_ colomon: what do you need besides 'rm -rf parrot_install; perl Configure.pl --gen-parrot && make ' ?
colomon hmm... guess that's a simple one-liner, eh? 19:24
mberends I'll leak a pre-publish outline of my design for Popolnik shortly, for everyone's perusal. I've gotten started with contributing to niecza, wrote a C translation of my p5 solution that is 2 million times faster than Rakudo, and am studying the Top Down Operator Precedence parser in Chapter 9 of "Beautiful Code" with a view to applying it to Perl 6. 19:25
Util If you want to examine changes in parrot *between* the changes in PARROT_REVISION, you need something more complex.
jnthn BTW, if anyone on non-Win32 can confirm that ctmo builds and passes all but one test for them, it can become master. It's got nice improvements...like pbeing able to write a compiler without needing PIR. :) 19:26
colomon Util: no need to do that unless / until I determine it's actually a parrot issue.
jnthn er, nqp/ctmo (branch)
tadzik jnthn: on it
jnthn Thanks :)
tadzik no, same error as before 19:27
jnthn tadzik: ? 19:28
tadzik pasing
jnthn hopes it's just a make program being silly again
tadzik nopaste.snit.ch/37775
masak mberends: does your 2-million-times-faster version apply to ASCII-only, or to (say) UTF-8?
jnthn tadzik: OK, I don't understand that failure at all. 19:29
mberends ascii, because it was little more than a proof of concept. I'm not bothered to extend it to UTF-8.
masak understandable.
tadzik This is Parrot version 3.1.0-devel built for amd64-linux.
jnthn Oh, not the PARROT_REVISION version?
masak mberends: in the best case, one could still have that speed for UTF-8, but eliminate false positives in a post-processing stage. 19:30
jnthn That may explain it.
mberends save as popolnik.xhtml if you want to see the inline SVG: pastebin.com/hmjcndwB
masak: yeah, but... ... 19:32
masak mberends: I'm not even sure, given how UTF-8 is laid out, that there will *be* any false positives. might be it'll just work straight off. 19:33
mberends: oh wait. no. you'll have to correct for character boundaries.
mberends masak: no, if you ignore UTF then a shorter substring could akshually masquerade as a longer one, and falsely displace the best find. 19:34
masak anyway, nice piece of engineering. :) 19:35
mberends :) Popolnik is gradually coming together in prototype form, and should be operational in a few weeks 19:36
masak \\o/ 19:37
Util I have had no feedback on last week's "Rakudo Star for Darwin binary preview"; comments and criticisms still welcome :) 19:38
s3.datasaw.com/Rakudo_Star_2011-01_...ment_3.dmg
masak '_experiment_3'? 19:39
Util The first two were based on earlier R*, in past months. 19:40
Same problem as package numbering in Linux distros; you have to distinguish between the version of the software you are packaging, and the version of the packaging itself. 19:41
masak makes sense. 19:44
Util Also, _experiment_ until the Rakudo pumpking approves it for upload to the repo Downloads area. 19:45
PerlJam Util: ask forgiveness rather than permission if you've got access to the downloads area yourself 19:54
jnthn has no problem with it 19:55
Anything that makes it easier for people to install/use Rakudo is nice :)
PerlJam indeed 19:56
Util PerlJam, jnthn: noted, and may-do; thanks!
PerlJam tries experiment 3 on his macbook 19:57
(well, work's macbook)
mberends experiment_3 reminds me of Lilo and Stitch ;) 20:04
PerlJam mberends: hopefully Util won't go all the way to experiement_626 20:16
mberends PerlJam: yeah :) 20:27
moritz_ jnthn: nqp ctmo branch tested fine (minus 02-modules.t) last I tried 20:50
jnthn moritz_: I guess you tested with the PARROT_REVISION rather than latest 20:51
moritz_: Will investigate more soon :)
moritz_ jnthn: yep, latest leads to build failure
jnthn aha, then it's probably an easy fix 20:55
tadzik jnthn: I've pasted that 20:56
21:25 masak left