perl6-projects.org/ | nopaste: sial.org/pbot/perl6 | evalbot: 'perl6: say 3;' | irclog: irc.pugscode.org/ | UTF-8 is your friend!
Set by moritz_ on 4 May 2009.
00:05 frew joined 00:15 eMaX left 00:17 justatheory left 00:30 DemoFreak left, bacek__ joined 00:33 ElectricHeavyLan joined 00:53 drbean_ joined 00:55 dalek left, amoc left, dalek joined
Tene pmichaud: pushed HLL changes. You can change RAKUDO_HLL after review. 00:55
00:57 PhatEddy left 01:02 d4l3k_ joined 01:04 aindilis left, bacek left, Infinoid left, bacek__ left, LylePerl left, rhr left, baest left, Khisanth left, ascent_ left, ingy left, cj left, nbrown left, H1N1[A] left, clkao left, kbrandt left, estrabd_ left, frew|work left, Helios left, gfldex left, stepnem left, raiph_ left, raiph left, japhb left, Aisling left, r0bby left, frew left, skids left, hercynium left, NoirSoldats left, PacoLinux left, sri_kraih_ left, hcchien left, rewt left, dukeleto left, allbery_b left, avar left, frodwith left, pdc303 left, silug left, cavelife^ left, simcop2387 left, REPLeffect left, Patterner left, araujo left, phenny left, drbean_ left, jan___ left, szabgab_ left, c9s left, cxreg left, revdiablo left, yahooooo left, pasteling left, rafl left, kcwu left, awwaiid left, spx2 left, lisppaste3 left, zev left, moritz_ left, gbacon left, arnsholt left, StephenPollei left, vx64z left, cdarroch left, jferrero left, drbean left, agentzh left, cosimo left, donaldh left, [particle]1 left, p6eval left, sbp left, LadyLuna2y left, cognominal left, elmex left, plu left, xinming left, mj41 left, kolibrie left, omega left, pjcj left, tarbo2 left, pugs_svn left, TimToady left, Trey left, Eevee left, PerlJam left, Caelum left, sunnavy left, dalek left, payload1 left, alester_away left, ElectricHeavyLan left, Kisu left, mikehh left, lambdabot left, orafu left, [particle]- left, ashizawa left, kirillm left, Gothmog_ left, scrottie left, felipe left, ilbot2 left, eiro left, hexmode left, mattp left 01:09 jnthn joined, Maddingue joined, buu joined, kirillm_ joined, AzureStone_ joined, dalek joined, drbean_ joined, ElectricHeavyLan joined, bacek__ joined, frew joined, nbrown joined, vx64z joined, donaldh joined, skids joined, payload1 joined, raiph_ joined, hercynium joined, LadyLuna2y joined, NoirSoldats joined, PacoLinux joined, alester_away joined, Kisu joined, jan___ joined, cognominal joined, REPLeffect joined, cdarroch joined, Patterner joined, mikehh joined, H1N1[A] joined, ashizawa joined, clkao joined, jferrero joined, LylePerl joined, sri_kraih_ joined, hcchien joined, lambdabot joined, orafu joined, [particle]1 joined, szabgab_ joined, aindilis joined, bacek joined, raiph joined, kbrandt joined, [particle]- joined, irc.freenode.net sets mode: +o jnthn, araujo joined, gbacon joined, drbean joined, xinming joined, gfldex joined, ingy joined, ascent_ joined, rhr joined, Infinoid joined, Aisling joined, cj joined, Khisanth joined, r0bby joined, baest joined, Helios joined, estrabd_ joined, frew|work joined, stepnem joined, japhb joined, allbery_b joined, avar joined, rewt joined, silug joined, dukeleto joined, pdc303 joined, cavelife^ joined, frodwith joined, simcop2387 joined, agentzh joined, c9s joined, kidd joined, AzureStone joined, Grrrr joined, antiphase joined, smtms joined, integral joined, broquaint joined, pmichaud joined, s1n joined, Woody2143 joined, cotto joined, hatseflats joined, Tene joined, mtve joined, literal joined, diakopter joined, phenny joined, cxreg joined, pasteling joined, kcwu joined, revdiablo joined, yahooooo joined, irc.freenode.net sets mode: +oo pmichaud Tene, awwaiid joined, rafl joined, elmex joined, cosimo joined, p6eval joined, kirillm joined, StephenPollei joined, arnsholt joined, mj41 joined, Gothmog_ joined, kolibrie joined, omega joined, pjcj joined, tarbo2 joined, pugs_svn joined, TimToady joined, sbp joined, Trey joined, scrottie joined, mattp joined, felipe joined, ilbot2 joined, eiro joined, plu joined, hexmode joined, irc.freenode.net sets mode: +o TimToady 01:11 moritz_ joined, spx2 joined, jiing joined, cls_bsd joined, jrockway joined, lisppaste3 joined, zev joined, irc.freenode.net sets mode: +o moritz_, jferrero left, dalek left, drbean left, d4l3k_ is now known as dalek 01:16 Caelum joined, meehav joined, PerlJam joined 01:17 kirillm left 01:19 IRSeekBot joined 01:20 AzureStone left 01:21 meteorjay joined 01:26 amoc joined 01:27 Caelum left, ray joined 01:31 Caelum joined 01:39 alc joined 01:40 dukeleto left 01:41 amoc left, amoc joined
dalek kudo: bf281cf | pmichaud++ | src/parser/ (2 files):
Add ability to recognize operator subs. The ability to define
 02:15
02:15 dukeleto joined 02:31 autin joined 02:51 dbezborodov joined 02:54 dbezborodov left 02:59 sri_kraih_ left, sri_kraih joined 03:08 nihiliad joined 03:20 donaldh left, donaldh joined 03:31 kst joined 03:33 cdarroch left 03:49 orafu left, orafu joined 03:54 sri_kraih left 04:08 dukeleto left 04:10 skids left 04:20 hercynium left 04:30 eternaleye joined 04:34 Jedai joined 04:37 sri_kraih joined 04:38 eternaleye left, eternaleye joined
s1n moritz_: s1n.dyndns.org/index.php/2009/05/14...confusion/ 04:49
05:01 sparc joined 05:18 frew left 05:19 vx64z left
Tene Ā»Ć¶Ā« | perl6-projects.org/ | nopaste: sial.org/pbot/perl6 | evalbot: 'perl6: say 3;' | irclog: irc.pugscode.org/ | UTF-8 is your friend! 05:31
Tene There, now we have Camelia to watch over us. 05:31
05:40 nihiliad left 05:53 ejs joined 06:00 raiph_ left 06:03 amoc left 06:05 amoc joined 06:09 ejs left 06:18 ElectricHeavyLan left, LadyLunacy joined
patmat Welcome LadyLunacy!! 06:22
06:27 LadyLuna2y left, ElectricHeavyLan joined, ElectricHeavyLan left
moritz_ good morning 06:29
06:31 mberends joined
mberends whoz op ? 06:32
patmat hey moritz_!
moritz_ OH HAI P6BUGS ;-)
06:34 ChanServ sets mode: +o Tene
Tene I'm op, what's up? 06:34
mberends ok Tene s op. how do u make your own chanle? 06:36
06:36 ab5tract joined
Tene I saw this before from masak I think... I think I missed something. 06:36
mberends im being polite, not using caplock. so how do u? 06:37
moritz_ ;-)
Tene: just a funny incident with somebody not used to IRC 06:38
Tene ah
OK
mberends Tene: but srsly, how is the blogserver doing? 06:39
moritz_ Tene: irclog.perlgeek.de/perl6/2009-05-12#i_1138627 if you're interested ;-) 06:40
Tene mberends: I haven't touched it since i put it up.
blogs.gurulabs.com/stephen/2009/05/...d-run.html
pleasedieinafire.net:2080/
mberends Tene: tried, that's why I was curious. Hmm, my post is gone :( 06:41
Tene I'd really enjoy feedback. :) 06:42
mberends: oh, I put it up earlier from a different server.
yeah, I see your post here in /tmp on my laptop. :)
mberends I was collaborating with masak on Yarn, any chance of merging the two sources? 06:43
Tene probably
mberends :) I'd like to add registration, authentication etc.
Tene Well, probably not. It wasn't really meant to be a serious blog. Mostly just an experiment.
06:44 DemoFreak joined
Tene I'll probably use it to test out features like that, and different ways of integrating them. 06:44
so, I don't have any real feelings of trying to work to make something good, yet. 06:46
mberends fine. On another topic, I tried mimicking a Net::SMTP and hd socket IO problems. Could you help troubleshoot the parrot bits? 06:47
*had # gah
Tene yes 06:48
mberends great, I'll reduce the failing code to a tiny test case later 06:49
Tene mberends: if you email it to me, I can take a look at it during the day.
I'll be offline during work 06:50
mberends thanks, in a few hours
Tene that will be great.
and it doesn't need to be *too* minimal, btw. 06:51
if you can't get it minimal, just send what you can.
mberends ok, but I have @family.life() & to multithread 06:53
pugs_svn r26824 | pmichaud++ | [t/spec]: Rakudo fudge a test we now regress.
Tene you've got 7 hours before I start waking up. :)
mberends that's doable 06:54
06:54 amoc left, ejs joined
mberends Tene: it was already all at gitorious.org/net-smtp/mainline/trees/master 06:58
Tene thx 06:59
07:01 dukeleto joined 07:03 ejs1 joined 07:10 mikehh_ joined 07:11 ejs left
Matt-W Morning 07:14
07:14 pnu left
mberends morning; afk & 07:15
07:16 iblechbot joined
pmichaud yay, I have operators in rakudo -- just need to run spectests and then push. :-) 07:17
Tene pmichaud++ 07:18
pmichaud: I will see about getting pugs' Set.pm running on rakudo
pmichaud we're still limited in that the tighter/equiv/looser traits aren't implemented yet. 07:19
I'll work on that tomorrow.
07:20 donaldh left, ElectricHeavyLan joined, donaldh joined
Matt-W pmichaud++ 07:24
07:25 mikehh left
dalek kudo: 8349d75 | pmichaud++ | src/parser/ (2 files):
Add ability to define operator subs.
 07:56
kudo: ad9e008 | pmichaud++ | (2 files):
Add ability to declare infix: subs.
kudo: 9dcaf47 | pmichaud++ | src/parser/actions.pm:
Update parse table both immediately and in saved code.
pmichaud rakudo: sub postfix:<!>($n) { [*] 1..$n }; say 7!; 07:58
07:58 jferrero joined
p6eval rakudo bf281c: OUTPUTĀ«Malformed routine definition at line 1, near "postfix:<!"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā» 07:58
pmichaud rakudo: say &infix:<+>(3,4);
p6eval rakudo bf281c: OUTPUTĀ«7ā¤Ā»
moritz_ rakudo: sub infix:<foo>($a, $b) { say $a + $b }; 3 foo 4 07:59
p6eval rakudo bf281c: OUTPUTĀ«Malformed routine definition at line 1, near "infix:<foo"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
pmichaud hasn't updated yet
moritz_ aye
anyway, pmichaud++ 08:00
pmichaud rakudo: say 'ready yet?'
p6eval rakudo bf281c: OUTPUTĀ«ready yet?ā¤Ā»
pmichaud (not yet.)
rakudo: sub postfix:<!>($n) { [*] 1..$n }; say 7!; 08:01
p6eval rakudo bf281c: OUTPUTĀ«Malformed routine definition at line 1, near "postfix:<!"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
pmichaud rakudo: say 'ready yet?'
p6eval rakudo bf281c: OUTPUTĀ«ready yet?ā¤Ā»
pmichaud (sigh)
moritz_ pmichaud: I'm starting the rebuild now
pmichaud Impatience is a virtue. 08:02
Doesn't the rebuild normally start by itself at the top of the hour...?
08:03 eMaX joined
pmichaud rakudo: sub postfix:<!>($n) { [*] 1..$n }; say 7!; 08:03
p6eval rakudo bf281c: OUTPUTĀ«./perl6: error while loading shared libraries: libparrot.so.1.1.0: cannot open shared object file: No such file or directoryā¤Ā»
moritz_ pmichaud: it needs to recompile parrot, takes a while on a single CPU machine
pmichaud goes back to playing solitaire. :-) 08:04
08:08 ab5tract left
pmichaud rakudo: sub postfix:<!>($n) { [*] 1..$n }; say 7!; 08:09
p6eval rakudo bf281c: OUTPUTĀ«sh: ./perl6: No such file or directoryā¤Ā»
pmichaud oh well. I need sleep, so will try it out tomorrow.
moritz_ it compiles gen_settings.pm right now
good night
rakudo: sub postfix:<!>($n) { [*] 1..$n }; say 7!; 08:11
p6eval rakudo 9dcaf4: OUTPUTĀ«5040ā¤Ā»
moritz_ pmichaud+++
08:17 autin left 08:18 autin joined, tombom joined
moritz_ rakudo: my $a; &infix:<=>($a, 4); say $a 08:19
p6eval rakudo 9dcaf4: OUTPUTĀ«4ā¤Ā» 08:20
moritz_ rakudo: my Str $a; &infix:<=>($a, 4); say $a
p6eval rakudo 9dcaf4: OUTPUTĀ«Type mismatch in assignment; expected something matching type Str but got something of type Int()ā¤current instr.: 'die' pc 16712 (src/builtins/control.pir:225)ā¤Ā»
moritz_ rakudo: my Int $a; &infix:<=>($a, 4); say $a
p6eval rakudo 9dcaf4: OUTPUTĀ«4ā¤Ā»
08:23 pnu joined
pugs_svn r26825 | moritz++ | [t/spec] unfudge assign.t, and extend a bit (you know, 318 assignment test just were not enough) 08:27
moritz_ pmichaud: the form with french quotes (infix:Ā«somethingĀ») also needs to be done ;-) 08:30
Matt-W rakudo: sub infix:<pointless>($a, $b) { $b ~ $a }; say "world" pointless "hello "
p6eval rakudo 9dcaf4: OUTPUTĀ«hello worldā¤Ā»
Matt-W that is so cool 08:31
pmichaud++
08:36 masak joined
masak I'm on my way to work. just stopping by to say pmichaud++ pmichaud++ pmichaud++ 08:36
08:36 masak left
pugs_svn r26826 | moritz++ | [t/spec] fudge sub.t for rakudo 08:36
dalek kudo: 1af7e29 | moritz++ | t/spectest.data:
we pass an operator definition/overloading test
 08:41
08:50 eMaX left
Matt-W @tell masak Clearly there is a lot of love for pmichaud++ at the moment 08:50
lambdabot Consider it noted.
08:52 baest left, baest joined
pugs_svn r26827 | moritz++ | [t/spec] unfudge proto.t for rakudo 08:53
08:54 meppl joined, kane_ joined 08:55 kane___ joined 08:57 wollmers joined 09:11 kane__ left, kane_ left
wollmers S01: 'Perl is written in Unicode.' 09:17
Yesterday IRC log: Text files are in Unicode. 09:18
moritz_ we certainly have a Unicode bias in here ;-)
wollmers There is no save way to autodetect UTF-8 versus UTF-16 for text files, for perl source - maybe, because source contains chars in the ASCII rabge. 09:20
*range
moritz_ wollmers: the presence of a BOM makes it rather safe 09:21
wollmers: and the spec is quite clear that if it's not UTF-16, UTF-8 is the fallback
i'm not 100% happy with it, but it seems to mostly work
lunch &
wollmers I cannot find it in the specs ... 09:22
Matt-W unicode is always a nasty mess it seems
wollmers Matt-W: Encoding is the most simple thing, but mostly underestimated as a main source of problems. 09:24
09:27 bacek__ left
jnthn H H 09:29
pmichaud++ # yay! 09:34
mikehh_ I've been getting failures in spectest for the last few days - both on Kubuntu Jaunty Amd64 and Ubuntu Jaunty i386 09:41
t/spec/S32-io/IO-Socket-INET.t fails test 2-3 - but when I run ./perl6 t/spec/S32-io/IO-Socket-INET.t it passes 09:42
t/spec/S09-subscript_slice/slice.rakudo passes all the tests says #Fudged and then segfaults 09:43
wollmers mikehh_: cd rakudo; make t/spec/S32-io/IO-Socket-INET.t 09:45
mikehh_: runs here on Debian/Sid
mikehh_ wolmers: that passes 09:47
09:47 autin left
mikehh_ it just seems that when I run make spectest the test fails - maybe a threading problem? 09:48
09:48 cognominal left 09:50 autin joined
mikehh_ it has consistently failed on my last few builds both on Amd64 and i386 09:51
sorry that should have been wollmers 09:56
09:56 Kyosuke_Kiryu joined 09:57 mikehh_ is now known as mikehh
jnthn mikehh: I see the slice.t one too - I looked into it a bit, but it ain't gonna be easy to fix, I fear. :-( 10:00
10:01 Kyosuke_Kiryu left
wollmers mikehh: IO-Socket-INET.t does ugly things (look into source); the reasons for the difference can be many ... 10:05
10:09 TRM_PEMROGRAMAN_ joined 10:10 TRM_PEMROGRAMAN_ left
mberends mikehh: ugliness is my bad. I'd like to improve it... 10:14
wollmers: the problem I had writing it, was determining an available port number on the fly. 10:15
mikehh jnthn: as far as I can see it in the exit code 10:18
all the other tests pass 10:19
jnthn mikehh: It doesn't make it through all of them for me, crashes along the way. 10:20
mikehh: I hunted it down a bit and it looks like a GC bug.
In which case it's liable to manifest itself in different ways in different places.
mikehh jnthn: thats what I think too 10:21
It seems to be in the gc cleanup
jnthn I thought so far that was mostly re-naming things, but maybe not. 10:23
mikehh anyway I'm trying to debug now - will let you know if I find anything
jnthn mikehh: Thanks. I did also write a ticket on this that you may wish to look up. 10:25
With some initial analysis.
mikehh: It was this one: rt.perl.org/rt3/Ticket/Display.html?id=65396 10:29
10:33 cognominal joined 10:37 iblechbot left 10:39 pastorn joined
mikehh jnthn: when I run as per the ticket I get maximum recursion depth exceeded then similar output 10:40
pugs_svn r26828 | moritz++ | [t/spec] unfudge postfix:<!> test for rakudo 10:42
mikehh with -G that is 10:43
jnthn: when I run the test -> ../parrot.t/parrot -G perl6.pbc t/spec/S09-subscript_slice/slice.rakudo the same stuff happens 10:49
jnthn: but it fails after ok 7 - slice from array slice, part 2, maximum recursion depth exceeded 10:51
jnthn: without the -G it gets past the say "# FUDGED!" then Segmentation fault 10:53
10:55 payload1 left 10:57 zamolxes joined
jnthn mikehh: I know. 10:57
10:58 amoc joined
jnthn mikehh: The ticket gives a smaller example too. 10:58
mikehh jnthn: ok and now for something completely different - or something to that effect :-} 10:59
10:59 masak joined
masak oh hai 11:00
lambdabot masak: You have 1 new message. '/msg lambdabot @messages' to read it.
masak whoz user-defined op?
@messages
lambdabot Matt-W said 2h 9m 27s ago: Clearly there is a lot of love for pmichaud++ at the moment
masak pmichaud++
moritz_ rakudo: sub postfix:<!>($x) { [*] 1..$x }; say 5!
p6eval rakudo 1af7e2: OUTPUTĀ«120ā¤Ā»
moritz_ rakudo: { sub postfix:<!>($x) { [*] 1..$x }; }; say 5! 11:01
p6eval rakudo 1af7e2: OUTPUTĀ«120ā¤Ā»
masak I have a few edge cases I'd like to discuss.
rakudo: sub infix:<,> { 42 }; say 5, 5
p6eval rakudo 1af7e2: OUTPUTĀ«55ā¤Ā»
masak I think this one should at least warn of overriding.
(secondly, it doesn't work.)
moritz_ it's not lexically scoped yet 11:02
Matt-W it just... doesn't do anything noticeable
masak moritz_: that is neither here nor there.
jnthn mikehh: I tried to debug it a bit but it's kinda hard. :-|
moritz_ masak: a warning isn't that easy...
Matt-W lunch &
moritz_ masak: because the setting will define a 'proto infix:<,>'
masak: so that any 'sub infix:<,>' is automatically a multi sub
masak moritz_: oh.
moritz_ masak: and adding a new multi never warns
11:02 zamolxes left
masak moritz_: well, then it shouldn't warn. 11:03
moritz_: but I bet it shouldn't not work either.
rakudo: sub infix:<#> { 42 }; say 5 # 5
p6eval rakudo 1af7e2: OUTPUTĀ«5ā¤Ā»
moritz_ masak: it should do an ordinary multi dispatch
masak along the same lines...
moritz_: aye.
moritz_: and probably ambig-fail.
moritz_ rakudo: sub infix:<,>(Int $x where 1, Int $y where 1) { 42 }; say 1, 1
p6eval rakudo 1af7e2: OUTPUTĀ«11ā¤Ā»
masak rakudo: sub infix:<+> { 42 }; say 5 + 5 # so should this 11:04
p6eval rakudo 1af7e2: OUTPUTĀ«42ā¤Ā»
moritz_ now *that* should be tighter, and give 42
masak submits rakudobug
11:04 zamolxes joined
moritz_ masak: a sub without a signature has slurpy @_ 11:04
masak oh, right. 11:05
moritz_ masak: so that it's not very specific, and will practically never match in the multi dispatch
masak rakudo: sub infix:<+>() { 42 }; say 5 + 5 # so should this
p6eval rakudo 1af7e2: OUTPUTĀ«42ā¤Ā»
moritz_ that should be 10, no?
masak I think so.
rakudo: sub infix:<+>($a, $b) { 42 }; say 5 + 5
jnthn I think the issues with overloading existing rather than defining new, may be an interaction between the existing ones using Parrot's MMD.
p6eval rakudo 1af7e2: OUTPUTĀ«42ā¤Ā»
jnthn And the new ones using Rakudo's. 11:06
masak aha.
jnthn: is there a plan to do something about that?
moritz_ rakudo: multi sub infix:<+>() { 42 }; say 5 + 5
p6eval rakudo 1af7e2: OUTPUTĀ«Null PMC access in find_method()ā¤current instr.: '_block18' pc 78 (EVAL_20:54)ā¤Ā»
moritz_ masak: ^^
masak bug
jnthn masak: Oh, actually, that bug is exactly coming up because of this issue. 11:07
masak: The other issue is that since we're not really defining a proto yet, your sub is just supplanting the existing operator.
masak: And yes, plan is to define all operators in Perl 6. (With inline PIR, no doubt.)
When that happens, the special-cased auto-threading froms can go away too. :-) 11:08
masak rakudo: sub postfix:<km>($n) { $n ~ ' kilometers' }; say 42\km 11:16
p6eval rakudo 1af7e2: OUTPUTĀ«Statement not terminated properly at line 1, near "\\km"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
masak rakudo: sub postfix:<km>($n) { $n ~ ' kilometers' }; say 42km 11:17
p6eval rakudo 1af7e2: OUTPUTĀ«42 kilometersā¤Ā»
masak \o/
but the first one should work too, right? 11:18
jnthn masak: Yes 11:19
rakudo: sub postfix:<km>($n) { $n ~ ' kilometers' }; say 42 \km
p6eval rakudo 1af7e2: OUTPUTĀ«Statement not terminated properly at line 1, near "\\km"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
masak submits a slam dunk
jnthn oh
rakudo: sub postfix:<km>($n) { $n ~ ' kilometers' }; say 42\ km
I meant that.
p6eval rakudo 1af7e2: OUTPUTĀ«42 kilometersā¤Ā»
masak oh, ah.
11:20 donaldh left
masak well, I un-submit then. 11:20
jnthn My previous one should not work.
masak no.
and the case of \ without ensuing ws is already in RT.
11:21 donaldh joined
jnthn aye 11:22
I think I know where to fix that too.
Ouch, LolDispatch abuses Rakudo guts.
moritz_ aye 11:23
jnthn That's gonna break in not too long. Happily, first I'm going to make the Right Way work. :-)
masak LolDispatch is cool. 11:25
but I was really surprised that what it does works already.
turns out it cheats. :P
jnthn masak: I'm going to write an example very much like loldispatch and make it work the proper way. 11:26
masak yay!
wollmers rakudo: sub infix:<cut> ($a, $b) { "Prolog: $a,!,$b." }; say 'a' cut 'b'; 11:30
p6eval rakudo 1af7e2: OUTPUTĀ«Prolog: a,!,b.ā¤Ā»
masak senses a new era in Rakudo hacking is about to begin... 11:31
jnthn needs to blog about some of the stuff he's been doing on his not-yet-approved grant at some point... 11:33
11:36 ejs1 left 11:39 ejs1 joined
jnthn woo, got cut-down LolDispatch-y thing working the Official Way. 11:50
With one annoying quirk that I can now fix hopefully...
masak ok, here's what I'd like: a Perl 6 project on github or wherever which took a Perl 6 source file as input, and gave a parse tree as output. it might be built on either STD or Rakudo, doesn't matter. the parse tree might be made of Match objects or just plain hashes, doesn't matter. 11:55
jnthn gist.github.com/111625 11:56
masak my main requirement is mainly that I don't have to do the dirty work of joining the piping together. :)
moritz_ once Perl6::Grammar returns a proper match object, that's a piece of cake.
masak indeed.
that's a legit solution too. 11:57
jnthn @tell Tene Here is the more proper way to do LolDispatch: gist.github.com/111625 - am comitting the changes to enable this Real Soon Now. Will not break the previous hacky way...yet.
lambdabot Consider it noted.
moritz_ jnthn: I don't think you need to worry about loldispatch
jnthn: if I understood Tene correctly it's meant as a throw-away thingy for now anyway
masak jnthn: is 'handler' in the signature a type? 11:58
jnthn masak: Yes
masak oh.
masak sees now
it's defined at the top.
jnthn See role handler { }
Yeah
But in Rakudo we're I think more liberal on the parse at the moment than STD
std: sub foo() is nonexistantthing { } 11:59
p6eval std 26828: OUTPUTĀ«ok 00:02 35mā¤Ā»
jnthn oh, maybe not. :-)
masak um. that feels like a STDbug.
moritz_ IMHO that should be an error
masak TimToady: is it?
12:00 iblechbot joined
jnthn Yeah, I was expecting STD to complain there, but I suspect it may more be out of pragmatism/incompleteness that it doesn't do so yet. 12:00
masak likes how the power of backlogging allows us to send messages through time
jnthn Or maybe it's to give the compiler a shot at handling special ones.
moritz_ oh wait
masak jnthn: maybe it's so that the subs can be defined before the handlers.
moritz_ it should be a compile time error, but does it have to be a parse time error?
masak moritz_: maybe not. 12:01
moritz_ I mean is thing after the 'is' always a type name?
jnthn moritz_: I'm not 100% clear on that.
S14 says
pmichaud good morning, #perl6
jnthn To define a trait handler for an "is xxx" trait, define one or
more multisubs into a property role like this:
Suggesting that xxx is a type name.
masak , for one welcomes our op overlord pmichaud
jnthn pmichaud: oh hai
12:01 mizioumt joined
masak s/s// 12:01
jnthn spectests his intial patch to get us started on the road to traits 12:03
12:05 payload joined 12:09 orafu left 12:10 orafu joined
pmichaud masak: in "say 5,5" the infix:<,> is never called. 12:12
say (5,5) would call infix:<,> however. 12:13
masak pmichaud: is that specced?
moritz_ and, more interestingly, why? ;-)
pmichaud I don't know if it's specced that way. But "say 5,5" passes two arguments to the say() function. 12:15
i.e., it never forms a list (it does form a capture)
moritz_ that's... weird
pmichaud "say (5,5)" sends one argument to the say() function; that one argument is the result of infix:<,> on two values.
jnthn doesn't find that surprising 12:16
It's like you'd not expect :($x, $y) to call it 'cus it's a signature.
I guess we know when we're expecting to parse a capture and when not. 12:17
Matt-W would say(5,5) pass two arguments?
jnthn yes
Matt-W good
masak jnthn: but isn't the ',' within the signature also an operator?
12:17 orafu left
jnthn masak: No. 12:17
It's just a parameter seperator
masak ok.
fair enough. 12:18
jnthn (As are ;; and :)
masak aye.
jnthn See param_sep in STD.pm
masak oh, oh, how does the thing with trailing commas in lists really work, by the way? 12:19
they can't be pure infix if they're trailing.
pmichaud PGE handles it by having a notion of a "null term" 12:21
I'm not sure how STD.pm handles it yet
moritz_ which implies they're probably a special syntactic form (in STD.pm)
masak rakudo: sub foo($a,) {}; say &foo.arity 12:23
p6eval rakudo 1af7e2: OUTPUTĀ«1ā¤Ā»
masak seems to work in sigs as well.
12:23 orafu joined
masak meeting & 12:24
jnthn Probably has to for method foo($inv:) {} to work.
Matt-W I'd expect so 12:26
unless : was excessively special-cased
12:36 elmex_ joined, elmex left 12:37 elmex_ is now known as elmex 12:42 payload left 12:47 kirillm_ is now known as kirillm 12:48 payload joined
jnthn -> slovak class, bbiab 12:49
12:53 lichtkind joined 13:01 ejs2 joined 13:11 ejs1 left 13:15 skids joined 13:18 iblechbot left 13:19 jferrero left 13:23 ruoso joined 13:27 mizioumt1 joined 13:28 ejs2 left
ruoso Hello! 13:31
masak ruoso: oh hai.
ruoso: I read through your Faz source code yesterday, before blogging. 13:32
ruoso masak, how does the trait handler works in LoLDispatch?...
masak, I've just read your blog post...
masak ruoso: very well, thank you? :)
dalek kudo: 705cb2b | pmichaud++ | src/parser/grammar.pg:
Fix use of ōæ½xAB ōæ½xBB and << >> in postcircumfix (and operator definitions). moritz++
masak ruoso: actually, it works by cheating, AFAIU.
ruoso: I had the impression Faz would use .reduce, but it does not seem to do that at all... 13:33
ruoso masak, we need .reduce to support different arity in the multies
but we can have a different dispatcher as well
13:34 exodist joined
masak ruoso: actually, Faz feels like a more syntax-rich LolDispatch. 13:34
ruoso masak, plus the chained dispatch
masak yes, I noticed the word 'chained'.
13:34 payload left
masak does that mean several things can match, in order? 13:35
ruoso yes
note the %*stash<posts> is populated at the root action's begin
which is re-used by both create and inde
*index
masak that seems to presuppose an application which survives single requests. 13:36
ruoso what do you mean? 13:37
masak, the regex of the chained actions are concatenated 13:38
so...
if root regex is / ^ /
and index regex is / \/? $ /
the full regex that will match index is
/^ \/? $/
then create also chains root... so, in the end, you get a regex like
masak ruoso: ah, nice. 13:39
that makes a lot of sense.
ruoso / ^ [ \/? $ { make $index } | \/create \/? $ {make $create} ] /
actually
there are some named captures in the way, so I can get parameters in the middle 13:40
so the capture of the regex for that specific action is used as the capture when calling the closures for that action
13:41 LadyLunacy left, plu left 13:43 LadyLunacy joined, plu joined 13:45 sjn joined, wollmers left 13:48 ElectricHeavyLan left, autin left 13:49 mizioumt left 13:54 pmurias joined
ruoso masak, I'm adding a view_post action in Faz' Yarn, so you can see the capture in action 13:54
hi pmurias 13:55
masak ruoso: nice. 13:56
pmurias ruoso: hi
ruoso pmurias, I think now we can just try to fix the tests, and the refactoring is over 13:57
pmurias we mostly have sigantures and perl 5 integration to restore 13:59
14:00 sjn left
pmurias ruoso: and we have a leak in re-smop :( 14:02
masak mberends: ping.
ruoso pmurias, a new one?
pmurias yes 14:05
mberends masak: OH HAI 14:06
masak mberends: oh hai, whoz op?
so I attempted the Ecosystem patch yesterday. 14:07
ran into a smallish number of interesting issues.
mberends git-pulls 14:08
masak haven't pushed anything from that yet.
the last issue is kind of a blocker...
mberends I'm going to distil yesterday's todo list into PIONEER and README, for your comment. What's the blocker? 14:09
ruoso masak, just pushed the version using the capture.... 14:10
masak mberends: PERL6LIB not set when compiling lib/Installer.pm
mberends: I think we need that full-fledged build system.
ruoso: ok, will check.
14:16 alester_away left
mberends lib/Installer.pm did not nest any 'use' commands, therefore it compiled anyway. You could insert PERL6LIB into proto:59. 14:17
moritz_ rakudo: say ?("ab" ~~ /a <!after b>/)
p6eval rakudo 705cb2: OUTPUTĀ«1ā¤Ā»
moritz_ rakudo: say ?("ab" ~~ /a <!before b>/)
p6eval rakudo 705cb2: OUTPUTĀ«0ā¤Ā»
moritz_ rakudo: say ?("ab" ~~ /<!after a> b/)
p6eval rakudo 705cb2: OUTPUTĀ«0ā¤Ā»
masak mberends: yes, I'll try doing that first. but I do think that a full build system is the long-term solution. 14:18
14:18 alester joined
mberends masak: agreed 14:18
'env -i ...' would be even better, rtfm 14:19
masak :) 14:20
yes, ISTR someone said I should use env, because the env-less form is bash-specific or some such.
mberends Bourne and Korn do it env-less too, but the -i is FTW 14:23
14:23 iblechbot joined
ruoso masak, btw... wait a bit more before checking Faz Yarn... I'm doing some other cool thing now 14:23
jnthn back
masak ruoso: :)
ruoso masak, it seems you're working on some kind of templating engine... is it? 14:24
14:24 jferrero joined
pugs_svn r26829 | moritz++ | [t/spec] unfudge infix:ƂĀ«...ƂĀ» test for rakudo 14:24
14:28 LadyLunacy left 14:30 LadyLunacy joined
masak is fascinated by the tests in t/spec/S01-perl-5-integration 14:31
jnthn livemocha++ # most interesting site I've found in a while 14:33
OK, back to debugging my test fail from my traits patch...
pugs_svn r26830 | pmurias++ | [re-smop] ported over Attribute
r26831 | pmurias++ | [re-smop]
r26831 | pmurias++ | started porting over DefaultBlockSignature
r26831 | pmurias++ | t/control_exception.t passes
pmurias ruoso: what is Faz Yarn? 14:34
ruoso pmurias, Faz is an Action Dispatcher in Perl 6
and Yarn is a blog engine masak started and that I later ported to Faz 14:35
masak there's also LolDispach Yarn, but it goes under the name "omgblog". :)
ruoso github.com/ruoso/faz/blob/a7a806ed8...ib/Yarn.pm 14:36
that's the latest version... that does fair use of Chained actions and of the different steps of the processing
masak jnthn: yes, livemocha seems really cool. will probably get a language partner on that site once my afk language partner has left town...
ruoso masak, now you can look at it ;) 14:37
masak ruoso: well, after my workday finishes, I can. :)
pmurias is remainded that he should be working on his project for uni rather than making smop pass tests... 14:39
ruoso the cool part is around line 91\
pmurias * reminded
jnthn masak: Yeah, I'm getting some additional Slovak practice with it. Seems to work well. :-) 14:40
masak jnthn: sounds promising.
pmichaud rakudo: sub postfix:<<!>>($x) { [*] 1..$x }; say 7!; 14:41
p6eval rakudo 705cb2: OUTPUTĀ«5040ā¤Ā»
masak wow, that trick never gets old. :)
pmichaud rakudo: sub postfix:Ā«!Ā»($x) { [*] 1..$x }; say 7!;
p6eval rakudo 705cb2: OUTPUTĀ«5040ā¤Ā»
moritz_ masak: it used to work only with <>, not with <<>> or Ā«Ā» 14:42
pmichaud rakudo: say Ā«hello worldĀ»;
masak oh!
p6eval rakudo 705cb2: OUTPUTĀ«helloworldā¤Ā»
masak pmichaud++
Matt-W wow. Livemocha looks really interesting.
moritz_ pmichaud: we have tests for Ā«...Ā», but not for <<...>> I think
pmichaud rakudo: my $foo = 'hello world'; say Ā«$fooĀ».elems
p6eval rakudo 705cb2: OUTPUTĀ«1ā¤Ā»
pmichaud oops.
oh well :-)
moritz_ it's not yet christmas ;-) 14:43
jnthn pmichaud: I added trait_auxiliary as a category, and it all Just Worked on the parsing side. Thanks! :-)
masak submits rakudobug
ruoso masak, please consider adding .redirect in the Web::Response module... 14:44
14:44 eternaleye left
masak ruoso: I've been wanting that too. so yes. 14:45
14:45 nihiliad joined
mberends nice, ōæ½xAB ōæ½xBB were bad in Pod::Parser, they might start working too 14:47
masak mberends: please check the encoding settings in your IRC client. :) 14:48
(quoting /topic: "UTF-8 is your friend!")
jnthn masak: They looked fine to me. ;-)
14:49 payload joined
pmichaud jnthn: (trait_auxiliary) I should probably fix it so that rakudo doesn't attempt to add those as tokens. 14:49
jnthn pmichaud: Oh? Would that be in the category action?
masak jnthn: ok. Emacs does the Right Thing often enough that I just tend to assume that it's right and other clients are wrong. :)
jnthn I didn't commit yet...
pmichaud jnthn: no, it's in the add_optoken sub (actions.pm)
mberends it's XChat, cannot see what might need changing
moritz_ they look fine here too, but my irssi is configured to do some latin1/utf8 autodetection 14:50
pmichaud you can go ahead and commit, I'll fix it up. I have other things to fix there also.
jnthn masak: If you used vi it'd do the right thing even more often. ;-)
jnthn runs off back to his Visual Studio
masak jnthn: I do use vi. just not for IRC.
jnthn: I do most of my Perl 6 coding in vim, actually.
practically all of it. 14:51
jnthn masak: I know what :q does ;-)
masak jnthn: it's, er, a smiley. :q
jnthn
.oO( sticking out a tongue with a tongue-ring )
masak heh.
jnthn Or "licking nose" perhaps 14:52
moritz_ on #perlde we have a bot that responds to :q ;-)
16:51 <@moritz> :q
16:51 <+dustpuppy> moritz: E37: No write since last change (add ! to override)
jnthn :!q ?
Matt-W moritz_++
But shouldn't it use the German translation?
Matt-W ponders setting his laptop to be in German and seeing how long it takes him to learn how to use it 14:53
moritz_ Matt-W: dunno, dustpuppy mixes the languages a bit
ruoso now needs to implement both Model and View in Faz... the controller part is set already...
moritz_ Matt-W: but I fear that most people wouldn't recognize the German translation ;-)
Matt-W moritz_: quite possible
14:54 skids left
Matt-W home & 14:54
14:54 amoc left
mberends dutch IT people hate using dutch-localized software 14:54
masak I'd say it depends among Swedes, but I sure don't enjoy running software in non-English localisation. 14:58
English is the lingua franca of software development. 14:59
moritz_ isn't that lingua anglica, or so? ;-)
Tene jnthn: please don't bother worrying about keeping loldispatch working. It was a 10-minute hack to see if I could do it. Thank you for making the right way work. 15:00
lambdabot Tene: You have 1 new message. '/msg lambdabot @messages' to read it.
masak moritz_: no, because 'lingua franca' was lexicalized at the time when the Frankish tongue was predominant. 15:01
moritz_: in the same way, my collegues order a Coke with their food and are happy with getting a Pepsi.
pmichaud ick, Pepsi. 15:02
15:02 skids joined
moritz_ masak: that wasn't entirely serious 15:02
masak moritz_: I know. :)
moritz_ (man smiley ;-)
pmichaud Whenever offered a Pepsi in lieu of a Coke, I choose water.
masak moritz_: I just felt like giving a serious answer.
moritz_ masak: I know that feeling
masak pmichaud: I tend go directly for the water. YMMV.
moritz_ masak: I also answer rhetorical questions sometimes, when I feel like 15:03
masak moritz_: aye. moi aussi.
dalek kudo: 90dbe0b | pmichaud++ | docs/spectest-progress.csv:
spectest-progress.csv update: 388 files, 11258 passing, 1 failing
masak moritz_: dictionary.com suggests that 'franca'/Frankish might have meant something more like 'European' at some point.
15:04 alc left 15:09 kidd left 15:10 kidd joined 15:12 mofino joined
mofino evalbot: 'perl6: say 3;' 15:13
moritz_ perl6: say 3
mofino oh
haha
p6eval elf 26831, pugs, rakudo 90dbe0: OUTPUTĀ«3ā¤Ā»
masak perl6: loop { print 'a' } 15:14
moritz_ masak: pfui!
masak was that bad? :P
moritz_ masak: "pfui" is what we say to a dog that does something forbidden ;-)
masak oh. we have 'fy!' in Swedish. must be parellel. 15:15
dalek kudo: 612bcf3 | moritz++ | docs/ChangeLog:
[docs] ChangeLog updates
jnthn Tene: It was not going to break too soon more out of because I have to refactor other bits first more than trying to keeo loldispatch working ;-) 15:16
alester Anyone else getting any traffic out of google/gmail/etc now?
masak alester: gmail works here. 15:17
15:20 donaldh left, donaldh joined
p6eval pugs, rakudo 612bcf: 15:20
..OUTPUTĀ«aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa...
..elf 26831: OUTPUTĀ«Can't call method "match_hash" on an undefined value at ./elf_h line 3238.ā¤Ā»
masak ok, I won't do that again, promise. 15:21
jnthn lol
Talk about delayed reaction...
pmichaud rakudo: say "masak: s{ 'u' x 100 }re you won't."
p6eval rakudo 612bcf: OUTPUTĀ«masak: suuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuure you won't.ā¤Ā»
masak :P 15:22
15:22 eternaleye joined
jnthn rakudo: sub infix:<ƶ>($x, $y) { say "flyyy" }; (1,2,3) Ā»Ć¶Ā« (4,5,6) 15:23
masak (why do I get the feeling that pmichaud might have been slightly sarcastic?)
p6eval rakudo 612bcf: OUTPUTĀ«Statement not terminated properly at line 1, near "\x{bb}\x{f6}\x{ab} (4,5,6"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
jnthn aww. 15:24
masak is that a new bug?
jnthn Yes. No.
moritz_ jnthn: I think meta-ops are pre-generated
masak I mean, apart from being a known butterfly.
jnthn (Can't parse unicode hypers is none.)
erm, known
masak ok.
moritz_ rakudo: sub infix:<ƶ>($x, $y) { say "flyyy" }; (1,2,3) >>ƶ<< (4, 5, 6)
p6eval rakudo 612bcf: OUTPUTĀ«Statement not terminated properly at line 1, near ">>\x{f6}<< (4, "ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
jnthn But as moritz_ said, we don't handle generating meta-op forms yet. 15:25
For user defined ops.
pmichaud ....but we might be able to do so.
moritz_ aka it's a "todo bug" ;-)
jnthn Oh, I'm sure we can.
It's just whether it's worth hacking it in now, or waiting for when we can parse them the more STD-ish way, if that's teh plan.
moritz_ heh, rakudo now parses STD.pm up to line 163 15:26
masak submits TODO bug
TimToady sounds like it's about time for a rakudo emitter for viv 15:27
alester is getting 500ms+ pings to google.com and gmail.com :-(
moritz_ but I can't figure out why it stops parsing there
pmichaud google.com was slow for me a bit earlier.
mofino alester, flush out your tubes
google is fine from the end 15:28
this
TimToady oh, bizarre
moritz_ there's nothing obvious advanced in that region
alester I am not the Tube Keeper. 15:29
TimToady when I ping google.com, it only returns every packet for which the seq# is !%5
but it's consistently 86 ms
pmichaud heh
mofino TimToady, yeah
pmichaud that is weird. I only get the packets where %5 == 1|2 15:30
mofino TimToady, it is weird
I suspect they treat ICMP with very low priority
TimToady but they've obviously got it running on some kind of commutator
maybe they're under ping attack :)
pmichaud oh, now I'm getting every packet.
TimToady so am I 15:31
weird
pmichaud from a different IP, though.
TimToady yes
jnthn TimToady: Should trait_auxiliary eventually be looking for a type name?
pmichaud 74.125.67.100 gives me 40% of the packets.
jnthn TimToady: So is Foo requires Foo to be a typename?
(or checking what <longname> hands back is a typename, perhaps.) 15:32
TimToady STD doesn't check trait_auxiliary:<is> for semantics currently 15:33
pmichaud jnthn: what about something like sub infix:<xyz> is equiv(&infix<+>) ...
TimToady it doesn't even care whether hte name exists
pmichaud would "equiv" need to be a type name there?
TimToady shouldn't care right now 15:34
std: sub infix:<xyz> is equiv(&infix<+>) {...}
p6eval std 26831: OUTPUTĀ«ok 00:03 47mā¤Ā»
TimToady std: sub infix:<xyz> is elephant(&infix<+>) {...}
p6eval std 26831: OUTPUTĀ«ok 00:03 47mā¤Ā»
pmichaud yes, but if we fix trait_auxiliary:<is> to require a typename, would that pose a problem?
s/fix/change/
or would we expect "equiv" to be a role/type name 15:35
jnthn pmichaud: I would expect that equiv would be yes.
pmichaud: Unless there are some that we magically recognize in the parser and don't try to dispatch.
ERm, in the actions...
And don't emit a dispatch for.
masak TimToady: do I recall correctly that forward gotos were problematic due to not-declared-yet-ness?
TimToady yes, we require goto "label" for forward goto 15:36
masak ah, quotes.
worksforme.
goto <label>
pmichaud jnthn: if "equiv" is a type or role name, does that pose difficulties with creating a sub named 'equiv' ?
i.e., are we eating up major portions of the namespace?
jnthn pmichaud: I think so; that's one of the things I'm bothered about. 15:37
pmichaud: But unless we special case some traits, I'm not quite sure how we handle this.
15:38 M_o_C joined
jnthn I'm just going on what S14 says for user-defined traits. 15:38
Which is that you do it by introducing a role or class of that name (though I figure any typename will do...)
TimToady S14 doesn't really connect all the dots 15:39
as you have noticed
typenames doesn't feel quite right to me for some reason
pmichaud same here.
jnthn TimToady: Are they connected elsewhere, or do you have a sense of how you want them connected? :-)
TimToady well, I've never thought about how they actually show up in your lexical scope 15:40
jnthn I'm fine with use not parsing a typename there. I do need to know what we should do with the <longname> we parse there though.
I'd guessed they were just roles that were in the setting, just like any other role. 15:41
TimToady well, but the intent was to do things beyond what roles normally do 15:42
pmichaud +1
(beyond what roles normally do)
afk for a bit
TimToady will try drinking more coffee and see if it helps...
jnthn So far as I can follow S14, the main point of introducing something is so you have a name to write for the type of the first param to trait_auxiliary 15:43
TimToady: Also while you are pondering over your coffee, I'm also curious in: 15:44
multi trait_auxiliary:<is>(xxx $trait, Class $container; $arg?) {...}
What Class is there.
TimToady I suspect that means any typename 15:45
jnthn Can it be a subset? 15:46
I can probably do it easily enough as one.
TimToady it's just intended to let you multi dispatch on different kinds of containers 15:47
jnthn Sure
TimToady where "container" is subject to negotiation
since pretty much any object could be considere a container of sorts
skids pmichaud++ operator subs booyah! 15:48
jnthn The name Class is just curious since there is no class class :-)
TimToady we didn't know that back then
jnthn Ah, OK.
TimToady that's actually very ancient doc
15:48 Psyche^ joined, ab5tract joined
TimToady pre protoobject, pre STD, almost prehistoric 15:48
jnthn I guess now you could write it as $container where { .BUILT }
erm 15:49
!(.BUILT)
To get type objects.
TimToady we're not trying to get a type object, I don't think
we're trying to get the actual container, like the array or hash object 15:50
since that's what has to be modified
jnthn What about the class Foo is Bar { } style thing though?
15:50 ingy is now known as ingynap
jnthn We'd like to be able to do a meta-class call to add_parent on Foo to add Bar. 15:50
15:51 eternaleye left
jnthn And I figured we'd call trait_auxiliary:<is>(Bar, Foo) 15:51
And Foo at that point is the type-object of the currently-under-construction class. 15:52
I maybe figured wrong though. :-)
TimToady well, but maybe a trait on class passes the metaobject, not the typeobject
15:52 Patterner left, Psyche^ is now known as Patterner
jnthn Well, at least in the smop world there is no meta-object. 15:53
TimToady because that's the actual container the trait is being applied to
jnthn So far as I understand it.
TimToady indeed, and that's why class Foo {...} fails there
jnthn The meta-class is more "singleton" rather than an instance.
Did you see the meta-class being something we have an instance of per class that gets declared? 15:54
That's how Rakudo currenlty does it, fwiw.
TimToady I think of it as an instance whose attribute values define the class
jnthn How does that play into the representation stuff? 15:55
TimToady I don't know what you're meaning by "play into" 15:56
jnthn OK, maybe a better question then
What does it mean to create a class based on a P5Hash?
ERm, an instance?
phone
back 15:57
TimToady I don't really care what it means, as long as it works :) 15:58
jnthn Yes, but I't not convinced I know what "works" means in this case. 15:59
Do you see a class based upon a P5Hash as using that hash for its attribute storage?
TimToady I'm not in a mental state this morning yet to generate the alternatives
as opposed to what?
and are you speaking of P5Hash as a type or a representation that .CREATE knows about? 16:01
jnthn I haven't really figured out how they separate out yet.
TimToady the argument to .CREATE is just a string, as far as I know
jnthn Is there something that basically says "here is how to use the type P5Hash for instance storage"? 16:02
Which allows it to work as a representation?
TimToady how that picks the right MOP is beyond my ken
it seems to be that different VMs are going to differ in how you write a RI, so this is beyond the design of P6 16:04
jnthn That's fair enough, I just want to try and make sure I've got the relationship between the parts worked out.
TimToady I don't guarantee that ruoso sees it the same way I do :) 16:05
but it stands to reason that the implementation of P5Hash is going to be very different on a platform which supports P5 than on one that doesn't... 16:06
16:06 justatheory joined
TimToady as a limiting case 16:06
jnthn On one it's non-existent? ;-)
pmichaud rakudo: say 11258-10467; 16:07
p6eval rakudo 612bcf: OUTPUTĀ«791ā¤Ā»
pmichaud not bad for May release. :-)
jnthn I hadn't known for sure up to now though, that you viewed the metaclass as "an instance whose attribute values define the class" 16:08
moritz_ I once calcualted that 690 or os is the average number of new passing tests per month
perlgeek.de/blog-en/perl-6/rakudo-r....writeback
TimToady just because I view it that way doesn't mean it's the right way forward, but the data about the class has to be stored somewhere
and I don't think it's the undefined type object
pmichaud moritz_: yes, but April was a bit of an anomaly, and I was thinking that we might see a slowdown after that spike. 16:09
jnthn A model where the meta-class knows this stuff fits well for me.
TimToady a given metaclass may well delegate a lot of this to the RI, of course
masak rakudo: say (16000 - 11300) / 715
p6eval rakudo 612bcf: OUTPUTĀ«6.57342657342657ā¤Ā»
masak we're running out of tests. o_O 16:10
moritz_ pmichaud: yes, but April wasn't part of may analysis
so the anomaly didn't make it into the average
pmichaud moritz_: sure; I just knew that we were averaging ~700 tests per month, and I had been expecting May release to be somewhat less than that. Looks like it won't be. :-)
moritz_ pmichaud: ;-) 16:11
mberends May + 6.5 months < Christmas ;)
jnthn TimToady: OK, that makes sense.
pmichaud rakudo: say 11258/16584
p6eval rakudo 612bcf: OUTPUTĀ«0.678847081524361ā¤Ā»
masak mberends: :) 16:12
rakudo: enum Month <Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec>; say May + 6.5
p6eval rakudo 612bcf: OUTPUTĀ«10.5ā¤Ā» 16:13
masak is there any way I can get a month back from that?
jnthn TimToady: So I guess if we have a way of knowing if something is a metaclass, we can write that into our trait_auxiliary signature.
moritz_ masak: NYI I think
masak moritz_: but theoretically possible?
moritz_ masak: the spec says Month(10.5.int)
masak moritz_: ooh, nice.
pmichaud rakudo: enum Month <Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec>; say Month((May + 6.5).int).name; 16:14
p6eval rakudo 612bcf: OUTPUTĀ«invoke() not implemented in class 'Perl6Role'ā¤current instr.: '_block30' pc 93 (EVAL_21:54)ā¤Ā»
pmichaud rakudo: enum Month <Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec>; say (May + 6.5).int;
p6eval rakudo 612bcf: OUTPUTĀ«10ā¤Ā»
masak submits TODO ticket 16:15
jnthn Need postcircumfix...
erm
postcircumfix:<( )>
And invoke in Parrot to be fixed.
Or at least, need to be able to pun...
And invoke the pun.
pmichaud I'll have to think about that one.
jnthn pmichaud: Yeah, I suspect we're blocked on doing anything implementation wise until Parrot's invoke overriding is fixed though. :-) 16:16
erm
:-(
pmichaud we could recognize it syntactically.
jnthn So I've really not given it a lot of thought yet.
Yeah, that's probably one way.
ruoso jnthn, TimToady, my vision for building an object with a Hash or P5Hash is that you're going to need a RI specific to support using that object as the representation
pmichaud In particular, I wonder if (Month)(10) works
(according to the spec) 16:17
TimToady it should
pmichaud okay.
Then yes, we'll need to fixup Parrot invoke.
jnthn: "fixing Parrot invoke" is one of those things we should put on our "important to Rakudo" list.
TimToady in fact, Month(10) is exactly the same, since Month doesn't look for args
ruoso TimToady, additionally, selecting the MOP is a different issue than selecting the representation... you can interchange them, in the way I see things (that's what the REPR API is all about) 16:18
TimToady (since types never do, or you couldn't say May + 6.5
jnthn ruoso: That's fine, the issue here is more that it looks like meta-classes need to be instantiable if they're to be used as "the thing we send over to trait_auxiliary".
TimToady )
ruoso jnthn, they can be... even in SMOP... 16:19
masak pmichaud: speaking of which -- and I don't mean to stress you out unnecessarily -- how's the ROADMAP coming?
pmichaud I'm thinking it'll be today's task.
ruoso the question is wether the type metadata will be stored in the HOW object or in the type object
masak yay!
ruoso in SMOP, it currently stores it in the type object
jnthn ruoso: I'm thinking that'll be up to the HOW.
pmichaud Since I got operators done yesterday, I feel like I've accomplished something code-wise .
ruoso jnthn, oh... certainly it is...
TimToady the HOW decides where to store it
ruoso yes
it's the HOW who receives the .add_method calls
jnthn Right, agree 16:20
But the issue I'm trying to get at is
ruoso but that's unrelated with choosing the represetnation
jnthn class Foo { method bar() { } }
argh
class Foo is Bar { }
ruoso you should be able to use whatever representation you want with whatever HOW you want
TimToady that doesn't seem obvious to me 16:21
jnthn When we dispatch on trait_auxiliary:<is>(Bar, Foo) what is passed for Foo.
If it's the meta-class that's fine, but in that case, and if the meta-class is delegating to the representation, how does that representation come to exist.
ruoso Foo.HOW.add_parent(Foo, Bar)
16:22 ejs joined
jnthn ruoso: No, you missed the point. 16:22
We need to do a multi-dispatch to trait_auxiliary:<is>
ruoso ok
oh... I see..
the problem is that usually you have a symbol for the trait 16:23
like
is export
or
is foo
jnthn Right, so it's a two fold issue.
ruoso in this case you have what usually would be an argument to the trait
jnthn 1) What do we pass for Foo - the thing that is currently under construction?
ruoso yes
jnthn 2) What do we write in the signature of the multi?
ruoso I guess there can be some catch in the syntax to realize that this is a specialized trait 16:24
and generate the equivalent to
16:24 payload left
jnthn Ugh, no. 16:24
ruoso is subclass_of Bar
jnthn Yeah but
That still doesn't solve the problem of users writing traits of their own.
TimToady syntax knows whether Bar is type or not 16:25
ruoso it doesn't?
jnthn class Foo is MyThingy { }
TimToady the syntax could map that differently depending on the typehood of MyThingy
jnthn multi trait_auxiliary:<is>(MyThingy $trait, ??? $class) { }
TimToady so the multi only sees non-type traits
16:25 NoirSoldats_ joined
TimToady is one possibility 16:26
16:26 NoirSoldats left
TimToady it's more in line with the original thought than to make all traits into types 16:26
jnthn TimToady: Here I'm not wanting to inherit from MyThingy, but write my own custom trait.
TimToady but doesn't solve the namespace problem
jnthn Which applies to classes.
ruoso I guess the syntax can reserve some uses for the "is"
pmichaud (jnthn: class Foo is mytrait(...) { ... } )
TimToady in that case, *don't* make it a type, and it dispatches to the multi
jnthn In which case I need to know what to write for ??? 16:27
16:27 FurnaceBoy joined
TimToady well, the incipient type object is sufficient to get at anything else 16:27
(in this case) 16:28
since the container is really mediated by .HOW
jnthn What does incipient mean?
TimToady "about to come into being"
jnthn ah, ok
TimToady but I still don't know what namespace mytrait sits in 16:29
16:29 alester left
ruoso I guess if it isn't a type, it's a sub 16:29
TimToady doesn't follow 16:30
ruoso (that's how it works everywhere else, afaics)
jnthn So this means that the thing we pass along is the incipient type object, and thus what goes in the signature is the same thing we use anywhere else to check if we've been passed a type object.
ruoso when you parse: Foo.bar
if there is a type with the name "Foo"
that's just a method call on the type object
jnthn ruoso: I'm not sure the problem is with it being a type, it's about namespace pollution.
ruoso but if there isn't a type,
TimToady std: rw
p6eval std 26831: OUTPUTĀ«Undeclared routine:ā¤ rw used at 1 ā¤ok 00:02 35mā¤Ā»
TimToady that is as it should be
there is no rw sub
ruoso ok
point taken
pmichaud and we don't necessarily want to preclude the definition of a "rw" sub. 16:31
TimToady it's got it's own namespace somewhere
a sigil or a subpackage
like export tags do
ruoso a sigil is more interesting...
because it's lexical
TimToady well, I'm always fascinated when a new sigil or twigil is proposed; we usually add it, and then remove it later :) 16:32
moritz_ we didn't remove the : sigil yet
ruoso well... the other option would be to accept a bare
jnthn ruoso: I guess it works if we can write role Ā£rw { } or something.
moritz_ although I've never seen it in use
erm, twigil
(except in tests)
jnthn ruoso: I'd prefer subpackage myself. 16:33
role TRAIT::rw { ... }
ruoso but then you loose it as being lexically scoped
TimToady moritz_: well, okay, some of them stick :)
why?
jnthn my role TRAIT::rw { ... } # :)
ruoso er
TimToady names containing :: can be lexically scoped these days
ruoso they can? 16:34
Trey msg [particle]-: hey there... quick q i thought you might know of the top of your head. if an uploaded CPAN module had no license information anywhere in it, either in the meta or the makefile.pl or the text docs, is there a default license? or can one just not use it legally in a redistributed product?
TimToady std: my class Foo::Bar {...}; my Foo::Bar $x;
p6eval std 26831: OUTPUTĀ«ok 00:02 36mā¤Ā»
ruoso and what if the traits are stored in the lexical scope without a sigil?
just as types are
TimToady then you'd get hash key collisions 16:35
symbol tables are just hashes
ruoso but do you really want to allow traits and types to have clashes?
jnthn I guess lexical names just act as flat names in the lexpad rather than packages per se though?
Trey oops, sorry about that missend. i need an extra slash... does TimToady, can you remove a slash from some feature so I can use it to message [particle]?
TimToady ruoso: huh? I'm trying to prevent clashes
jnthn: no, they make subpackages 16:36
jnthn Trey: You use @tell on this channel
ruoso TimToady, I mean... is it ok to define both a type MyTrait and a trait MyTrait?
jnthn ruoso: No, thus the suggestion of TRAIT::MyTrait instead.
ruoso so my point is to store both in the lexpad without sigils... so you can't define both 16:37
TimToady oh, no that level. but generally classes will be uppercase and traits lowercase, so I'm not worried about type clashes much
but sub clashes are more likely
ruoso subs are stored with &
so no clash
TimToady right, in which case they might as well be types
hmm 16:38
ruoso but they aren't types
so they get parsed different
jnthn huh?
ruoso their use get parsed different
is <type> # parses one way 16:39
jnthn ruoso: A trait name is just so far as I can tell something you can look up and then use in a multi-dispatch.
ruoso exactly
jnthn But in a signature
ruoso is <somethingelse> # parses other way
jnthn :(Foo $x) # Foo should be a typename here
ruoso right 16:40
TimToady but if they're not types, why not just give them their own sigil in the symbol table so as to prevent confusion with sub rw
(when called without &)
ruoso but nobody looks in the lexpad without the &
masak two questions: is someone willing to be grant manager for my Collections grant proposal? are there any general guidelines for knowing what amount of money to request?
TimToady otherwise parsing a bare rw has to figure out what rw is
ruoso except when the parser knows it is a Type
and I'm proposing that it also looks when it's a trait 16:41
jnthn ruoso: I don't understand "when it's a trait".
TimToady I think a sigil would be cleaner here
ruoso jnthn, is something # something is either a type or a trait
but I'm ok with a sigil as well... 16:42
jnthn TimToady: So supposeing we chose Ā£ as the sigil (I guess we won't ;-)) then
is rw # actually looks up Ā£rw 16:43
16:43 [particle] joined
ruoso yes 16:43
TimToady going up the lexical scope till it finds it, or doesn't
jnthn And you could write a role Ā£rw { ... } or whatever you wanted to introduce that label?
TimToady or maybe it's just exported that way, maybe 16:44
ruoso bare Ā£rw;
jnthn ?
ruoso since it's just used for identity
my bare rw; # declares Ā£rw
jnthn Well, if you read S14 note that you often want to introduce a role so you can mix in it as part of applying the trait. 16:45
In fact, well-behaved traits generally should.
16:45 plu left
jnthn I guess you could declare a separate role too, but when where would it go? 16:46
It's probably better to hang everything off the same name.
TimToady the question is whether the role itself needs to be exported, or just the Ā£rw that knows about it
a trait isn't required to use a role 16:47
jnthn TimToady: Sure, S14 says a class is just fine too.
TimToady it doesn't have to use a class either
jnthn I guess you can alias whatever to Ā£rw
TimToady Ā£rw just knows how to call the trait_auxiliary multi, and that's what has to be visible in the user's scope 16:48
ruoso it's just used for identity
jnthn TimToady: OK, but it needs to be something you can write in a signature, no?
TimToady it?
ruoso you can write a bare value in a signature
TimToady there are too many its floating around
as a constraint? yes 16:49
ruoso std: multi foo(1) { }; multi foo(2) { };
p6eval std 26831: OUTPUTĀ«ok 00:02 38mā¤Ā»
TimToady and Ā£rw is a bare value, you're saing
*saying
ruoso yes
we can even omit the sigil
jnthn I'd seen more Ā£rw as an alias to something. 16:50
ruoso std: multi trait_auxiliary:<is>(export, $code)
p6eval std 26831: OUTPUTĀ«##### PARSE FAILED #####ā¤Malformed block at /tmp/5lgy1jOZHl line 1:ā¤------> multi trait_auxiliary:<is>(export, $code)ā¤ expecting any of:ā¤ parameterā¤ signatureā¤ type_constraintā¤ typenameā¤ whitespaceā¤FAILED 00:02 37mā¤Ā»
ruoso right... it wasn't meant to be sent to std
ruoso 's brain--
TimToady that only works if "export" is a type, or an enum-like value 16:51
which parses like a type
16:51 alester joined
TimToady but happens not to be undefined 16:51
ruoso my point is that it would look for bare names as well
TimToady I'd still like to keep bare trait names out of the user's normal namespace, if possible 16:52
generally it can only lead to confusion
since most people aren't going to be writing trait handlers
ruoso it is out of the normal namespace
only certain things look for it
TimToady you can't write bare export in a signature without it being there in the namespace, it seems 16:53
jnthn You could write Ā£export perhaps, though. 16:54
ruoso yes... but it only look for "bare" in a few specific sport
*spots
TimToady list of exceptions, bah
ruoso it's the same list of exceptions that Types have
jnthn Aye, just seeing that we are about to parse a trait_auxiliary should not make us parse the signature differently...
16:55 cdarroch joined
TimToady shower & 16:58
jnthn cooking dinner & 16:59
17:00 [particle]- left 17:02 ejs left 17:08 DemoFreak left 17:09 DemoFreak joined
masak what'll happen to the Perl 6 spec after Christmas? 17:09
moritz_: how's the implementation of infix:<...> going? 17:11
17:20 hercynium joined
jnthn masak: iirc it neeeded laziness too 17:21
s/e// 17:22
17:22 zamolxes left
masak jnthn: oh, right. 17:23
couldn't laziness be faked in this case? 17:24
(implemented in pure Perl 6 in some way)
jnthn Perhaps. 17:26
Well, I don't know exactly what moritz_ has in mind. :-)
Anyway, I hope laziness ain't *that* far off. 17:27
We've put it off for quite a while already. ;-)
rakudo: my @a = -1,2,-3; for @a { .= abs }; say @a.perl; 17:29
p6eval rakudo 612bcf: OUTPUTĀ«Unable to parse block; couldn't find final '}' at line 1, near ".= abs }; "ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
jnthn rakudo: my @a = -1,2,-3; for @a { .=abs }; say @a.perl;
p6eval rakudo 612bcf: OUTPUTĀ«[1, 2, 3]ā¤Ā»
jnthn masak: Not sure if you missed it, but that works now. :-)
(remember you asking about it)
Though of course this is shorter...
rakudo: my @a = -1,2,-3; @a>>.=abs; say @a.perl; 17:30
p6eval rakudo 612bcf: OUTPUTĀ«[1, 2, 3]ā¤Ā»
masak jnthn: that's great. I'll update the pun README, then. 17:33
jnthn: well, pun can only use the first variant.
jnthn oh wow, PayPal finally caught up with Slovakia having joined the euro... 17:35
Only took a few months. :-|
BTW, anyone who has requests for Rakudo day tomorrow, please let me know. 17:39
Other than any "help with moving to .HLL" tasks, I don't have anything planned yet.
masak browses through the recent RT tickets
jnthn: if you have a minute, I'd appreciate if you'd skim through this: gist.github.com/111790 17:40
mberends jnthn: with luck, there might be a first cut of Temporal.pm to add to setting
jnthn mberends: Ooh, I'm happy to help on that. 17:41
(On getting it into the setting, that is.)
mberends :)
jnthn I guess it'll win us some spectests too?
TimToady at some point we need to move rakudo's setting to lexically outside of the user's program
dunno how close that is to doable 17:42
jnthn masak: Ah, I can do that today. In fact, now, amongst dealing with preparing dinner...
masak excellent.
I must detach soon to do likewise, but I will backlog.
mberends jnthn: the Pugs time.t is P5 style obsolete, I'm making new tests 17:43
jnthn mberends: Great. :-)
17:43 justatheory left
jnthn TimToady: It's probably not so bad in some senses. But none of the built-ins are lexical yet. 17:43
Plus we didn't get lexical importation working.
S11 may need a little clarification on the route to that. 17:44
masak jnthn: I wouldn't mind if you took a look at [perl #65208], on the theory that the less broken Match objects are the more pleasant they are to work with. OTOH, maybe it's pmichaud's area of expertise, I dunno.
jnthn I've been working on some issues to get us closer - lexical multis now essentially work.
And lexical subs etc. Lexical classes/roles shouldn't be much harder. 17:45
But we probably gotta fix the "can't see lexicals outside a class" issue...
17:46 [particle]2 joined
masak jnthn: [perl #64876] is slightly disturbing. 17:46
jnthn: but, to be honest, I don't actually have any wtf bugs right now, not that I can remember. 17:47
oh, [perl #64656] is "unfortunate".
masak ponders building his own nomenclature of bug severity 17:48
anyway, food &
17:48 masak left
jnthn Oh, 64876 I think I know how to fix. 17:51
rakudo: module A { sub foo() { return 42 }; say foo } 17:52
p6eval rakudo 612bcf: OUTPUTĀ«Null PMC access in find_method()ā¤current instr.: 'parrot;A;foo' pc 200 (EVAL_22:99)ā¤Ā»
moritz_ Matt-W: it still waits on lazy lists
jnthn rakudo: module A { sub foo() { 42 }; say foo }
moritz_ sorry, meant masak
p6eval rakudo 612bcf: OUTPUTĀ«42ā¤Ā»
moritz_ Matt-W: ignore me ;-)
jnthn Aye, I know about that bug.
(or why we have it, at least)
17:56 Morpheus^ joined
pmichaud 16:40 <ruoso> but nobody looks in the lexpad without the & 18:01
fwiw, rakudo doesn't store its subs with the & (there are a variety of parrot-related reasons for this)
18:01 [particle] left
pmichaud it will probably convert to using the &, but at present it does not. 18:02
ruoso but that's not correct in terms of spec, is it?
pmichaud I don't believe the spec is specific about that point, other than one needs to be able to look up subs using the sigil in a namespace or symbol table.
ruoso rakudo: sub foo { say OUTER::<&foo>.WHAT }; foo();
pmichaud But an overload of postcircumfix:<{ }> could achieve this.
p6eval rakudo 612bcf: OUTPUTĀ«Could not find non-existent sub OUTERā¤current instr.: 'foo' pc 107 (EVAL_18:71)ā¤Ā»
ruoso rakudo: sub foo { say CALLER::<&foo>.WHAT }; foo();
p6eval rakudo 612bcf: OUTPUTĀ«Could not find non-existent sub CALLERā¤current instr.: 'foo' pc 107 (EVAL_18:71)ā¤Ā»
ruoso meh
jnthn Aye, I'm not sure the spec is specific on that either. 18:03
jnthn -> nom
TimToady the & must be used with .{} to look up a sub name in a symtab 18:04
pmichaud correct.
that part I knew about.
ruoso hm?
you mean it's different than other variables?
TimToady no, it's not 18:05
OUTER::<$foo>
OUTER::<&foo>
just the same
ruoso ah
ok
TimToady S11 even gives some examples
ruoso but... 18:06
sub foo { }; say MY::<foo>; should not return the sub
TimToady that is correct
pmichaud that's my understanding as well. 18:07
TimToady if foo is not a type or a named value, it shouldn't find it
ruoso but if you store the sub without the '&'
TimToady then you have to patch up that side too
pmichaud correct. 18:08
TimToady so certainly we'll be pushing parrot in that direction
pmichaud I'm not saying that rakudo is taking the most straightforward or easiest approach here -- but for a variety of parrot-related reasons I'm not entirely at liberty to suddenly switch everything to using the &'s
ruoso right... so conceptually, rakudo does store with the '&'
TimToady at least for Perl 6, if not for other langauges
pmichaud and in some languages we can't use the &
TimToady conceptually, rakudo implements Perl 6 :P 18:09
18:09 justatheory joined
[particle]2 in theory, theory and practice are the same. only in practice do they differ. 18:12
18:12 [particle]2 is now known as [particle]-
TimToady only if you practice enough, in theory 18:13
but sometimes you can get lucky
justatheory TimToady: I am not a woodshed.
TimToady no, you just need to be taken out to one :)
[particle]- just a framework
TimToady while we're refactoring traits, I still wonder if we should also rename them... 18:15
[particle]- loads his shotgun
...couldn't afford an electron gun...
TimToady just shoot yourself 18:16
justatheory TimToady: I saw something nasty in the woodshed.
TimToady well, if you'd patched the roof like I asked you to, there wouldn't be fungus growing on the woodpile 18:17
[particle]- well, heck, if we get some mold growing, we won't even have to paint it
justatheory heh 18:18
ruoso I like mold... I even deprecated slime in favor of it in SMOP...
TimToady woodsheds mold character
so maybe we should rename traits to characteristics 18:19
just to discourage people
characteristic_auxiliary:<is>
justatheory or distinctions
ruoso I think the _auxiliary:<is> part is discouraging enough
jnthn <- nom 18:20
justatheory www.imdb.com/title/tt0112701/quotes
pmichaud masak: (RT #65208) I know why .print isn't working on Match objects... I'm not quite sure how to fix/resolve it, though. 18:25
jnthn @tell masak I just looked through your grant proposal - looks good overall.
lambdabot Consider it noted.
pmichaud .print on match objects is conflicting with the predefined <print> subrule. 18:27
jnthn TimToady: I wouldn't be against re-naming them.
ruoso likes trait as well 18:28
18:28 mofino left
pmichaud print(A.parse('foo')) works, however. 18:28
jnthn pmichaud: What's the <print> subrule? 18:29
pmichaud matches a printable character
like <upper> matches an uppercase character
jnthn ah, ok
pmichaud also, A.parse('foo').Str.print works, though, as does (~A.parse('foo')).print 18:30
jnthn No shortage of work-arounds then. :-)
pmichaud (updated ticket) 18:36
moritz_ just held a talk about Perl 6 regexes/grammars 18:44
called "regexes on crack"
it was received quite well
pmichaud excellent. Maybe I can borrow some notes/slides for my talk(s) on regexes/grammars?
18:45 plu joined
moritz_ pmichaud: they are quite rudimentary, but it might give you some inspiration 18:46
oh, and they are in German ;-)
pmichaud well, I've given talks on regexes/grammars before, but it's always good to grab examples/ideas from others
German is no problem, I think I can read the code at least.
and I can often decipher my way through the german
18:47 M_o_C left
moritz_ moritz.faui2k3.org/tmp/slides.vroom 18:47
you can view them with vim + Vroom::Vroom ;-) 18:48
the theme is "why don't you write parsers in regular expression?"
pmichaud got it, thanks! 18:49
18:49 mizioumt joined
moritz_ you're welcome 18:49
pmichaud heh, I like this part:
+* Backtracking
moritz_ ;-) 18:50
each + makes that line appear on the next slide 18:51
pmichaud right, similar to Spork.
(which is what I typically use for my presentations)
18:51 masak joined
masak radical suggestion: how about renaming the subrule "printable"? 18:51
lambdabot masak: You have 1 new message. '/msg lambdabot @messages' to read it.
moritz_ I should look into that too 18:52
masak @messages
lambdabot jnthn said 26m 28s ago: I just looked through your grant proposal - looks good overall.
masak jnthn: thank you.
moritz_ masak: what's the URL?
I can look throug it too, if you want
pmichaud masak: I think the subrules were original intended to correspond with the ctype function names
masak moritz_: gist.github.com/111790
pmichaud like isalpha, isupper, isspace, isprint, isgraph, etc.
masak pmichaud: I'd be happy if you have a look at gist.github.com/111790 too :)
pmichaud: I see. 18:53
literal you're applying for a TPF grant?
masak literal: a Hague grant.
literal cool
masak pmichaud: well, collision with an already defined method might or might not be considered a sufficient reason to deviate from the rule... 18:54
[particle]- <:print> would be better, maybe, or <printable> or <isprint>
pmichaud masak: Overall I agree; I'm just not sure what @Larry will settle on. :-)
masak pmichaud: understood. :) 18:55
moritz_ masak: (re grant, Buf) you could mention that Buf is the thing that keeps binary data / blobs in Perl 6 18:56
masak: I'm not sure how well the people now Perl 6 and Str
and know about decoded and undecoded stuff etc. 18:57
18:57 Morpheus^ left
pmichaud "the people" tends to be rdice and the folks he passes the proposal to 18:57
18:57 plu left
masak moritz_: based on your previous comments, I added a bit about the deconding/encoding. I could add something about blobs and binary data too. 18:57
pmichaud it would be good to mention blobs/binary data
moritz_ masak: I revise my previous comment ;-) 18:58
masak :)
pmichaud especially since I just switched open() to default to text/utf8
18:58 mizioumt2 joined
pmichaud it would be good to have some way to treat files as blobs again :-) 18:58
masak I'll probably also add a few details about what we know about Set et al.
pmichaud: oh, I'll mention that. thanks.
pmichaud (and yes, I'll happily serve as grant manager for this project) 18:59
masak \o/
pmichaud: I forgot to ask; making you read the proposal draft wasn't a circumspect way of asking. :P 19:00
pmichaud masak: it's no problem, really :-) 19:01
moritz_ masak: you could also mention your impressive number of submitted tickets in your bio
masak: IMHO it proves some of your qualification
masak moritz_: good idea.
by last count (before Oslo) I had submmitted 314 tickets. 19:02
should be over 350 by now.
moritz_ masak: you also had 76 commits to t/ and 85 to docs/ in pugs, an 40 in the rakudo repo ;-) 19:03
masak moritz_: thank you. :) 19:04
moritz_ git log|grep|wc -l is a nice combination ;-)
masak is it 40 already on Rakudo? I wouldn't have guessed.
moritz_ (which makes you the top 9 commiter in rakudo)
pmichaud overall the grant proposal looks pretty good to me. 19:05
masak moritz_: could you paste that list? sounds interesting.
moritz_ masak: nopaste.snit.ch/16549
(no attempt was made at unifying multiple email addresses) 19:06
Infinoid There's also www.ohloh.net/p/rakudo/contributors
moritz_ you can also run `perl tools/commit-stats.pl' in the rakudo repo
Infinoid But I think that's affected by rakudo's history in the parrot repo.
Tene jnthn: defining a multi sub in a class fails
I know why, but can't fix now
masak moritz_: thank you. I like to make similar statistics, but I've mainly attempted it on RT and the different Perl 6 projects. 19:07
moritz_: ooh!
masak does so
moritz_ Infinoid: sure it is
Tene rakudo: class A { multi sub foo { ... } }
p6eval rakudo 612bcf: OUTPUTĀ«No method named 'foo' to remove in class 'A'.ā¤current instr.: '!TOPERL6MULTISUB' pc -5310275 ((unknown file):-1)ā¤Ā»
Infinoid Which explains why I'm on there, despite not being sure whether I have a rakudo commit bit. :P
moritz_ masak: ticket!
Tene can someone make ticket?
I must teach now
masak rakudo: class A { multi foo() {} } 19:08
p6eval rakudo 612bcf: OUTPUTĀ«No method named 'foo' to remove in class 'A'.ā¤current instr.: '!TOPERL6MULTISUB' pc -5310275 ((unknown file):-1)ā¤Ā»
masak makes a ticket!
19:08 mizioumt left
masak uuh, flaky connection... 19:09
jnthn Ah, yes. We actually re-bless blocks now so that's an easy fix too (wasn't really fixable back when multis were first put in...)
19:12 mizioumt1 left, ejs joined 19:13 masak left 19:17 justatheory left 19:20 donaldh left, donaldh joined 19:23 masak` joined, masak joined, masak left 19:24 [particle]1 left 19:31 masak` left 19:54 nihiliad left, nihiliad joined
TimToady pmichaud: I believe the distinction between Cursor and Match objects will solve the .print problem 20:00
20:00 pmurias left, pmurias joined
TimToady $cursor.print is <print>, whereas $match.print is Any.print 20:01
a Match is basically a "dead" Cursor
with not much more interface than Capture 20:02
and it's the .parse boundary that kills the returned Cursor and makes it a Match 20:03
pmichaud would it be (more) correct to say that we build a Match from the returned Cursor ? 20:04
TimToady possibly
pmichaud okay.
I like that distinction. It makes several things much easier.
TimToady maybe the Match is just a Capture plus link to the returned Cursor
to which we delegate .from and .to
and it makes it possible to continue a match from a Match object using Match.parse, maybe 20:05
pmichaud currently PGE has Match.next (but Match.parse would be fine)
20:05 mib_5ckh6m joined
TimToady well, whatever, we just want to keep the accidental collisions down 20:06
pmichaud Agreed, and that explanation (plus the one you gave last week) is a huge help.
20:06 ingynap is now known as ingy
TimToady anyway, that seems to be closer to the STD model of Cursor vs Match 20:06
jnthn I think I even sort of get it now. :-)
TimToady: I'm guessing on the traits, it's best if I leave you to ponder exactly what to do there for a while, and look out for some updates to S14, before digging in with implementation. 20:07
TimToady yeah, still puzzling that out
jnthn OK, thanks. :-)
pmichaud I'm wondering how I can get the Cursor/Match distinction into PGE without breaking too much existing code. 20:08
ingy 5hola
pmichaud I guess I should start by (re)writing it the way I want it to ultimately look, and then I'll figure out what shims I need for backwards compat
ingy hmmm
TimToady pmichaud: you could rebless for now
we've been ingified! 20:09
ingy hi gang
pmichaud well, I can certainly get Rakudo to do something close to the right thing, by giving Rakudo its own Match class (which it has) that treats PGE's Match class as if it is a Cursor
TimToady what's cookin'?
ingy y'all headed to YAPC?
jnthn hi ingy
TimToady which one? 20:10
jnthn wh...? :-)
pmichaud instead of making Rakudo's Match class as a subclass of PGE's Match
ingy I was thinkng PA
jnthn Ah, not me.
pmichaud I'm headed to ::NA and (hopefully) ::EU
TimToady YAPC::PA haven't heard of that one
jnthn I'm doing ::EU and ::Asia.
ingy YAPC Perl America
TimToady ::Asia is the only doubtful one currently
20:11 justatheory joined
TimToady well, not even doubtful, just a bit doubted 20:11
ingy jnthn: awesome. look forward to seeing you in tokyo
TimToady just haven't planned autumn at all yet
ingy TimToady: you are BIG in Japan
TimToady only among the grannies
pmichaud yapc::asia would be fun. 20:12
TimToady the youngsters tend to be larger
ingy TimToady: I been cooking on this testml.org/spec/
TimToady ah, yes
pmichaud I don't have a September trip planned yet :-). So far I've had a trip every month this year.
jnthn pmichaud: BTW, I plan to go wonder around Japan and South Korea for a few weeks after YAPC::Asia.
TimToady I try to keep it to once a month
pmichaud (and have one-per-month through August)
jnthn pmichaud: So don't expect too much of me on Rakudo in sep. ;-) 20:13
ingy pmichaud: do tokyo!
pmichaud jnthn: okay, I'll try not to expect too much.
TimToady jnthn: you won't be done by then?
pmichaud ingy: depends on funding. :-)
ingy I'll buy you some soba and beer
jnthn TimToady: Depends on the spec being done by then. ;-) 20:14
TimToady oh well...
20:14 mib_5ckh6m left
ingy seen masak 20:14
moritz_ @seen masak 20:15
lambdabot I saw masak leaving #perl6 51m 6s ago, and .
ingy thx
pmichaud That dangling conjunction leaves a lot of room for imagination... :-)
jnthn pmichaud: I'm having some bits of vacation in June and July too, but be doing bits when I'm around. :-)
pmichaud jnthn: Good.
vacation++
jnthn Yeah, will want it. 20:16
ingy is on permanent vacation
pmichaud I'm doing vacations in June/July as well. After YAPC::NA Paula and I are doing some travelling in PA
jnthn Ah, that overlaps with when I'm away too.
moritz_ is away most of June 20:17
in beautiful norway
jnthn moritz_: You must learn to hate beer before you travel there.
TimToady or hate money
moritz_ jnthn: I already do.
pmichaud jnthn: due to quality or cost? ;-) 20:18
jnthn pmichaud: Cost. The quality is pretty OK. ;-)
(And sometimes very OK.)
dalek kudo: bd9fafc | pmichaud++ | src/builtins/match.pir:
Update make() to give a slightly more useful warning when $/ not set.
ingy beer is frickin expensive in .no
jnthn yeah...makes Sweden look like good value. 20:19
ingy :)
20:19 Eevee joined
pmichaud hey, dalek, where's the other commit?!? 20:19
Hmmpf. 20:20
jnthn dalek has selective interests.
moritz_ pmichaud CAN HAZ COMMIT, BUT DALEK EATED IT
20:20 wtm joined
jnthn oh, hmm, I think I forgot to push some stuff I comitted earlier.. 20:21
20:21 wtm left
jnthn oh yes, I did. fail 20:21
Tene pmichaud: any chance you can get export working on ops?
pmichaud Tene: it doesn't work? 20:22
Tene not as of this morning
pmichaud I'm surprised. Seems like it shouldn't make a difference.
can you post a simple example, when you get a chance?
jnthn fwiw I exported a trait_auxiliary and it worked... 20:23
moritz_ it works here
jnthn Tene: Is it a new operator or an overload of an existing one?
moritz_ (for a new operator)
dalek kudo: 52d8f28 | jnthn++ | src/builtins/guts.pir:
First crack at applying custom user traits to blocks. Hopefully we can eventually get rid of the special cases. A spectest fails if we blow up on unknown triats, so for now we'll just warn.
 20:24
kudo: ce81644 | jnthn++ | src/parser/ (2 files):
Start parsing and emitting trait_auxiliary definitions.
jnthn dalek. I totally comitted those in the _opposite_ order.
pmichaud dalek often gets them backwards.
moritz_ nopaste.snit.ch/16550 20:25
Tene Foo.pm: multi sub infix:<Ā±> { ... }
20:25 ejs left
Tene test.pl: use Foo; 1 Ā± 2 20:25
even with an "is export" which is implied by multi 20:26
moritz_ Tene: look at my nopaste above
pmichaud I don't know if Rakudo has "multi implies export" yet.
Tene moritz_: can't conveniently.
TimToady do you have it in a module or a role? 20:27
Tene oh, it works for infix:<xxxxx>
but not for Ā±
moritz_ module Foo { multi sub postfix:<!>(Int $x) is export(:DEFAULT) { ... } };
Tene but works for Ā± if deffined in the same file 20:28
moritz_ Tene: Unicode doesn't work on -e ... 20:29
jnthn Tene: You ain't pre-compiling at all?
Tene moritz_: no -e
files
no pre-compiling
moritz_ ah, I can reproduce it here
pmichaud I can reproduce it also. But it doesn't quite make sense to me.
Tene I'm playing with set ops... :) 20:30
jnthn pmichaud: Maybe it's losing its unicode-ness somewhere during the export?
(Don't ask me where... :-|)
(But it wouldn't entirely surprise me...)
pmichaud I'm wondering if this is more of that weird Parrot bug we once found
jnthn The hashes one? 20:31
pmichaud yes.
jnthn Namespaces are based on hashes.
pmichaud but it's also not working for me with infix:<abc>
jnthn So entirely possible.
pmichaud (which doesn't have any unicodeness)
No applicable candidates found to dispatch to for 'infix:abc'
moritz_ but it parses infix:abc, no? 20:32
pmichaud yes.
jnthn pmichaud: What are you defining? a multi in a class|role|module?
pmichaud yes, in a module. 20:33
jnthn OK
pmichaud gist.github.com/111891
jnthn looking 20:34
pmichaud: Well no wonder 20:35
You wrote an arity 0 multi candidate.
pmichaud multis require a signature? 20:36
okay.
jnthn multi sub infix:<abc>($x, $y) is export(:DEFAULT) { 1 }
# works fine
No, but if you're going to define an infix operator you need to define parameters to tkae the two operands, surely?
pmichaud no, with no signature it should default to (*@_, *%_) 20:37
jnthn ugh
pmichaud ....which might not be properly implemented yet
because I still haven't finished fixing @_ and %_
jnthn rakudo: sub foo { }; say &foo.signature.perl
pmichaud (which are seriously broken)
p6eval rakudo 52d8f2: OUTPUTĀ«:()ā¤Ā»
jnthn Ah
moritz_ rakudo: sub foo { }; foo(3)
pmichaud rakudo: sub foo { @_ }; say &foo.signature.perl
jnthn I think it's jsut that they're not defining a signature.
p6eval rakudo 52d8f2: ( no output )
rakudo 52d8f2: OUTPUTĀ«:(Any *@_)ā¤Ā»
pmichaud see? @_ is broken :-) 20:38
moritz_ that's wrong.
why the Any?
anyway
moritz_ works on tests for exporting operators
jnthn pmichaud: Once those get signatures the MMD should be happy.
pmichaud because subs assume Any on their parameters. I agree it should be Object.
okay, so we know why the infix:<abc> was failing and now works. Now for unicode. 20:39
gist.github.com/111898 20:40
jnthn heh, my first attempt got me
Malformed UTF-8 string 20:41
pmichaud I'm almost certain it's the hash issue again.
oh, perhaps not. Hmmm. 20:42
moritz_ rakudo: sub infix:<a>($a, $b) { $a + $b }; say eval '3 a 5'
p6eval rakudo 52d8f2: OUTPUTĀ«8ā¤Ā»
ruoso often still types $foo.'bla' instead of $foo ~ 'bla'
moritz_ ruoso: you don't write enough Perl 6 ;-)
jnthn pmichaud: OK, now after getting my editor not to save it was the wrong thing it works... :-) 20:44
erm
damm
it gives the same error as your'e seeing
pmichaud "name" => "infix:\x{c2}\x{b1}", 20:45
moritz_ that loos so wrong. 20:46
*looks
pmichaud yes, it's saving the utf-8 version.
jnthn ah. 20:47
pmichaud I don't understand why/how, though.
jnthn pmichaud: Is that in the call to the parser to register the token?
ruoso does rakudo tries to load .pbc files first when in "use"?
pmichaud ruoso: yes. 20:48
ruoso cool
pmichaud jnthn: no
here's the call to register the token (that is in the Foo PIR)
$P45."newtok"(unicode:"infix:\x{b1}", "infix:+" :named("equiv"))
but there are two calls.
moritz_ perl6: say chr(:16('b1'))
p6eval pugs, rakudo 52d8f2: OUTPUTĀ«Ā±ā¤Ā»
..elf 26831: OUTPUTĀ«Unknown rule: rad_numberā¤It needs to be added to ast_handlers.ā¤ at ./elf_h line 2850ā¤Ā»
pmichaud there's a call to add the token to the optable immediately after parsing the signature 20:49
and then of course the generated code has one that occurs at :load :init
jnthn Yes, and one in the PIR for the PBC.
OK, and the one to do it immediately is wrong? 20:50
pmichaud I can't see how. If it was wrong, then it wouldn't work in the case where we're defining a token in the same file.
jnthn pmichaud: If I pre-compile Foo.pm to Foo.pir it works.
pmichaud yes.
but there must still be something else happening here. 20:51
when I do 'use Foo.pm', it should still be invoking the :load :init calls that are generated in the PIR
and that should be adding the token to the optable.
jnthn Right, 'cus they are :init
pmichaud er, "use Foo" (from the .pm)
jnthn pmichaud: It couldn't be that the second time causes a problem because it fails to detect the token is already registered or something? 20:52
pmichaud as far as the optable is concerned, they should be different tokens.
jnthn Wait, that still wouldn't make sense...hmm
pmichaud because they have different names.
pugs_svn r26832 | moritz++ | [t/spec] importing operators 20:54
pmichaud what really confuses me is that it does all seem to work for Ā« and Ā» 20:59
20:59 tombom left
dalek kudo: fe54884 | moritz++ | t/spectest.data:
test importing of operators
 20:59
moritz_ rakudo: sub infix:< ab >($a, $b) { say $b, $a }; 3 ab 4; 21:00
p6eval rakudo fe5488: OUTPUTĀ«43ā¤Ā»
jnthn pmichaud: Those are also in Latin-1 - wonder if there's some confusion somewhere...
moritz_ rakudo: sub infix:Ā« ab Ā»($a, $b) { say $b, $a }; 3 ab 4;
p6eval rakudo fe5488: OUTPUTĀ«43ā¤Ā»
pmichaud \xb1 is latin-1 also, though. 21:01
jnthn ah, true
21:01 ZuLuuuuuu joined
jnthn pmichaud: In perl6.pir we do 21:03
## tell PAST::Var how to encode Perl6Str and Str values
We're not somehow missing this setup (or other encoding related setup) when we invoke PAST::Compiler, are we?
pmichaud that just tells PAST that it needs to escape strings.
it's global, and by the time PAST::Compiler is run from actions.pm it's been long setup. 21:04
jnthn OK.
pmichaud if that was missing, then Perl6Str and Str would end up completely unquoted (and we'd have PIR syntax errors)
jnthn Ah, true.
pmichaud i.e., the pir output would say (bare) infix:+- instead of unicode:"infix:\x{b1}" 21:05
jnthn Right. 21:06
Hmm.
pmichaud Oh!
I might know the problem.
21:07 skids left 21:08 frew|work left
pmichaud a-ha! 21:08
result = 'evalfile'(realfilename, 'lang'=>'Parrot')
it's not reading it as utf-8
(this is inside of 'require', which is called by 'use')
jnthn Ah! 21:09
That'd do it.
pmichaud in particular 21:10
.tailcall compiler.'evalfiles'(filename) 21:11
I think HLLCompiler needs some attributes to indicate what its preferred input encoding and transcoding are
instead of requiring them from the options 21:12
jnthn That could be a neater way, yes.
pmichaud well, 'evalfile' takes a :lang parameter, and other languages might not assume utf8, so we really shouldn't be forcing it.
[particle]- you should look it up in the lang, then 21:13
pmichaud no, just let the compiler itself decide.
for a Perl 6 compiler, it would default to utf8 21:14
actually, each HLLCompiler should probably have a hash of any defaults it wants to set.
ruoso hmmm... interesting... even after having everything compiled down to pbc... it still takes a huge time to start...
pmichaud instead of separate attributes for each possible option.
ruoso: parsing speed is no longer our only bottleneck. 21:15
ruoso but I mean... it still takes a huge time...
oh wait...
the Web.pm files are not compiled
pmichaud gist.github.com/111930 # now works! 21:17
moritz_ pmichaud: I also committed a test file for that 21:18
jnthn pmichaud: Nice!
moritz_ 20 minutes ago or so
ruoso pmichaud, :DEFAULT is now the default in the spec
pmichaud I'll go ahead and push this change without doing a spectest first -- it's really minor.
ruoso (or did someone revert that?)
pmichaud ruoso: yes, :DEFAULT may be the default -- I was just being extra-explicit to reduce number of potential variables. 21:19
jnthn ruoso: Shouldn't need to write :DEFAULT in Rakudo.
ruoso ok then
[particle]- ruoso: he was testing unicode operators there, not export tags, in case you missed it
ruoso ah... 21:20
pmichaud yes, it works without :DEFAULT 21:22
(running a complete spectest now, too)
dalek kudo: c223b06 | pmichaud++ | src/builtins/eval.pir:
When reading Perl 6 files, evalfiles needs to assume utf8 encoding.
 21:23
kudo: f7dbcce | pmichaud++ | :
Merge branch 'master' of [email@hidden.address]
moritz_ wow, compiling perl6 with optimization really needs much memory 21:25
like, 1.3G or so
21:26 exodist left
moritz_ pmichaud: you'll see a plan failure in imported-subs.t 21:27
pmichaud: that's my error, no need to worry ;-) 21:28
pugs_svn r26833 | moritz++ | [t/spec] unfudge and fix imported-subs.t
ruoso hits some weird bug again 21:32
21:32 hercynium left, hercynium joined 21:36 frew joined
moritz_ assign.t segfaults after 166 tests here 21:36
rakudo: say (1..10).reduce: &infix:<+> 21:39
p6eval rakudo f7dbcc: OUTPUTĀ«Null PMC access in find_method()ā¤current instr.: 'parrot;P6object;WHAT' pc 83 (runtime/parrot/library/P6object.pir:114)ā¤Ā»
moritz_ rakudo: say &infix:<+>(3, 4)
p6eval rakudo f7dbcc: OUTPUTĀ«7ā¤Ā»
moritz_ rakudo: say &infix:<+>(3, 4, 5)
p6eval rakudo f7dbcc: OUTPUTĀ«too many arguments passed (3) - 2 params expectedā¤current instr.: 'infix:+' pc 22572 (src/builtins/op.pir:292)ā¤Ā»
moritz_ rakudo: say (1..10).list.reduce: &infix:<+> 21:40
p6eval rakudo f7dbcc: OUTPUTĀ«Null PMC access in find_method()ā¤current instr.: 'parrot;P6object;WHAT' pc 83 (runtime/parrot/library/P6object.pir:114)ā¤Ā»
ruoso getting invoke() not implemented in class 'ResizablePMCArray' when trying to invoke a method in a object... any idea of what it can be? 21:41
ruoso getting that line twice, which is the most strange part
moritz_ I'm sure I've seen that before...
ruoso: are junctions involved? 21:42
ruoso no...
21:42 LylePerl left 21:43 hercynium left, jnthn left, buu left, Maddingue left
moritz_ rakudo: my $p = &infix:<+>; say (1..10).list.reduce: $p; 21:43
p6eval rakudo f7dbcc: OUTPUTĀ«Parameter type check failed; expected something matching Code but got something of type Array() for $expression in call to reduceā¤current instr.: 'die' pc 16712 (src/builtins/control.pir:225)ā¤Ā»
ruoso it's a plain method call, and I have a line with say $action.WHAT just before the failure
moritz_ rakudo: my $p = &infix:<+>; say $p.PARROT 21:44
p6eval rakudo f7dbcc: OUTPUTĀ«Perl6Scalar->Perl6Arrayā¤Ā»
moritz_ rakudo: my $p = &infix:<+>; say $p.perl
p6eval rakudo f7dbcc: OUTPUTĀ«[{ ... }, { ... }, { ... }, { ... }, { ... }, { ... }, { ... }, { ... }]ā¤Ā»
moritz_ rakudo: my $p = &infix:<+>; say $p.map: { .WHAT }
p6eval rakudo f7dbcc: OUTPUTĀ«Code()Code()Code()Code()Code()Code()Code()Code()ā¤Ā»
moritz_ rakudo: my $p = &infix:<+>; say $p.map: { .signature.perl }
p6eval rakudo f7dbcc: OUTPUTĀ«undefundefundefundefundefundefundefundefā¤Ā»
moritz_ "Huh." 21:45
21:45 jnthn joined, Maddingue joined, buu joined, irc.freenode.net sets mode: +o jnthn, ab5tract left
pasteling "ruoso" at 189.97.48.190 pasted "failure when trying to invoke a method...." (2 lines, 133B) at sial.org/pbot/36622 21:47
pmichaud rakudo: say Multi ~~ Code; 21:48
p6eval rakudo f7dbcc: OUTPUTĀ«1ā¤Ā»
pmichaud rakudo: say &infix:<+>.WHAT;
p6eval rakudo f7dbcc: OUTPUTĀ«Null PMC access in find_method()ā¤current instr.: '!dispatch_method' pc 17954 (src/builtins/guts.pir:113)ā¤Ā»
pmichaud rakudo: say &infix:<+> ~~ Code; 21:49
p6eval rakudo f7dbcc: OUTPUTĀ«Null PMC access in find_method()ā¤current instr.: 'parrot;P6object;WHAT' pc 83 (runtime/parrot/library/P6object.pir:114)ā¤Ā»
21:49 iblechbot left
pmichaud I'm guessing that infix:<+> isn't blessed as a Perl6MultiSub 21:50
21:51 amoc joined
frew if I have a class method, (A.bar vs $a.bar) how do I distinguish which is being called in the method? 21:51
moritz_ if self === A { class method } else { instance method } 21:52
but usually you don't need any distinction
ruoso commits his last version of Faz+Yarn and pushes...
frew I was just curious 21:53
21:53 nihiliad left
moritz_ curiosity is good ;-) 21:53
frew and is there a modifier to do that with dispatch?
pmichaud also potentially use defined
ruoso decommute &
21:53 ruoso left
pmichaud if .defined { class method } else { instance method } 21:53
er, backwards
if .defined { instance method } else { class method } 21:54
to do it in a signature, perhaps
moritz_ pmichaud: $.defined, not .defined
frew gotcha
pmichaud method bar($_ where .defined:) { ... }
moritz_ pmichaud: uhm, can there be any more constraints on the invocant?
rakudo: class A { multi method bar ($_ where .defined:) { say "a" } }; A.new.bar 21:55
p6eval rakudo f7dbcc: OUTPUTĀ«aā¤Ā»
moritz_ rakudo: class A { multi method bar ($_ where .defined:) { say "a" } }; A.bar
p6eval rakudo f7dbcc: OUTPUTĀ«Could not locate a method 'bar' to invoke on class 'A'.ā¤current instr.: 'die' pc 16712 (src/builtins/control.pir:225)ā¤Ā»
pmichaud :-)
frew I'm surprised that there's no sugar for class vs instance methods
moritz_ frew: that's because it's usually not needed
frew but that's still multi dispatch, so in my mind, still turtles
yeah, definitely 21:56
moritz_ frew: if one needs it, there's always the option of writing a module for that (for example via a trait)
frew I don't think I've actually used Class methods at work ever
moritz_ frew: I'm sure you did ;-)
frew sure, and I'm all for that. I never said it should be in the language
moritz_ frew: constructors are usually class methods, in some sense
frew ok, those don't count
moritz_ ;-)
frew although I think the config stuff I do probably is 21:57
21:57 payload joined
pmichaud I've used class methods a fair bit, but rarely in situations where I needed them to be completely separate from instance methods. 21:57
There was discussion last week about method, metamethod, and meta to distinguish the invocants
frew interesting
moritz_ "I never metamodel not obsessed with reflection" -- chromatic
pmichaud irclog.perlgeek.de/perl6/2009-05-08#i_1128584 21:58
frew well, the where thing is fine for starters and if it becomes a need using an attribute would prorbaly be fine
that doesn't parse...should it?
moritz_ frew: what?
frew what chromatic said 21:59
it seems like it's missing a word or something...
moritz_ frew: metamodel and met a model
frew: it's an intional pun/misparse
perlgeek.de/blog-en/perl-6/custom-o....writeback
frew got it 22:00
Tene I'm going to implement "is autodispatched" for methods for s1n. 22:02
pmichaud Tene: hmmm?
Tene afk, driving.
lichtkind moritz_: why you dont use macro in the example in >perlgeek.de/blog-en/perl-6/custom-o....writeback ? 22:05
moritz_ Tene: if you have some spare tuits, I'd know a good task: 'class Foo is also' is now called 'augment class Foo', and we're regresssing on a few tests there.
lichtkind: because macros are not yet implemented
22:05 plu joined
lichtkind moritz_: gute antwort :) 22:05
pmichaud I do have to say that changing 'is also' to 'augment' is (to me at least) very non-trivial.
frew plus, isn't defining an operator cleaner than a macro? 22:06
rakudo: say [*] 1..8
p6eval rakudo f7dbcc: OUTPUTĀ«40320ā¤Ā»
moritz_ pmichaud: that's why I ask Tene to do it, and don't try myself ;-)
frew rakudo: say [*] 1..3
p6eval rakudo f7dbcc: OUTPUTĀ«6ā¤Ā»
frew rakudo: say [+] 1..3
p6eval rakudo f7dbcc: OUTPUTĀ«6ā¤Ā»
frew rakudo: say [~] 1..3
p6eval rakudo f7dbcc: OUTPUTĀ«123ā¤Ā»
lichtkind frew: macros are sub called at compiletime
frew ah, got it 22:07
I can see how that could be good I guess
rakudo: say (1..3).reduce({ $^a + $^b })
p6eval rakudo f7dbcc: OUTPUTĀ«6ā¤Ā»
frew rakudo: say (1..3).reduce({ $^a * $^b })
p6eval rakudo f7dbcc: OUTPUTĀ«6ā¤Ā»
frew how does it know the initial value?
moritz_ macros can also have more syntactic freedom than operators 22:08
pmichaud frew: it's (1 + 2) + 3
moritz_ frew: it just takes the first item as initial value (reduce, at least)
pmichaud frew: and (1 + 2) * 3
er,
(1 * 2) * 3
moritz_ rakudo: say [+] ()
p6eval rakudo f7dbcc: OUTPUTĀ«0ā¤Ā»
moritz_ rakudo: say [*] ()
p6eval rakudo f7dbcc: OUTPUTĀ«1ā¤Ā»
moritz_ those are hard-coded for now 22:09
pmichaud oh, for the reduction operators -- yes, they're defined by the spec.
moritz_ the spec says we just call nullary infix:<+>()
which returns the initial value
frew that's cool
moritz_ so that when you define new operators, you can also provide your own initial value
frew rakudo: my $c = 3; say (1..3).reduce({ $^a + $^b * $c-- }) 22:10
p6eval rakudo f7dbcc: OUTPUTĀ«13ā¤Ā»
moritz_ rakudo: say 1 + (2*3) + (2*2)
p6eval rakudo f7dbcc: OUTPUTĀ«11ā¤Ā»
moritz_ rakudo: say 1 + (2*3) + (3*2) 22:11
p6eval rakudo f7dbcc: OUTPUTĀ«13ā¤Ā»
jnthn (backlogging) (operators not Perl 6 multi subs) I figured that'd come when we could define the operators in Perl 6
moritz_ it helps not to mis-type ;-)
frew and if I wanted to set the first one?
pmichaud jnthn: I'm experimenting with registering MultiSub as Multi
moritz_ frew: what do you mean by "set the first one"? 22:12
frew the initial value, sorry
just change the list?
moritz_ yes
frew hmmm
pmichaud but even then it doesn't get us as far as we'd probably like 22:13
multi sub a(Int $x, Int $y) { return 1; }
multi sub a(Str $x, Str $y) { return 'foo'; }
say <1 2 3>.reduce(&a);
Unknown introspection value 'pos_required'
frew the reason was that I was thinking I could refactor this: rafb.net/p/DP9Eqm13.html to use a reduce
and in p5 it actually turned out to be harder to use a reduce than a regular for loop 22:14
but I guess that is really because it has side effects
jnthn (still backlogging) (augment) I've been thinking that gets cleaner with context vars
pmichaud jnthn: (context vars) yes, but we have to be able to define and set them in the grammar. 22:15
moritz_ frew: there's something better for Perl 6, I guess..
pmichaud sure, the checkdigit is easier in perl6
moritz_ rakudo: say (1, 0, 1, 0 Ā»*Ā« 1..4).perl
frew ...
p6eval rakudo f7dbcc: OUTPUTĀ«Multiple Dispatch: No suitable candidate found for 'cmp', with signature 'PP->I'ā¤current instr.: 'parrot;Range;!to_test' pc 9563 (src/classes/Range.pir:250)ā¤Ā»
jnthn pmichaud: Registering multisub (parrot's) as multi probably won't help.
frew moritz_: is that really equiv? 22:16
pmichaud [+] @digits[*] >>*<< 1..@digits
moritz_ rakudo: say (1, 0, 1, 0 Ā»*Ā« (1..4).list).perl
p6eval rakudo f7dbcc: OUTPUTĀ«[1, 0, 1, 0, 0, 0, 0]ā¤Ā»
22:16 lichtkind_ joined
pmichaud my @digits = <1 2 3 4>; say @digits >>*<< 1..@digits; 22:16
rakudo: my @digits = <1 2 3 4>; say @digits >>*<< 1..@digits;
moritz_ uhm... that looks wrong
p6eval rakudo f7dbcc: OUTPUTĀ«Non-dwimmy hyperoperator cannot be used on arrays of different sizes or dimensions.ā¤current instr.: 'die' pc 16712 (src/builtins/control.pir:225)ā¤Ā»
jnthn pmichaud: Either we can move operators to the setting, or we need to do the equivalent of giving the blocks a loadinit kinda thing.
pmichaud rakudo: my @digits = <1 2 3 4>; say @digits[*] >>*<< 1..@digits; 22:17
p6eval rakudo f7dbcc: OUTPUTĀ«Non-dwimmy hyperoperator cannot be used on arrays of different sizes or dimensions.ā¤current instr.: 'die' pc 16712 (src/builtins/control.pir:225)ā¤Ā»
pmichaud rakudo: my @digits = <1 2 3 4>; @digits[*].perl.say
jnthn pmichaud: That calls !TOPERL6MULTISUB
p6eval rakudo f7dbcc: OUTPUTĀ«["1", "2", "3", "4"]ā¤Ā»
moritz_ rakudo: say 1, 2, 3 >>*<< 3, 4, 5
p6eval rakudo f7dbcc: OUTPUTĀ«12945ā¤Ā»
moritz_ rakudo: say 1, 2, 3 >>*<< 3..5
pmichaud jnthn: for the problem I'm trying to resolve, even converting to Perl6MultiSub doesn't help.
p6eval rakudo f7dbcc: OUTPUTĀ«Multiple Dispatch: No suitable candidate found for 'cmp', with signature 'PP->I'ā¤current instr.: 'parrot;Range;!to_test' pc 9563 (src/classes/Range.pir:250)ā¤Ā» 22:18
jnthn pmichaud: I think I missed the problem?
pmichaud jnthn: suppose we want @list.reduce(&infix:<+>)
frew rakudo: say (1..3) >>*<< (3..5)
p6eval rakudo f7dbcc: OUTPUTĀ«3815ā¤Ā»
pmichaud there's not presently a way to determine .arity of &infix:<+>
(or of any Multi)
jnthn rakudo: multi foo() { }; multi foo($x) { }; say &foo.candidates>>.arity 22:19
moritz_ rakudo: my @a = 4, 2, 3, 1; say @a Ā»+Ā« (1..@a)
22:19 LadyLunacy left
p6eval rakudo f7dbcc: OUTPUTĀ«01ā¤Ā» 22:19
rakudo f7dbcc: OUTPUTĀ«5465ā¤Ā»
jnthn pmichaud: Like that.
pmichaud jnthn: in which case .reduce needs to know to call >>.arity instead? 22:20
moritz_ rakudo: my @a = 4, 2, 3, 1; say [+] @a Ā»+Ā« (1..@a)
p6eval rakudo f7dbcc: OUTPUTĀ«20ā¤Ā»
moritz_ frew: see above ;-)
jnthn pmichaud: What does reduce need to know?
pmichaud: The arities of the available candidates?
pmichaud how many arguments to pass to the sub provided as an argument
frew moritz_: it doesn't seem to be doing what I expect, but I'm playing with it
either way, I see what you mean
jnthn pmichaud: Even if you define .arity on multi, there's no one answer so I guess you'd get a junction back. 22:21
pmichaud jnthn: right.
rakudo: say &infix:<+>.candidates>>.arity;
p6eval rakudo f7dbcc: OUTPUTĀ«Null PMC access in find_method()ā¤current instr.: '!dispatch_method' pc 17954 (src/builtins/guts.pir:113)ā¤Ā»
jnthn pmichaud: Thing is, you can't easily know even then because what if the multi has an arity 2 that was incompatible with the types you were offering?
pmichaud jnthn: right, I'm saying it's a place where the spec doesn't quite dwim (for various values of "im") 22:22
i.e., people will expect one thing but not get quite what they expect.
jnthn pmichaud: The other thing about a multi in reduce is that if there's different types in the list things are gonna get fun too.
frew what does @f >>*<< @f[*] do? 22:23
jnthn You won't even dispatch to the same variant each time then...and again, in the worst case, the arity available would depend on types...
frew $x * index_of_$x ?
pmichaud frew: gives you an array of the squares
jnthn frew: Shouldn't it be the same as @f >>*<< @f?
frew I don't know 22:24
moritz_ rakudo: my @a = 1..5; @aĀ».sqrt; say @a.perl
pmichaud my @a = <1 2 3 4>; say [+] @a >>*<< (1..@a)
p6eval rakudo f7dbcc: OUTPUTĀ«[1, 2, 3, 4, 5]ā¤Ā»
moritz_ rakudo: my @a = 1..5; @aĀ».=sqrt; say @a.perl
p6eval rakudo f7dbcc: OUTPUTĀ«[1, 1.4142135623731, 1.73205080756888, 2, 2.23606797749979]ā¤Ā»
pmichaud rakudo: my @a = <1 2 3 4>; say [+] @a >>*<< (1..@a)
p6eval rakudo f7dbcc: OUTPUTĀ«30ā¤Ā»
pmichaud rakudo: my @a = <4 3 2 1>; say [+] @a >>*<< (1..@a) 22:25
p6eval rakudo f7dbcc: OUTPUTĀ«20ā¤Ā»
moritz_ bed &
frew what if I want to do this:
rakudo: [+] @a >>*<< (@a..1)
p6eval rakudo f7dbcc: OUTPUTĀ«Scope not found for PAST::Var '@a' in ā¤current instr.: 'parrot;PCT;HLLCompiler;panic' pc 146 (src/PCT/HLLCompiler.pir:105)ā¤Ā»
pmichaud frew: @a >>*<< (1..@a) produces ($a[0]*1, $a[1]*2, $a[2]*3, $a[3]*4, ...)
rakudo: my @a = <4 3 2 1>; say [+] @a >>*<< (1..@a).reverse 22:26
frew right, what if I want the reverse of that, so $a[0] * @a.elems
p6eval rakudo f7dbcc: OUTPUTĀ«30ā¤Ā»
frew ahh
nice
pmichaud rakudo doesn't implement it yet, but one could also do @a..1 :by(-1)
frew for some reason the check digit isn't working... 22:27
22:27 LadyLunacy joined
frew I need to play with it some thing 22:27
thanks guys
pmichaud have an example input?
frew yeah
773218 should check to 5
and rafb.net/p/NsOkuA12.html 22:28
that's the algorithm we use at work
pmichaud rakudo: my @digits = '773218'.split(''); @digits.perl.say;
p6eval rakudo f7dbcc: OUTPUTĀ«["7", "7", "3", "2", "1", "8"]ā¤Ā»
pmichaud rakudo: my @digits = '773218'.split(''); say [+] @digits >>*<< 1..@digits;
p6eval rakudo f7dbcc: OUTPUTĀ«Non-dwimmy hyperoperator cannot be used on arrays of different sizes or dimensions.ā¤current instr.: 'die' pc 16712 (src/builtins/control.pir:225)ā¤Ā»
jnthn is a little confused by that error
pmichaud rakudo: my @digits = '773218'.split(''); say [+] @digits >>*<< (1..@digits);
p6eval rakudo f7dbcc: OUTPUTĀ«91ā¤Ā» 22:29
frew rakudo: my @digits = '773218'.split(''); say [+] @digits >>*<< (1..@digits) % 10;
p6eval rakudo f7dbcc: OUTPUTĀ«6ā¤Ā»
pmichaud rakudo: my @digits = '773218'.split(''); say [+] @digits.reverse >>*<< (1..@digits);
p6eval rakudo f7dbcc: OUTPUTĀ«105ā¤Ā»
frew rakudo: my @digits = '773218'.split(''); say [+] @digits.reverse >>*<< (1..@digits) % 10;
p6eval rakudo f7dbcc: OUTPUTĀ«6ā¤Ā»
frew should be 5
pmichaud rakudo: my @digits = '773218'.split(''); say [+] @digits >>*<< (1..@digits).reverse; 22:30
p6eval rakudo f7dbcc: OUTPUTĀ«105ā¤Ā»
pmichaud rakudo: my @digits = '773218'.split(''); say ([+] @digits >>*<< (1..@digits).reverse) % 10;
p6eval rakudo f7dbcc: OUTPUTĀ«5ā¤Ā»
frew rakudo: my @digits = '773218'.split(''); say [+] (@digits.reverse) >>*<< (1..@digits) % 10;
p6eval rakudo f7dbcc: OUTPUTĀ«6ā¤Ā»
frew wow
er
rakudo: my @digits = '773218'.split(''); say [+] @digits >>*<< ((1..@digits).reverse) % 10;
p6eval rakudo f7dbcc: OUTPUTĀ«6ā¤Ā»
frew oh
parens
pmichaud you want to %10 the entire sum, not just the last part 22:31
22:31 eternaleye joined
frew right 22:31
rakudo: my @digits = '773218'.split(''); say ([+] @digits.reverse >>*<< (1..@digits)) % 10;
p6eval rakudo f7dbcc: OUTPUTĀ«5ā¤Ā»
frew wow
that is awesome
eternaleye perl6: sub infix:<...> ( Object @starter, Code &generator ) { -> { my Object @tmp = (); for 1 .. &generator.arity() -> $i { @tmp.unshift( @starter[-$i] ); }; return( &generator.( |@tmp) ); }; }; .say for ( 1, 1 ... { $^a + $^b } ); 22:32
p6eval rakudo f7dbcc: OUTPUTĀ«Null PMC access in get_string()ā¤current instr.: 'parrot;Role;perl' pc 255242 (src/gen_metaop.pir:23)ā¤Ā»
..pugs: OUTPUTĀ«1ā¤<SubPointy(<anon>)>ā¤Ā»
..elf 26833: OUTPUTĀ«/home/evalenv/pugs/misc/STD_red/match.rb:141:in `block in to_dump0': undefined method `to_dump0' for true:TrueClass (NoMethodError)ā¤ from /home/evalenv/pugs/misc/STD_red/match.rb:140:in `each'ā¤ from /home/evalenv/pugs/misc/STD_red/match.rb:140:in `map'ā¤ from
../home/evalenv/pugs/mi...
22:32 skids joined
frew rakudo: my @digits = '773218'.split(''); say ([+] @digits.reverse >>*<< (1..@digits)) % 10 == 5; 22:32
p6eval rakudo f7dbcc: OUTPUTĀ«1ā¤Ā»
22:32 lichtkind left
eternaleye perl6: sub infix:<...> ( Object @starter, Code &generator ) { -> { my Object @tmp = (); for 1 .. &generator.arity() -> $i { @tmp.unshift( @starter[-$i] ); }; return( &generator.( |@tmp) ); }; }; .().say for ( 1, 1 ... { $^a + $^b } ); 22:32
p6eval rakudo f7dbcc: OUTPUTĀ«Null PMC access in get_string()ā¤current instr.: 'parrot;Role;perl' pc 255242 (src/gen_metaop.pir:23)ā¤Ā»
..elf 26833: OUTPUTĀ«/home/evalenv/pugs/misc/STD_red/match.rb:141:in `block in to_dump0': undefined method `to_dump0' for true:TrueClass (NoMethodError)ā¤ from /home/evalenv/pugs/misc/STD_red/match.rb:140:in `each'ā¤ from /home/evalenv/pugs/misc/STD_red/match.rb:140:in `map'ā¤ from
../home/evalenv/pugs/mi...
..pugs: OUTPUTĀ«*** Cannot cast from VInt 1 to Pugs.AST.Types.VCode (VCode)ā¤ at /tmp/z2qne2sG1B line 1, column 189-197ā¤Ā»
eternaleye Hm. I've got most of it figured, but how can I get it to return a lazy list? 22:33
Maybe... 22:34
pmichaud eternaleye: note that in order for that to work, infix:<...> would have to have lower precedence than infix:<,>
eternaleye Ah
pmichaud you could parenthesize, though.
eternaleye perl6: sub infix:<...> is looser infix<,> ( Object @starter, Code &generator ) { -> { my Object @tmp = (); for 1 .. &generator.arity() -> $i { @tmp.unshift( @starter[-$i] ); }; return( &generator.( |@tmp) ); }; }; .say for ( 1, 1 ... { $^a + $^b } );
p6eval rakudo f7dbcc: OUTPUTĀ«Malformed routine definition at line 1, near "infix:<..."ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
..pugs: OUTPUTĀ«*** ā¤ Unexpected "infix"ā¤ expecting trait or blockā¤ at /tmp/kApkTwJlws line 1, column 27ā¤Ā»
..elf 26833: OUTPUTĀ«Parse error in: /tmp/xWhYjKJvb1ā¤panic at line 1 column 16 (pos 16): No previous operator visible to adverbial pair ([#<Match:0x81c80b0 @on_str="sub infix:<...> is looser infix<,> ( Object @starter, Code &generator ) { -> { my Object @tmp = (); for 1 .. &generator.arity() -> $i {
.....
eternaleye perl6: sub infix:<...> is looser infix:<,> ( Object @starter, Code &generator ) { -> { my Object @tmp = (); for 1 .. &generator.arity() -> $i { @tmp.unshift( @starter[-$i] ); }; return( &generator.( |@tmp) ); }; }; .say for ( 1, 1 ... { $^a + $^b } );
p6eval rakudo f7dbcc: OUTPUTĀ«Malformed routine definition at line 1, near "infix:<..."ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
..pugs: OUTPUTĀ«*** ā¤ Unexpected "infix"ā¤ expecting trait or blockā¤ at /tmp/6u5aLhNXHk line 1, column 27ā¤Ā»
..elf 26833: OUTPUTĀ«Parse error in: /tmp/Yi989QbUpwā¤panic at line 1 column 16 (pos 16): No previous operator visible to adverbial pair ([#<Match:0x81c7624 @on_str="sub infix:<...> is looser infix:<,> ( Object @starter, Code &generator ) { -> { my Object @tmp = (); for 1 .. &generator.arity() -> $i
..{...
[particle]- eternaleye: use rakudo: instead of perl6: to reduce noise 22:35
eternaleye Okay
pmichaud and rakudo doesn't implement is looser/tighter/etc yet.
rakudo: multi sub a() { 1; }; multi sub a($x) { $x; }; say &a ~~ Callable; 22:36
p6eval rakudo f7dbcc: OUTPUTĀ«0ā¤Ā»
pmichaud rakudo: multi sub a() { 1; }; multi sub a($x) { $x; }; sub b(&x) { &x() }; b(&a); 22:37
jnthn :-( That's a ticket. 22:38
p6eval rakudo f7dbcc: OUTPUTĀ«Parameter type check failed; expected something matching but got something of type Multi() for x in call to bā¤current instr.: 'die' pc 16712 (src/builtins/control.pir:225)ā¤Ā»
jnthn And that's the upshot of the first problem.
pmichaud: If you file, I can fix that tomorrow.
pmichaud files ticket.
jnthn pmichaud: Do you expect we'll move operators to the setting, once is tighter and is looser traits are in place? 22:39
eternaleye rakudo: sub infix:<...> ( Object @starter, Code &generator ) { return( gather { while Bool::True { my Object @tmp = (); for 1 .. &generator.arity() -> $i { @tmp.unshift( @starter[-$i] ); }; take( &generator.( |@tmp) ); } }; ); }; .say for ( 1, 1 ... { $^a + $^b } ); 22:40
p6eval rakudo f7dbcc: OUTPUTĀ«Statement not terminated properly at line 1, near "( gather {"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
pmichaud jnthn: operators don't depend on tighter/looser
but beyond that, I'm getting very concerned about our startup time. 22:41
eternaleye rakudo: sub infix:<...> ( Object @starter, Code &generator ) { return( -> { gather { while Bool::True { my Object @tmp = (); for 1 .. &generator.arity() -> $i { @tmp.unshift( @starter[-$i] ); }; take( &generator.( |@tmp) ); }; }; }; ); }; .say for ( 1, 1 ... { $^a + $^b } );
p6eval rakudo f7dbcc: OUTPUTĀ«Statement not terminated properly at line 1, near "( -> { gat"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
pmichaud (operators do depend on equiv) 22:42
eternaleye rakudo: sub infix:<...> ( Object @starter, Code &generator ) { -> { gather { while Bool::True { my Object @tmp = (); for 1 .. &generator.arity() -> $i { @tmp.unshift( @starter[-$i] ); }; take( &generator.( |@tmp) ); }; }; }; }; .say for ( 1, 1 ... { $^a + $^b } );
p6eval rakudo f7dbcc: OUTPUTĀ«Null PMC access in get_string()ā¤current instr.: 'parrot;Role;perl' pc 255242 (src/gen_metaop.pir:23)ā¤Ā»
eternaleye Well that's something different.
pmichaud anyway, I'd be okay with beginning to move selected operators into the setting to see how it goes.
eternaleye rakudo: sub infix:<...> ( Object @starter, Code &generator ) { -> { gather { while Bool::True { my Object @tmp = (); for 1 .. &generator.arity() -> $i { @tmp.unshift( @starter[-$i] ); }; take( &generator.( |@tmp) ); }; }; }; }; .().say for ( 1, 1 ... { $^a + $^b } );
p6eval rakudo f7dbcc: OUTPUTĀ«Null PMC access in get_string()ā¤current instr.: 'parrot;Role;perl' pc 255242 (src/gen_metaop.pir:23)ā¤Ā»
pmichaud I'm still a little uncertain about how some of the basic operators will fare. We can certainly try them. 22:43
jnthn pmichaud: Are you thinking of keeping grammar-oper.pg?
pmichaud only to define the base precedence tokens. 22:44
jnthn pmichaud: And having that define the precedence?
OK.
pmichaud Either that or we add "is precedence(...)" to the setting.
jnthn *nod*
On startup time...
I have a few ideas for that.
pmichaud I think I somewhat prefer the precedence identifiers to remain in grammar-oper.pg for now. That most closely matches STD.pm's approach, I think. 22:45
frew rakudo: sub foo ( $bar where { $bar ~~ 'test' } ) { say 'test'; }; foo('test')
p6eval rakudo f7dbcc: OUTPUTĀ«testā¤Ā»
frew rakudo: sub foo ( $bar where { $bar ~~ 'test' } ) { say 'test'; }; foo('tes')
p6eval rakudo f7dbcc: OUTPUTĀ«Parameter type check failed; expected something matching all(Any, { ... }) but got something of type Str() for $bar in call to fooā¤current instr.: 'die' pc 16712 (src/builtins/control.pir:225)ā¤Ā»
jnthn 1) Can we collapse all of the loadinits into a single Parrot sub rather than having one for everything?
pmichaud jnthn: we can start to move that direction, yes. Having them separate makes for slightly easier debugging, though.
I might be able to add annotations to simplify it when in loadinit. 22:46
jnthn pmichaud: We could have a "turn it off" flag somewhere.
e.g. it's an opt
Yes, or annotations.
pmichaud annotations would probably be sufficient. Or even :inline('# comment identifying block being loadinited')
frew rakudo: 1 ~~ /(\d)/;
p6eval rakudo f7dbcc: ( no output )
jnthn That would mean we don't have to fall in and out of the runloop but also means we get to unpack a bunch less Sub PMCs on load.
frew rakudo: if 1 ~~ /(\d)/ { say 'true' };
p6eval rakudo f7dbcc: OUTPUTĀ«trueā¤Ā»
jnthn And our .pbc would shrink too.
frew rakudo: if 1 ~~ /(\d{1,8})/ { say 'true' }; 22:47
p6eval rakudo f7dbcc: OUTPUTĀ«Statement not terminated properly at line 1, near "~~ /(\\d{1,"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
frew rakudo: if 1 ~~ /(\d<1-8>)/ { say 'true' };
p6eval rakudo f7dbcc: OUTPUTĀ«Statement not terminated properly at line 1, near "~~ /(\\d<1-"ā¤ā¤current instr.: 'parrot;PGE;Util;die' pc 129 (runtime/parrot/library/PGE/Util.pir:86)ā¤Ā»
pmichaud I had tested combining the loadinits into one previously (with only a few problems) -- I can see about doing it again.
22:47 ZuLuuuuuu left
jnthn Another thing is trying to work out what takes time during the loadinit, since I suspect that's our main source of performance issues. 22:47
Startup performance issues, I mean. 22:48
pmichaud I suspect it's largely that there's a huge amount of work to be done :-|
a profiler would help. :-) 22:49
er, :-(
skids oprofile seems to work OK
jnthn I'd kind of hoped we'd not have to approach the issue too soon, but in the long run I liked to hope that we'd not construct classes and signatures at startup eventually, but instead freeze them as PMCs.
pmichaud I've never quite figured out how that's going to work 22:50
jnthn :immediate can run a chunk of code and then it replaces the sub PMC with the PMC that the sub returned.
so a .const (in theory at least) can get at it. 22:51
pmichaud I understand that part
but how does it maintain references to the things that aren't part of the class?
for example, how does the frozen PMC keep the fact that it 'isa P6object' ?
22:52 justatheory left
jnthn Aye, freezing the parents list for parents outside of the assembly gets tricky. 22:52
pmichaud and it's not just parents -- it might have references to other things
jnthn I don't think Parrot gives us what we need to do this today.
pmichaud that aren't actually loaded/constructed until runtime 22:53
namespace entries come to mind.
jnthn And we might need to be selective in what we can do.
eternaleye I can't figure out how to make the coderef '&generator' be triggered by an access beyond the lazy list's bounds. Is there some way to tie accesses to a list to a callback which extends the list?
pmichaud eternaleye: we don't implement laziness yet.
22:53 justatheory joined
jnthn pmichaud: I suspect a bunch of our expense at the moment is in signature object creation. 22:54
frew is $1 etc implemented in rakudo?
jnthn frew: yes, should be
frew and I still capture with () ?
pmichaud I agree with that. Even there, I'm not sure exactly how to freeze those, since they depend on a (dynamically-created) Signature class.
eternaleye Drat. I can't wait until we have laziness; I'm looking forward to the Coolest Perl6 Operator ( AKA infix:<...> )
pmichaud capture with () yes, but the first one is $0 22:55
jnthn pmichaud: Worse, if you have a type that is a Class, then to freeze the sig you'd need a frozen reference to that.
pmichaud jnthn: exactly.
frew cool
pmichaud jnthn: which is why I've been concerned that the ability to freeze PMCs is far more limited than anyone had expected.
I should rephrase that 22:56
frew should it be giving warnings about uninitialized values because of $0 et al?
pmichaud jnthn: which is why I've been concerned that the ability to freeze ['parrot';'Object'] (and descendents) ....
frew: if you use $0 and it hasn't been set, you get the warning, yes.
frew hmmm 22:57
I'm trying to validate my input in the signature constraint
is that a bad idea?
pmichaud depends on the use case; in general, having pattern-match signature constraints is supposed to be supported, yes. 22:58
frew and being able to get at the $0 from the match?
inside the method?
pmichaud that probably won't work, since it's a different lexical scope. 22:59
frew ah, ok
pmichaud oh, it might be made to work somehow. I'd have to play with it a bit.
jnthn pmichaud: To construct a signature at the moment takes quite a bunch of method calls. And those - including parameter passing - ain't awessomely fast. :-( 23:00
pmichaud jnthn: well, this is one of the reasons I wasn't a big fan of signature objects :-|
(I agree we need them. But they're expensive at startup.)
jnthn Aye. :-| 23:02
pmichaud if we can come up with a constant string-or-other implementation that could be quickly decoded to build the signature, or at least big parts of it, that might be useful. 23:03
jnthn The main things we can't freeze are the type references. 23:04
pmichaud even nicer would be if we could build signatures lazily. :-)
jnthn Being able to build them without a bunch of method calls would help too...
pmichaud well, we could certainly turn those method calls into (private) sub calls. 23:05
but I don't know if that's more or less expensive.
jnthn Avoids the method dispatch cost but not the parameter passing cost.
pmichaud and I don't think the method dispatch cost ought to be that high here. 23:06
looking up a method shouldn't be much more expensive than searching lexical+package+global scopes.
(which is what a sub call would do) 23:07
jnthn Right, especially as they are defined right in the child hash anyway so would be a direct hit on the method table.
pmichaud exactly.
otoh, we could look up the sub once (at the beginning of the loadinit) and then use it thereafter
so there'd be no lookups, just argument passing. 23:08
jnthn ooh, that's true.
pmichaud of course, I guess we could do that with the method as well :-)
jnthn find_method ... ;-)
Yeah!
pmichaud anyway, I suspect method dispatch (in this case single dispatch) is negligable to the argument passing cost
jnthn Me too. 23:09
Argument passing is known to be slow.
pmichaud anyway, in the next few days I'll see about moving the loadinits to one sub and see what we get.
jnthn It can't hurt, and it may make a reasonably sized difference.
pmichaud well, if it fails it can hurt :-| 23:10
it shouldn't fail, but...
TimToady well, setting everything up when you compile CORE.pm should have much the same effect
when you dump the setting, it should have precompiled everything it can 23:11
pmichaud TimToady: agreed in theory; in practice, Parrot doesn't give us a lot of options for storing precompiled data structures, so we end up building them at load-time. 23:12
TimToady well, that's gonna bite
on arity, we can probably take minimum arity from the proto 23:13
pmichaud Agreed. I've been bristling about it for some time but haven't see any relief on the horizon
TimToady also &[x] form is defined to be arity 2 regardless of the underlying operaotr
pmichaud for example, it bugs me that the way we currently build operator precedence tables is by loading the tokens one-at-a-time. 23:15
(at program startup)
TimToady e.g. &[~] is concatenate two things even though ~ itself is list assoc
pmichaud IWBNI Parrot gave me a way to easily serialize data structures in PIR or some other representation.
23:16 eternaleye left
pmichaud we might need to look at adding libyaml or something like that to do that. 23:16
anyway, dinner time here, $wife is calling
bbl
TimToady chow
jnthn pmichaud: At startup, Parrot loops through all of the subs looking for loads or inits or whichever it wants. When it finds one, it calls do_1_sub_pragma which then works out what needs to be done and runs the sub. Since in theory it may be a :immediate it invokes it with Parrot_runops_fromc_args which involves playing the calling conventions game. We then call runops_args which along the way creates a return continuation PMC plus a context.
23:17 eternaleye joined
jnthn ...so doing that once rather than for every single method/sub in the setting might just be cheaper. ;-) 23:17
23:17 mberends left
jnthn (I approximate us currently to do this 150 times from a quick grepping.) 23:18
(Plus once for every class.)
TimToady one begins to see why many Smalltalk implementations just provide a way to dump/restore the whole memory image... 23:19
23:20 donaldh left
TimToady the setting notion isn't so far off of that 23:20
23:20 donaldh joined
jnthn TimToady: Aye, the challenge is how that dump is linked up with objects from other dumps. 23:20
23:21 pmurias left
TimToady just compile, say, CORE.pm, and all its dependencies, then dump everything visible to the lexical scope at YOU_ARE_HERE 23:21
jnthn TimToady: We already to compile, the problem is the data structures rather than the code (e.g. signatures).
TimToady it all ends up in the one dump, because the use made everything in the other dump visible
23:22 payload left
jnthn Sadly, while it's fairly easy to say that, achieving it is slightly harder. ;-) 23:22
TimToady all data structures are visible to the setting's lexical scope, because GLOBAL is just a part of CORE 23:23
details, details :)
jnthn Yeah, but it does back to what pm said - Parrot's support for freezing the data structures at the moment isn't giving us what we want.
Or rather, what we need. 23:24
TimToady but anyway, it's one of the reasons officially installed modules are considered immutable
so we don't have to worry about recompiling them every time
jnthn To allow pre-compilation?
Right.
TimToady and not just pre-comp of themselves, but their incorporation into CORE
the *use* is also precompiled
anyway, that's the ideal 23:25
jnthn Aye.
TimToady the current approach is a bit like not taking invariant code out of the loop when you could
23:26 DemoFreak left
jnthn I agree with you on the ideal. I just know the work involved in getting there. 23:26
And I mostly hold to, it'll be better to do all of Perl 6.0.0 and then optimize, rather than run a subset fast. 23:27
TimToady well, sure, up until the point where people are depending on the bugs for production code
Tene jnthn: did you see my comments earlier about how rakudo is slower when in .HLL 'perl6' ? 23:28
jnthn Tene: I saw them. Do you have much thought about why?
TimToady things in parrot that prevent atomicity of startup will tend to prevent other things as well
Tene none whatsoever.
jnthn Tene: OK. Did you measure how much slower? 23:30
TimToady I wonder how long it would take to (pseudo-)fork a parrot interpreter that was preloaded with the compiler... 23:31
jnthn TimToady: Parrot certainly has issues.
Tene jnthn: nopaste.snit.ch/16544 23:32
'make test' times compared similarly
jnthn TimToady: Much as I would love to dig in and help be a part of solving them sooner, trying to commit to Parrot as I do with Rakudo right now just ain't going to happen. At least, not without burning me out in the process. 23:33
And that'd probably be worse. 23:34
Tene: Looking.
Tene: Ouch. Notable. 23:35
Tene: Think we'll have to just accept it. 23:36
For now anyways.
When I finally stop procrastinating and implementing fun things, I'll eventually get around to the refactor that method dispatch needs, and we'll be less slow on quite a few bits... 23:39
23:39 mizioumt2 left
jnthn Tene: BTW, the trait version of your LolDispatch should run just fine now on Rakudo. :-) 23:40
Impressive cheat though!
Infinoid jnthn: Are there parrot tickets for the things rakudo needs? 23:41
jnthn Infinoid: Yes, some, though we could maybe do with somehow tagging them.
Infinoid: Some are just more general though. Like, calling performance.
Infinoid true, that effects everyone 23:42
23:42 frew left
jnthn I'm hoping that Allison's latest work on this will deliver us performance wins. 23:42
Parrot seems to have got people in the right order. 23:43
We had somebody how came along and optimized the hell out of stuff. Then we had someone come along and work out the architecture for 1.0. :-|
(Yes, I know it's more complex than that really. :-))
23:43 bacek_ joined
TimToady Fire! Aim! Ready! is good XP 23:44
jnthn ŠŸŃ€ŠøŠ²ŠµŃ‚, bacek :-)
jnthn uses XP every day. 23:45
;-)
bacek_ jnthn: с Š“Š¾Š±Ń€Ń‹Š¼ утрŠ¾Š¼! 23:46
jnthn bacek_: ŠšŠ°Šŗ Š“ŠµŠ»Š°? 23:47
.oO( my Russian probably runs out of steam in another sentence or two :-( )
meppl good night 23:49
23:50 meppl left
s1n Tene: 'is autodispatched' sounds like a good idea :) 23:51
pmichaud: ping 23:52
pmichaud s1n: pong
Tene s1n: "method foo(...) is autodispatched { ... }" meaning "There's a 'sub' form of foo that looks up 'self' in the caller and calls the 'foo' method on it" sound about right?
s1n pmichaud: what was wrong with $.method? 23:53
pmichaud s1n: nothing, as it turns out, other than rakudo has a known bug with it.
s1n oh okay
$.method is much better 23:54
pmichaud but from a Perl 6 perspective, $.method is just fine.
bacek_ jnthn: Š¾Ń‚Š»ŠøчŠ½Š¾!
s1n Tene: sounds like a decent idea
Tene s1n: so I'm misremembering you having an objection to $.method ?
s1n pmichaud: yeah, it's much better than self.method, but I like Tene's idea
Tene: well, it's just not fully huffmanized :) 23:55
that idea of yours sounds somewhat reasonable
Tene "All is fair if you predeclare."
pmichaud 23:36 <jnthn> Tene: Think we'll have to just accept it.
s1n Tene: yeah, i'll agree with that
pmichaud Given that, I'm fine with switching Rakudo over to .HLL 'perl6' at your convenience.
Tene pmichaud: OK, I'll commit now then. 23:56
pmichaud be sure to spectest with latest changes :-)
s1n Tene: i saw your LolDispatcher
i wanted to ask about that, it looks pretty amazing, can you explain how that works to me?
pmichaud (or prepare to hear lots of flames if spectest breaks :-)
Tene s1n: sure. lemme move it to use the real syntax... 23:57
pmichaud "No!! Not THE REAL SYNTAX!!!" :-P
dalek kudo: b41e13b | tene++ | perl6.pir:
Migrate Rakudo to use .HLL 'perl6'
Tene pmichaud: spectest is clean for me... this is the only way to get anyone else to actually test it. :)
jnthn pmichaud: It sucks to perform even worse, but putting off something that needs to happen when we have a patch ready to make it happen seems like a potential missed opportunity. 23:58
pmichaud I have to disappear for a bit, I'll run a spectest while gone.
Tene jnthn: there are tests for sub traits now?
pmichaud jnthn: I agree -- I only wanted to get your opinion before pulling the trigger.
jnthn Tene: No, I didn't get to writing the tests yet, mostly because TimToady++ wants to review traits anyway and I've decided to hold fire on further traits stuff until he's done so. 23:59
Tene++ for HLL stuff. :-)
pmichaud Indeed.
jnthn As in, *really* Tene++. :-)
pmichaud Now we should start trying 'hll_map' :-)

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!