»ö« | perl6.org/ | nopaste: paste.lisp.org/new/perl6 | evalbot usage: 'perl6: say 3;' or rakudo:, alpha:, pugs:, std:, or /msg p6eval perl6: ... | irclog: irc.pugscode.org/ | UTF-8 is our friend!
Set by lichtkind on 5 March 2010.
00:03 ruoso left 00:20 ruoso joined 00:33 gbacon joined
lue ohai everyone! 00:40
00:45 rgrau left 00:48 yinyin joined 00:50 sakib left 00:51 isBEKaml left
dalek p-rx: 3fcffdb | pmichaud++ | src/ (5 files):
Revert regex-interpolation branch changes back to 3ad311, fresh
 01:00
p-rx: 183d47b | pmichaud++ | src/Regex/Cursor.pir:
890caca | pmichaud++ | (2 files):
01:00 dalek left, dalek joined
lue afk 01:01
01:09 masonkramer_ joined 01:16 whiteknight left 01:22 ash__ joined 01:40 ash__ left 01:44 justatheory left 01:57 hercynium joined, bpalmer joined 02:01 [mark] joined 02:19 meppl left 02:25 mantovani left, mantovani joined
p-rx: b81ec42 | pmichaud++ | src/stage0/ (3 files):
Update bootstrap.
02:38 BrowserUk left 02:52 ash__ joined 02:58 JimmyZ joined 03:03 gbacon left 03:07 masonkramer_ left 03:14 alester joined
diakopter chirp 03:31
lue rechirp
I'm so annoyed with the USPTO, I wish to patent "A system for safeguarding the concepts and ideas created by persons" 03:43
Or something of the sort.
03:50 masonkramer_ joined, masonkramer left, masonkramer_ is now known as masonkramer 03:55 TiMBuS joined 04:03 sorear joined 04:25 patspam left 04:34 baest left
sorear hello. 04:36
04:36 ash__ left 05:05 XaeroOne joined
lue TimToady: ping [quick question about romaji] 05:08
TimToady ポン 05:13
lue Could I write rakudadou as rakudadō ? 05:14
TimToady sure, that's one way
05:14 bpalmer left
TimToady there's more than one system of romaji 05:14
05:14 xomas_ left
TimToady the more disturbing thing to English speakers is to write /ti/ for "chi" and /tu/ for "tsu" 05:15
though of course they'll also tend to mis-interpret o vs ō as tenseness rather than length 05:16
lue I've always viewed a macron (¯) as meaning say the letter, not its sound
ā = eh-ee, ī = ah-ee, etc. 05:17
TimToady we call them "long" vowels in English, because they *used* to be that way
but o and ō are exactly the same vowel sound in Japanese, only the second is two mora long instead of one
lue I've read about the slight sound differences, so got to watch out for that [or sound like a foreigner] 05:18
TimToady as an English speaker, just thing of saying o twice in a row
without a dipthong
*ph
so don't say owow
lue The "u" and "r" sounds would seem to be troublesome... 05:19
TimToady also, since they have only a five-vowel system, none of the vowels land in your mouth exactly where you find natural
lue Don't we have a five-vowel system as well? [discounting y] 05:20
TimToady the u is maid more with the back of your throat rather than with lip rounding as in English
05:20 tootroot joined
TimToady *made 05:20
no, we have a (lemme count) 05:21
lue I think I've watched enough japanese game shows to have the sound at least somewhere in my head :D
tootroot argh that topic messed up my irssi D:
TimToady 10 or 11 at least
05:21 kaare joined
lue Well, with the semivowels and such :) 05:21
05:22 kaare is now known as Guest99099
sorear As someone learning to speak Japanese, I've discovered that I can't actually hear the difference between o and oo. 05:23
lue I blame my interest in (natural&programming) languages on being born German & English :)
sorear The other long vowels are fine
lue [If i can find spots in conversation where I could say "way of the camel", I could say "駱駝道"] 05:24
05:24 sorear left, sorear joined
TimToady seek, sick, sake, sock, sack, suck, suk, (for)sook, soak, LASIK (schwa) for starters, and that's not counting dipthongs 05:24
sorear As someone learning to speak Japanese, I've discovered that I can't actually hear the difference between o and oo. 05:25
The other long vowels are fine
TimToady that's because you're listening for the wrong difference
maybe
what is your native dialect? 05:26
sorear Well, it's not like I can dictate English without knowing the words
None; Japanese is entirely foreign
TimToady are you a native English speaker?
sorear Yes
(US southwest)
TimToady okay, so basically western, same as me
lue [I, having been born in Germany, find it amazing people cannot say 'ch', as in 'Ich bin ein Berliner'] 05:27
TimToady well, not that Germans can agree on that either :)
a Berliner doesn't say it like a Münchner 05:28
lue Lustiges Taschenbuch!
Southern germans wear the lederhosen. They're weird :)
TimToady but the difference between o and oo will (to a western English speaker) be something like the difference between the o sounds in 'dough' and 'dough-owner' 05:31
it really is just length in time
lue [Glottal stop is just like lisp; just by saying it tells us if you have it.] 05:32
TimToady but since Japanese is syllable timed rather than stress times, you have to learn to hear the syllables as equal in time, not equal between stresses, as we do in English
pmichaud My glottal stopped a long time ago.
TimToady lue: well, only if you're cockney and say glo'l stop :) 05:33
lue I do love literal transalation of ja->en . They always sound like proverbs :)
or Estuary! Estuary accent has glottal stop as well!
TimToady the rest of us tend not to use glotals except at the beginnings of words with vowels
lue or in uh-oh
TimToady it's the *absence* of a glottal stop that we hear as an 'h', more or less 05:34
lue [I've read up on Estuary because the 10th Doctor spake it :D]
TimToady yes uh-oh is an exception
lue Language is fun. Ever since I picked up a Visual Basic book at age 11.... :) 05:36
TimToady well, you got be started on phonetics/phonemics, which is my linguistic speciality, but a bit OT here...
lue This is #perl6. We were OT when I asked about romaji :)
TimToady I think part of the reason I'm good at phonetics is because I'm somewhere on the autistic spectrum, and can hear things on the detailed/unanalyzed level if I choose to 05:37
neurotypicals usually have a harder time not filtering everything through the phonemic system of their native language 05:38
lue rakudo: autistic ~~ smart
p6eval rakudo 10a321: OUTPUT«Could not find sub &smart␤current instr.: '_block14' pc 29 (EVAL_1:0)␤»
lue but what about autistic? That was the first one!
TimToady Every now and then, I am the smartest person you will ever meet. Pity about the rest of the time...
lue rakudo: say licks[TootsiePop]
p6eval rakudo 10a321: OUTPUT«Could not find sub &licks␤current instr.: '_block14' pc 29 (EVAL_1:0)␤» 05:39
lue rakudo: say @licks[TootsiePop]
TimToady Maybe I should have called it the Autistic License...
p6eval rakudo 10a321: OUTPUT«Symbol '@licks' not predeclared in <anonymous>␤current instr.: 'perl6;PCT;HLLCompiler;panic' pc 152 (compilers/pct/src/PCT/HLLCompiler.pir:108)␤»
lue I once saw a Death and Restitution license
basically, you can't use the software if you are alive, or a ghost or an angel :) 05:40
05:40 simcop2387 left
TimToady angels aren't alive?!? 05:40
lue I know; there was a part about being punished if you used it if you were a ghost/angel or something 05:41
Where I found it [thru slashdot], you could either pick the BSD or D&R license
I want to patent patents; if it ever gets through the USPTO and gets approved, then, well, enough said about the USPTO :) 05:43
TimToady Copyright (2054) Larry Wall, you may use this if I'm still alive, or if you're dead...
sorear lue: you know the wheel is currently patented in australia, right? 05:44
lue Or something else absurd, like goverment, or CO2
No.
After patenting patents, I can sue the USPTO, who approved the patent, for infringing upon my patent. It's PERFECT! \o/ >:) 05:45
sorear No.
It doesn't work that way.
lue That's what I *hope*.
sorear Patent lawsuits tend to result in the patent getting thrown out
TimToady Interestingly, the contract for my timeshare in Hawaii is valid only until the last living descendent of Ronald Reagan or Queen Elizabeth dies.
lue If it actually worked, I'd be very frightened 05:46
sorear The power of a patent is in scaring people who don't want to enter a lawsuit /at all/
pmichaud mmmmm, Hawaii
TimToady Hawaii doesn't allow contracts in perpetuity, it seems...
pmichaud already got my plane tickets for hawaii :)
TimToady which island?
pmichaud Maui
sorear TimToady: Does it specify which Queen Elizabeth? :D 05:47
TimToady lotsa nice things on Maui
II
lue TimToady: for that contract, is that or as in xor ?
TimToady no
lue Darn; then the contract would only be temporarily invalid :)
"Copyright (c) Doomco 32 B.C. All rights (and then some) reserved. Any violation will be prosecuted to the fullest extent ABOVE the law" 05:49
05:51 uniejo joined
TimToady you would not have liked it back then, since you are a loose cannon. 05:52
loose catapult just doesn't have the same panash
lue But what would those darn Golganfrinchans know? :) 05:53
TimToady or panache even
lue Adams' Theory of Evolution
TimToady I prefer the Addams
lue Douglas Nöel Adams, RIP [Towel Day, in honor of his death, is tomorrow! (May 25)] 05:54
TimToady um, that's today there... 05:55
lue Well, here in the good ol' US of A, it's still 1 hour 4 minutes until the 25th 05:56
s/US of A/PD of T/
TimToady only the good new US of A :) 05:57
the old bits are on the other end
well, except for the really old bits, which are in New Mexico 05:58
but it wasn't very U yet back then...
lue NQU? :) 05:59
TimToady well, it wasn't Less Than America, I suppose...
pmichaud there's some really old bits here in the Republic of Texas, too. :-) 06:00
TimToady yes, but you were anti-united back then
pmichaud I think just dis-united :-)
lue More like the PMC of the USA back then :)
[or was that something else?] 06:01
eternaleye TimToady: (re autistic spectrum making it easier to differentiate sounds) That explains it for me. I thought it was that I grew up hearing Hebrew from my mom, Arabic from my dad, and Japanese from anime, so I heard a large subset of available phonemes early enough that they made sense
pmichaud mostly people kept forgetting ice cream with their pie, I think.
The motto became "Remember the a-la-mode!"
lue
.oO(A Perl6-ish representation of the US)
TimToady well, being multilingual certainly doesn't hurt
such as in Texas, where they speak both English and Texan 06:02
eternaleye lue: (death and restitution) ITYM Death and Repudiation, it's used as a dual-license with BSD in the sole project that uses it; it was referenced in a semirecent slashdot story
lue It's an awesome license, and it smells faintly of dead parrots and silly frenchmen :) 06:03
06:03 simcop2387 joined
lue [That one reason I started learning french this year; to get a good frenchman-speaking-english accent] 06:03
afk and good night (off to 楽土, aka bed) :) 06:04
eternaleye All in all, Japanese is the natural language I like the most (having also taken French in Junior High). It's relatively sane in terms of having rules and sticking to them, and it sounds beautiful. 06:05
TimToady お休みなさい。
sorear I like 'em all 06:06
eternaleye Besides, 私のホバークらーフトはうなぎでいっぱいです！ 06:11
(hiragana overusage due to not knowing relevant kanji)--
Although if I took the time to open it Kiten would help with that (VERY nice japanese dictionary and radical lookup tool for KDE) 06:13
sorear well it's not like I'd *understand* the relevant kanji
eternaleye "My hovercraft is full of eels" -> watashi no hobaakurafuto wa unagi de ippai desu
I had an extra ー too 06:14
TimToady are you sure they were うなぎ and not あなご?
eternaleye Every time I've seen it romanized, it's with 'u', so... 06:15
TimToady well, they distinguish two different kinds of eels
unagi are freshwater eels, and anago are sea eels
eternaleye Ah. The one on the 'multiple translations of hungarian sketch' page used unagi 06:16
TimToady I keep track of this, being allergic to seafood, but not freshwater
sorear do they distinguish a common superclass?
or just "round fish"
TimToady I am not aware of a superclass 06:17
eternaleye rakudo: class うなぎ　{ }; class ホバークラフト does Positional[うなぎ] { }; 06:19
p6eval rakudo 10a321: ( no output )
TimToady I think a hovercraft is more like a Bag[うなぎ] 06:20
eternaleye I ♥ Unicode 06:21
sorear pmichaud++ dealing with my problems without me
eternaleye TimToady: Maybe - it depends on if you have the eels sit in the seats :D 06:22
TimToady ⟅うなぎ⟆ 06:23
dalek p-rx: becdc98 | pmichaud++ | (2 files):
Change tests for isa ['Sub'] to does 'invokable'.
 06:24
p-rx: 0b3416f | pmichaud++ | src/stage0/ (4 files):
Update bootstrap.
TimToady looks like they're on the bottom to me
eternaleye rakudo: sub infix:<♥>( $lhs as Str, $rhs as Str ) { say $lhs ~ " love " ~ $rhs }; "I" ♥ "Perl 6";
p6eval rakudo 10a321: OUTPUT«I love Perl 6␤»
eternaleye rakudo: sub infix:<♥>( $lhs as Str, $rhs as Str ) { say $lhs ~ " love " ~ $rhs }; "I" ♥ $*VM; 06:25
p6eval rakudo 10a321: OUTPUT«Contextual $*VM not found␤current instr.: 'perl6;Perl6Exception;throw' pc 15629 (src/builtins/Associative.pir:32)␤»
eternaleye rakudo: sub infix:<♥>( $lhs as Str, $rhs as Str ) { say $lhs ~ " love " ~ $rhs }; "I" ♥ $?VM;
p6eval rakudo 10a321: OUTPUT«Symbol '$?VM' not predeclared in <anonymous>␤current instr.: 'perl6;PCT;HLLCompiler;panic' pc 152 (compilers/pct/src/PCT/HLLCompiler.pir:108)␤»
eternaleye Awww
sorear wow, I'm now caught up on everything 06:27
I was expecting 4 days of backlog to take more than 3 hours 06:28
you guys are such SLACKERS
06:28 kel__ joined
pmichaud <-- SLACKER 06:28
sorear reads pmichaud's REPL implementation 06:29
pmichaud I haven't done the REPL part yet
TimToady sorear: we're only doing this for you, and when you wander off, we lose all motivation...
06:30 Su-Shee joined
pmichaud but I've figured out how I think it should be factored 06:30
I was going to finish it up this morning, but got sidetracked on the regex interpolation fixes
06:30 alester left
TimToady never gets sidetracked, but only because everything is a sidetrack to begin with... 06:31
06:31 masak joined
sorear pmichaud: do you still have the baton for this or do you want me to try and rejoin? 06:31
masak good morning, #perl6!
phenny masak: 24 May 18:16Z <Tene> tell masak ping
sorear good morning masak
pmichaud sorear: let me keep the baton until tomorrow noonish, if you don't mind
masak Tene: pong
sorear sure.
pmichaud I think it'll just take me a couple of hours to get nqp to have a repl that saves lexicals 06:32
then updating rakudo (or any other hll) will be a fairly simple matter
06:33 redicaps joined 06:37 JimmyZ_ joined 06:38 JimmyZ left, JimmyZ_ is now known as JimmyZ 06:44 redicaps left
sorear Is $*CURPAD.<!IN_DECL> another gimme5ism? 06:55
moritz_ masak: iwbn if you brought dalek to report your commits to the buf branch (or bribe Infinoid or diakopter or so to do it for you) 06:58
masak *nod*
that would indeed be a boon. 06:59
07:09 rgrau joined, rgrau_ joined 07:11 rgrau_ left 07:12 rgrau left 07:13 rgrau[work] joined, iblechbot joined 07:20 k23z__ joined 07:23 ethel left
sorear k23z__: Hi. 07:29
You wanted me?
k23z__ sorear, yes but it depends on what you mean by "wanted" , for example some women might want you, and I would like to clarify that I don't want you the same way :) 07:30
07:31 ejs joined
masak now that we've cleared that up... 07:32
07:38 ejs left 07:39 gfx joined 07:43 ejs joined
cognominal is there any way to trace function calls in Perl 6? 07:44
masak cognominal: are you asking for more backtraces than those that already plague us? :P 07:45
cognominal :)
I am not asking backtrace here (which become rarer) but plain trace 07:46
masak guess you could emulate tracing by pushing and popping a contextual array variable.
it's a bit of manual work in each function, but it'd do the trick. 07:47
sorear I think what you're looking for is pir::trace(4)
cognominal thx, I shoud learn more about parrot 07:48
masak clearly, so should I :)
cognominal For my presenetion I am just documenting the Q::Pir and pir:: and other interfaces between rakudo and parrot. 07:49
sorear oooooh presentation?
cognominal sorear++ # for blizkost
jnthn++ # for initiating it
sorear it needs to be better
cognominal sorear, I want to use blizkost, so be sure I will whine loudly. 07:50
sorear what are you using it for? 07:51
sorear is using it for POSIX & DBD::SQLite, but the project is design-stalled, so not much testing is being done
cognominal in Calais, we are lacking tacks (hint, hint) so we will have a big Perl 6 track. Work for me :(
07:51 snarkyboojum joined
moritz_ FPW? 07:52
sorear Calais?
cognominal I intend to use it for code serving web pages hopefully to call Perl 6 code (for demo)
yes, french Perl workshop at Calais
anglophone speakers welcome. 07:53
mberends will probably talk.
pmichaud
.oO( "We have ways of making him talk...muwahahahahah" )
 07:54
cognominal and we dirnk Chartreuse the evening. jnthn++ will attest it nice (also to make people talk). 07:55
besides being a nice liquor, chartreuse is TimToady's favorite color.
masak not many things are both a liquor and someone's favorite color. 07:57
cognominal I strap people in a chair and can dress like Lawrence Olivier ("is this safe?") to make people talk 07:58
*I can strap people to a chair
pmichaud masak: .... um, burgundy? ;-) 08:01
or champagne?
masak champagne is a color?!
moritz_
.oO( only for women :-)
pmichaud en.wikipedia.org/wiki/Champagne_(color)
moritz_ like, my girlfriend als thinks that "lavender" is a color 08:02
*also
for me it's a herb
08:02 viklund joined
masak wow. 08:02
moritz_: you're probably alluding to blog.xkcd.com/2010/05/03/color-survey-results/
or you should be :)
08:06 snarkyboojum left
pmichaud well, time for sleep here 08:06
I have the refactor of regex interpolation almost done in the 'ri2' branch of the rakudo repo. 08:07
Just have three failing test files, if anyone wants to play with it and perhaps fix some of them. :-)
cognominal moritz_, if you went to Provence to admire a lavender field, you would know for sure that lavender is a color (and an odor too).
pmichaud see you all a bit later 08:08
moritz_ cognominal: odor I well believe :-) 08:11
perlgeek.de/blog-en/perl-6/rakudo-s...-list.html lolihazblogged 08:14
masak \o/ 08:19
mathw That's a good list
I'd really like those backtraces
08:19 mathw sets mode: +o masak
masak I should make a corresponding list of bugs I'd like to see fixed :) 08:19
mathw Yes 08:20
masak the 'April list' is slowly running out of open bugs :P
mathw \o/
Better add some more then
cognominal we should clone pmichaud and jnthn (that's for the rakudo side).
masak probably won't add to the original blog post -- more likely to create a new one. 08:21
moritz_ cognominal: ... and properly clone their lexpads too :-)
masak actually, moritz_++'s list is already quite a good list of bugfixes rather than features.
cognominal moritz_++ # context always helps 08:22
moritz_ and I find it's really the bugs that bug me, not the missing features
mathw but then we'd also have to clone masak, because otherwise they'd be at risk of emptying the bug queue
moritz_ at least 90% of the time
mathw For Rakudo * we can accept missing features - after all, we've got a lot of features already
But the ones that are there should work... 08:23
masak I definitely agree on closures, match objects, and iterators. I'd like to add "more mature List/Iterator implementation" and "Short (Rakudo-only) backtraces"
moritz_: missing closing paren on the line starting 'General Whatever-currying'. 08:24
moritz_ masak: thanks, fixing... 08:25
08:27 k23z__ left, dakkar joined
pugssvn r30779 | sorear++ | [viv] Implement declarative or in regex compiler 08:46
masak sorear++ # cool! 08:47
sorear rakudo: say "π" ~~ /\w/ 08:48
p6eval rakudo 10a321: OUTPUT«π␤»
masak π is an alphanumeric.
sorear rakudo: say "π" ~~ /<[\0x3c0]>/
p6eval rakudo 10a321: OUTPUT«␤»
08:49 jhuni left
mathw Mmmm pie 08:49
sorear rakudo: say "π" ~~ /<[\u03c0]>/
masak pie pie pie pie pie pie pie.
p6eval rakudo 10a321: OUTPUT«␤»
sorear rakudo: say "π" ~~ /<[\U03c0]>/
p6eval rakudo 10a321: OUTPUT«␤»
sorear rakudo: say "π" ~~ /<[\x03c0]>/ 08:50
p6eval rakudo 10a321: OUTPUT«␤»
moritz_ rakudo: say "π" ~~ / <[\x[03c0]]> / 08:51
p6eval rakudo 10a321: OUTPUT«regex assertion not terminated by angle bracket at line 11, near "]> /"␤current instr.: 'perl6;HLL;Grammar;panic' pc 501 (ext/nqp-rx/src/stage0/HLL-s0.pir:327)␤»
moritz_ std: say "π" ~~ / <[\x[03c0]]> /
p6eval std 30778: OUTPUT«ok 00:01 116m␤»
moritz_ rakudo: say 'c' ~~ /<[\x03c0]>/ 08:53
p6eval rakudo 10a321: OUTPUT«c␤»
moritz_ rakudo: say '0' ~~ /<[\x03c0]>/ 08:55
p6eval rakudo 10a321: OUTPUT«0␤»
09:10 ive joined, Trashlord left 09:11 proller left
pugssvn r30780 | sorear++ | [viv] Implement || 09:14
09:15 proller joined
eternaleye moritz_: (re blog) As to $/ in s///, might a workable hack be to make s/$lhs/$rhs/ delegate to s/$lhs/{ "$rhs" }/ using sorear's textual macros and thus manage it using the .subst closure semantics? 09:15
sorear I doubt it 09:16
eternaleye Eh, just a random idea
moritz_ eternaleye: that's basically how it works at the moment, but it still has the problem that $str.subst() doesn't set $/ for the RHS
sorear the infrastructure for sanely handling lexicals through macros doesn't really exist
the parser has to knoe the lexical referencing environment at parse time
09:17 k23z__ joined
eternaleye Huh, I thought .subst( /foo/, { code($0) } ) worked 09:17
moritz_ rakudo: my $x = 'foo'; $x ~~ s/(o)/$_[0]$_[0]/; say $x
p6eval rakudo 10a321: OUTPUT«Method 'postcircumfix:<[ ]>' not found for invocant of class ''␤current instr.: '!postcircumfix:<[ ]>' pc 11584 (src/builtins/Cool.pir:15)␤»
moritz_ :(
alpha: say 'foo'.subst(/(.)/, { uc $0 }, :g) 09:18
p6eval alpha 30e0ed: OUTPUT«FOO␤»
moritz_ it worked in alpha
HOW?
eternaleye rakudo: say 'foo'.subst(/(.)/, { uc $0 }, :g)
p6eval rakudo 10a321: OUTPUT«Method 'postcircumfix:<[ ]>' not found for invocant of class ''␤current instr.: '!postcircumfix:<[ ]>' pc 11584 (src/builtins/Cool.pir:15)␤»
eternaleye Huh
moritz_ that's what you get for accessing $0 if $/ isn't defined 09:19
rakudo: $0
p6eval rakudo 10a321: OUTPUT«Method 'postcircumfix:<[ ]>' not found for invocant of class ''␤current instr.: '!postcircumfix:<[ ]>' pc 11584 (src/builtins/Cool.pir:15)␤»
eternaleye moritz_: Yeah, but in your first example you used $_ not $/
09:21 lestrrat is now known as lest_away
pugssvn r30781 | sorear++ | [viv] Implement [ ] 09:21
moritz_ eternaleye: that's because $/ is passed as the first positional parameter to the closure that .subst receives
09:21 ruoso left
sorear moritz_: I did a clever trick in the macros implementation by adding $/ as a placeholder parameter 09:22
09:24 jaldhar left
eternaleye moritz_: github.com/rakudo/rakudo/blob/alpha...r.pir#L842 09:24
09:24 jaldhar joined
eternaleye lexpad['$/'] = match 09:24
moritz_ huh 09:25
will have to try that
sorear TimToady: How does Cursor know when protoregexes have changed (composition, &c) and need to be recalculated? 09:30
09:32 XaeroOne left, gfx left 09:35 ruoso joined 09:49 Trashlord joined 09:51 yinyin left
eternaleye Also re blog, could whatever-currying use some of the code from junctions? As far as I can see, both do something special when no multi that accepts them is defined 09:56
moritz_ yes, I think the idea is to do it in the dispatcher
rakudo: say Whatever ~~ Any 09:57
p6eval rakudo 10a321: OUTPUT«1␤»
moritz_ that's a problem though
most built-ins have a very general candidate that accepts two Any's
rakudo: say *~*
p6eval rakudo 10a321: OUTPUT«!whatever_dispatch_helper!whatever_dispatch_helper␤»
moritz_ rakudo: say (*~*).arity 09:58
p6eval rakudo 10a321: OUTPUT«Method 'arity' not found for invocant of class 'Str'␤current instr.: '_block14' pc 29 (EVAL_1:0)␤»
moritz_ so it doesn't fail
which is where junction dispatch kicks in
eternaleye Hm, maybe respec Whatever as ~~ Mu && !~~ Any?
Makes it easy then
moritz_ and makes other things harder
eternaleye drat 09:59
moritz_ but that's something that TimToady++ has to decide
I don't see all the implications involved
oh wait...
eternaleye And again the question becomes 'how did alpha do it?'. I'll dig in alpha again.
moritz_ maybe the solution is not to have an infix:<~>(Any, Any) but rather (Cool, Cool) 10:00
jnthn morning 10:01
moritz_ \o/
jnthn No, I don't think we want whatever stuff to fall out of junction dispatch.
(if that's the suggestion) 10:02
moritz_ it was, yes
jnthn afaik, we're mostly meant to set up the auto-currying at compile time. 10:03
eternaleye jnthn: That seems... Difficult. Like, 'must know whether all multi candidates explicitly accept Whatever at compile time' halting problem difficult. 10:05
s/all/any/ 10:06
10:06 k23z__ left
eternaleye So range's 1..* vs 2+* would be... interesting. 10:07
If whatever always curried it'd be trivial
moritz_ fsvo "trivial" 10:08
eternaleye Well, insert a 'relatively'
Perl 6 seems to enjoy thoroughly nuking 'trivial' on the part of implementations from orbit so as to preserve it for users. 10:09
moritz_ yes 10:10
masak sushi &
moritz_ "tormenting the implementor on behalf of the user", TimToady++ called it once
eternaleye mmmsushi
jnthn eternaleye: But we *do* know all the multi candidates in a given lexical scope at compile time, by spec. 10:11
eternaleye: Granted it's the halting problem in general, yes.
eternaleye jnthn: Mmm
jnthn eternaleye: I think that right now, though, I'd rather have auto-currying and say that it's tough luck for anyone who writes a multi that takes Whatever for an operator, rather than not have the auto-currying. 10:12
(Since not having it is helping nobody right now.)
eternaleye jnthn: I think lots of people would miss 1..* though, and it would be a test regression 10:13
As well as 1...*
jnthn eternaleye: We can easily eliminate operators known to have special handling too. 10:16
moritz_ or maybe a cheating trait
multi sub foo(...) is not_whatever_curried { ... } 10:17
jnthn moritz_: That's no easier to get in place than looking at the signature really, though.
moritz_ hm, ok 10:18
10:20 slavik left 10:33 ruoso left 10:34 XaeroOne joined 10:50 JimmyZ left, slavik joined 10:58 masonkramer left 11:12 iblechbot left 11:13 masonkramer joined 11:14 synth joined 11:24 MOAR-f00li5h joined, f00li5h left 11:27 MOAR-f00li5h is now known as f00li5h
masak moritz_++ gets karma from uasi on Twitter for JSON::Tiny -- twitter.com/uasi/status/14684657673 11:31
11:32 redicaps joined
moritz_ so, what does he write? :/ 11:32
google translate says "The JSON:: Tiny is now helpless Rakudo like bugs around and take a regular expression, but I was still rewriting the work Nde."
to which I can mostly say "huh?" 11:33
masak I read that as having to do with the alpha->master transition, and JSON::Tiny being 'helpless' due to bugs in master. 11:34
but I might read too much into it -- I'm a bit biased :)
moritz_ there's a lot of truth in there
jnthn thought JSON::Tiny worked in master now
moritz_ jnthn: nope - uses too many Match features 11:35
jnthn Ah.
moritz_ the parsing works, except for character classes containing \t
_sri how is the socket support in rakude these days?
*rakudo 11:36
moritz_ not changed in the last few weeks
jnthn Socket support v Rakude je fajn. ;-)
.oO( maybe I wasn't meant to parse that typo as a case declension, but it was spot on correct :-) )
 11:37
_sri :D
jnthn _sri: The socket code in master is pretty much what we had in alpha. It was capable of various things, though I guess it probably had some issues. But people managed to write servers and clients for various things using it, so I guess it's at some level usable. 11:38
11:40 envi^home joined
_sri does that include non blocking sockets? 11:43
moritz_ don't think so
_sri was thinking about starting to port his http 1.1 client and server libraries
moritz_ but I think parrot recently got support for those
or somebody's working on them 11:44
jnthn Once Parrot supports them, it should be easyish to expose it in Rakudo. 11:45
_sri perl5 network programming is so awful (compared to ruby and python), i guess with a little effort rakudo could already beat it 11:47
11:47 redicaps left 11:48 redicaps joined
_sri well, keeping an eye on it *crosses fingers* 11:50
masak so many wishes, so few tuits :-/ 11:53
jnthn Aye. :-/ 11:55
jnthn has had a tuit drought for the last week or so. 11:56
moritz_ I failed to backport the code that sets $/ in the code block of .subst()
it does work, but gives a NPMCA in the $str.=subst form
jnthn moritz_: Feel free to gist the patch.
moritz_: Who knows, maybe I get a round tuit.
moritz_: I bet it's because caller in the case of .= isn't doing quite what you want though. 11:57
(because the caller is the implementation of .=)
moritz_ jnthn: gist.github.com/413051 11:58
aye, that's what I suspected
but I don't know how to fix it
masak good interview with Damian Conway: www.oreillygmt.co.uk/2010/05/damian...uture.html 11:59
lots of Perl 6 questions and answers.
moritz_ and for s/// there will be yet another caller
colomon moritz_++ # for getting that far
jnthn moritz_: Does it make things work more than they did before, btw? 12:00
masak "I've switched most of my development to Perl 6, and it's like stepping out of a prison I didn't even know I was in" -- Damian++
jnthn Wow!
colomon I was noticing in the YAPC:NA listings that he seemed to be doing a lot of p6 work -- where is it?
moritz_ jnthn: with this patch 'str'.subst(/(.)/, { uc "$0" }) DWIM 12:01
jnthn moritz_: 12:03
I just have to wonder if we can get away with:
pir::find_dynamic_lex__ps('$/') = $m; 12:04
moritz_ now that's an awesome idea
jnthn Rather than all the caller lookup which depends on an explicit number of lexpads being there.
Well, only if it works. :-) 12:05
moritz_ I'll try
I think I've tried something similar before, but it involved stored_dynamic_lex 12:06
which didn't work out too well
12:09 lest_away is now known as lestrrat
moritz_ jnthn: the error goes away, but the improvement too :( 12:09
do find_dynamic_lex search the current lexpad too? 12:10
s/do/does/
masak colomon: Damian's MO seems to be lone hacking followed by a presentation and a first release.
colomon masak: still seems like this would be a good time to be providing the feedback you get from developing actual code in p6... 12:11
jnthn moritz_: Oh, it probably does
colomon but I hope that means there are presentations coming next month...
jnthn moritz_: We maybe need an op like find_lex_skip_current
But a dynamic variant of it.
masak colomon: I've been known to withhold code in the interests of making a splash, so I'm not one to throw stones. :)
colomon: if you ask me, we have a large proportion of people sharing already. what we need isn't a larger proportion, but simple more people coding Perl 6. 12:12
colomon was getting all excited to run git bisect for $work, but then discovered there were only about 3 commits in the relevant time period...
masak: more people coding in Perl 6 would be Very Good 12:13
12:13 [mark] left
masak colomon: and Rakudo Star is the chance many people are waiting for to start. 12:13
moritz_ jnthn: I can't find find_lex_skip_current in current... are you sure that's the name? 12:14
masak including those who see it as "the Perl 6 release".
colomon masak: sure enough. but R* would be better if more people started now.
12:14 jaldhar left
colomon hmmm.... probably too late at this point, I suppose. 12:14
moritz_ I remember seeing such an opcode, but I can't find it :(
pmichaud good morning, #perl6
moritz_ good morning pmichaud
masak R* might be better, but I really do understand people who want to wait until the worst bugs are squashed. 12:15
colomon o/
masak mo', pm.
12:15 BrowserUk joined
jnthn pmichaud: morning! o/ 12:15
moritz_: No, I'm not sure that's the name...let me check. 12:16
moritz_: oh, no, I was right
=item find_lex_skip_current(out PMC, in STR)
moritz_: btw, Rakudo dynop
Not Parrot op.
moritz_ ah
that's why I didn't find it in the parrot repo :/ 12:17
jnthn :-)
moritz_ I'm curious, doesn't NQP need those?
jnthn Wow, prototyping garbage collection in Perl 6! 12:18
trac.parrot.org/parrot/wiki/GCMassacre 12:19
moritz_: Not atm.
pmichaud ...nobody took me up on the challenge of cleaning up regexes in the ri2 branch :-) 12:20
okay, I'll work on 'em
mathw jnthn: I like that the rough outline is written in NQP :)
moritz_ pmichaud: at leat I looked at the failures :-)
pmichaud moritz_++ 12:21
it's probably something trivial-ish
I do like the new formulation much better. Also, it knows how to properly do LTM on array interpolations.
moritz_ we have proper LTM? 12:22
that's news :-)
pmichaud i.e., my @xyz = <b bc bcd>; 'abcde' ~~ / @xyz /; # properly finds bcd
right now, only in interpolations, and only if backtracking isn't involved
jnthn protoregexen with constant prefixes also do LTM-ish stuff too, iirc? 12:23
pmichaud jnthn: yes
we definitely don't have "proper LTM" yet, though. that's still a significant challenge
jnthn Aye.
Turns out we don't have proper multi handling again yet either after latest spec changes. :-/ 12:24
jnthn doesn't look forward to that refactor all that much 12:25
masak jnthn: why not -- apart from the obvious, that you wrote one working thing already, and now it changed?
jnthn masak: It took more than one attempt to get the current one working as per the old spec too. 12:26
masak: But also because multi dispatch is pretty central to all kinds of things. 12:27
So when you go tweak it, there's a lot to break.
Including building the setting.
masak so, definitely post-R*. 12:28
jnthn Oh, yes.
And probably not immediately after that.
masak jnthn: do you think we'll have time to hackathon once more before R*? I'd be happy to pop down to Lund this time.
jnthn masak: Perhaps.
masak: We need to squeeze it in amongst the workshops.
;-)
masak oh right :) 12:29
jnthn I'm going nowhere from here before the Ukrainian Perl Workshop.
Well
Perl Mova + YAPC::Russia to give it its full name.
mathw masak: the Form.pm push commences this weekend. I can't do it sooner as I have to make up some aikido classes this week.
jnthn It's historically cute that YAPC::Russia is in Kiev. :-)
masak mathw: oh well, then I'll simply have to find something else to do until then. :P 12:30
moritz_ masak: tough call :-)
12:30 iblechbot joined
masak I'll live it down. :) 12:30
12:30 masonkramer left
jnthn masak: You thinking of popping down for a weekend/long weekend kinda thing? 12:30
masak jnthn: yes, think so. 12:31
jnthn 5th/6th June would work, if it ain't too soon.
The weekend after that I'm in Kiev \o/
masak no can do :(
jnthn Aww
Next weekend is free but, well, soon.
masak I have a shareholders meeting on Lyr.
pmichaud rakudo: say '3' ~~ /<[\x03c0]>/
p6eval rakudo 10a321: OUTPUT«3␤» 12:32
mathw masak: you can always do some yourself, I'm sure I can still find more to do. I was planning to do it this week but then I realised that I don't want to flunk testing this month
masak mathw: *nod*
mathw: I might have a brief look, but I probably wont do leaps and bounds without your supervision.
might focus, say, on my GSoC work instead.
or the last mile of Web.pm.
or porting GGE to master. 12:33
or Emmentaler.
or, you know. something else. :)
mathw Yeah it's not like you're stuck for things to do :)
jnthn masak: I'd love to have the basic bits of Web.pm working on master.
(Request/Response)
mathw But I am looking forward to figuring out what's going on with Form.pm because it's weird
masak jnthn: porting Web.pm to master is definitely post-grant. but it might still be pre-R*.
12:34 ejs left
jnthn
.oO( or porting master to Web.pm )
 12:34
masak dang, forgot Yapsi on that list! release is a week away.
BrowserUk HI all
masak HI
jnthn o/ BrowserUk
masak: If I wanted to get Request/Response to work, should I branch Web? 12:37
12:37 masonkramer joined
jnthn masak: So I don't break things on alpha? 12:37
masak jnthn: for dual-Rakudobranch projects, I tend to prefer if there's either master/ng-compat, or alpha-compat/master. that ties in well with what Emmentaler will do. 12:39
Su-Shee
.oO(how many rakudo stars are you planning?!)
12:40 redicaps1 joined, ruoso joined 12:41 iblechbot left, redicaps left
moritz_ as many as needed 12:41
12:41 iblechbot joined
masak and no more than necessary. :) 12:41
Su-Shee yeah but which one is which? :) 12:42
masak likely the first one is number one. etc. :)
man, today's questions are easy. :P 12:43
jnthn masak: OK, ng-compat it will be, if I get to it. 12:44
phenny: show web
grr 12:45
hugme: show web
hugme jnthn: the following people have power over 'web': PerlJam, TimToady, [particle], colomon, jnthn, masak, moritz_, pmichaud. URL: github.com/masak/web/
jnthn yay
12:45 masonkramer left
moritz_ it means you're a meta committer, not automatically a committer 12:45
12:46 jaldhar joined
colomon but if you're a meta committer, you can make yourself a committer? 12:46
moritz_ sure
12:46 M_o_C joined
moritz_ however hugme can't do that automatically, because irc nicks != github nicks 12:47
[Coke] reviewingly sees TimToady go off on a cotangent.
masak jnthn: you're now a committer to Web.pm 12:48
jnthn masak: w00t
masak hugs jnthn - welcome aboard!
jnthn \o/
masak #rs tonight, no? 12:51
mathw hugs everyone, just 'cause
jnthn Yes.
Clockan nio. 12:52
masak Klockan nio, ja. :)
jnthn meh, speling.
masak om typ sex timmar. 12:53
arnsholt Swedish spelling is weird =p 12:57
masak compared to what, Norwegian spelling? :P 12:58
PerlJam good morning #perl6
masak PerlJam: \o
moritz_ o/ 12:59
jnthn o/ 13:00
13:00 perlygatekeeper joined
mathw No, English spelling is weird 13:00
moritz_ encoding sounds in visual symbols is weird all in itself 13:01
pmichaud what time is #rs?
arnsholt masak: Norwegian spelling is about the same I suppose. We also have the most delightfully messed up phonology =D
jnthn pmichaud: Now + 6 hours
pmichaud: Well, give or take some minutes. 13:02
moritz_ which is UTC 19, if my math hasn't left me
pmichaud okay, 19:00 utc then
[Coke] I can put that on the comp.lang.parrot calendar if people want. (or give someone here write access to it.)
(having jumped through the hoops to make the stuff on there already UTC-based)
jnthn was like, "pm musta missed it when masak said so", then noticed masak wrote it in Swedish :-)
masak :) 13:03
pmichaud I thought that was just some rot13 encoded text or something.... :-P 13:04
13:04 nehal joined 13:05 nehal is now known as Guest73293 13:06 Guest73293 left 13:07 rgrau joined
mathw I've developed a bad habit of expecting my friends to be able to understand amusing German song lyrics, but most of them never studied German... 13:07
moritz_ never studied German, and still understands it :-) 13:08
[Coke] conversely, I studied for 3 semesters, and don't! 13:09
ich kannst nicht Deutsche sprechen? 13:10
13:10 JimmyZ joined
moritz_ [Coke]: nicht ganz richtig, aber verstaendlich :-) 13:10
like, my English usualy (not quite correct, but understandable)
[Coke] ... it's wrong but you understand me? close enough. =-)
jnthn *usually ;-) 13:11
[Coke] heh
moritz_ jnthn: exactly my point :-)
13:11 ejs joined
[Coke] I keep hoping I'll have a business need to learn a new language. whenever I do it for fun it stops being fun after a bit and I have no one to chat with. 13:11
jnthn values communication over correctness 13:12
PerlJam [Coke]: you should find some like-minded people and form a "language learning club" :)
mathw My company actually pay for online language training 13:13
I can take any of the courses I like
But I've never quite got around to it
Juerd I can understand most German. It's pretty similar to Dutch :) 13:14
mathw I can understand some Dutch for the same reason
Correspondingly less though, as I'm not a fluent German speaker :)
Juerd Most Germans can't understand Dutch though :)
mathw Odd
I always thought they looked very similar
Juerd They do, but Dutch people are exposed to German more than the other way around. 13:15
mathw Point
Which is another reason why English people are awful at other languages
Juerd The Dutch tend to switch to English all the time
Germans however will try German first, even abroad.
jnthn Juerd: That's another problem English speakers can have learning another language. 13:16
mathw English people will just try English again, more loudly
jnthn Juerd: People switch to English quite naturally in a lot of places.
Juerd jnthn: I know. I've heard several people complain about this habit.
mathw Yeah
On my German exchange we asked someone for directions
And stood there discussing the implications
and he came back and repeated them in English 13:17
We had actually understood them...
we were just arguing about where to go first :)
moritz_ Juerd: I think it's impolite to try German first anywhere except 'de'|'ch'|'at'
Juerd I've forced myself to speak Dutch at the Perl Monger meetings
Even when others were speaking English. 13:18
mathw Trying German first in England is unlikely to be successful
Juerd Finally we're back to Dutch as the primary language for our meetings :)
mathw yay!
English is a terrible language really
Juerd Unfortunately, some non-Dutch mongers no longer attend the meetings
They weren't here to learn our language... 13:19
mathw I think we've got a goal for Perl 7
Make it so expressive you can also use it as a natural, spoken language
Juerd Incorporate Esperanto
13:19 SmokeMachine joined
mathw Then all Perl mongers meetings can be conducted in Perl 7 13:19
Problem solved! 13:20
masak I've been learning English for almost twenty years. there are still big gaps in my knowledge.
Juerd Mi pensas ke uzi Esperanton estus pli facile ol verki lingvon kiu estas kaj komputila kaj homa.
mathw Not an uncommon problem, even among native speakers
PerlJam masak: Americans have been learning English all their lives. There are still big gaps in their knowledge too :)
Juerd Tiu lingvo *jam ekzistas* 13:21
mathw PerlJam: That's because they speak it *wrong*
masak UR DOIN IT RONG
Juerd Preskaux neniu volas lerni lingvon sole por renkontigxoj 13:22
PerlJam mathw: wrong relative to what? Usage defines language; there are no absolutes. :)
moritz_ isn't "there are no absolutes" and absolute by itself, and therefor self-contradictory? 13:23
s/and/an/
jnthn Absolutely.
mathw PerlJam: 'colour' has a 'u' in it :P
PerlJam heh!
moritz_ that's why Perl 6 has aliases for named parameters 13:24
so that you can write :color(:$colour)
jnthn We even have *that* example in book. :-)
moritz_ aye
pmichaud shouldn't that be :$colou?r :-P
Juerd mathw: Strangely, 'color' does not :)
masak moritz_: sure, but the proposition 'there are no absolutes' might still be 'true' from a vantage point outside of the formal system we're in, even though it's paradoxical from the inside. 13:25
jnthn pmichaud: There will be modules... :-P
masak pmichaud: loul!
moritz_ masak: I'm well aware that that's the case here
masak moritz_: thought so :) 13:26
PerlJam this is certainly the right channel for joy and humor^W er, humour ;)
13:27 ejs left
jnthn takes a fag break ;) 13:28
moritz_: btw, did you have any luck with find_lex_skip_current? 13:31
moritz_ jnthn: I looked at find_dynamic_lex in the C code, and it does appear to start at the caller's lexpad 13:32
jnthn moritz_: Ah
moritz_ jnthn: unless I'm mislead by the pcc function names used
jnthn moritz_: No, sounds likely.
pmichaud correct, find_dynamic_lex does not look in the current lexpad
jnthn moritz_: We could make it a friend that dkips current.
pmichaud: oh, it doesn't?
pmichaud it's analogous to find_outer_lex, which also does not look in the current lexpad 13:33
jnthn rakudo: my $*x = 42; say $*x; # wouldn't that be a nuisance for this?
p6eval rakudo 10a321: OUTPUT«42␤»
jnthn Well, it ain't because we do !find_contextual I guess.
pmichaud partially
jnthn I thought NQP just used that op directly, though. 13:34
pmichaud but it's also not-a-nuisiance because presumably the compiler knows about the lexicals in the current lexpad already
jnthn True.
pmichaud ...and :scope<contextual> is smart enough to check the block's symbol table to determine which opcode to emit
moritz_ anyway, it didn't help me in setting $/ for the closure of .subst
pmichaud (either that or NQP is doing it.)
jnthn moritz_: Aww. OK.
moritz_ alpha: my $x = 'foo'; $x.=subst(/(o)/, { uc "$0" }); say $x 13:35
p6eval alpha 30e0ed: OUTPUT«Use of uninitialized value␤fo␤»
pmichaud and yes, it's generally trivial to have find_dynamic_lex look in the current lexpad by simply calling a sub that does the find_dynamic_lex for you :-P
moritz_ alpha: my $x = 'foo'; say $x.subst(/(o)/, { uc "$0" }); 13:36
p6eval alpha 30e0ed: OUTPUT«fOo␤»
pmichaud (whereas the reverse is emphatically not true :-)
moritz_ so it never worked in alpha either, it just didn't barf as loudly as my attempt
pmichaud ...didn't work?
it looks like it worked in alpha to me 13:37
or am I looking at somethign different?
moritz_ pmichaud: the in-place variant didn't work
pmichaud oh
moritz_ pmichaud: and with my patch, the in-place variant dies loudly, but the normal one works as in alpha 13:38
jnthn guesses he could hack up !dispatch_.=
moritz_ gist.github.com/413051 this one
13:38 gbacon joined
jnthn But that won't help us with s/// 13:38
moritz_ correct too
pmichaud oh, .subst can almost certainly be made to work
jnthn pmichaud: The thing is, it goes looking in the immediate callers lexpad. Always.
pmichaud jnthn: right, that's fixable.
jnthn pmichaud: right, I was just expecting find_dynamic_lex to perhaps do it. 13:39
moritz_ but there's still a problem
jnthn pmichaud: That is, find the first $/
Becuase !dispatch_.= sure don't declare one.
moritz_ right now we assume that the closure passed to .subst comes from the same lexical scope
that doesn't have to be the case 13:40
(though maybe it's safe to ignore that problem)
13:40 M_o_C left
pmichaud it's safe to ignore that problem, I think. 13:40
13:41 viklund left
jnthn pmichaud: Did using find_dynamic_lex completely break it, or just not fix the problem? 13:42
er
moritz_:
^
:-)
moritz_ jnthn: it made it not barf, but didn't set $/ as it should 13:43
pmichaud I have just one more fix to make for interpolated regexes, then I can help out with .subst
moritz_ and I think I know why (speculation ahead)
becaue I did it inside the loop
and the loop body is compiled as a subroutine to PIR 13:44
jnthn Ah
moritz_ so the caller of that sub is method subst
jnthn *nod*
yeah. :-(
moritz_ but that might be fixable
13:45 baest joined
moritz_ just need to do the lookup in the outer scope, and do some binding cheating magic at PIR level 13:46
pmichaud oh, wait. my failure in in .subst 13:47
so I guess I get to look at it now.
where is .subst defined?
moritz_ Cool-str.pm
in src/core/
around line 30
13:50 SmokeMachine left, redicaps1 left
pmichaud $result ~= ~($replacement ~~ Callable 13:50
?? $replacement($m)
!! $replacement);
....really? $replacement($m) ?
that looks.... wrongish
moritz_ why?
pmichaud do we expect the match to also be in $_ ?
moritz_ yes
pmichaud hmmmmm 13:51
moritz_ however I realize that doing the call as '$m ~~ $replacement' would be more robust 13:53
because that code handles zero-arg closures too
pmichaud well, no matter what it's going to end up having to be helperfunction($replacement, $m)
where helperfunction does the switching around of $/ 13:54
(or otherwise mucking with $replacement's outer context)
13:56 mikehh joined
moritz_ mucking with $m's outer seems better than with self's caller 13:58
pmichaud that's what alpha did.
it modified $/ in the replacement's outer context
13:58 BrowserUk left
jnthn That'd in theory avoid the .= problem. 13:58
13:59 pyrimidine joined
moritz_ pmichaud: not if I understand that old code correctly 13:59
pmichaud ...assuming that the replacement's outer context was the same as the one calling .subst
moritz_ github.com/rakudo/rakudo/blob/alpha...r.pir#L834
$P0 = getinterp lexpad = $P0['lexpad';1]
pmichaud right 14:00
moritz_ ...
pmichaud that gets the caller's lexpad
and then later
14:00 alester joined
pmichaud lexpad['$/'] = match 14:00
changes the meaning of $/ in the caller's lexpad
moritz_ right
pmichaud the problem in this case is that the caller isn't replacement's outer 14:01
(the caller to .subst is !dispatch.=
and in the more general case, it's probably unwise to assume that the caller is the replacements outer 14:02
moritz_ right; so I thought we could call $replacement.get-outer.lexpad['$/'] = $M
or so
fiddled into parrot semantics
pmichaud that's a fair bit of fiddling
unfortunately parrot doesn't expose the outer context. 14:03
moritz_ :(
it's available at the C level
pmichaud it exposes the outer sub, though, which if cloned properly can do what we want/need.
jnthn That sounds potentially fragile. 14:04
pmichaud it is.
jnthn I'd prefer a dynop that goes off and does the right thing.
Then at least if Parrot changes its structures around we'll blow up when compiling
And know to go and fix it
Than have an obscure bug.
pmichaud let me think on this just a sec
jnthn Aye
moritz_ no hurry 14:05
jnthn needs a little break too, bbs
pmichaud actually, the part that bugs me at the moment is the fact that $/ always defaults to OUTER::<$/> 14:06
I almost wish there was a way to make it do something more useful/different.
moritz_ like, have a way to initialize it on function call?
or make it default to the caller's $/? 14:07
14:07 pyrimidine left
pmichaud or make it always default to the dynamic $/ 14:07
14:07 ruoso_ joined
masak std: my $a; {{{{{{{ $a; my $a }}}}}}} 14:07
p6eval std 30781: OUTPUT«===SORRY!===␤Lexical symbol '$a' is already bound to an outer symbol (see line 1);␤ the implicit outer binding at line 1 must be rewritten as OUTER::OUTER::OUTER::OUTER::OUTER::OUTER::OUTER::<$a>␤ before you can unambiguously declare a new '$a' in this scope at
..…
moritz_ that sounds potentially harmful
pmichaud but that probably causes issues if we start cloning closures with things like $0 and the like in it.
14:08 isBEKaml joined
isBEKaml hallo hexers! :) 14:09
pmichaud okay, here's a bizarrish approach
masak isBEKaml: \o
14:09 SmokeMachine joined
isBEKaml I was bit surprised to find a new branch, ri2 when I git pulled.. What's up? 14:10
pmichaud what if we wrap $replacement with a dynamically-created sub that has its own $/ (that we can set) but that otherwise takes $replacement's outer context as its own?
14:10 ruoso left
moritz_ isBEKaml: simplification of regex interpolation 14:10
pmichaud in some sense it really bugs me to be rebinding $/ in $replacement's outer context 14:11
isBEKaml moritz_: ah, cool. one of the items on your Ponie. :)
pmichaud action-at-a-distance, and all that.
moritz_ aye, it's not a nice solution 14:12
pmichaud I wonder if it could be done as a named argument somehow. 14:13
(which defaults to OUTER::<$/> when not supplied)
14:13 ive left
pmichaud much like what currently happens for $_ 14:14
then we could have $replacement($m, :MATCH($m)) 14:17
and $/ would bind to $m instead of OUTER::<$/>
moritz_ or even call that thing :$/
pmichaud how would we invoke it? (I'm forgetting my syntax at the moment) 14:18
moritz_ which we then can fill with |%( '/' => $m)
pmichaud a little icky, but do-able.
moritz_ less icky than colliding with \w+ names, IMHO 14:19
pmichaud it would mean that all of our blocks end up with named parameters, though.
not so sure I like that.
moritz_ maybe we can filter out immediate blocks at compile time?
pmichaud oh, that's somewhat true 14:20
this would likely be only for blocks w/o an explicit signature
well, at the moment I think I'm leaning toward the dynop. :-| 14:24
although that's a bit messy also, as $replacement is likely to be a wrapped sub 14:25
although I suppose we can de-wrap it before handing it off to the dynop (sounds cleaner that way) 14:26
or, perhaps the better dynop is something that lets us get the current context from a sub 14:27
which we can then muck around with directly.
then it's just $replacement.real_sub.get_outer.get_ctx.get_lexpad['$/'] := $m :-P 14:28
although that's not really right either :-(
really what we want is the outer_ctx attribute of $replacement 14:29
hey, waitaminnit! 14:32
14:36 snarkyboojum joined
pmichaud let me ask around on #parrot a bit 14:36
snarkyboojum bonjour perl6 hackers 14:37
moritz_ bonjour snarkyboojum
isBEKaml snarkyboojum: yo!
snarkyboojum :)
isBEKaml rakudo: enum Stuff <A B C D E>; say E.WHAT; say E.key; 14:39
p6eval rakudo 10a321: OUTPUT«Int()␤Method 'key' not found for invocant of class 'Integer'␤current instr.: '_block18' pc 29 (EVAL_2:0)␤»
14:40 jaldhar left
snarkyboojum moritz_: I noticed day8 and day20 aren't in the spectest list yet - do they need reviewing/changing? 14:40
isBEKaml std: enum Stuff <A B C D E>; say E.WHAT; say E.key; 14:41
p6eval std 30781: OUTPUT«ok 00:01 115m␤»
pmichaud moritz_ / jnthn: I'm going to see if we can get better introspection on Parrot subs; we can then use that to handle the $/ binding from .subst
I've come across other places (REPL) where I really feel I need to have this, also. 14:42
moritz_ setting the sub name by a different API than $P0 = $S0 also wouldn't hurt
jnthn pmichaud: +1
masak snarkyboojum! \o/
snarkyboojum masak-san! o/
:) 14:43
moritz_ wonders what robots get attached to their name, must be different from -san
isBEKaml rakudo: enum Stuff <A B C D E>; say E.WHAT; say Stuff::.keys;
p6eval rakudo 10a321: OUTPUT«Int()␤Could not find sub &Stuff␤current instr.: '_block18' pc 29 (EVAL_2:0)␤»
moritz_ rakduo: enum Stuff <A B C D E>; say Stuff.keys 14:44
rakudo: enum Stuff <A B C D E>; say Stuff.keys
p6eval rakudo 10a321: OUTPUT«ABCDE␤»
moritz_ after about two years I still can't spell 'rakudo' reliably :-)
isBEKaml :)
habity of the ginfers... :) 14:45
oh, well. ;)
moritz_: the Enum's .key is NYI or a known bug? 14:46
moritz_ isBEKaml: known NYI
14:47 rokoteko left
isBEKaml moritz_: club the two unknowns, you get a known! ;) 14:48
I'm trying to search through RT, but the search is a little, er, primitive? Can't do a detailed search? 14:49
Results are out from perl5, parrot and perl6, when I only want perl6. :| 14:50
moritz_ isBEKaml: are you logged in?
it makes a huge difference 14:51
isBEKaml moritz_: I don't have an account yet. Should I create one?
moritz_ yes
14:56 ash__ joined 14:57 ash__ left
XaeroOne in perl6 do you use . instead of -> in perl5? 15:00
moritz_ aye 15:01
at least in the case of calling stuff
when dereferencing array or hash refs, you don't need anything explicit at all
15:01 k23z__ joined
moritz_ rakudo: my $h = { a => 42, b => 23 }; say $h{'a'} 15:02
p6eval rakudo 10a321: OUTPUT«42␤»
15:02 hercynium left
XaeroOne are there perl6 modules somewhere like cpan? 15:03
moritz_ XaeroOne: proto.perl6.org/ 15:05
jnthn proto.perl6.org/
ah, moritz_ beat me :-)
XaeroOne :P
lot has changes in perl6, i wonder if python to perl6 will be easier than perl5 to perl6 :-?
masak note though that proto is *not* like CPAN :)
moritz_ XaeroOne: it depends on which level you want to rewire your brain
15:05 IllvilJa joined
XaeroOne xD true 15:06
it would invole some level of unlearning of perl5 right?
15:07 orafu left
moritz_ yes 15:07
15:07 orafu joined
moritz_ or maybe co-learning Perl 6 :-) 15:07
isBEKaml well, not only perl5 unlearning. You'd have to unlearn some to learn some more. :)
moritz_ it's not like you need to forget how Perl 5 works - I use both 5 and 6 almost every day
IllvilJa what about dropping knowledge regarding 'Lost' in favor of Perl6 ;-) 15:08
That way, you can keep Perl5 AND Perl6 in your head, at the same time!
moritz_ and "lost" isn't such a big loss
snarkyboojum moritz_: me too - it's more like keeping the knowledge of p5 and p6 separate 15:09
XaeroOne dammit I stopped watching Lost after season1
[Coke] XaeroOne: I'm netflixing season 2 now. =-) 15:10
IllvilJa snarkyboojum: isn't Perl6 a bit too viral to stay away from Perl5 in your head?
isBEKaml p5 and p6 can really co-exist. They share a... symbiosis ;)
IllvilJa Watch what the Moose/Catalyst crowd is doing... they are slowly being assimilated into Perl6 (sort of) by letting Perl6 infect Perl5 more and more...
isBEKaml I wouldn't use the word infect... ;) 15:11
IllvilJa BTW, is duke leto around in this channel?
15:12 plobsing left
moritz_ IllvilJa: nope, but usually on #parrot (irc.perl.org) network as dukeleto 15:12
[Coke] moritz++
IllvilJa moritz_: thanks!
[Coke] ponders where the advent stuff sits. +1 to everyone who's hacked on that this week. 15:13
moritz_ only 5 files left without tests 15:14
might be that not all of them need tests in the end
isBEKaml I find them too vague to come up with tests... ## Maybe, next advent should focus more on code than concepts ?
TimToady ⟅うなぎ⟆ 15:15
moritz_ I think we had a good mixture of code and concepts
isBEKaml moritz_++ 15:16
[Coke] I basically just cut and pasted the calendar page in, and made sure that all the code that was presented compiled and did what it said.
except where it didn't. :P
I think MONKEY_TYPING was the only spec change I found in the pages I did.
masak TimToady: I might have missed your answer to my question about the spec's definition of (Unicode) 'character', and about .chars -- or perhaps you're still mulling over it?
15:17 XaeroOne left, XaeroOne joined
isBEKaml yes, I think that way too. I was more inclined towards each day's coverage should give a balanced treatment to code and concepts. 15:17
than, the whole advent series. :)
snarkyboojum [Coke]: some of the posts without tests are a little fuzzier, and require supporting code etc 15:18
or maybe not :) 15:19
isBEKaml [Coke]: case in point, day 10.
[Coke] (require supporting code) yah. still doable, though. 15:20
TimToady masak: at one point, character was supposed to mean the largest available abstraction under use graphs/codes/bytes
but that is not terribly practical with method dispatch, so .chars basically means graphemes these days 15:21
snarkyboojum definitely doable, but not as clean as some of the others
TimToady (except that it really means codepoints in rakudo still)
moritz_ but bytes have a separate data type these days
snarkyboojum ah [Coke] beat me to it
moritz_ so having .bytes in class Str is just confusing
m6locks whoa, building rakudo was like a lightning strike compaared to what it was in the beginning of this year, i see some significant improvement in performance
moritz_ ... and startup time has gone down too
m6locks kudos goes out to y'all 15:22
moritz_ it's currently at about 0.38s
TimToady also, if we get to some kind of NFG, then even Str is represented by a list of integers (but some of them are out-of-line indexes)
moritz_ used to be more like 0.7s-ish
[Coke] chromatic++ # et al.
TimToady so Str and Buf are more like each other than they used to be
so it becomes not quite so evil to use numbers as character positions
isBEKaml [Coke], snarkyboojum: how about tacking regexes and stuff like that without tests with a common set of test data? Rather than using random test data in the specific test files... 15:23
snarkyboojum isBEKaml: common set of test data?
moritz_ TimToady: so... have .graphs on Str and .bytes on Buf, and in each case .chars re-dispatch as appropriate?
masak TimToady: thanks. that makes it easier to think about. I'll see if I think some of the spec needs rewording to be clearer. 15:24
TimToady maybe we need a Cod string type :)
pmichaud looks like I have regex interpolation working again... just need to catch all of the repos up with the changes.
TimToady rakudo: say '𠱠'.chars 15:25
p6eval rakudo 10a321: OUTPUT«1␤»
moritz_ doesn't like Cod, or fish in general
[Coke] isBEKaml: the advent tests aren't for features, they're just to make sure the sample code we have put out there still works.
pmichaud (and run lots of tests)
TimToady doesn't like UTF-16, or codepoints in general
[Coke] anything more complicated than that should be tested outside of t/spec/integration.
masak TimToady: then we could refer to slices of Cod, and Codpieces.
[Coke] pmichaud++
(that other guy)++ 15:26
pmichaud
.oO( t/spec/complicated ?)
15:26 BrowserUk joined
dalek p-rx: c69f6f6 | pmichaud++ | src/Regex/Cursor.pir:
Avoid "Cannot take substr outside of string" exceptions on interpolated literal
15:26
p-rx: 64ff6a1 | pmichaud++ | src/stage0/ (4 files):
Update bootstrap.
moritz_ next parrot, then rakudo, then the world! 15:27
moritz_ -> decommute
pmichaud yes, testing parrot now
isBEKaml [Coke]: ah, not for features. I was thinking more of broader tests like regexes.
Maybe, we could just make some tests common across them. like variations of strings, rather than repeating them through test files.. 15:28
15:30 rob___ joined
[Coke] isBEKaml: I think the regexp tests are already setup like that. (at least the parrot ones are.) 15:30
masak aye. the PGE tests are essentially big tables of situations. 15:31
a script parses them and runs them as TAP tests.
isBEKaml Err, feel free to ignore them if not suitable. :)
pmichaud okay, merged locally to master, spectesting 15:32
isBEKaml (i'm not that conversant in perl6 yet to come up with concrete examples)
My perl6fu is still in monosyllables. ## or even primitive. ;) 15:33
pmichaud well, in perl 6 those monosyllables can take you a lot farther :-) 15:34
isBEKaml pmichaud++ ## Too true. :) 15:35
S05-nonstrings/basic.t is happpy ## Clue: Zaphod. :) 15:36
15:37 rob___ left 15:39 JimmyZ left 15:43 Ross joined
lisppaste3 cognominal pasted "I expected that this subclassing would worl" at paste.lisp.org/display/100464 15:44
cognominal probably pmichaud or jnthn are the most qualified to help 15:45
Apparently I cargo culted something...
jnthn Where are you trying to run this? Rakudo?
cognominal yup
jnthn Maybe it's HLL issues 15:46
cognominal probably I aped npq code.
jnthn (PCT::Dumper lives in parrot I guess)
try like
$P0 = get_root_global ['parrot';'PCT'], 'Dumper' 15:47
And 'parent'=>$P0
may work
You may want a class HTMLDumper { ... } above too, just to keep the parser happy when it sees the augment. 15:48
oh, otoh the BEGIN block will put the thingy in the namespace
So it'll work anyway I guess.
cognominal jnthn++ # Can't augment class HTMLDumper without 'use MONKEY_TYPING' 15:49
I am getting there :)
jnthn oh, you need that too, yeah :-)
cognominal some errors are less frightning to me ) 15:50
jnthn When they're less frightening, that means they're good ones. :-)
cognominal I hope to make perl6 code navigage using the web eventually 15:51
*navigable
ideally, "eventually" means for fpw
jnthn, you have a talk cut for you "crossing boundaries between rakudo and parrot" 15:52
snarkyboojum typo in S02-bits.pod line 1087 "in the sense that it always calls any closure it finds *it its list" should be "in" :) 15:54
cognominal now, Class HTMLDumper already registered! 15:55
ho, nope :( 15:59
16:00 SmokeMachine left
snarkyboojum oh wait, I might have commit access 16:00
cognominal i mistyped parent, back to Parent isn't a Class 16:01
pugssvn r30782 | snarkyboojum++ | [S02] typo fix
16:04 SmokeMachine joined
cognominal f* me, read the error message!! 16:06
pmichaud yay! all tests pass!
moritz_ \o/ 16:07
isBEKaml pmichaud++ \o/
pugssvn r30783 | pmichaud++ | [t/spec]: Fudge out some sequential-alternation tests.
pmichaud (big) push coming :)
snarkyboojum :) 16:08
cognominal pmichaud++ 16:09
16:09 hercynium joined
colomon \o/ 16:09
jnthn pmichaud++ 16:10
jnthn builds latest
dalek kudo: d034418 | pmichaud++ | (8 files):
Revert "Regex Interpolation"
 16:12
kudo: 3362b1a | pmichaud++ | src/Perl6/ (2 files):
Refactor <codeblock>-related assertions.
kudo: f9a12d2 | pmichaud++ | src/Perl6/ (2 files):
Add variable assertions.
kudo: a48d4df | pmichaud++ | build/PARROT_REVISION:
Bump PARROT_REVISION to get nqp's "does invokable" fixes.
kudo: 263a905 | pmichaud++ | (4 files):
baebe1f | moritz++ | (2 files):

Fixes RT #68482. Inspired by patrickas++ Also adds three more test files (of which two are unrelated)
masak pmichaud++ \o/
masak builds too
moritz_ recommends to merge with --no-ff
cognominal --no-ff? 16:13
moritz_ no-fast-forward
sorear why?
moritz_ because that way it's much easier to see what belonged to the branch
and for example bisecting along the master branch shows the branch as one commit, not as many 16:14
cognominal too late...
XaeroOne does perl6 have something like XS?
moritz_ it's no big deal, just something which I'd like to propose as best practise
XaeroOne: github.com/jnthn/zavolaj
XaeroOne thanks 16:15
sorear XaeroOne: Not really. zavolaj handles foreign binding pretty well, but we don't have anything for the other use cases 16:16
jnthn sorear: That said, there's a bunch more introspectable stuff from Perl 6 land too. 16:18
sorear: And other ways to extend things. 16:19
So hopefully many of those bits of XS won't need dropping to C.
Or...whatever Perl 6 happens to be running on...
16:19 simcop2387 left
sorear I'm a member of the "profile, then rewrite 10% in a lower level language" school of optimization 16:20
moritz_ would kinda love a C::function__sig(args) interface like we have with pir::
sorear ooooooh
Yes.
moritz_ "powered by macros", maybe 16:21
sorear compiler_psuedo_package is absolutely doable as a macro in my new (unimplemented) framework
16:21 mberends joined
pmichaud when NCI gets cleaned up a bit better in Parrot, then c::function(...) seems very doable. 16:22
jnthn spotted the new NCI proposal today
moritz_ sorear: implement it! :-)
sorear how does it differ from the old proposal?
pmichaud lunchtime here, then repl 16:23
jnthn sorear: Well, differs from current reality more than an old proposal.
sorear \o/
jnthn sorear: Seems there's syntax for structs that can be declared right in there.
Plus adds missing types
It's on Parrot mailing list.
I didn't give it a good look yet, just glanced through it earlier and was happy to see someone working on it. :-) 16:24
sorear same here.
16:25 TiMBuS left
cognominal whate the difference between NCI and zavolaj? is this between a spec and animplementation? 16:25
16:25 clintongormley joined
jnthn cognominal: NCI (as I was talking about it here) is the Parrot layer. 16:25
cognominal: Zavolaj is a Perl 6 module written atop of that so that you may write Perl 6 subs with Perl 6 signatures, mark them as native and coming from a library, and then Zavolaj introspects the signature and builds a Parrot NCI signature and makes the sub a call forwarder. 16:26
16:26 patspam joined
jnthn cognominal: See the mysql example to get an idea of some of what it can do. 16:27
16:28 justatheory joined
pmichaud jnthn: I'm thinking of a new dynop: find_outer_lexpad <pmc>, <str> 16:28
jnthn <str> ?
pmichaud given a Sub pmc, it follows the outer contexts until it finds the lexpad that contains <str>
jnthn oh
pmichaud in our case, we would have
jnthn OK, makes sense. 16:29
pmichaud lexpad = find_outer_lexpad $replacement, '$/'
jnthn The name find_outer_lexpad_containing ...
...is too long
:-)
Though clearer. :-)
pmichaud well, all of the other find*lex always include a symbol :-)
16:29 rgrau left
jnthn It didn't say any of those were well named. :-P 16:29
pmichaud there is that :-)
jnthn Anyway, +1. It's a nice general mechanism. 16:30
pmichaud right
jnthn Plus, if you can get hold of the lexpad, you can look the thing up *and* bind into it.
So no need for a matching store op.
pmichaud right
jnthn Plus get the $_ in the same scope, if that was ever gonna be interesting. Or the $!. Or something. :-)
TimToady sorear: re recalc, it's done in _AUTOLEXgen about line 1937; uses moose to find all visible methods and calls each one's PEEK to get its sublexer. 16:31
jnthn pmichaud: Actually it may be that this makes callsame and friends neater...
sorear TimToady: How does it know to rerun in subclasses?
jnthn pmichaud: Or maybe covers some or all of what the dynop involved in that business does.
pmichaud I've also considered find_dynamic_lexpad and find_lexpad variants as well
TimToady it's called JIT
on the actual call into the subclass's method
pmichaud anyway, I'll do find_outer_lexpad a bit later, if nobody beats me to it. :)
TimToady of course, the recalc is expensive, so we try to avoid it if possible 16:32
16:32 gbacon left
TimToady two main ways 16:32
pmichaud time for lunch here -- bbiah
TimToady first, caching, both in-memrory and on-disk
16:32 pyrimidine joined 16:33 simcop2387 joined
TimToady second, when we know it's an anonymous language such as from a macro, we actually look at its parent's lexer, and if none of the fates mention the current category, we know that we can just copy it down, since nothing will be recalculated 16:33
that'd done at line 1818
it's kinda convenient that the fates also track the dependencies 16:34
16:35 SmokeMachine left
sorear oh I see, $::LEXERS is keyed on the *class* 16:36
TimToady we could probably do more with that, though it's possible it could make things worse, if the 'peek' on the superclass forces lexer generation where we happened to have never needed the lexer
and, in fact, that could be one of the things that slows down a list of operator overloads that aren't actually used
yes, ref $self, and the //= is the typical "I'm a cache" idiom 16:38
obviously, requires 5.10 :)
and then after we gen the lexer, we cache linked lists for fates, since they can share tails 16:39
and then we load up the trie for constant prefixes
if we had a better nodal representation of the regexes, we could extend that trie into a parallel nfa 16:40
currently, all the non-constant regexen are just evaluated at the position where they make that transition 16:41
16:43 ilogger2 joined
TimToady well, we could, instead of node pointers, just keep offsets into the regex strings, but that's icky 16:43
moritz_ std: / <[ \xdefg ]> /
p6eval std 30783: OUTPUT«ok 00:01 115m␤» 16:44
16:44 stephenlb joined
TimToady oh, I have a fix for the \u thing there 16:44
moritz_ is that the same as <[ \xdef g ]> ?
or are hex chars always supposed to come in pairs?
TimToady yes, \x, \o, \b etc always eat as much as they can
16:45 SmokeMachine joined
TimToady which is not much of a burden when you can always put whitespace 16:45
or use []
moritz_ ok
just wanted to know how to implment it
pugssvn r30784 | lwall++ | [STD] fix up backslashes in character classes to allow \s etc, and reject \u etc. 16:46
r30785 | sorear++ | [viv] Implement 'proto' regexes
TimToady er, not \b 16:47
sorear TimToady: what is the arrow munging around gimme5 line 430 for?
TimToady it's just installing a 'sym' parameter before the --> 16:49
16:49 am0c joined
TimToady derived from the :sym<op> 16:49
sorear what's the -->?
TimToady the return type comes after it
sorear ...what? in perl5?
TimToady the sig isn't really processed into p5 until down about line 509 16:51
sorear oh 16:52
TimToady and the return type is the first thing stripped off there
so the line you were asking about eventually produces a p5 line such as: local $::sym = $args{sym} // 'twigil'; 16:53
16:55 snarkyboojum joined, cdarroch joined, cdarroch left, cdarroch joined
TimToady this poking into the sig somewhat predates the ability to declare contextuals within a regex, which is why it ended up being "declared" as a fake parameter 16:55
it's a bit of a fossil these days
nobody is really expected to pass :sym as an argument 16:56
it's just abusing the named parameter mechanism to fake a declaration of $*sym 16:57
which used to be used by the <sym> rule as a default 16:58
when we actually had a <sym> rule
nowadays gimme5 just inlines the check for the literal symbol 16:59
and also inlines any 'endsym' check such as for a word boundary
sorear what's the standard way to pull a value out of a colonpair in compile time code? 17:00
17:02 SmokeMachine left 17:06 ruoso_ joined 17:08 ash___ joined
TimToady there is no standard yet, it's all ad hoc, and some of it is ad hoccer than others 17:12
the most complete one is probably in the slightly misnamed add_macro
it will be very much simpler when we can just eval some AST
17:15 Trashlord joined
TimToady I think maybe I'll rename add_macro to add_categorical 17:15
17:16 ash___ left
pugssvn r30786 | moritz++ | [t/spec] unfudge complex number comparison tests for rakudo, colomon++ 17:16
17:18 Trashlord left 17:19 Trashlord joined
moritz_ rakudo: say 'a' ~~ /(o) | (.)/; say $1 17:20
p6eval rakudo 52a137: OUTPUT«a␤Any()␤»
moritz_ rakudo: say 'a' ~~ /(o) | (.)/; say $0
p6eval rakudo 52a137: OUTPUT«a␤a␤»
moritz_ uhm
is that correct?
I thought (o) would be the $0 capture, and (.) the $1
moritz_ re-reads S05 17:21
TimToady no, alternatives restart the count
it's in there somewhere
pugssvn r30787 | lwall++ | [STD,Cursor] rename add_macro (which it doesn't) to add_categorical (which it does)
TimToady S05:3017 17:22
first hit on /restarts
moritz_ just found it 17:23
it's easy if you know what you need to look for :-)
TimToady and p6 is easy to implement if you know what code to write :P 17:24
moritz_: on putting Whatever outside of Any, I proposed that once upon a time, but these days * op 42 is supposed to be recognized by the compiler, by spec, so the exact dispatch semantics are a bit less crucial 17:27
but we could still probably put it outside of Any, alongside Junction and the conjectural Each 17:28
17:29 snarkyboojum left
TimToady maybe someday those will all be derived from Concept, which is derived from Mu 17:29
17:29 SmokeMachine joined
TimToady maybe there's a better word than Concept 17:31
pugssvn r30788 | sorear++ | [viv] implement 'multi' regexes
TimToady I guess they're all meta-programming notions 17:33
sorear how is the ENDSYM translation supposed to work? 17:34
TimToady it inlines the call right after it inlines the sym 17:35
sorear examples of use?
oh, I see
is it a STD.pm6 specific thing? I don't see anything on it in S05 17:36
TimToady it's probably more generally useful than just for STD, I suppose 17:38
but it's probably going to fall out of how we've redefined protos
and be specified inside the proto's block instead of a a trait 17:39
sorear is STD's usage fossily?
17:40 SmokeMachine left
TimToady so the proto does something like { :my $*endsym = 'nofun'; <callsame> } 17:40
yes, somewhat fossily
so instead of just throwing away the current <...> and replacing it with the proto's subdispatcher 17:41
that's just substituted in for the <callsame>/<nextsame>
17:41 jbwiv joined
TimToady that's the direction I'm headed, one of these days 17:41
so the proto gets full "wrapper" privileges
and can even usurp the category entirely 17:42
so you can override an entire category and make it not inherited in a subgrammar using proto token mycategory { <?> } 17:43
or put an error message in there saying "You can't do that in this DSL!?!?!" 17:44
pmichaud back 17:45
TimToady anyway, as usual, traits are a smell
so endsym will be going away in favor of a contextual, or some such
hmm, 'dynamic variable' is just too hard to say 17:46
how soon till we shorten it to dynavar?
pmichaud I think it just happened. 17:47
TimToady starvar
pmichaud although I've sometimes done "dynvar"
TimToady 'course, some starvars don't var
pmichaud "constants aren't, variables won't"? 17:48
TimToady dynathingy...
if it's small, it could be a dynamite
sorear dyvar 17:49
TimToady: so <sym> has to depend on a contextual at run time?
unneccessary dynamic dependencies are also a smell to me...
TimToady troo
so maybe it stays a trait, but we're still making proto bodies useful (if normally inlined) 17:51
so the dynamic mechanism will be there if we want it someday
sorear punts on endsym handling for now 17:52
what is _SYM in Cursor for? 17:53
TimToady it's the rule that matches <sym> if we can't inline it 17:57
gimme5:1964
(which is unreachable code) 17:58
sorear is there any full perl6 situation where <sym> can't be inlined? 17:59
lisppaste3 cognominal pasted "contents of Q:PIR in a BEGIN is executed twice" at paste.lisp.org/display/100472
sorear yes, BEGIN blocks are executed twice 18:00
cognominal why?
TimToady it would appear that the only place that calls it anymore is Cursor.pmc:2393
cognominal th 18:01
sorear if we don't run them immediately during parsing, we're going to parse the rest of the file with wrong assumptions
cognominal I mean normal stuff in a BEGIN is not executed twice. It is proper to stuff in Q:PIR apparently
jnthn sorear: Aye, though we're meant to only run them once in any given execution of Rakudo.
sorear oh
TimToady sorear: if we were to force the setting of $*sym before the proto call, then a proto could match <sym> <endsym> generically 18:02
sorear ineresting.
jnthn rakudo: BEGIN { Q:PIR { say "hi" } }
p6eval rakudo 52a137: OUTPUT«hi␤hi␤»
jnthn wtf
I thought I'd fixed that.
rakudo: BEGIN { say 42 }
p6eval rakudo 52a137: OUTPUT«42␤»
jnthn ...
huh.
sorear perl -p -e 'next if /^---/../\A\w+\Z/;' -e 's/\A[ \t]+//;' STD.pm5 >STD.pmc
what?
jnthn So what's special about Q:PIR ...
TimToady rakudo: BEGIN { say 42 }; 42 42 18:03
p6eval rakudo 52a137: OUTPUT«42␤Confused at line 11, near "42 42"␤current instr.: 'perl6;HLL;Grammar;panic' pc 501 (ext/nqp-rx/src/stage0/HLL-s0.pir:327)␤»
TimToady rakudo: BEGIN { Q:PIR { say "hi" } }; 42 42
p6eval rakudo 52a137: OUTPUT«hi␤Confused at line 11, near "42 42"␤current instr.: 'perl6;HLL;Grammar;panic' pc 501 (ext/nqp-rx/src/stage0/HLL-s0.pir:327)␤»
cognominal hum it is exectued twice when there is Q:PIR including stuff outside the Q:PIR
TimToady only one 'hi' before the parsefail 18:04
cognominal I am sure jonathan will fix that in a jiffy :)
pmichaud get_hll_global $P35, ["Perl6";"Actions"], "%BEGINDONE"
set $P36, $P35["BEGINDONE_12"]
what is that for?!?
TimToady sorear: that's just removing leading whitespace so p5 can parse it faster, but not removing the whitespace from the embedded yaml 18:05
18:05 SmokeMachine joined
jnthn pmichaud: What I just said - making sure we don't run it during the parse *and* at runtime. 18:06
pmichaud jnthn: fire_phasers is supposed to take care of that already
you should just be able to fire the BEGIN phaser and it'll be prevented from firing again later
TimToady BEGIN ought to simply remove any reference to itself once it is run, I'd think
cognominal your phasers are firing before aiming :)
TimToady the whole point of BEGIN is to fire before aiming, and not after :) 18:08
cognominal but twice?
TimToady since BEGIN is the fire that is creating the cannon :)
I said *not* after 18:09
pmichaud with %BEGINDONE there's also a huge potential for conflict among loaded modules that also have BEGIN blogs (that might also be called "BEGINDONE_12")
TimToady you don't want to melt your cannon again after you've made it
pmichaud anyway, I suggest using the built-in fire_phasers mechanism rather than inventing another wheel. if fire_phasers isn't doing the right thing, we can fix that.
TimToady that tends to have deleterious results
pmichaud wonders about the KABOOM phaser.
TimToady does fire_phasers have a :delete-on-completion option? 18:10
pmichaud TimToady: We could add that, yes. More precisely, fire_phasers keeps track of everything it's fired and refuses to fire it again.
TimToady seems like it'd be useful for any kind of only-do-once semantics
if you delete it, you *know* you can't fire it again, unless someone keeps a reference to it sekritly 18:11
pmichaud that's the problem -- Parrot tends to have lots of sekrit references
so, we delete it *and* we keep track of it (but not by reference) so that it's not fired again.
(it's kept track of by uuid)
TimToady I suppose if you deleted parrot you'd also delete some references you want to keep :) 18:12
pmichaud "!begin_unless_begun"("BEGINDONE_12", $P38)
ugh
TimToady maybe you need a fire_and_forget command :) 18:13
well, I suppose that metaphor doesn't really work, since have to wait for it to hit its target 18:14
*since you
cognominal fire and forget, like in Doctor StrangeLove
no problem with supporting users anymore 18:15
jnthn pmichaud: I guess I musta taken the "this is wrong" comment above the begin phaser impl to mean it was wrong to be using the fire phasers mechanism. 18:17
I guess we could tweak that to do the righ tkinda thing.
pmichaud looks
jnthn TimToady: The problem kinda is that if somebody writes BEGIN { @*INC.push('blah') }
TimToady: Then pre-compiles to a bytecode file... 18:18
...they'll probably be surprised when @*INC doesn't contain blah when they run the compiled bytecode.
pmichaud jnthn: "this is wrong" comment?
jnthn: I would think that BEGIN could work with 18:20
jnthn pmichaud: The one that used to be above the begin phaser. 18:21
Which I was attempting to deal with.
pmichaud method statement_prefix:sym<BEGIN>($/) { add_phaser($/, 'BEGIN'); Q:PIR { '!fire_phasers'('BEGIN') }; }
or something like that
I'll go look at previous code to see if I can remember the genesis of the comment
jnthn pmichaud: What does that do with the return value?
Also, does it immediately compile the code? 18:22
pmichaud it can do so, if we want it to
jnthn OK
18:22 gbacon joined
jnthn It just felt like quite a lot of changes for the fire phasers stuff. 18:22
pmichaud at one point, I had !add_phaser so that a phase of 'BEGIN' automatically invoked the item 18:23
jnthn Could it return some "handle" to then emit into the code and access the value at runtime?
cognominal afk&
pmichaud jnthn: yes, that would seem to make sense 18:24
jnthn Anyway, if your intention is that fire_phasers handles that, I can re-work things to do so.
pmichaud we probably need that for CHECK / INIT also, anyway
jnthn It just wasn't clear to me at the time that was the way to go.
pmichaud yes, my intention is that fire_phasers is used to handle the various phases of execution 18:25
jnthn OK
Then sorry for mis-guessing. :-)
pmichaud including BEGIN, if possible
oh
(looked at the comment)
# XXX BEGIN isn't correct here, but I'm adding it along with this
that one?
jnthn Yes 18:26
pmichaud I merely meant that the implementation of BEGIN there was incomplete, not that it shouldn't be using fire_phasers
bad comment on my part.
jnthn "along with this" I guess I took as "along with the add_phaser-y htings"
18:26 meppl joined
jnthn Whereas I guess you meant, along with this patch, as a starting point. :-) 18:26
pmichaud right
jnthn Ah well 18:27
'tis fixable.
pmichaud anyway, I'd like it all to go through the add_phaser/fire_phaser mechanism, if possible
and yes, we probably need a way to retrieve the return value from any fired phaser
it'd be fine with me if fire_phasers stored that also 18:28
then have some other function to look up return values by phaser
jnthn OK.
Feel free to leave that with me.
I'll re-work it.
pmichaud currently %!PHASERS_FIRED just has a hash which keeps a reference to the phaser. Would make perfect sense to have it keep the return value of the fired phaser. 18:29
jnthn Aha.
Yes.
pmichaud phasers.pir:67
18:30 patrickas joined
patrickas sabaho perl6! 18:31
TimToady jnthn: re @*INC, that's why you need to have a freeze/thaw model of the proto-data constructed by the compiler, and not try to figure out which bits of the construction code need to be redone later
moritz_ ni hao patrickas
patrickas moritz_ I think I got another patch for a LHF :-) 18:32
moritz_ patrickas: \o/
patrickas I'll mail you the diff ...
moritz_ si tu veux
patrickas running spectest now ... should be two extra passing tests 18:33
jnthn TimToady: That's quite a requirement on a runtime.
18:33 Trashlord left
TimToady it's vastly simpler 18:33
patrickas hehehe je veux je veux.
jnthn I guess it's muchly what is wanted for the setting, though. 18:34
patrickas Actually I am not sure what is the best practice ... sending diff by email seems really low tech... and I still am not very confortable with git
jnthn Goes without saying I don't see Rakudo on Parrot being able to get this right all that soon.
18:34 Trashlord joined
jnthn Though somebody did mention on working on the freeze/thaw bits. 18:34
plobsing++ iirc. 18:35
moritz_ patrickas: the high tech solution is to fork rakudo on github, and push your changes there
TimToady it's not just BEGIN, it's all code that needs to run at compile time, like constant and enum initializers
moritz_ so you get version control, but if you screw something up you still don't harm the "canonical" rakudo repo
TimToady lexical variables with proto-values
jnthn -> dinner, back in 20 for #rs. 18:36
TimToady but a data structure that is decided in only one spot is inherently more stable than one that has to be redecided and reconstructed later 18:37
patrickas moritz_: I was trying to merge back your process-cmd-args fork to practice that but I failed miserably ... so I decided to stick to old tech for now :-)
TimToady this is something I discovered (to my dismay) in Perl 4, which is why Perl 5 does constant folding with the same run-time engine that the normal expressions use
Perl 4 had its own constant folder, and it was always out-of-sync with the actual expresion evaluator 18:38
but making the run-time engine available to the compiler in Perl 5 implies a promise to keep around all the data-structurey side effects of running any code at BEGIN time 18:39
18:40 hercynium joined
TimToady which doesn't matter so much when you just run the program after CHECK time in the same process, like p5 does 18:41
but it does put the freeze/thaw requirement for separate compilation
sorry if it sounds like I'm repeating myself every year or so. :) 18:43
pmichaud if you are, it's worthwhile repetition.
Parrot's freeze/thaw capabilities just don't quite measure up to the task at present, I fear.
TimToady I just figure there are some things worth saying too many times rather than too few...
with parrot, you might be able to fake it by setting up the namespace/lexical scope skeleton, and then poking the data in after the fact, but that data should probably be harvested by a tree walker, not by trying to dup the original init code 18:45
since it cannot be proven idempotent 18:46
pmichaud yes, the challenge is to know when to stop walking the tree
18:46 tlb joined
TimToady basically, the freeze/thaw provides a *transaction* 18:47
or some kind a sequence point where you know what has happened before, and what has not, and the state is known to be consistent
18:48 kfo_ joined
TimToady STD's .syml files are a primitive version of that, since STD doesn't have an evaluator, so can't produce "real" data, but it does in fact dump all the lexical scopes and packages 18:49
adding data to those would be fairly trivial, if I had any data. :)
18:50 Trashlord left 18:51 kfo left, Trashlord joined, kfo_ is now known as kfo
TimToady wanders off to see if there's a suction cup that will fit his hot button so he can pull it back up... 18:52
18:53 clintongormley joined, masak joined
moritz_ #rs in 5? 18:54
masak aye.
moritz_ with our first gsoc report from masak?
18:55 plainhao joined
masak er, um, sure. 18:55
colomon \o/
moritz_ speaking of planning stuff... any good ideas for this week's challenge? 18:56
colomon "your challenge is to figure out a good challenge for this week, and then do it."
masak recursion!
jnthn omnomnomköttbullar 18:58
jnthn is back 18:59
masak 'perl 6 cookbook?' use.perl.org/~iaw4/journal/40366 19:00
jnthn ;-)
masak: About things Perl 6 devs cook? :-)
moritz_ I know of none
moritz_ made a tasty Lasagne on Sunday 19:01
jnthn didn't hear about one
Su-Shee whut? I just tweeted my bread recipe!
masak compiling a cookbook is probably too big a week challenge. :P
moritz_ and that barbecue yesterday was quite nice, too
jnthn I'm glad I just ate, rather than came here hungry. :-)
masak jnthn: you had meatballs? I had a yummy pasta salad.
jnthn masak: I make them about once a week now. 19:02
med lingonsylt och smörgåsgurka ;-)
masak I hardly ever eat them. my parents refused to cook typically Swedish food, so I never got into the habit of eating it at home. :) 19:03
19:03 quietfanatic joined 19:06 ShaneC1 joined
patrickas just encuntered %h1<b>:delete and is kind of confused (so is rakudo) ... I guess means %h1.delete('b') ... where can I read about it ? 19:06
19:06 Trashlord left
moritz_ it's the official adverb syntax 19:06
S03 probably, or so
19:06 ShaneC1 left
masak NYI. 19:07
19:07 smash_ joined
smash_ hello everyone 19:07
masak smash_! \o/
patrickas cause the t/spec/S32-hash/delete.t test file uses this syntax so none of the tests are being run even though delete itself is implemented 19:09
19:09 Trashlord joined 19:10 clintongormley left, Trashlord is now known as Guest35543 19:15 quietfanatic left 19:25 SmokeMachine left
masak I see you've been discussing freeze/thaw between the compile phase and the run phase in the backlog. fwiw, that's the model Yapsi is drifting towards. I only wish I knew a blazingly fast way to thaw things with Rakudo. 19:25
(preferably not involving &eval)
I don't understand TimToady's comment about the constant folder in Perl 4 being 'out-of-sync with the actual expression evaluator', though. that sounds interesting. 19:26
19:28 plainhao left
pugssvn r30789 | patrickas++ | Modified the delete.t file to use the normal form of delete, this way it can be added to the spec tests. 19:30
r30789 | Created delete_adverb.t which uses the adverbial form that was used in the original file the file should not be added to the spec tests since adverbs seem to be NYI
19:36 tlb left 19:38 Guest35543 left
patrickas I think S32-array/delete.t can now be safely added to rakudo's t/spectest.data 19:38
19:38 Siddy joined, Siddy is now known as Trashlord
moritz_ tries it 19:39
19:40 tlb joined
moritz_ patrickas: btw I have a nicer patch for RT #68482: return |@deleted 19:42
19:42 masak left
moritz_ that way we don't need any special-casing of one-item deletes 19:42
patrickas :-)
moritz_++, I love one character patches ... 19:43
moritz_ there's nothing wrong with special-casing, it just feels wrong :-)
19:44 masak joined
patrickas yea it does :-( but on the bright side it worked :-) 19:44
19:47 SmokeMachine joined 19:48 masak left 19:49 masak joined
patrickas moritz_: |@deleted is a capture ? 19:49
moritz_ patrickas: it's "interpolation into arg list"
rakudo: sub f($a, $b) { say "a: $a; b: $b" }; my @a = 1, 2; f(@a) 19:50
p6eval rakudo 52a137: OUTPUT«Not enough positional parameters passed; got 1 but expected 2␤current instr.: 'f' pc 241 (EVAL_1:44379420)␤»
moritz_ rakudo: sub f($a, $b) { say "a: $a; b: $b" }; my @a = 1, 2; f(|@a)
p6eval rakudo 52a137: OUTPUT«a: 1; b: 2␤»
moritz_ in the first example we have just one positional argument, which is an array
in the second, the array is interpolated/dissolved into the argument list
patrickas ok I see ... thx. 19:51
TimToady think of the | as the place where you slice the argument list open with a knife and insert the extra arguments
patrickas hahaha so I get to be a muse for patches now :-P
19:53 masak left
[particle] i'm so confused. i'm at a conference, and "the guy in the second row" is sitting in the last row. 19:55
patrickas TimToady: thx for the image, for some reason it reminds me of Why's (Poignant) Guide to Ruby...
TimToady [particle]: someone just shuffled the rows before you got there--it's still the second row 19:58
the last row is actually #4 right now
[particle] $seat.pick 19:59
lisppaste3 pmichaud pasted "<&var> now works" at paste.lisp.org/display/100475 20:00
jnthn pmichaud: \o/ 20:01
lisppaste3 pmichaud pasted "<xyz=&var> now works" at paste.lisp.org/display/100476 20:02
pmichaud spectesting, then committing.
pugssvn r30790 | moritz++ | [t/spec] tests for <&foo> interpolation 20:03
r30791 | moritz++ | [t/spec] some random fudges
r30792 | pmichaud++ | [t/spec]: more lexical regex tests 20:06
r30793 | moritz++ | [t/spec] test renaming of lexical regexes with other name 20:09
20:13 _ilbot2 left 20:14 _ilbot2 joined
moritz_ pmichaud: do you need more tests? 20:17
20:23 patrickas left
moritz_ -> bed 20:23
TimToady n8 20:24
pmichaud moritz_: no, this is plenty of tests 20:27
20:28 pragma_ joined
dalek kudo: 2404007 | pmichaud++ | (2 files):
Add /&var/ and /<&var>/ regex interpolation syntax, and enable tests.
 20:42
sorear !!! pmichaud++ 20:45
pyrimidine pmichaud++ # very nice! 20:56
PerlJam $ perl6 21:06
> 34+8
42
*that's* nice
pmichaud++ 21:07
jnthn sorear++ too :-)
sorear jnthn: what did I do??? 21:08
jnthn sorear: Worked on auto-print support as well?
sorear oh, I blinked 21:09
I thought you were still on about <&var>
jnthn :-)
PerlJam although ... this isn't nice at all ...
> my $a = 5;
5
> say $a;
Segmentation fault
jnthn core blimey!
sorear PerlJam: known issue (exception handling doesn't work)
$a needs to be $GLOBAL::a anyway 21:10
no lexical persistence (yet)
pmichaud has given himself until 20 hours from now to fix that
[Coke] I've sort of been volunteered to coordinate any parrot issues rakudo needs for the june release. Feel free to assign any trac tickets to me or ask me to follow up on statuses for the next month or so. 21:11
though the latest I heard is "just don't break anything" 21:12
jnthn Please don't. ;-)
[Coke] would probably be good to setup a continuous build to insure that.
jnthn [Coke]: I guess the GC changes that are being started now will land after it, anyways.
[Coke] is anyone watching the rakudo smolder reports? 21:13
jnthn: I think some of those might land sooner.
jnthn [Coke]: Ah, OK
[Coke]: Cleanup or functional?
[Coke] bacek is usually pretty good about doing anything invasive on a branch, so we'll have time to test. 21:14
and... I don't know. =-)
jnthn Ok. :)
[Coke] bacek says "It will be faster." 21:15
. o O (whee!)
bacek arrive in flames and smokes
jnthn \o/
bacek I will try to make new GC ready for R*
It's my main goal/driving reason.
PerlJam "make a new GC" sounds like "break things" to me ;-) 21:16
bacek It is not. Ok. Actually - it will not :) 21:17
[Coke] if it breaks anything, it won't go in. no worries. 21:18
21:19 pmurias joined
diakopter ______________ 21:20
[Coke] oh no, he's flatlined.
pmurias diakopter: hi
[Coke] diakopter: please assure me that C# is really cool once everything is installed and working. :P 21:21
(that's not perl6 related)
21:21 iblechbot joined
diakopter /^\ /^\ 21:21
[Coke]: er, hee?
by "everything" do you mean mono 2.6.latest? :)
diakopter confused still 21:24
[Coke]: funny you mention "cool" since "Cool" was the internal codename for C# at microsoft in the late nineties
wolverian projectfortress.sun.com/Projects/Co...ingPartOne I like fortress more and more 21:25
diakopter C-style Object Oriented Language 21:26
diakopter [Coke] I think I missed your joke, if intended :) 21:27
/me , /msg , same thing 21:28
[Coke] diakopter: no joke. I'm trying to do some C# work for $DAYJOB, and when it works, it's awesome. when there's an error, I'm completely stumped. =-)
diakopter o 21:29
21:31 gbacon left
dmpk2k Do conditionals like if or unless return a value? 21:31
sorear TimToady: STD.pm6 parses <sym> as an identifier capture. Is this correct?
21:32 dj_goku joined
jnthn dmpk2k: The value of the executed block. 21:33
diakopter but on falseness of the conditional, I think TimToady said Nil
[it returns] Nil
jnthn Ah, yes, if there's no else clause.
diakopter rakudo: say 4 if (if 4 { say 5 }) 21:35
p6eval rakudo baebe1: OUTPUT«5␤4␤»
diakopter rakudo: say 4 if (if 0 { say 5 })
p6eval rakudo baebe1: ( no output )
diakopter rakudo: say(if 0 { say 5 })
p6eval rakudo baebe1: OUTPUT«Unable to parse postcircumfix:sym<( )>, couldn't find final ')' at line 11␤current instr.: 'perl6;Regex;Cursor;FAILGOAL' pc 1971 (ext/nqp-rx/src/stage0/Regex-s0.pir:1050)␤»
dmpk2k jnthn: thanks. That makes me happy to hear. Why are they called "if statements" then, and not "if expressions"?
diakopter rakudo: say (if 0 { say 5 })
p6eval rakudo baebe1: OUTPUT«0␤»
sorear Because they're not expressions.
dmpk2k But they return a value.
sorear rakudo: say if 0 { say 5 }
p6eval rakudo baebe1: OUTPUT«Confused at line 11, near "say if 0 {"␤current instr.: 'perl6;HLL;Grammar;panic' pc 501 (ext/nqp-rx/src/stage0/HLL-s0.pir:327)␤»
jnthn std: say if 0 { say 5 } 21:36
p6eval std 30793: OUTPUT«===SORRY!===␤Unexpected block in infix position (two terms in a row, or previous statement missing semicolon?) at /tmp/23vt08zCts line 1:␤------> say if 0 ⏏{ say 5 }␤ expecting any of:␤ bracketed infix␤ infix or meta-infix␤Other
..potential dif…
sorear expression ::= '(' statement ')'
statements return values
but aren't legal in expressions unless parenthetized
diakopter rakudo: say [if 0 { say 5 }]
p6eval rakudo baebe1: OUTPUT«0␤»
diakopter or otherwise bracketed
jnthn Essentially, it's a difference in where one can be parsed. 21:37
And circumfixes tend to parse a statementlist within them.
21:37 iblechbot left
dmpk2k Okay. Thanks for the clarification. 21:38
Are there things in Perl6 that cannot return a value? 21:39
diakopter what else would they return? :P 21:41
sorear Gather blocks?
Infinite loops?
Die statements?
jnthn sorear: Gather blocks return an iterator. That's a value. :-) 21:42
pmichaud s/iterator/list/ :-)
(it's a lazy list, but still a list.)
jnthn list *containing* an iterator ;-)
But yes
jnthn is too used to That Rakudo Leakage.
pmichaud return an Iterable :-)
21:43 SmokeMachine left
diakopter dmpk2k: I was punning; does anything never *return*? nothing that doesn't also exist in nearly every other language. (as sorear implied, I suspect) 21:44
bkeeler Heya folks, sorry I missed the #rs. I'll read over the logs...
jnthn o/ bkeeler 21:45
dmpk2k diakopter: I'm sorry, I'm having trouble parsing the second sentence. ^^;
diakopter oops, too many negnegnegatives 21:46
infinite loops & things that cause the entire language environment to exit. 21:47
(as sorear said)
dmpk2k Thanks. I just wanted to know if things like conditional statements can return a value since I'm quiet a fan of the idioms that makes possible. 21:49
sorear Also, comments. 21:50
[particle] hee
return unless # foo
sorear dmpk2k: perl6 has a ternary operator
[particle] comments returning values would be very silly 21:51
sounds like a great module
sorear [particle]: search.cpan.org/~chorny/Smart-Comme...omments.pm 21:52
dmpk2k sorear: ternary operators alone returning a value is a bit limiting. I think the more functional approach where (almost) everything is an expression is more interesting. 21:55
21:55 ruoso_ left
dmpk2k Or, in Perl6's case, wrap parens around it. 21:56
pugssvn r30794 | sorear++ | [viv] Implement <sym> rule 21:58
22:05 whiteknight joined, stephenlb left, stephenlb joined
sorear TimToady: What is _SYM for? 22:17
TimToady: What are tweaks? 22:18
TimToady _SYM is for matching <sym> when we don't know its literal value, iirc 22:19
tweaks are grammar mixins
sorear what is %impute used for?
TimToady don't recall, lemme look
sorear %impure
sorear is about to make the viv regex compiler much more similar to gimme5's 22:20
TimToady it used to be that way, but that's less extensible, I think
%impure says that this method has no PEEK, and will therefore terminate LTM 22:24
sorear The current viv regex compiler is entirely my creation 22:25
and it sucks
I have a VAST->RE_ast walker and a VAST->perl5 walker
this makes for DRY problems
I'm going to change it into a VAST->RE_ast walker and a RE_ast->perl5 walker 22:26
22:26 spinclad joined
sorear the latter component of which will be able to take code directly from gimme5 22:26
TimToady okay, that seems sensible 22:27
22:27 pmurias left
TimToady always sees "parser" when people say "compiler", and it's a hard habit to break :) 22:28
sorear current viv is powerful enough to translate "sigil" from STD.pm6 22:30
22:32 smash_ left
TimToady as in proto token sigil and all the multis? 22:33
sorear yes 22:35
TimToady: what is 'min' in RE_ast used for? 22:36
TimToady ultimately, left-recursion detection in LTM calculation
the only important values are really 0, 1, and 'many'
sorear this is something I do not understand at the moment 22:37
22:37 whiteknight left
sorear attempts to understand the LTM calculator 22:39
TimToady the important line from RE_ast.pm is $Cursor::fakepos++ if $self->{'min'};
sorear what does that do?
TimToady if the fake position doesn't advance on revisiting a token rule, we've got infinite recursion 22:40
sorear Is infinite recursion always an internal error?
What is the magic of :: at Cursor.pmc +1806? 22:41
TimToady used to be an error, but I think it just cuts off LTM at that spot these days, and the token in question just has a shorter LTM
so it can only match if longer things don't
sorear Under what circumstances is it legal for a left recursive grammar to be generated? 22:42
Is left recursion in perl 6 rules legal?
TimToady technically, yes, it's allowed to write an infinite loop :) 22:43
but the situation doesn't arise often in practice in Perl 6, at least
because nearly all such loops are mediated by EXPR, which requires a term
sorear Why doesn't cursor just die if you write a loop? 22:44
Perl automatically breaking while 1 { } would be suprising
TimToady there might be valid reasons for writing a left recursion that is stopped by a constraint further on in the pattern
sorear If the constraint is further on in the pattern, how can it break a loop at the beginning? 22:45
TimToady well, the state could have changed somehow in the recursion, I should have said 22:46
accidental recursions are typically fairly deep before they get back to the original rule
sorear Cursor uses pos(). Heh. 22:47
TimToady so it's not so much that the deeper rule looks farther, but with (potentially) more information
yes, Cursor goes back and forth between matching on the $::ORIG string and the @::ORIG array of ords 22:48
when checking @::ORIG we can just stick with the .pos from the Cursor, but any time we dip down in p5 regex of unknown length, we have to do the pos() dance 22:49
hopefully the new pos is not far from the old one, so p5's pos cache doesn't have to scan utf-8 very far 22:50
diakopter was someone referring to the p-o-s acronym? 22:54
sorear Cursor uses tries to optimize the literal prefix case of the LTM hierarchy? 22:55
TimToady it uses a trie match for literals, yes
I talked about that earlier today, maybe you missed it
sorear And at the ends of the tries are big lists of regexes that are used for LTM matching
Maybe, or maybe I just didn't understand it 22:56
TimToady there's only regexes at the end if there was a common prefix of the same length
usually there's 0 or 1 regexes at the end of each chain 22:57
my point earlier is that we could evolve the current trie into a DFA-ish graph if we could manage pointers into the middles of those regexes that currently hang off of various points in the trie 22:58
22:58 cdarroch left
sorear which is to say, if we had our own regex compiler instead of relying on p5 22:58
diakopter giggles childishly
TimToady yes, but just having *a* regex compiler isn't good enough, since there are lot of unsuitable ones out there already :) 22:59
22:59 Psyche^ joined
sorear if I look in a lexer, say lex/STD/P6/term, why do I see 231 regexes? 23:00
TimToady we need an approach more like egrep uses, where it's managing a set of active states based on the current input
sorear is the "usually 0 or 1" thing done after storage?
TimToady: Direct NFA execution?
sorear is not familiar with the egrep guts
TimToady I don't know what "direct" means 23:01
sorear not translating to DFA first
TimToady yes, more of a parallel NFA is wanted here, I think, but with good caching characteristics based on the tokenization problem space 23:02
since as you traverse your parallel NFA, you're continuing to build up the trie as a set of cached NFA decisions so you don't have to do set theory again
sorear Cursor.pmc +1806 23:03
justatheory does set TimToady
sorear what's so magical about :: ?
TimToady you mean (?#::) ?
23:03 tlb left
sorear yes 23:03
the comment makes me think :: matters 23:04
23:04 Psyche^ is now known as Patterner
sorear or do I need to read perlre again 23:04
TimToady that's a p5 comment that I'm using to mark a transition from declarative to procedural
the :: is mnemonic for the backtracking control
but any side-effect-ish pattern will add the marker 23:05
and then any rule that uses this as a subrule knows that it should not add any more "tokenstuff" to it
on the end, that is
it's fine to add prefix
in lex/STD/P6/term, it's a very "square" or relational layout 23:06
with no attempt to gather common prefixs or literal prefixes
that's calculated after it's slurped in and turned into trie
23:06 snarkyboojum joined
TimToady but the file needs to be human readable for debugging purposes 23:06
the Storable form of cache has this precalculated, however 23:07
to the extent we do the trie
once we actually start caching partial parallel NFA results in the trie, it might be good to run the parser over an amount of typical program to extend the tries with common names, such as come from CORE 23:09
and cache those in Storable as well
currently only the literal prefixes are cached in the Storable trie 23:10
sorear perl -MRE_ast -MStorable -MYAML -e 'print Dump retrieve("lex/STD/P6/term.store")' 23:11
eternaleye ly
gah, wrong window
sorear Wow. 23:12
This is completely clear now.
TimToady the real performance killers are the ~~ regexes at the beginning of the pattern; those need to be distributed into the trie as we learn how each of those patterns responds to each individual character 23:13
but currently it just reruns them on each <term> match 23:14
and sorts them by LTM ruls
sorts their results by LTM rules, rather
in this case, the yaml elucidates the structure rather nicely, for a change :) 23:15
but you can see why I haven't touched the LTM in a while, because it's in a metastable state of sorts 23:16
and it does the semantics more or less correctly, if slowly
but there's room for improvement, certainly
and, actually, in terms of the PEG tiebreaker rules described at the beginning of S05, it really only does 1, 2, and 4. 23:17
I'm hoping to push some of these tiebreaking determinations back into the lex/STD/P6/term files as additional columns 23:19
the length of the literal prefix can be known in advance, and some of the orderings can reflected in the order of the file in question
if not in a specific collation datum 23:20
sorear um, should . generate 0x10FFFE trie nodes? 23:22
TimToady example? 23:23
sorear it's a regex 23:24
you feed it every possible character
then....?
or is this what you were talking about when you said lazy dfa generation
TimToady okay, on something like .*? in a nibbler, we'll have to be careful to cut off caching 23:25
but yes, we only build that subset of DFA reflecting the states for characters that have actually been seen 23:26
23:26 alester joined
TimToady in egrep, this cached set of states is limited to, oh, 10,000 or so, and older unused states are dropped off the end, and reconstructed if needed 23:26
sorear I think the solution for .* is to limit the size of the ... yeah 23:27
TimToady I think of the trie as a fixed set of states for anchored patterns
and then longer ones cut loose and do something more like egrep after N characters
or at that point we revert to the current ~~ algorithm, as a stopgap 23:28
it's also a matter for research how much of the trie/dfa info can be shared across different languages, especially if everything is linked listed and there are common tails 23:29
at the moment, the sharing is all or nothing with a parent grammar's lexer for a particular category 23:30
23:30 whiteknight joined
TimToady again, that's determined within the first 20 or so lines of _AUTOLEXgen, looking at the superlexer for this category (but only for anonymous languages, since they don't get Storabled) 23:31
otherwise we'd need a way to name all anonymous languages universally to avoid collisions, and that's...hard 23:32
I suppose practically we'd just name them all with a large random hash of some sort
but you still have the garbage collection problem
or lex/ would fill up your disk 23:33
eventually we'll also have to solve the global lex/ vs per-user lex/ issue
sorear how much of a win is sending lex/ to disk? 23:34
TimToady but I've specifically been shying away from anything resembling productizing
it certainly helps for the 'make snaptest' on the t/ files, which have a lot of commonality
even STD.pm6 compiles faster the second time through 23:35
s/compiles/parses/ # grr
sorear $text =~ s/([[:alpha:]])/'[' . $1 . chr(ord($1)^32) . ']'/eg if $self->{i}; 23:36
erm...
TimToady but it's not clear how much it helps once STD is compiled with most of the common patterns
yes, there are ASCII assumptions in various spots still 23:37
23:37 ive joined
eternaleye (re dyvar) my @*sparse = ( 1, 2, Nil, 4) # I can has Holey Dyvar? 23:39
TimToady um, not that way 23:41
Nil mean ()
and () in list context disappears 23:42
23:43 stephenlb left, whiteknight left
TimToady sorear: I think that code is a bit of a fossil from when I was trying to feed these patterns to other regex engines that didn't support either utf8 or case insensitivity 23:45
but certainly if I know I'm targeting p5 patterns I can just use uc/lc, and maybe ucfirst for titlecase 23:46
or, actually, it can just use p5's embedded (?i) ability 23:47
but I was targeting TRE or some such a the time, and it got very upset about those sorts of thigns 23:48
I gave up on TRE when I realized it was allocating many megabytes on the stack for *every* token lookup 23:49
23:49 ive left
TimToady it doesn't scale well in # of parallel patterns, turns out 23:49
sorear TRE? 23:51
let me explain an incomplete thought of mine
TimToady Tagged Regular Expressions
libtre, etc 23:52
sorear when you compile a NFA to a DFA, you're forcing the regex to express its entire state in some complex encoding of one word
TimToady yes, but that's an abstract idea
sorear a NFA expresses the state as a bit array, but needs to loop over it for every character
TimToady the word can be spread out in various spots, physically
sorear I'd like to see some middle ground 23:53
where the regex's state space is multi-dimensional, so less storage is required, but there's less of a looping problem
TimToady the trie cache can get the subset of the state space for the current match
so in essence your dimensional lookup is keyed on what we've seen so far 23:54
sorear I don't like the cache.
It feels like an awkward crutch.
TimToady it precomputes some probabilistically likely state sets 23:55
and it could conceivably be quite fast, for something that feels "lame" :) 23:56
especially as the redundancy in your program goes up 23:57
it starts working more like a compression algorithm
but hopefully the compression is on the state space more than the output :) 23:58
sorear My goal is to output something that looks vaguely like a hand-written recursive descent parser
TimToady but anyway, if you have something that feels less lame, by all means it's worth a shot
sorear Not really.
Just expressing my concerns in case you had insights.
TimToady well, recursive descent doesn't guarantee LTM easily 23:59

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!