»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'p6: say 3;' or rakudo:, std:, or /msg camelia p6: ... | irclog: irc.perl6.org | UTF-8 is our friend!
Set by moritz on 25 December 2014.
00:21 brrt left 00:27 skids joined 00:41 virtualsue left
masak 'night, #perl6 00:42
00:43 virtualsue joined
japhb o/ 01:01
01:08 dj_goku left 01:13 dj_goku joined, dj_goku left, dj_goku joined
adu \o 01:14
or is it "o/" I'm not sure anymore
parametric extensions, I have to remember that 01:15
IMHO, I would have called 6pe "Perl6 Generics" 01:16
01:23 japhb left 01:24 virtualsue left, japhb joined
japhb He is in fact extending the 6model API with parameterization, so that name is apropos. 01:24
dalek kudo-star-daily: 173b35c | coke++ | log/ (9 files):
today (automated commit)
 01:35
[Coke] is finally done with work and exercise and dinner, and hopes he can get the release out before the day ends locally. :) 01:36
skids ++ to whoever made all three Order enums 4 characters long. Little touches like that make code fun. 01:46
01:50 vendethiel left
b2gills gtodd: I couldn't help myself, I added a comment to that SO question 01:52
[Coke] m: Order.perl.say 02:00
camelia rakudo-moar 98db03: OUTPUT«Order␤»
[Coke] m: Order.keys.say
camelia rakudo-moar 98db03: OUTPUT«␤»
raydiak m: Order::.keys.say 02:06
camelia rakudo-moar 98db03: OUTPUT«Less Same More␤»
TimToady skids: yeah, well, it also bothers me that Complex isn't Cpx or some such :) 02:08
adu japhb: parameterized by types? or arbitrary parameterization? 02:09
skids: yeah 02:10
raydiak: hi
raydiak hi there adu
adu raydiak: I got really close to making a new AST
but I forgot to git push
raydiak heh well it's still somewhere, right? 02:11
adu ok merged 02:12
[Coke] bah. we know there's a leap second this year, but the site we pull from isn't updated yet. 02:13
adu raydiak: all of the unit tests passed, so I think I did good
raydiak adu: nice, I'll try to check it out later tonight...did you sort out your syntactic vs semantic conundrums? 02:14
02:14 vendethiel joined
adu raydiak: yes, I made something called TypeOp.new() that is generic enough to represent both 02:15
TypeOp is kind of a misnomer, since it represents types and declarations
TimToady b2gills: actually you've got some superstitious parens in there 02:16
dalek kudo/nom: ad1dc19 | coke++ | docs/ (2 files):
First pass at 2015.01 releasework.
 02:21
raydiak adu: I see...awesome, looks like a good solution to me 02:22
adu raydiak: I think I can use the same AST for C++ 02:23
raydiak adu: did you know you have tabs on a couple lines in lib/C/AST.pm6 ? (and those lines miss trailing semicolons too, if you're worried about those)
b2gills TimToady: It's all of the switching back an forth between Perl 5 and 6 I've been doing
adu raydiak: I am very pedantic about formatting, but I just don't have the time to worry about that 02:24
dalek p: 739d3d7 | coke++ | VERSION:
bump VERSION to 2015.01
adu I also have Rust projects, oh, and my job
[Coke] I changed the release announcement to have real quotes. 02:25
japhb "real quotes"?
[Coke] no, “real quotes” 02:26
skids m: my @s; @s = 0..Inf; for @s { .say; exit if $_ > 4 } 02:27
camelia rakudo-moar 98db03: OUTPUT«0␤1␤2␤3␤4␤5␤»
skids m: my @s; @s <== 0..Inf; for @s { .say; exit if $_ > 4 }
camelia ( no output )
adu raydiak: fixed
dalek kudo/nom: 48083f5 | coke++ | docs/release_guide.pod:
whoops, fix release name.
02:30 regreg left
raydiak adu: method direct-abstract-declarator in lib/C/Parser/Actions.pm6 is giving me "Could not find symbol '&Var'" 02:33
afk 02:34
adu I think I did s/Var/Name/ 02:36
sorry about that
raydiak: fixed, kind of 02:38
japhb [Coke]: I thought someone did a CL way back to convert the quotes the other direction. Perhaps I am mistaken ... 02:43
[Coke] I will not be offended if someone does it again, but I suspect we'll just be lazy and keep copying them. 02:51
02:53 chenryn joined
jercos m: "whatever".reverse 02:58
camelia ( no output )
jercos m: say "whatever".reverse
camelia rakudo-moar 98db03: OUTPUT«whatever␤»
jercos some things never change :p
(.comb.reverse.join works fine, but lol)
02:58 vendethiel left
[Coke] tag only pushes not reported by dalek? 03:00
japhb m: say "whatever".flip 03:06
camelia rakudo-moar 98db03: OUTPUT«revetahw␤»
japhb jercos: Is that ^^ what you wanted? 03:07
dalek kudo/nom: 91aeeb7 | coke++ | tools/build/NQP_REVISION:
[release] bump NQP revision
 03:11
kudo/nom: eabaa68 | coke++ | VERSION:
[release] bump VERSION
[Coke] oh, right, it takes 10 hours to build a release! 03:21
moritz++ for providing a build box. 03:25
03:26 japhb left 03:27 japhb joined
[Coke] one or two roast failure, no problema 03:42
03:47 vendethiel joined 03:52 noganex joined 03:56 noganex_ left
[Coke] AAAAAAAAAAAAAAAAGH 04:00
lost connection to hack during the LONGEST running step here.
04:08 vendethiel left 04:14 anaeem1 joined 04:15 anaeem1 left, anaeem1_ joined 04:19 [Tux] left 04:21 kaleem joined
japhb [Coke]: Not in a screen or tmux session, I take it? 04:25
04:27 [Tux] joined
dalek kudo/nom: bac0b62 | coke++ | docs/announce/2015.01.md:
Pull in actual items from Changelog
 04:28
raydiak adu: now I get "Default constructor for 'Name' only takes named arguments", with the "abstract?!?" being passed in and all :) 04:33
adu damnit 04:35
I need to rethink the abstract versions of the direct declarators
raydiak I see...I assume for now it would do whatever you were trying to just by passing it as :name<abstract?!?> instead? 04:37
adu insert name => and it should work 04:40
but I'm not sure if I want to refactor the grammar, or refactor the actions, I would like to combine the direct and abstract versions if possible 04:41
but I'm just now sure how yet
raydiak should I just put some of the failing lines I run into in a gist for you to check out at your leisure? 04:49
also, did you say __attribute__ should work now, or am I remembering something different? haven't tried it recently 04:50
afk dinner
adu sure 04:52
my email is andjrob(AT)gmail.com
skids m: < foo < 1 >.perl.say # not sure whether this is intended behavior 04:55
camelia rakudo-moar eabaa6: OUTPUT«===SORRY!=== Error while compiling /tmp/XhT4OF25o2␤Unable to parse expression in quote words; couldn't find final '>'␤ ␤at /tmp/XhT4OF25o2:1␤------> t sure whether this is intended behavior⏏<EOL>␤ …»
[Coke] we don't have any releases claimed for the rest of the year. 05:05
05:07 vendethiel joined
dalek kudo/nom: 53c80a1 | coke++ | docs/release_guide.pod:
Claim another release.
 05:08
05:10 BenGoldberg left 05:19 kaleem left
dalek kudo/nom: cd87c47 | coke++ | docs/announce/2015.01.md:
Fix date in announcement.
 05:29
[Coke] so: the announcement file in the repo as of the tag has the wrong date in the text of the announcement; Since I've already pushed the tags and this is a minor issue, I'm not going to re-push the tags. 05:30
the email will have the correct date.
Release done. 05:31
Thanks to everyone who did the actual work of improving perl 6 this month.
jdv79 [Coke]: i forget if i asked already - why the brackets on your nick?
TimToady [Coke]++ 05:33
05:34 muraiki_ left
jdv79 yes of course; [Coke]++ # as always 05:35
[Coke] jdv79: because Coke was already taken 05:47
at the time I started using freenode, it was the common way to get a slightly different nick.
05:49 telex left, ggoebel111111119 joined 05:50 vendethiel left, telex joined, ggoebel111111118 left 06:03 gtodd1 joined 06:04 gtodd left
b2gills m: role Val { has Str $.Str }; sub val ($s) { $s.Numeric but Val($s) }; my $a = val('5e0'); say $a; say $a ~~ Num # fake NumStr 06:06
camelia rakudo-moar 53c80a: OUTPUT«5e0␤True␤»
06:18 agentzh joined
agentzh hi, i'm trying out rakudo today and i'm wondering if it supports automatic error message generation when my grammar parrser fails to parse the user input :) 06:19
i've noticed that Rakudo itself is using "{ $/.CURSOR.panic(...) }" to generate parse fail messages but it does not work with the Perl 6 code input it seems. 06:22
according to S05, i'm supposed to use "$¢" instead but rakudo apparently dislikes it.
S05 seems to talk about the automatic error messages generated by the lexer and the :dba() mechanism but I'm not sure how to fetch such error messages. are they going to "$!"? 06:26
sorry if i've missed anything obvious :) 06:27
moritz yes, they are thrown as exceptions 06:28
agentzh moritz: is it something already implemented in rakudo? 06:30
moritz agentzh: :dba isn't, ~ throwing exceptions is 06:31
agentzh i'm playing bad inputs to interpreters/RPN.pl in perl6-examples and get no error messages.
(but the default "This is not an RPN expression") 06:33
okay, so the automatic exceptions are only generated for the '~' operator? 06:37
is it possible to do something like this? search.cpan.org/~dconway/Regexp-Gra...en_parsing 06:38
06:40 rurban joined 06:42 bjz joined 06:47 bjz left
PerlJam agentzh: I'm not sure which specific part you mean by "this" but the answer is bound to be "yes" :) 06:49
agentzh PerlJam: basically i'm wondering how to provide friendly error messages with position details when the user supplies something bad into my parser implemented in Perl 6. 06:50
ideally rakudo can generate automatic error messages similar to P::RD and P::G's <error> notation.
06:51 kurahaupo1 left
agentzh PerlJam: failing that, i hope i can use my own subroutines to generate such messages with "$¢" 06:51
or something like that.
because i need the current position where the parse fails.
position in the input string. 06:52
but it seems that Rakudo hates "$¢". 06:53
using the latest git master of rakudo from github :)
TimToady because for a while the parser ran much slower with Unicode, I suspect
06:53 kurahaupo joined
TimToady we could probably switch to $¢ one of these days 06:54
but $/.CURSOR is still the cursor that contains the positional information
agentzh TimToady: $/.CURSOR is the NQP only thing?
s/the/an/ 06:55
TimToady m: grammar Foo { token TOP { ... { say $/.CURSOR.from } } }; Foo.parse("abc")
camelia rakudo-moar cd87c4: OUTPUT«0␤»
TimToady m: grammar Foo { token TOP { ... { say $/.CURSOR.to } } }; Foo.parse("abc")
camelia rakudo-moar cd87c4: OUTPUT«No such method 'to' for invocant of type 'Foo'␤ in regex TOP at /tmp/_r9lkTS9uQ:1␤ in method parse at src/gen/m-CORE.setting:14583␤ in block <unit> at /tmp/_r9lkTS9uQ:1␤␤»
TimToady anyway, .CURSOR is available in user code too 06:56
at least in rakudo, for now
agentzh interesting
thanks for your help :) 06:57
TimToady m: grammar Foo { token TOP { ... { say $/.from } } }; Foo.parse("abc")
camelia rakudo-moar cd87c4: OUTPUT«0␤»
TimToady m: grammar Foo { token TOP { ... { say $/.to } } }; Foo.parse("abc")
camelia rakudo-moar cd87c4: OUTPUT«3␤»
TimToady and positional info is in $/ as well
m: grammar Foo { token TOP { ... { say $/.orig } } }; Foo.parse("abc")
camelia rakudo-moar cd87c4: OUTPUT«abc␤»
agentzh cool 06:58
06:58 Rounin joined
agentzh does "..." mean matching anything? 07:00
PerlJam any 3 chars :) 07:01
agentzh oops, silly me :)
PerlJam: thanks for the hint :) 07:02
thought it was some special operator as a whole.
TimToady in normal code it would indicate a stub 07:06
m: sub foo { ... }; foo; 07:07
camelia rakudo-moar cd87c4: OUTPUT«Unhandled exception: Stub code executed␤ at <unknown>:1 (/home/camelia/rakudo-inst-1/languages/perl6/runtime/CORE.setting.moarvm:throw:4294967295)␤ from src/gen/m-CORE.setting:14148 (/home/camelia/rakudo-inst-1/languages/perl6/runtime/CORE.setting.…»
agentzh yep, i was aware of the yada yada operator in that context :)
that's why it caught my eyeballs :)
TimToady sorry I was so misleading :)
agentzh that's completely fine :)
it's the fun to ponder various fun operators in perl :) 07:08
i'm trying to construct some nontrivial compilers in perl 6 or NQP.
NQP looks ugly though it seems much faster.
TimToady the gap should narrow as our optimizers get better 07:09
agentzh great
i've been away from the perl 6 world for several years already. need to reread the synopses :) 07:10
TimToady indeed, welcome back
agentzh thanks :)
TimToady we're fine with people coming and going as they please :) 07:11
agentzh i've been deep into nginx and luajit in the last few years.
TimToady well, we're currently using luajit, or at least bits of it
agentzh but when coming to compiler construction, i really want to use perl and especially perl 6 :)
oh really? which bits?
TimToady uh, you'd have to ask over on #moarvm to get a sensible answer :) 07:12
agentzh okay, got it :)
TimToady though scanning #moarvm logs for luajit would probably be informative
agentzh i have had the dream to build a perl 6 compiler using similar approaches in luajit.
but maybe just a core subset of it, like NQP.
TimToady it seems to work okay, though doesn't really produce optimal code yet 07:13
bit of an impedance mismatch between lua's worldview and Perl's...
agentzh yeah sure.
luajit's vm is specific to lua's semantics.
from ground up.
my $work requires speed so i had to switch from perl to luajit for online logic. 07:14
TimToady anyway, brrt and jnthn are probably the people who know most about the jit, and timotimo
agentzh thanks for the info, thanks :)
TimToady (brrt++ did the jit last year for GSoC) 07:15
agentzh TimToady: I directly listened to your talk at YAPC::NA 2013 for the first time btw :)
TimToady oh, well, that one was kinda pathetic :)
agentzh TimToady: i enjoyed it :) 07:16
TimToady oh, wait, 2013
I was thinking of 201
2014
agentzh i missed 2014 :)
TimToady I probably should've :)
agentzh but i'll try to find my way to 2015 :)
TimToady \o/
Salt Lake this year 07:17
agentzh yep
not far from my place.
TimToady where are you?
agentzh i moved to the bay area in 2012 :)
with my wife.
TimToady ah, well, I'm just down in Mountain View
agentzh got tired of Beijing :)
wow, very near.
TimToady we finally got to Beijing last year
agentzh i live in Millbrae.
oh really? 07:18
TimToady food was wonderful
agentzh heh
yep i miss the food there too :)
TimToady: just out of curiosity, are you still coding in C these days? ;) 07:19
TimToady occasionally, but not often
agentzh i see :)
TimToady worked on the NFA engine some lately 07:20
agentzh wow
i was working on an NFA-based regex engine two years ago: github.com/openresty/sregex
the goal is to support streaming parsing like in an nginx output filter. 07:21
but it is sloooow
meant to optimize it at some point.
TimToady yeah, we'd like ours to run faster too...
agentzh :)
TimToady though it's really only about 10% of the parser overhead currently 07:22
agentzh for rakudo?
TimToady nod
agentzh got it.
for my use case, it can be 90% :)
TimToady 'course we're parsing a fairly complicated languge
agentzh indeed.
TimToady the regex are all integrated in p6, so it could well be 90% for some tasks 07:23
agentzh i'm really glad to see rakudo is much faster than a few years ago :)
yes, sure.
TimToady those are the ones it might be worthwhile to do a DFA for
agentzh yeah
do you think it's possible to do submatch capturing with a DFA without going backward in the input stream? 07:24
TimToady for P6 lexing it probably only makes sense to cache the first set of NFA states, since the possibilities drop rapidly after the first char is recognized
oh, yes, it's certainly possible
agentzh wow.
TimToady but there's a price, of course
agentzh the re2 author has to reverse the regex and match backward in the input string to get the *beginning* pos of each submatch captures AFAIK. 07:25
TimToady if you look up "tagged dfa" you should find TRE
which is a tagged dfa implementation
agentzh okay, so TRE is the way to go?
TimToady not exactly
agentzh cool, i'll have a closer look at it.
i was just aware of its existence :) 07:26
TimToady when I tried it in the early days of prototyping STD, I found it allocated 17MB on the stack for every token
agentzh ouch!
TimToady which doesn't really work out
so it's a bit of a toy in my mind
agentzh yes, i have some vague impression of that while reading the perl 6 meetup notes.
TimToady not really suited for very wide alternations
and token recognition is nothing if not wide
agentzh indeed. 07:27
TimToady regex always find some way to reach out and bite you :) 07:28
agentzh for my use case, i just hope to support a good subset of perl 5 regexes and can make the regex engine yield at any point in the input stream and resume later without going back.
so that we can filter out data by chunks. 07:29
one chunk at a time and never buffer anything more than needed.
TimToady: yes, i already got the pain while building sregex.
TimToady well, at least we support p5 regexen already in p6, for the most part
agentzh TimToady: yep. 07:30
TimToady: my toy was built for the nginx world :)
TimToady and we do concurrency now too, so one could run the parser in a different thread and not worry about continuations too much
agentzh TimToady: the state of art is to use ragel to generate nginx C modules to do the response body processing.
TimToady: but ragel is much more painful than perl regexes :)
TimToady: is the "thread" OS-like threads or just lightweight threads like coroutines? 07:31
07:31 rindolf joined
TimToady yeah, I keep seeing references to nginx, but haven't had any direct exposure 07:31
it's more like Go's promises and channels, with green threads over a built-in scheduler that maps to worker threads 07:32
kinda the best of both worlds
agentzh i see. so it's collaborative?
TimToady yes, but fairly transparently 07:33
agentzh cool
i built a coroutine scheduler for lua in the nginx event model :)
it's still single threaded though. 07:34
TimToady there's also a whole reactive programming system, called 'supplies'
agentzh never want the pain dealing with multiple threads :)
url? 07:35
leedo ls
whoops
TimToady design.perl6.org/S17.html
agentzh cool thanks :) 07:36
having to do all the synopsis recap :)
TimToady: nice to talk to you! gotta get some sleep. G'night! 07:37
TimToady night!
07:38 vendethiel joined 07:42 FROGGS joined, agentzh left 07:43 virtualsue joined 07:44 rindolf left 07:46 bjz joined, fhelmberger joined, kaare_ left 07:47 bjz left 07:49 kaare_ joined 07:54 abraxxa joined 07:55 bjz joined 07:56 xfix joined
moritz [Coke]++ # release 07:59
08:05 prime left 08:10 darutoko joined
dalek pan style="color: #395be5">perl6-examples: 05669d7 | (Yichun Zhang (agentzh))++ | interpreters/RPN.pl:
improvements and fixes in RPN.pl

  * use .list to fix "for" iteration over an array reference.
  * properly handle the case when no input string argument is specified.
  * suppress the "Useless declaration of a has-scoped method" warnings
   from the latest rakudo.		 08:11
pan style="color: #395be5">perl6-examples: 9cea16f | FROGGS++ | interpreters/RPN.pl:
Merge pull request #17 from agentzh/RPN-fixes

improvements and fixes in RPN.pl
08:12 jmark joined 08:14 prime joined 08:22 zakharyas joined, vendethiel left 08:33 xinming left 08:34 xinming joined 08:45 fhelmberger_ joined, mr-foobar left 08:46 vendethiel joined, rurban1 joined 08:47 fhelmberger left 08:55 chenryn left 09:02 kaleem joined 09:08 vendethiel left 09:13 virtualsue_ joined 09:14 virtualsue left, virtualsue_ is now known as virtualsue 09:20 vendethiel joined 09:21 xfix left 09:23 virtualsue_ joined, virtualsue left, virtualsue_ is now known as virtualsue 09:24 Kristien joined 09:29 molaf__ joined 09:30 fhelmberger_ left, kjs__ joined
jnthn morning, #perl6 09:30
Kristien hola
jnthn [Coke]++ # 2015.01 release
nwc10 good *, jnthn 09:31
09:32 molaf_ left 09:34 pecastro joined, Sir_Ragnarok joined
lizmat [Coke]++ # 2015.01 release and everything else! 09:41
09:42 vendethiel left
Kristien do Perl 6' text I/O APIs always assume UTF-8 by default? 09:48
09:48 vendethiel joined
FROGGS Kristien: aye 09:49
moritz Kristien: the specs say that they should do some limited form of encoding autodetection that detects UTF-{16,32}, and falls back to UTF-8 09:50
Kristien: rakudo doesn't do that, and always assumes UTF-8
Kristien OK
So it doesn't look at the system locale, which is good. 09:51
09:55 denis_boyun_ joined
Kristien m: say (.codes, .chars) given "a\x030A" 09:59
camelia rakudo-moar cd87c4: OUTPUT«2 2␤»
Kristien interesting
I expected 2 1, since U+030A is a combining-character.
moritz grapheme level is not yet implemented 10:00
arnsholt That's graphemes, not characters
Kristien oh ok 10:01
10:03 virtualsue left
FROGGS arnsholt: well, .chars will be about graphemes one day 10:03
10:03 fhelmberger joined
arnsholt Ah. I thought that was .graphs =) 10:05
10:07 fhelmberger left 10:08 fhelmberger joined
Kristien can arguments default to $_? 10:10
10:11 rindolf joined, vendethiel left
jnthn Not easily; we don't really do "takes $_ by default" in Perl 6. Those situations tend to be handled by the .foo or .=foo syntax 10:14
lizmat m: sub a($a = CALLER::<$_>) { say $a }; $_ = "foo"; a # doesn't work :-(
camelia rakudo-moar cd87c4: OUTPUT«Nil␤»
FROGGS m: sub foo ($a = OUTER::DYNAMIC::<$_>) { say $a; say OUTER::DYNAMIC::<$_> }; $_ = 42; foo() # wonders why
camelia rakudo-moar cd87c4: OUTPUT«Nil␤42␤»
FROGGS err
m: sub foo ($a = CALLER::DYNAMIC::<$_>) { say $a; say CALLER::DYNAMIC::<$_> }; $_ = 42; foo() 10:15
camelia rakudo-moar cd87c4: OUTPUT«Nil␤42␤»
jnthn m: given [1..10] { .=grep * %% 2; .=map 'a' x *; .say }
camelia rakudo-moar cd87c4: OUTPUT«===SORRY!=== Error while compiling /tmp/R_MD_K6055␤Two terms in a row␤at /tmp/R_MD_K6055:1␤------> given [1..10] { .=grep * %% ⏏2; .=map 'a' x *; .say }␤ expecting any of:␤ infix stopper␤ …»
jnthn m: given [1..10] { .=grep: * %% 2; .=map: 'a' x *; .say }
camelia rakudo-moar cd87c4: OUTPUT«aa aaaa aaaaaa aaaaaaaa aaaaaaaaaa␤»
Kristien jnthn nic
nice*
10:15 rurban1 left
FROGGS .oO( jnicethn? ) 10:15
jnthn lizmat: Probably 'cus the thunking of the default gets in the way. 10:16
lizmat ah, more levels you mean ?
jnthn Yeah
And here's a nice bug: 10:17
lizmat m: sub a($a = CALLER::CALLER::<$_>) { say $a }; $_ = "foo"; a # indeed
camelia rakudo-moar cd87c4: OUTPUT«foo␤»
jnthn m: sub foo($a = CALLER::CALLER::<$_>) { say $a }; given 42 { foo(1); foo() }
camelia rakudo-moar cd87c4: OUTPUT«1␤No such method 'dynamic' for invocant of type 'Int'␤ in method at_key at src/gen/m-CORE.setting:11131␤ in sub postcircumfix:<{ }> at src/gen/m-CORE.setting:3090␤ in sub foo at /tmp/MgALuGqWgZ:1␤ in block <unit> at /tmp/MgALuGqWgZ:1␤␤»
lizmat hehe 10:18
rakudobuggable ?
jnthn Yeah
lizmat done: #123660 10:22
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=123660
jnthn Thanks 10:23
dalek kudo/nom: 2b2de11 | lizmat++ | docs/release_guide.pod:
Claim February's release
 10:32
kudo/nom: a05dc90 | lizmat++ | src/core/ (2 files):
Make EnumMap a first class citizen, skids++

This reverts commit a8e384cfa1d1d5414d367416a7aca227a5729739.		 10:33
10:41 virtualsue joined
jnthn lizmat: If you've chance, you may also try a build/test run on OSX with MoarVM master; I bumped the default optimization level yesterday. 10:46
lizmat will do
jnthn lizmat: Tested it with gcc and clang, though only on Linux.
10:46 vendethiel joined
jnthn lizmat: Hopefully clang on OSX is equally fine 10:46
lizmat: I guess you build with defaults, so this might also get you a time saving on builds/spectests :)
lizmat yeah, let's see :-) 10:47
jnthn Measuring the GCC build with callgrind showed 10% less instructions at Rakudo startup, for example.
lizmat building now
10:48 andreoss joined
andreoss m: my $n = 2; $n = ( 2 * $n ) % 10_000_000_000 for 1 .. 7830456; say $n; 10:49
lizmat jnthn: definitely not warning free
camelia rakudo-moar cd87c4: OUTPUT«(timeout)» 10:50
jnthn lizmat: Oh?
lizmat: What warnings?
andreoss with native int $n it's even slower
why?
lizmat jnthn: a *lot* in unicode.c
jnthn wtf 10:51
It was just two about pthread_yield for me yesterday...
lizmat src/strings/unicode.c:41600:20: warning: initializing 'unsigned char *' with an
expression of type 'char *' converts between pointers to integer types
with different sign [-Wpointer-sign]
unsigned char *bname = MVM_string_ascii_encode(tc, block, &size);
jnthn I fixed that? :/
lizmat is it not a generated file ?
the other one: 10:52
src/6model/serialization.c:327:34: warning: '&' within '|'
[-Wbitwise-op-parentheses]
buffer[offset++] = value & 0x7F | 0x80;
~~~~~~^~~~~~ ~
src/6model/serialization.c:327:34: note: place parentheses around the '&'
expression to silence this warning
jnthn lizmat: Yeah, I thought I fixed the thing generating it too
lizmat I thought I did too, at one time :-(
jnthn m: say nqp::isbig_I(10_000_000_000)
camelia rakudo-moar cd87c4: OUTPUT«1␤»
masak good antenoon, #perl6 10:53
jnthn andreoss: 'cus it considers - maybe wrongly - that literal too big for a native int, so the % is being done with Int math instead, so the native is getting boxed all the time.
andreoss i guess it fits in int64 10:54
jnthn o/ masak
andreoss: Yeah, it would, thus my "maybe wrongly" :)
andreoss: Workaround is to assign that value into an int variable, I guess.
10:56 xfix joined
andreoss jnthn: yep. this way it is faster 10:56
still not as fast as p5 (20 times faster) 10:57
jnthn What if you write it with a loop instead of a for? 10:58
andreoss now it's 13 times slower 11:01
m: loop (my int $x; $x < 10; $x++) {}
camelia rakudo-moar cd87c4: OUTPUT«Cannot assign to an immutable value␤ in sub postfix:<++> at src/gen/m-CORE.setting:5080␤ in block <unit> at /tmp/OXPvF1trKs:1␤␤»
lizmat andreoss: $x = $x + 1, natives don't handle ++ yet
jnthn is working on that right now, afaik :-) 11:02
jnthn I get 3.03s with it written as a loop, vs 0.82s in Perl 5
xfix Hm, interesting. Perl 5 just added /n modifier to regexes which makes parentheses not capture. Why not just add Perl 6 regexes?
jnthn m: say 3.03 / 0.82
camelia rakudo-moar cd87c4: OUTPUT«3.695122␤»
jnthn So rather closer here.
Still surprised we don't do better there, though.
Ah, it's allocating somewhere, for some reason... 11:03
Yeah, I think it's an optimizer bug 11:04
'cus if I re-write it as 11:05
my int $n = 2; my int $lim = 10_000_000_000; loop (my int $i = 0; $i <= 7830456; $i = $i + 1) { $n = 2 * $n; $n = $n % $lim }; say $n;
Then I get the answer in 0.51, which is faster than Perl 5.
Especially when 0.25s of that is startup. 11:06
andreoss i run this code ideone.com/AOapBS
11:07 BenGoldberg joined
andreoss and it's still faster in perl5 and in perl6 with ints 11:07
*than in perl6
jnthn Looks close enough to what I ran. 11:08
andreoss % time perl6 prob097.pl 11:09
real0m9.281s
user0m8.959s
sys0m0.311s
% time perl prob097.pl5
jnthn Yeah but...which perl6?
andreoss real0m2.056s
user0m2.052s
sys0m0.003s
rakudo
jnthn Yes, but which backend?
andreoss moar
jnthn OK, and on x64? 11:10
andreoss yes
jnthn Hm, then you should have JIT
11:11 BenGoldberg left
andreoss also without C-style for in perl5 it reduces to 1.35s 11:11
jnthn Did you run exactly the code I pasted? 11:13
andreoss I see 11:16
dalek line-Perl5: 8fc9014 | (Stefan Seifert)++ | lib/Inline/Perl5.pm6:
Pass on arguments to the P5 constructor
 11:17
jnthn If so, and it's still slower, I've little idea what's going on, though running with --profile may be informative.
andreoss jnthn: it's faster, thanks
11:22 andreoss left
lizmat Files=936, Tests=34394, 177 wallclock secs ( 9.10 usr 3.15 sys + 1050.45 cusr 138.03 csys = 1200.73 CPU) # jnthn, wow! 11:26
Files=936, Tests=34394, 229 wallclock secs ( 9.33 usr 2.86 sys + 1228.85 cusr 139.25 csys = 1380.29 CPU) # before switch to master 11:27
jnthn Damn, that's quite a win. :) 11:28
lizmat startup time for me went from ~ .29 to .23
that's about 20% better :-)
jnthn And no new test failures?
lizmat none, so I think a bump is in order 11:29
panda also bootstraps fine on it
shall I bump ?
jnthn Go for it; will get it wider testing.
lizmat or wait for you to fix the unicode warnings ?
jnthn Can wait too 11:30
We've got a whole month :)
lizmat ok 11:31
dalek p: a05ddba | lizmat++ | tools/build/MOAR_REVISION:
Bump Moar revision to get better optimizations
 11:33
nwc10 sorry to be a bit laggy, but what particular bit of awesome was responsible for that?
lizmat default opt level of Moar, afaik 11:34
11:34 Kristien left
dalek p-js: 41e92e7 | (Pawel Murias)++ | src/vm/js/nqp-runtime/sixmodel.js:
Set the correct invocant when calling $call on a method during method call.
 11:34
p-js: fcf4d5f | (Pawel Murias)++ | t/nqp/18-associative.t:
Add a test for nqp::hash(...) to 18-associative.t.
11:35 pmurias joined
pmurias people seem to really love benchmarking arithmetic 11:36
vendethiel no idea why :-) 11:37
jnthn nwc10: Bumped default opt level to O2, from O1
We used to get into problems with that
Maybe the warnings fixes helped nail the issues, maybe it was other improvements (like being stricter about various things when doing the BE stuff) 11:38
rurban Released parrot-7.0.1 11:40
lizmat rurban++
dalek kudo/nom: 3e97dc9 | lizmat++ | tools/build/NQP_REVISION:
Bump NQP to get better optimized MoarVM
 11:43
nine What would be the correct syntax for something like my $handle_tag = sub($tag, $num, *@attrs) { ... }? 11:55
jnthn m: my $handle_tag = sub($tag, $num, *@attrs) { ... } 11:59
camelia rakudo-moar a05dc9: OUTPUT«===SORRY!=== Error while compiling /tmp/yv6tWcNIO9␤Variable '$tag' is not declared␤at /tmp/yv6tWcNIO9:1␤------> my $handle_tag = sub($tag⏏, $num, *@attrs) { ... }␤ expecting any of:␤ postfix␤»
jnthn m: my $handle_tag = sub ($tag, $num, *@attrs) { ... }
camelia ( no output )
jnthn Need the space there; foo(.... is always a function call.
nine Oh, thanks!
12:05 anaeem1__ joined 12:07 anaeem1_ left 12:08 andreoss joined 12:12 zakharyas left
lizmat ok, so I just built rakudo on parrot with 7.0.1, and it build and spectested clean 12:15
so I guess a bump there is in order as well, right rurban?
rurban yes, please 12:17
there are two improvement branches for nqp and rakudo open, called parrot-rpa 12:18
this is optional if you want better speed on parrot 12:19
lizmat one thing at a time :-)
moritz pmichaud didn't see too happy about removing qrpa
dalek p: 4589d81 | lizmat++ | tools/build/PARROT_REVISION:
Bump Parrot to 7.0.1
 12:21
rurban yes, what can I do. 12:22
I took his qrpa code and improved it 12:23
12:24 frew left 12:26 tinyblak joined 12:28 abraxxa left 12:29 Kristien joined, rurban1 joined
lizmat hmmm.... that last NQP bump did not work out for parrot: it doesn't build 12:31
with rakudo now
rurban: could you verify that RELEASE_7_0_1 is the correct tag ? 12:32
lizmat is a complete parrot noob
moritz lizmat: yes, that's correct 12:36
12:41 yeahnoob joined
lizmat ok, seems to build now 12:46
guess I did something wrong
12:48 rindolf left 12:49 rindolf joined
psch hi #perl6 \o 12:51
raiph++ # jvm interop testing
rurban1 lizmat: yes, RELEASE_7_0_1 is good.
psch apparently the old marshalling (for longnames and not-overloaded methods) is still broken
and it's annoying to fix, because i have to write bytecode that loops and unboxes... 12:52
lizmat rurban1: spectesting now
rurban1 I am working on an 10% faster parrot right now, with an improved StringBuilder. benchmark at github.com/parrot/parrot-bench#parrot-bench
psch .tell raiph i apparently missed fixing marshalling of List-y types for non-overloaded methods. that's the error you're seeing with "method/update/([B)V" 12:53
yoleaux psch: I'll pass your message to raiph.
12:57 skids left, anaeem1__ left
lizmat Files=876, Tests=33426, 451 wallclock secs (10.51 usr 3.55 sys + 2916.91 cusr 410.29 csys = 3341.26 CPU) # before bump 13:00
13:00 Kristien left
lizmat Files=876, Tests=33426, 474 wallclock secs (10.78 usr 3.76 sys + 3023.09 cusr 424.04 csys = 3461.67 CPU) # after bump for parrot 13:00
could be noise, but the tendency seems towards being slower
dalek kudo/nom: 912a7fa | lizmat++ | tools/build/NQP_REVISION:
Bump NQP for parrot 7.0.1
 13:01
13:05 tinyblak left, tinyblak joined, kaare_ left 13:09 tinyblak_ joined 13:12 tinyblak left 13:13 zakharyas joined 13:15 rindolf left 13:31 H2O1 joined 13:32 H2O1 left
jmark Hi, all! I'm skimming the official perl6 doc and wondering if "sized low-level types" (will) lead to any optimization comparable to cython? 13:33
moritz jmark: low-level types generally are much faster
jmark: though there aren't any optimizations for the sized ones in particular, I think 13:34
jmark: so far they are mostly implemented to make it easier to interface with libraries written in C
jmark I would be satisfied when I do just some number crunching with ints and floats I could get near the speed of light. 13:37
moritz we're working on it :-) 13:38
jmark Without using C::Inline and such.
nice!
[Coke] Util++ rurban++ # parrot releases 13:42
FROGGS moritz: I was under the impression that operations with native types are jitted well 13:44
jnthn Within the next month we should have natively typed arrays also, which will make the native types stuff a whole lot more useful.
FROGGS and we'll have write accessors on native attributes /o/ 13:45
nwc10 and a pony? 13:46
jnthn Well, if I catch another winter cold I might be a little horse again...
nwc10 if so, be careful, lest you end up in Tesco 13:48
13:49 andreoss left
[Coke] nwc10: the pony's dead, you know that. 13:55
nwc10 yes, very true that.
moritz notices that "the pony's dead" is a wonderful text for this tune: de.wikipedia.org/wiki/Der_Hahn_ist_...28Kanon%29 13:56
FROGGS *g* 13:57
[Coke] moritz: Den Hunn ass fräkt 14:05
jmark Just read something about "PDL support". Is this what you mean by "natively typed arrays"? 14:06
14:07 Kristien joined
moritz jmark: it's a step in that direction 14:08
jmark: when people read "PDL", they expect to be able to multiple matrices, for example
14:08 anaeem1_ joined
moritz jmark: that's not part of the core language; storing a matrix as compact as PDL does is, though 14:08
jmark moritz: why not implementing some basic LinAlg, too? 14:09
One operator more or less wouldn't hurt perl6, right? ;) 14:10
moritz jmark: why not do it in a library? :-)
14:11 anaeem1_ left
Kristien jmark: Noo! It would invalidate everyone's Period Table of the Operators posters! 14:11
moritz jmark: doing linear algebra efficiently doens't quite require the same skill set as implementing a compiler
jmark: so I'd leave it to other folks
also, scope creep
14:11 anaeem1_ joined
Kristien there are already fast linear algebra libraries, like Eigen. use them 14:11
jmark Tooo sad: Just checked ... ** is still free yet, isn't it? 14:12
Kristien: But I want to do some LinAlg in perl6. Eigen is big and clumpsy.
-p 14:13
Kristien I mean an Eigen binding for Perl 6 :P
psch m: say &infix:<**>.perl
camelia rakudo-moar 912a7f: OUTPUT«sub infix:<**> ($?, $?) { #`(Sub+{<anon>}+{Precedence}|51016560) ... }␤»
masak I agree. Perl 6 library + custom operators + fast bindings to something exist sounds just right.
Kristien or BLAS
jmark Alright, you convinced me. 14:14
moritz uhm, Eigen is quite small in comparison to many other linear algebra libraries
14:14 ptc_p6 joined
Kristien SciPy uses BLAS IIRC 14:14
moritz it just doesn't compile very fast, because it does template magic
14:14 ptc_p6 left
moritz (if we're talking about the same library here :-) 14:14
Kristien yeah that one 14:15
jmark moritz: You can bypass that by precompiling specializations.
Kristien it uses expression templates to simplify expressions at compile-time
moritz jmark: but then there's not too much point in using the library, no?
Kristien expression templates is one of its features, not the only one
moritz TBH I don't remember too much about it; just that it was less of a fuss to use as other libraries, and had an active and friendly community 14:16
jmark jmark: hmm? But what when I just need to do some bigger float matrix multiplications but being still in the compile run cycle of my application.
moritz hasn't done too much linalg lately
14:19 anaeem1_ left 14:20 anaeem1 joined
jmark Could anyone recommend me an elaborate article/howto about "roles" besides the official doc and chromatics "The Why of Perl Roles"? 14:21
14:22 anaeem1 left
masak jmark: random ddg search: blogs.perl.org/users/ovid/2011/01/h...roles.html blogs.perl.org/users/ovid/2014/12/u...asses.html strangelyconsistent.org/blog/ovid-i...re-awesome 14:22
I think those might net you a nonzero understanding delta.
er, a positive one, even ;)
moritz jmark: the academic research calls them "traits", might also help searching in that direction 14:23
masak jmark: that reminds me: github.com/perl6/specs/issues/80 will be interesting reading to you, too. it has links to the two relevant papers. 14:24
14:27 ptc_p6 joined
Kristien doc.perl6.org/language/modules mentions only GitHub 14:28
does that mean you can't publish modules without putting them on GitHub?
14:28 skids joined
moritz Kristien: currently our infrastructure is very github-centric 14:29
Kristien: contributions to diversify that are very welcome 14:30
Kristien OK.
14:30 kaleem left 14:31 Kristien left
jnthn I guess any Git repo will actually do it? 14:32
moritz jnthn: I'm afraid not 14:33
jnthn: there are some parts (like generating the module list) that check for github URLs
jnthn Oh
moritz jnthn: though I'm not entirely sure
skids Is there a standard way yet to tell panda "use this commit for vX.Y.Z"? 14:42
14:44 clkao left, clkao joined 14:48 rurban left 14:50 Rounin left 14:55 booly-yam-6137__ left
moritz skids: have you tried to create a tag in git? 14:55
14:56 booly-yam-1010 joined
PerlJam Does panda have a way to checkout a tag/ 14:57
?
moritz dunno
PerlJam doubts
moritz but skids' question sounded as if it has a way to checkout a commit
and if it does, it'll probably support tags too 14:58
15:06 PZt joined
FROGGS AFAIK panda does not support installing a certain commit/tag 15:08
but, one can `panda look Foo::Bar`, then do a git checkout to the desired comit/tag, and then do `panda install .` 15:09
15:11 telex left 15:12 telex joined 15:13 tinyblak_ left 15:14 kurahaupo left 15:19 ptc_p6 left, ptc_p6 joined 15:22 kaleem joined
PerlJam Here's a diff (untested) that will support adding @tag (or other committish things) to the end of the github URLs -- gist.github.com/perlpilot/d3001269e80f26915a27 15:22
15:23 kaleem left
PerlJam \w probably needs to change though since that won't match . 15:24
15:27 Kristien joined 15:29 kurahaupo joined 15:30 adu left
PerlJam updated with a better committish matcher 15:32
15:33 rurban1 left
hoelzro I was playing around with type coercion (ex. MyType('')), and I noticed that if I implement postcircumfix:<( )> for MyType, that method receives a Capture as its single argument, rather than the value being coerced. Is there a reason for this? 15:41
moritz \S+ as a first approximation?
hoelzro: something bootstrappy, iirc 15:42
hoelzro hmm
moritz: thanks for the input
jnthn The bootstrappy reason is gone-ish these days, though. 15:43
hoelzro I'm going to try building/spectesting with passing the value itself, rather than a capture 15:44
for experimental purposes
jnthn I'm a bit bothered about coercion stuff overall, fwiw.
We have a parsing problem or even ambiguity on coercion type syntax vs. actually performing a coercion. 15:45
I guess we can say that Int(Str) is a coercion type specification because the inner thing and outer thingy are both types. 15:46
And Int($foo) is a coercion, not a function call, because we don't see a type on the inside of it. 15:47
But then we try and overload invocation on the type object to mean coercion, except if that's not defined we fall back to calling a .TypeName method 15:48
Except if you have a multi that only accepts calls on instances then you get a dispatch error and not falback to the method call 15:49
15:53 Kristien left, adu joined
lizmat www.google.com/safebrowsing/diagnos...=perl6.org # *phew* :-) 16:02
hoelzro jnthn: I'm not a fan of it either 16:03
16:04 booly-yam-1010 left
skids jnthn: Wasn't Type() (with no args) on the road to deprecation, though? 16:04
jnthn
.oO( The road to deprecation is paved with melting features )
 16:07
I thought Type() was kinda short for Type(Mu) or maybe Type(Any)
16:08 booly-yam-9299 joined 16:09 regreg joined
skids Oh now I think I understand. So really it is just a matter of when people define their own coercers, ensuring they provide a path to the fallback? 16:10
16:11 rindolf joined 16:14 pecastro left
lizmat m: class A { method Int { say "foo" } }; Int(A) # just about 16:15
camelia rakudo-moar 912a7f: OUTPUT«foo␤»
16:16 gfldex joined
skids So is this just a case of "This hurts doctor"/"So don't do that" which could be "patched" through enculturation? 16:17
16:17 pecastro joined
skids I guess it would not be possible to warn like "Defining Int coercer without a :D ill advised" because the type named the same as the coercer might not yet exst... and at runtime that would be a long search to track them all down. 16:19
hoelzro ok, doing that experiment broke a few tests 16:21
7, to be exact
so not bad
16:22 kjs__ left
hoelzro it's not spec'd how to override MyType(Str), is it? 16:22
just that MyType(Str) looks like a function call, and it's a coercion?
lizmat m: class A { method Int { say "foo" } }; A(Int) # the other way around
camelia rakudo-moar 912a7f: OUTPUT«Cannot find method 'A'␤ in block <unit> at /tmp/iDu7MxkyRo:1␤␤»
lizmat that goes looking for a method A in Int 16:23
16:23 rurban joined
lizmat m: class A { multi method A(Int:U:) { say "foo" } }; A(Int) # does this work? 16:24
camelia rakudo-moar 912a7f: OUTPUT«Cannot find method 'A'␤ in block <unit> at /tmp/0cn73E9U5z:1␤␤»
hoelzro lizmat: I tried implementing postcircumfix:<( )> on MyType, that *works*, in that the function is called
but a Capture is passed, rather than the value itself 16:25
which is fine, except it's not spec'd (iirc), and I think that would throw a new user of Perl 6
lizmat yeah, but that approach feels wrong to me
hoelzro I would honestly just like a multi method coerce($src-type, $dest-type)
skids m: class A { method Int { say "foo" } }; Int.(A) # that is worse, appearence-wise.
camelia rakudo-moar 912a7f: OUTPUT«foo␤»
PerlJam hoelzro: maybe spelled COERCE because it's something that Perl will call on your behalf? 16:26
hoelzro so if I want to tell Perl 6 how to coerce Strs into MyTypes, I need only implement multi coerce(Str $value, MyType) { ... }
PerlJam: oh, good call
(no pun intended)
jnthn hoelzro: Aside from the fact it will have an epic number of candidates, a multi COERCE ain't so bad.
hoelzro jnthn: epic indeed 16:27
we could also have a fallback multi COERCE(Any $value, Any:U $type) { my $name = $type.^name; $value."$name"() } 16:28
so that .Str and friends still works
I think it should be easy, when defining a new type, to control how values of other types may be converted into my new type, whether or not I control those other types 16:29
(I suppose, given the nature of Perl 6, "control" is a murky term =/)
PerlJam Most of the "control" seems like it would be the compiler carping about "I don't know how to convert TypeA into TypeB, perhaps you should write a multi COERCE(A,B) routine" 16:33
hoelzro right, in this case, the control would be black magic-free 16:34
instead of, for example, having to monkey patch TypeA 16:35
skids Don't I remember reading something about if the source type doesn't have an opinion on the matter, then go see if the destination type does? 16:36
16:37 kaleem joined
jnthn skids: That's the basic idea of the current policy, yes, but having an opinion is currently a bit all-or-nothing 16:37
hoelzro I think it's actually the other way around; I think the destination type is checked first, then the source type is consulted 16:38
16:38 denis_boyun_ left
jnthn Oh? 16:38
Quite possibly. :)
hoelzro I *think* so
my Rakudo fu is quite weak =)
skids Not according to lizmat++s example above.
jnthn Ah, yeah, destination is the invoke-y one
16:39 panchiniak joined
hoelzro github.com/rakudo/rakudo/blob/nom/....nqp#L2644 16:39
16:39 FROGGS left
hoelzro if you can invoke the type, try it; otherwise, try $value."$typename" 16:39
lizmat fwiw, a COERCE method feels in line with things like ACCEPTS 16:40
hoelzro jnthn, you mentioned that the dispatch table for such a multi would be huge; how bad are the implications of that? 16:41
skids Is there a way to split a proposed COERCE up to reduce the candidates?
jnthn hoelzro: With the present way things work, it depends how many of them you hit in the time a program is running. 16:42
hoelzro: If we totally fill up the multi-dispatch cache then dispatches not in it get slow and spesh can't do much 'cus it looks at the cache. 16:43
We could of course make the cache growable.
hoelzro ah ha
jnthn Or re-focus the caching around the callsite, not the callee 16:44
skids That might be an interesting generic thing to do once a high water mark gets hit. 16:45
japhb jnthn: This reminds me, how will spesh react to a class Foo { has $.bar handles * }, when $.bar changes every so often (say, every 1000 method calls to a Foo instance)?
Well, not just spesh, but all method-call optimizations, really 16:46
jnthn japhb: Changes *object* or changes type of object?
japhb Changes the object instance, but $.bar always has the same type.
jnthn Spesh cares about types 16:47
So if it does something there it'll likely have a guard on the type $.bar gives back
japhb So if the type were allowed to vary, then what?
jnthn Probably if you're stable for a while it'll specialize assuming it'll always be that type
And then when it changes, you'll hit the deopt path every time. 16:48
japhb Forever more? Or will it notice a "new stable" and try again?
jnthn In the long run, we'll keep stats on deopt rates and retire specializations that work out badly.
Forevermore. It's kinda the generational hypothesis for optimization: either things vary in type a lot, or keep the same type, and things in the middle are rarer 16:49
japhb wonders if "punctuated equilibrium" is pessimal for r-m right now
nodnod
jnthn Yes, we likely handle that less well at the moment.
16:49 rurban_ joined, jluis_ joined
jnthn Though if you have just a couple of places in your program like that, it's not the end of the world. 16:50
japhb How is 'handles *' actually implemented? Does Foo's method cache just fill up with entries from $.bar over time?
true
jnthn No, I doubt it's actually very optimal at all at the moment.
PerlJam Could there be an op exposed to Rakudo that could trigger a "respecialization"? (so that you could manually control it for the puntuated equilibrium that japhb mentioned) 16:51
jnthn PerlJam: I think better to let the VM keep stats and learn, tbh.
TimToady wonders how long the world stayed optimized for dinosaurs after the asteroid... 16:53
or rather, deoptimized for everyone...
japhb ISTR reading something that the world was already beginning to deopt for existing species before the asteroid hit, and that just put the cap on it. 16:54
TimToady too many megamorphs running around
skids COERCEASAURUS
japhb Yeah, here it is: arstechnica.com/science/2014/12/mas...rs-demise/ 16:55
skids That was just the earth breaking into a sweat because it knew it was about to be beaned on the forehead :-) 16:56
adu metamorph?
japhb heh
16:56 Otterpocket joined 16:57 Otterpocket left
adu Kyodai Ka? 16:57
dalek rl6-roast-data: 30fa912 | coke++ | / (5 files):
today (automated commit)
 16:58
TimToady 強大も。
16:59 zakharyas left, kjs__ joined
adu ⿰弓⿱口虫 17:01
17:01 rurban left
adu 巨大化 17:01
Megawandler 17:02
pmurias jnthn: Can setinvokespec be called multiple times on the same type? 17:03
jnthn pmurias: Um...not sure we refuse it, but the behavior isn't very defined. :) 17:05
In practice, it doesn't happen anywhere I can think of.
17:05 yeahnoob left, Otterpocket joined 17:11 vendethiel left 17:13 vendethiel joined 17:16 Kristien joined
Kristien hola 17:19
17:19 LLamaRider joined
TimToady halo 17:20
[Coke] m: sub Int() { say "3" } ; say Int(); say Int(10); 17:24
camelia rakudo-moar 912a7f: OUTPUT«0␤10␤»
[Coke] Should the sub declaration there complain ?
japhb If type coercion will always win, then yes. 17:26
17:27 Otterpocket left
japhb (There's an argument to be made that an our sub might actually be usable, but at the very least this should be a warning.) 17:27
17:29 kaleem left
jnthn One does start to wonder if we might be hanging things off the wrong peg somehow with coercion... 17:31
TimToady where have I heard that metaphor before... :) 17:32
TimToady thinks of type names as new reserved words, basically 17:33
[Coke] m: say Int("three"); say Complex("Sausage")
camelia rakudo-moar 912a7f: OUTPUT«Cannot convert string to number: base-10 number must begin with valid digits or '.' in '⏏three' (indicated by ⏏)␤ in method Int at src/gen/m-CORE.setting:14137␤ in method Int at src/gen/m-CORE.setting:6340␤ in block <unit> at /tmp/Ca0NYzo2Jd:1…»
[Coke] m: say Complex("Sausage")
camelia rakudo-moar 912a7f: OUTPUT«Cannot find method 'Complex'␤ in block <unit> at /tmp/ukzPHD4Vbs:1␤␤»
[Coke] the Int error makes sense. the Complex one does not. (also, should I be testing this on master?) 17:34
er, nom?
17:34 vendethiel left
TimToady were you thinking of that as a coercive type? 17:34
I don't think 6pe is merged quite yet, though I still have some backlog... 17:35
17:35 tgt joined
[Coke] Yes. I was expecting I could pick arbitrary core classes and try to coerce to them. 17:35
ah, ok. I'll wait until post merge before poking more, I guess. or switch to that branch and poke. 17:36
jnthn At the moment, 6pe would give you all the same answers to the things you tried. 17:37
It only parses a coercion type when typename is hit, which for now is only in explicit calls toit.
TimToady we need to teach the parser that Int() and Int(SomeType) are coercive types
jnthn *to it
*nod*
TimToady or is that there already?
17:38 vendethiel joined
jnthn No, I didn't do that 17:38
I guess that goes in term:identifier
But that now has 3 cases
1) Coercion type
2) Coercion operation
3) Function call
TimToady basically we stole two syntaxes from ... what you said
jnthn If we hang it there.
TimToady doesn't matter so much where a type is already expected 17:39
moritz let's find another use for the colon
TimToady but as an rvalue...
moritz Int:(Any) # as coercion type
moritz only 20% serious
jnthn Yeah, it was no problem in typename, it's the rvalue case...
TimToady but I don't think it's semantically ambiguous
which is why I picked that in the first place
Int(Any) is really a kinda generic for Int($any) 17:40
jnthn I think we can differentiate it without WATs...
TimToady and the arg has to be literally a type or ()
jnthn Right.
What a coercion operation compiles into is really the one I'm not comfortable with at the moment.
TimToady you mean Int($foo)? 17:41
jnthn Yeah
At the moment it's a call to an internal function that does a bunch of introspection. 17:42
Which is, uh, LTP.
TimToady ...parsimonious?
jnthn Performant is what I had in mind...
TimToady ...peggy?
...puresmelling 17:43
17:43 FROGGS joined, rmgk_ joined, rmgk left, rmgk_ is now known as rmgk
TimToady it really is sort of an instantiation of the generic coercion type 17:43
moritz ...probabilistic
jnthn TimToady: That's pretty much how it's been implemented :) 17:44
17:44 kjs__ left 17:46 booly-yam-9299 left
jercos japhb: haha, yes, that was what I was looking for. Thanks :) 17:46
m: say "whatever".WHAT 17:47
camelia rakudo-moar 912a7f: OUTPUT«(Str)␤»
jercos m: say "whatever".methods.grep({$_.gist ~~ /^fl/})
camelia rakudo-moar 912a7f: OUTPUT«No such method 'methods' for invocant of type 'Str'␤ in block <unit> at /tmp/RYnbpgIk9L:1␤␤»
jercos erm
m: say "whatever".^methods.grep({$_.gist ~~ /^fl/})
camelia rakudo-moar 912a7f: OUTPUT«␤»
17:47 booly-yam-9299_ joined
jercos Oh, it's on Cool, haha, got it. 17:48
jnthn .^methods(:all)
moritz m: say 'whatever'.^methods(:all).grep({.name ~~ /^fl/}) 17:49
camelia rakudo-moar 912a7f: OUTPUT«floor flip flat flatmap␤»
17:49 denis_boyun joined
jercos Well yes, that totally works, but doesn't tell me where it's actually from. 17:50
And gives a rather huge list to look through on a Str, if I didn't already know what I was looking for
jnthn m: say 'whatever'.^methods(:all).grep({.name ~~ /^fl/})>>.package.say 17:51
camelia rakudo-moar 912a7f: OUTPUT«(Cool) (Cool) (Any) (Any)␤True␤»
moritz jercos: doc.perl6.org/type/Str nicely tells you where which method comes from
jnthn m: say 'whatever'.^methods(:all).grep({.name ~~ /^fl/}).map({ "{.name} in {.package.^name}" }) 17:52
camelia rakudo-moar 912a7f: OUTPUT«floor in Cool flip in Cool flat in Any flatmap in Any␤»
17:52 pmurias left
jnthn
.oO( That's flipin cool! )
 17:52
jercos bril :D
I'm still sorta on the fence about Cool in general though :p 17:53
jnthn So...you're a bit cool about it?
jercos Most of the methods it bears seem to act like either self.Numeric.something or self.Str.something
jnthn Yes
TimToady that's what it's for :) 17:54
jnthn It's a strongly typed implementation of weak typing...
jercos But without any indication from the method itself which is which
tony-o_ hoelzro: i think i have a fix for the bug you discovered in html-parser-xml
moritz doc.perl6.org/type/Cool has the table of which method coerces to which
moritz feels a small mental reward each time he can point to a point of doc.perl6.org to answer a question that folks actually ask 17:55
jercos Indeed it does. And most of them are pretty obvious.
I don't think anyone would expect .cos() to be anything but Numeric :p 17:56
jnthn moritz: Yes, doc.perl6.org increasingly seems to have the answers to things :)
moritz jercos: except when they expect cos play :-)
jercos And of course the return type is programmatically accessible as well, I suppose.
moritz doc.perl6.org/type/Signature#method_returns 17:57
:-)
m: say Cool.^find_method('cos').signature.returns.^name 17:58
camelia rakudo-moar 912a7f: OUTPUT«Mu␤»
moritz but it seems to be not declared :(
17:58 virtualsue left
moritz m: say Cool.^find_method($_).signature.returns.^name for <chop abs floor> 17:58
camelia rakudo-moar 912a7f: OUTPUT«Mu␤Mu␤Mu␤»
moritz :(
TimToady m: say Num.^find_method('cos').signature.returns.^name 17:59
camelia rakudo-moar 912a7f: OUTPUT«Mu␤»
TimToady that one really oughta say Num anyway
moritz the methods in Cool don't seem to declare return types right now
17:59 Rounin joined
moritz TimToady: I think the sentiment so far has been to declare return types where they give a performance boost, which is just for subs so far 18:01
m: say IO::Handle.^find_method('tell').signature.returns.^name
camelia rakudo-moar 912a7f: OUTPUT«Mu␤»
[Coke] moritz: should they need to if they have a single return point and that item declares a return? (I'm assuming yes because that level of introspection is hard)
moritz [Coke]: you are correct 18:02
TimToady m: say Cool.^find_method('ord').signature.returns.^name
camelia rakudo-moar 912a7f: OUTPUT«Int␤»
moritz $ git grep 'method \w.*returns'|wc -l
9
TimToady m: say ord(42) 18:03
camelia rakudo-moar 912a7f: OUTPUT«52␤»
moritz m: say chr 52
camelia rakudo-moar 912a7f: OUTPUT«4␤»
18:04 denis_boyun left
TimToady some kinds of optimiztions depend on knowing the return type, so I think we should work towards getting return types declared and performant 18:05
I realize that's more of an AOT concern than a spesh concern
but we really don't want to penalize folks for declaring their return types
jnthn I don't think there's much penalty at all 18:06
TimToady maybe there was at some point in the past?
jnthn Yes
But today should really not be
Spesh probably strips a bunch of those checks out when it can prove they're not needed also
TimToady goodgood
I assume spesh can also deal with the optimization of Foo(Any) coercions when we start combining multis that currently are there only to provide Any,Any parameters that map to Foo,Foo? 18:10
lizmat recently removed a return type because it could also return a Failure 18:11
jnthn At the moment, Foo(Any) $x is deconstructed in the signature into the same thing that Any $a as Foo has turned into 18:12
lizmat and that broke
TimToady lizmat: yes, that's also a concern
jnthn Which is a nice fast, cachable, nominal type match on Any, and then there's an istype guard against Foo
If spesh sees it already is a Foo, then it turns the branch into a goto and the coercion becomes dead code and vanishes. 18:13
TimToady but that's on the callee end, whereas the multies might get duomorphically cached on the caller end? 18:14
so maybe not equiv
I don't imagine we do per-caller speshing like that 18:15
well, unless you've inlined first...
jnthn Well, it's done at the callee
Uh
Specializations are per-callee
But the caller might be specialized to call a particular specialization.
That's a pre-req for inlining, in fact. 18:16
TimToady so we don't lose that information?
jnthn Inlining is "just" going one step further when you've decided you know exactly where the call is going.
No, if when specializing the caller we find the callee has a specialization that fits, we just emit an invoke op that points directly to the specialization. 18:17
TimToady suppose you have 10 call sites that are monomorphic on the call, but collectively are megamorphic. do we deopt?
18:17 Hor|zon left
jnthn looks confusedly at the question :) 18:18
TimToady I can well imagine a generic having that behavior when instantiated 10 different ways
jnthn We only deopt locally if we fail a guard, or globally if we have a mixin or some other such action-at-a-distance change. 18:19
TimToady locally meaning in the caller?
jnthn So far as 6model parametric stuff and spesh interact so far, each parameterization can have its own specialization.
Locally meaning the frame we're in when the guard fails 18:20
Note that if we are optimizing a call and there's no specialization of the callee that type-matches what we have, then we just leave the slower-path call in place.
Though we may cache the code object if it's a lexical that isn't going to change. 18:21
The thing is structured to specialize on frame *exit*, not entry, so your callees of lesser or similar size will always have been specialized before you.
TimToady nod 18:23
18:23 vendethiel left
TimToady I love the response 'nod' because it conveys anything from "I hear you" to "I agree completely", and in general conveys greater intelligence than is felt. :) 18:24
jnthn nod 18:25
18:27 tgt left
TimToady :D 18:27
[Coke] returns Int:F # normally an Int, but potentially a Failure. 18:32
perhaps?
Kristien F# :o 18:33
18:33 FROGGS_ joined
TimToady Int() would coerce failure to Int :) 18:34
18:36 Sqirrel left, FROGGS left
ab5tract jnthn: here's a smaller example which seems to trigger segfaults and unhandled exceptions gist.github.com/ab5tract/34408eaa7b5c64dc84a7 18:38
i think it is the same behavior i saw previously when playing around with raw thread locking in perl 6 18:39
18:42 Hor|zon joined
japhb [Coke]: Perhaps 'Int?' instead of 'Int:F'? I don't think we use 'Type?' syntax yet .... 18:44
jnthn ab5tract: Yes, I've got a collection of issues like that to look into... 18:45
dinner &
ab5tract enjoy! (the dinner)
18:48 Kristien left 18:51 vendethiel joined
dalek c: 9d785d6 | moritz++ | lib/Type/ (2 files):
Document run, shell, Proc::Status
 18:52
c: 5efa369 | moritz++ | lib/Type/Metamodel/MROBasedMethodDispatch.pod:
Document .^find_method_qualified
c: 64385e5 | (Geoffrey Broadwell)++ | htmlify.p6:
Improve preamble boilerplate for type graph SVG image
 18:53
c: e87e534 | (Geoffrey Broadwell)++ | lib/Type/ (3 files):
Merge branch 'master' of github.com/perl6/doc
[Coke] japhb: I would expect Int? to mean it could return Nil, I think. 18:56
hoelzro tony-o_: hooray!
[Coke] m: sub bar() returns Int { Nil }; my Int foo = bar(); say "y" 18:57
camelia rakudo-moar 912a7f: OUTPUT«===SORRY!=== Error while compiling /tmp/FS5VUU6Yx3␤Malformed my (did you mean to declare a sigilless \foo or $foo?)␤at /tmp/FS5VUU6Yx3:1␤------> ub bar() returns Int { Nil }; my Int foo⏏ = bar(); say "y"␤»
hoelzro I was going to start just using the XML module =)
[Coke] m: sub bar() returns Int { Nil }; my Int $foo = bar(); say "y"
camelia rakudo-moar 912a7f: OUTPUT«Type check failed for return value; expected 'Int' but got 'Any'␤ in any return_error at src/vm/moar/Perl6/Ops.nqp:649␤ in block <unit> at /tmp/4FnwFk8TnL:1␤␤»
TimToady Int? would be a Maybe type, not an Error type, to put it into Haskellian terms
japhb Yeah, fair enough. 18:58
TimToady it's also ambiguous currently with optional params
japhb Oh, for when you don't even specify a sigil?
TimToady m: sub foo (Int?) { say "here" }; foo()
camelia rakudo-moar 912a7f: OUTPUT«===SORRY!=== Error while compiling /tmp/qG1V2W0UIq␤Missing block␤at /tmp/qG1V2W0UIq:1␤------> sub foo (Int⏏?) { say "here" }; foo()␤»
TimToady o_O
japhb I guess it's not ambigious in Rakudo ... ;-) 18:59
[Coke] m: sub foo (Int $?) { say "here" }; foo()
camelia rakudo-moar 912a7f: OUTPUT«here␤»
19:00 Sqirrel joined
TimToady regarding coercion priority, it's neither of the things suggested at irclog.perlgeek.de/perl6/2015-01-23#i_9992611 19:01
see S13:198
synopsebot Link: perlcabal.org/syn/S13.html#line_198
19:02 ashleydev left, ashleydev joined
TimToady you just want to avoid the situation where the functional and method forms are both deferring to each other 19:02
maybe the right way to do that is to have both forms write a multi COERCE that is the fallback, with poisoning if both ends try to define it 19:04
19:08 Hor|zon left, virtualsue joined 19:09 rindolf left 19:14 vendethiel left
dalek kudo/nom: 25da7a9 | moritz++ | src/core/Cool.pm:
Declare a few more return types
 19:14
19:15 LLamaRider left 19:17 fhelmberger left, Kristien joined
japhb moritz: How often is doc.perl6.org regenerated? 19:19
19:20 vendethiel joined
Kristien I just noticed the 6 in the wing of Camelia Ö 19:20
moritz japhb: 17,47 * * * * ./update-and-sync > update.log 2>&1 19:21
hobbs Kristien: how about the P? :) 19:24
Kristien :O
Arguably the P is a rotated 6. 19:25
japhb It would have to be reflected as well, since 9 is a rotated 6.
moritz: thx
Kristien Rotated on the Y axis. :P 19:26
japhb s/on/about/, but yes, I get your point. :-)
japhb is happy that the type graph at doc.perl6.org/type/Cool is actually quite informative -- though better at 200% zoom on my monitor 19:29
Kristien I found a nice use-case for Perl 6: my compiler I originally intended to write in Haskell. :D 19:32
hoelzro tony-o_: thanks for the fix; I'll try it out later 19:35
raydiak good morning #perl6 19:36
19:36 booly-yam-9299_ left
raydiak m: multi foo (num) { say "num" }; multi foo (Any) { say "Any" }; foo num # ? 19:36
camelia rakudo-moar 912a7f: OUTPUT«Any␤»
moritz japhb: it would be even nicer if a click on the SVG would zoom it to fill the whole page 19:38
\o raydiak
raydiak: native type constraint always constrain to instances
19:39 bjz left
raydiak o/ moritz 19:39
m: multi foo (num.new(1)) { say "num" }; multi foo (Any) { say "Any" }; foo num.new(1)
camelia rakudo-moar 912a7f: OUTPUT«===SORRY!=== Error while compiling /tmp/dgjCf1_OCV␤Missing block␤at /tmp/dgjCf1_OCV:1␤------> multi foo (num⏏.new(1)) { say "num" }; multi foo (Any) ␤»
raydiak m: multi foo (num $x) { say "num" }; multi foo (Any) { say "Any" }; foo num.new(1) 19:40
camelia rakudo-moar 912a7f: OUTPUT«Any␤»
raydiak moritz: I'm somewhat unclear what that means, but trying to decypher it :)
moritz m: say num.^name 19:41
camelia rakudo-moar 912a7f: OUTPUT«num␤»
raydiak m: multi foo ($x where num) { say "num" }; multi foo ($x) { say "other" }; foo num.new(1) 19:42
camelia rakudo-moar 912a7f: OUTPUT«num␤»
moritz raydiak: think of sub foo(num $x) { } as a very low-level thing. It doesn't accept the num type object, only "real" doubles (pun intended)
japhb moritz: as a popover, or a new tab, or replacing the current page?
moritz japhb: popover or new tab, IMHO. But I'm open to suggestions :-) 19:43
19:43 vendethiel left, jluis_ left
japhb New tab is probably easiest, popover *might* be better UX, not sure. 19:44
19:44 vendethiel joined, avalenn_ left
raydiak imo a lightbox would help, but getting those typegraphs to come out at less extreme aspect ratios would be even better 19:44
moritz well, replacing the current tab would just be an ordinary link
raydiak: the type graphs are pretty well optimized already, as far as dot output goes 19:45
japhb raydiak: Unfortunately, our type heirarchies tend to be wide but shallow. Most aren't quite as extreme as Cool, but Any and Exception are so bad I had to special case them to rotate 90 degrees.
I spent a lot of time tuning the layout heuristics (which is why the current method *usually* looks OK, if not always). 19:46
raydiak oh that's why a few look different :)
moritz finds it quite fortunate that our type hierarchies tend to be not too deep :-)
japhb moritz: Fortunate for programming, unfortunate for pictures. ;-) 19:47
tony-o_ hoelzro: when you get around to trying it out, can you let me know how it goes and i'll close out the github issue
hoelzro tony-o_: I will let you know!
japhb raydiak, if you'd like to experiment with a lightbox, be my guest. I'm supposed to be reviewing design docs at $dayjob anyhow .... 19:48
raydiak japhb: probably not today, but I'll do it some time soon if nobody else does, b/c you're right that some are kinda a nuissance as is 19:50
19:50 Hor|zon joined
raydiak is waiting for that oh-so-elusive thing called "inspiration" to strike for the syn redesign 19:51
yesterday's work was decidedly uninspired and basically just a waste of time :P
japhb Sometimes it just takes learning lessons from a lot of 19:52
"uninspired" days to have one truly inspired one.
moritz start by removing everything that's ugly?
.oO( color: white; background-color: white )
japhb
.oO( Cut off his head! )
moritz: Or just convert the entire page to encoded whitespace as per Acme::Bleach .... 19:53
19:54 wilx joined
TimToady likes flat org charts... 19:54
19:56 Jefffrey joined
Jefffrey Hello 19:56
Kristien Hello Jefffrey
raydiak Helllo o/
19:57 spider-mario joined, lucas joined
Jefffrey So... 19:58
Let's talk about Perl 5.
hoelzro when we were talking about COERCE earlier today, there seemed to be a fair amount of support in favor. How does one get such a change into the language?
should I file a PR against perl6/spec?
PerlJam hoelzro: works for me. 19:59
moritz hoelzro: the strongest case you can make for it is implement it in rakudo, and push your changes to perl6/specs 20:00
(and tests, and docs, if you're really good :-)
hoelzro =)
PerlJam yeah, that's even better
lucas m: my Nil $x; my True $y; my NaN $z; say 'alive'
camelia rakudo-moar 912a7f: OUTPUT«alive␤»
lucas ^^^ It seems like I can initialize a variable with some odd values. Is this ok? I would expect only typeobjects to be allowed in that position.
hoelzro ok, I will endeavor to do that!
PerlJam hoelzro: but even TimToady often just specs stuff with no implementation ;) 20:01
TimToady hoelzro: so what do you do if both endpoints define contradictory functions?
hoelzro I was thinking "spec first" so that people could discuss in case the idea was not well recieved widely
TimToady: for example?
moritz m: my NaN $z; say $Z
camelia rakudo-moar 912a7f: OUTPUT«===SORRY!=== Error while compiling /tmp/T0JPa1nWb0␤Variable '$Z' is not declared␤at /tmp/T0JPa1nWb0:1␤------> my NaN $z; say $Z⏏<EOL>␤ expecting any of:␤ postfix␤»
moritz m: my NaN $z; say $z
camelia rakudo-moar 912a7f: OUTPUT«NaN␤»
TimToady well, did you look at S13:198 like I pointed to earlier? 20:02
synopsebot Link: perlcabal.org/syn/S13.html#line_198
TimToady that's current spec on it
hoelzro I did not!
hoelzro looks
TimToady and it allows both ends to specify how they think the coercion should go
and it allows the user to pick one or the other in the case that they conflict
moritz lucas: yes, we allow values in some situations that are usually reserved for types
TimToady I suggested we earlier we might do COERCE as the mutual fallback though 20:03
moritz m: multi f(0) { 1 }; multi f(Int $x) { $x * f($x - 1) }; say f 4 # for lucas++
camelia rakudo-moar 912a7f: OUTPUT«24␤»
TimToady that would be one way to avoid infinite regress
20:03 denis_boyun joined
Kristien m: my ($s, $t) = (Supply.new, Supply.new); my ($p, $q) = ($s.Promise, $t.Promise); $t.emit(1); $s.emit($q); say ($p.result, $p.result.result); 20:04
camelia rakudo-moar 912a7f: OUTPUT«Promise.new(scheduler => ThreadPoolScheduler.new(initial_threads => 0, max_threads => 16, uncaught_handler => Callable), status => PromiseStatus::Kept) 1␤»
Kristien No assimilation, good.
TimToady lunch &
hoelzro interesting
one thing that bothers me about $value."$typename"() is what about types like X::MyException (/me has no idea if : is legal in method name) 20:05
20:06 Jefffrey left
psch m: class Foo { }; Foo.^add_method("it's a silly::test", method (Foo:U:) { say "dunno?" }); Foo."it's a silly::test"() 20:06
camelia rakudo-moar 912a7f: OUTPUT«dunno?␤»
lucas m: constant Foo = 42; my Foo $x; say $x 20:07
camelia rakudo-moar 912a7f: OUTPUT«42␤»
hoelzro but also, let's say I have two types with the same base name (ex. module Trangressions { class Sin { ... } } module Trig { class Sin { ... } }), and I define .Sin on some other class. Which one am I coercing to?
lucas It seems like if I have any bareword, I can use it as a typename 20:08
PerlJam hoelzro: whichever one is in scope
lucas moritz: Thanks for the example!
moritz PerlJam: in scope where the coercion method is defined? or where it's called?
hoelzro PerlJam: what if I import the other class into another module, and call .Sin, having imported neither Transgressions nor Trig?
moritz jnthn: is there anything holding off the 6pe merge? 20:09
PerlJam moritz: if you're not typing your return values, then either :) 20:10
lizmat hoelzro: then that Sin will be visible in that scope
or the import will fail if there is already a Sin there
skids m: class A { method y { "yes".say } }; class B is A { }; B.new.A::y() # hoelzro, methods syntactically can have ns colons and such is often used like this. 20:11
camelia rakudo-moar 912a7f: OUTPUT«yes␤»
hoelzro let's say I have class MyType { method Sin { #`( ...magic...) } }
jnthn moritz: Various things :)
hoelzro and I import MyType into Foo.pm, and that's all I import. Then I call MyType.new.Sin. Which Sin should it try coercing to?
moritz hoelzro: I'm well aware of the the problem, what's missing is a solution 20:12
PerlJam hoelzro: also, that's one of the things I liked about COERCE, that you have to declare the types of both sides
hoelzro granted, that's up to the person writing the method in MyType, but then can you create a coercion from MyType to the other Sin class?
jnthn moritz: For one, it'd bust the Parrot build until I port it there. And there are some spectest regressions, and so on.
hoelzro moritz: right, I'm pointing out what bothers me about the current state of affairs 20:13
moritz jnthn: oh
hoelzro skids: thanks for the demo
20:13 denis_boyun left
moritz the problem with COERCE(source-type, dest-type) is that it type-checks the wrong 20:13
(co vs. contra variant)
dalek ast: 8fc7d02 | usev6++ | S32-str/sprintf.t:
Add test for RT #122907
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=122907
moritz uhm, no, ignore me 20:14
moritz has his head all knotted up
20:14 agentzh joined
japhb wonders what it means for a relatively compact, mostly convex structure to be knotted 20:15
hoelzro I don't mind implementing COERCE for people to try it out, but I don't want to spend time implementing something unless I have a fair amount of confidence that it would eventually be used
hoelzro backlogs to see what TimToady said before hoelzro asked again 20:16
PerlJam hoelzro: such is the implementor's risk. (How else do we get implementations to vet the ideas of the spec?)
hoelzro PerlJam: point taken =)
jnthn gets back to work on 6pe stuffs 20:17
20:17 lucas left 20:18 darutoko left
hoelzro I just saw lizmat's comment on removing a returns trait because that routine could return a Failure, which intrigues me 20:22
the spec allows for definite return values (ex. sub mysub() returns 3 { ... }), so that mysub always returns 3...
...expect (iirc) the spec *also* allows mysub to return a Failure!
er, except
hoelzro needs coffee
lizmat m: sub a(-->Int) { fail }; a
camelia rakudo-moar 912a7f: OUTPUT«Type check failed for return value; expected 'Int' but got 'Failure'␤ in any return_error at src/vm/moar/Perl6/Ops.nqp:649␤ in sub a at /tmp/abIECMht3v:1␤ in block <unit> at /tmp/abIECMht3v:1␤␤»
hoelzro I find it interesting that the spec allows Failures to sneak past the returns type in the case of definite return specifications, but not definite ones 20:23
moritz hoelzro: does it, still? I thought that was removed 20:24
lizmat m: sub a(-->Int:D) { fail }; a
camelia rakudo-moar 912a7f: OUTPUT«Type check failed for return value; expected 'Int' but got 'Failure'␤ in any return_error at src/vm/moar/Perl6/Ops.nqp:649␤ in sub a at /tmp/95UbbofjMN:1␤ in block <unit> at /tmp/95UbbofjMN:1␤␤»
hoelzro moritz: the part about definite return specifications?
I see what TimToady was saying now 20:25
it's kind of orthogonal to whether or not COERCE exists 20:26
moritz hoelzro: the part about allowing failure to sneak past return type declarations
hoelzro S06:672 20:27
synopsebot Link: perlcabal.org/syn/S06.html#line_672
hoelzro neat, it worked =)
20:27 vendethiel left
hoelzro end of the first paragraph 20:27
actually, that sentence implies that sneaking should happen for indefinite return specs as well
moritz and it stil points to perlcabal.org :( 20:28
tadzik: didn't I fix synopsebot? :(
hoelzro I'm guessing if any routine may return a Failure, that could muck with optimization =/
20:29 bjz joined
hoelzro unless we check for "fail" while compiling a sub, and annotate the code object with that information 20:29
moritz that's why rakudo doesn't implement it
hoelzro moritz: sorry, doesn't implement which?
PerlJam hoelzro: the sneaking
moritz hoelzro: allowing failures to sneak past
hoelzro ah 20:30
just checking =)
20:30 vendethiel joined
PerlJam See lizmat's offering to rakudo-m a few minutes ago 20:30
hoelzro yeah, I was just questioning whether or not the spec made sense, and then after I re-read that paragraph, whether or not getting Rakudo to conform with the spec there would impact optimization 20:31
lizmat jnthn has stated the other day, that that would be almost impossible to not have that impact performance significantly 20:33
moritz perl6/doc htmlify.p6 doesn't work with current rakudo + current Pod::To::HTML :( 20:34
raydiak ouch 20:35
FROGGS_ hmmm, my code has not suffered...
moritz: what's wrong?
moritz perlpunks.de/paste/show/54c2b083.700a.2b6
FROGGS_ moritz: hmmm
github.com/perl6/Pod-To-HTML/blob/...ML.pm#L161 20:36
looks like a rakudobug if you ask me
m: my @a; say @a[*-1] 20:37
camelia rakudo-moar 25da7a: OUTPUT«Index out of range. Is: -1, should be in 0..Inf␤ in method at_pos at src/gen/m-CORE.setting:9827␤ in sub postcircumfix:<[ ]> at src/gen/m-CORE.setting:2837␤ in sub postcircumfix:<[ ]> at src/gen/m-CORE.setting:2937␤ in block <unit> at /tmp/iskh…»
FROGGS_ star-m: my @a; say @a[*-1]
camelia star-m 2014.12: OUTPUT«Calculated index (-1) is negative, but Array allows only 0-based indexing␤ in method gist at src/gen/m-CORE.setting:13837␤ in sub say at src/gen/m-CORE.setting:16595␤ in block <unit> at /tmp/bgxDUS6alV:1␤␤»
FROGGS_ star-m: my @a = 1; say @a[*-1]
camelia star-m 2014.12: OUTPUT«1␤»
FROGGS_ m: my @a = 1; say @a[*-1]
camelia rakudo-moar 25da7a: OUTPUT«1␤»
FROGGS_ hmmm
ohh
tony-o_ does p6 build for 32b os/jvm? .. i'm getting "main" java.lang.NoClassDefFoundError on 32b but it builds fine for 64b same version 20:38
moritz star-mm: my @a; say @a[*-1] // 0
raydiak maybe like this one: irclog.perlgeek.de/perl6/2015-01-10#i_9916445
(the infinte array bug, that was for)
moritz star-m: my @a; say @a[*-1] // 0
camelia star-m 2014.12: OUTPUT«0␤»
FROGGS_ moritz: that's it!
m: my @a; say @a[*-1] // 42
camelia rakudo-moar 25da7a: OUTPUT«Index out of range. Is: -1, should be in 0..Inf␤ in method at_pos at src/gen/m-CORE.setting:9827␤ in sub postcircumfix:<[ ]> at src/gen/m-CORE.setting:2837␤ in sub postcircumfix:<[ ]> at src/gen/m-CORE.setting:2937␤ in block <unit> at /tmp/ap_2…»
FROGGS_ m: my @a; say @a[*-1].WHAT
camelia rakudo-moar 25da7a: OUTPUT«Index out of range. Is: -1, should be in 0..Inf␤ in method at_pos at src/gen/m-CORE.setting:9827␤ in sub postcircumfix:<[ ]> at src/gen/m-CORE.setting:2837␤ in sub postcircumfix:<[ ]> at src/gen/m-CORE.setting:2937␤ in block <unit> at /tmp/xPuB…»
FROGGS_ star-m: my @a; say @a[*-1].WHAT
camelia star-m 2014.12: OUTPUT«(Failure)␤»
FROGGS_ lizmat: ^^
20:40 denis_boyun_ joined
lizmat FROGGS_: yes, did that behaviour change? seems ok to me 20:40
ah, you mean it throws rather than just fail ?
ah, yuck 20:41
yes
argh
FROGGS_ yeah, it explodes too early
lizmat argh, why doesn't the spectest catch that :-( 20:42
this is probably something we need to do a hotfix for, I would say :-(
FROGGS_ my thought... seems we're missing a test here
lizmat investigates
FROGGS_ I dunno if there is much fallout
I don't think this is a too common idom 20:43
idiom
tony-o_ does p6 build for 32b os/jvm? .. i'm getting "main" java.lang.NoClassDefFoundError on 32b but it builds fine for 64b same version 20:44
FROGGS_ tony-o_: I have no knowledge that somebody tried it in a 32bit jvm 20:45
on*
jnthn It's the first time I've heard of a 32-bit issue
bartolin lizmat: isn't that the same thing as with the test fudged here: github.com/perl6/roast/commit/22ef35ba3f
jnthn I'm sure in the early days I built it on a 32-bit JVM
tony-o_: Which OS, ooc?
tony-o_ do you want a gist?
debian 7 20:46
jnthn Hm, nothing unusual
And no errors along the way?
lizmat bartolin: yeah, but that test is weird
jnthn Yeah, feel free to gist the build output
tony-o_ outside of that, i didn't see any fly by on my screen - it's a fresh copy of rakudo
jnthn Did the build fail with that, or it just fails to run after building?
tony-o_ the build fails with that 20:47
lizmat FROGGS_: testing a fix
ab5tract tony-o_: i'm running a build right now
20:47 denis_boyun_ left
jnthn tony-o_: OK, gist me the last bit of the build output. 20:48
ab5tract hadn't realized just how waysided i686 has become in gnu/linux
tony-o_ okay, i'll give you whats in the buffer and full gist when the build finishes 20:49
FROGGS_ lizmat++
hoelzro lizmat: what about the idea about "can fail" annotations on the code objects? would that be good enough
?
tony-o_ jnthn: not much in my scrollback bufer - gist.github.com/tony-o/ca71572bc9fe0e25f28d
hoelzro oh, wait 20:50
lizmat jnthn: what I understand of the typechecking logic, is that that would not help much
hoelzro you *could* do return $obj.could-fail() from a routine, and then you don't know
jnthn It's fine if we have some way to indicate "can't fail" and "can fail" in the returns. 20:54
It's just that anything marked as "can fail" isn't going to allow the static optimizer to trust what comes back
tony-o_: What on earth... 20:55
psch suspects badly set up jre
20:55 beastd joined
psch tony-o_: how are you running the 32bit build? on a seperate 32bit debian? 20:56
raydiak do we allow junctions as return type declarations? e.g. --> Int|Failure
jnthn No 20:57
raydiak darn
jnthn That's a worse kettle of fish :)
tony-o_ psch: yea 20:58
moritz well, at least it'd only make it worth for those routines marked as Int|Failure
not for all of them
psch tony-o_: but «java -version» works?
tony-o_ psch: i have a fresh debian 7 (32b) with open jdk installed
and the build fails, i'm comparing it to the same version 64b jdk installed on a different 64b installation of debian 20:59
lizmat is github having problems?
raydiak well all I was thinking is that I hate it when stuff like failures leak/"sneak" through by default...would rather it gets thrown before/at the return if I specified a return type which doesn't allow it 21:00
so if not junctions maybe an "is failable" pragma or something?
dalek kudo/nom: f120204 | lizmat++ | src/core/ (6 files):
< 0 exceptions should fail rather than throw

Spotted by FROGGS++ . Not sure of the ecosystem fallout of this: could warrant a hot fix update of the compiler release before R* ?
lizmat FROGGS_ ^^^ 21:01
moritz lizmat++
hoelzro raydiak: but the whole point of failures is that you can handle it at the callee, and if you don't, it croaks, right? 21:02
moritz well, if a routine throws an exception, it avoids violating the return type contract 21:03
TimToady thinks we should replace the caller/callee terms because they're too confusing
moritz simply by not returning
dalek p/6pe: a28b48e | jnthn++ | src/vm/parrot/QAST/Operations.nqp:
Map nqp:: ops for 6model parametrics on Parrot.

Note they're not implemented yet.		 21:04
p/6pe: 70fcb2c | jnthn++ | src/vm/parrot/ops/nqp.ops:
Stub in ops.

t/nqp/87-parametric-6model.t now compiles now, though of course fails.
moritz wonders if it's worth parallelizing htmlify 21:07
writing the type documents at least should be quite easy to do 21:08
tony-o_ jnthn: here is the full log gist.github.com/tony-o/2b5e1f4d71311cc8756b
agentzh sorry if it's something obvious, but is it possible to precompile perl6 scripts to bytecode (or something like that) to save parsing the source? 21:09
tony-o_ stackoverflowerror..think that might be too little memory (it has 1.5G)
raydiak hoelzro: if I'm reading things right, there are a number of ways that user code (the caller) might inadvertantly receive such an object and dumbly shove it through the pipeline, no?
moritz agentzh: no, just modules
agentzh moritz: thanks for the reply.
jnthn agentzh: Modules, yes. Scripts, no (well, more "not yet", it's on the wish list)
raydiak hoelzro: like checking the return value for definedness, or it being received in lazy list context? 21:10
agentzh jnthn: no problem, i'll try to make my boilerplate script minimal :)
jnthn tony-o_: Well, it tends to mean specifically too little stack memory
raydiak admits he has not much first-hand practice with our Failures and how they work, is just reading as he goes 21:11
jnthn tony-o_: That is configurable, though...
hoelzro raydiak: yeah, I think you might end up with a little bit of the "nil problem" (I call it that, because Lua uses nils for error conditions) 21:12
tony-o_ option on the jvm or in the Configure script?
dalek c: 24ea783 | moritz++ | htmlify.p6:
htmlify: fix traits. Closes #41
 21:13
21:13 vendethiel left
moritz tony-o_: you might need to patch tools/build/create-jvm-runner.pl and tools/build/Makefile-JVM.in 21:13
tony-o_: search for -Xss or -Xmx 21:14
21:14 booly-yam-4652 joined
tony-o_ for create-jvm-runner.pl i'm only seeing it for the 'install "perl6-eval-server", "java" line - i'll check the makefile 21:15
raydiak hoelzro: yeah precisely...what I'm hoping to avoid is that nasty feeling I get from stuff like javascript where mistakes end up making NaNs that go through and cause weird, totally unrelated errors later on and force you to bisect your pipeline to find it
which admittedly is slightly different, but similar enough for my analogy (I think) 21:16
hoelzro do Failures describe where they come from?
that would at least help
moritz jnthn: that doc commit should fix the routine-trait 404 link on doc.perl6.org you reported yesterday(-ish)
raydiak yes
moritz hoelzro: they have a backtrace
raydiak well, they wrap an exception which does
hoelzro ok, that's something =)
skids r: my $f = fail("foo"); $f; 21:17
camelia rakudo-parrot 25da7a: OUTPUT«foo␤current instr.: 'throw' pc 501346 (src/gen/p-CORE.setting.pir:206811) (gen/parrot/CORE.setting:12506)␤called from Sub 'sink' pc 544716 (src/gen/p-CORE.setting.pir:222609) (gen/parrot/CORE.setting:14089)␤called from Sub 'MAIN' pc 396 (src/gen/pe…»
..rakudo-moar 25da7a: OUTPUT«Unhandled exception: foo␤ at <unknown>:1 (/home/camelia/rakudo-inst-1/languages/perl6/runtime/CORE.setting.moarvm:throw:4294967295)␤ from src/gen/m-CORE.setting:14149 (/home/camelia/rakudo-inst-1/languages/perl6/runtime/CORE.setting.moarvm:sink:29)…»
raydiak so the damage is mitigated, but if I already said "--> Int" somewhere earlier and still ended up with this failure thing getting out, I'd be kinda...unhappy if it was one of my first P6 experiences
moritz jnthn: erm, nope, it doesn't :(
dalek c: 0b7b112 | moritz++ | htmlify.p6:
Hide link that is a 404 now
 21:18
jnthn tony-o_: On JVM, and yeah, I think the Makefile is the place to look
tony-o_ jnthn: i changed Xss from 1m to 500m , i didn't touch Xmx because the JVM doesn't get even close to that number 21:21
stil the same error
moritz tony-o_: did you reconfigure?
tony-o_ the box has over a gig of memory left when make pukes
21:21 kaare_ joined
tony-o_ moritz: i do the configure and then change the make file 21:21
jnthn tony-o_: Xss is the one I think, but 500m is crazy...try more like 2m or 5m 21:22
hoelzro raydiak: I can understand that
tony-o_ jnthn: i tried 5m first with same error so i went nuts with it
jnthn tony-o_: ohh, but now I notice
/home/perl6/bash-perl6-burner/wa/p6/../sixes/2014.12/bin/nqp-j
It's that shell script that means an -Xss I guess 21:23
agentzh could anyone help with this little NQP program? gist.github.com/agentzh/dc53d8f415ac0c6a0260 I keep getting the "cannot stringify this" error. 21:24
but the equivalent works in rakudo.
seems like $match<Expr><Item> is a "null object" somehow. 21:25
moritz agentzh: string do you pass on the command line?
agentzh i passed.
tony-o_ jnthn: same error still with 1G allocated to the nqp jvm and -Xss set to 5m
moritz agentzh: token Expr { [<Item> <ws>]+ } 21:26
agentzh: the + means that the <Item> capture is a list/array
agentzh: try say($match<Expr><Item>[0])
jnthn tony-o_: OK, then...I'm confused. :( 21:27
tony-o_ :(
moritz agentzh: in rakudo, lists know how to coerce themselves in string context
agentzh: nqp isn't that forgiving
21:27 booly-yam-4652 left
agentzh moritz: i see, thanks! 21:27
dalek p/6pe: 98c0b61 | jnthn++ | src/vm/parrot/ (2 files):
Add parametricity bits to STable.
agentzh really like NQP's speed :) 21:28
moritz too 21:29
though rakudo-m is also not bad
agentzh yeah, but it still has a startup hit of 200ms on my laptop :)
while nqp-m is just 30ms :) 21:30
moritz aye
tony-o_ jnthn: we'll see if it happens on BSD 32b later too 21:31
agentzh will try to port some of my slooow Rarse::RecDescent compilers over to nqp and rakudo.
and see how it goes.
moritz lizmat: re rakudo bugfix release, that might make sense. At least Pod::To::HTML is badly broken with 2015.01.
TTFN
lizmat moritz: does my fix just now, solve the problem? 21:32
moritz lizmat: yes
lizmat then I'm afraid we'll need a hotfix release, right ? 21:33
what's the procedure for that?
21:33 vendethiel joined
moritz lizmat: there's not much of a procedure for it 21:33
TimToady minimally, put a new tag, and reference it 21:34
moritz lizmat: I'd start a new branch based on the 2015.01 tag
21:34 kurahaupo left
timotimo hello everybody 21:34
what did i miss? :)
TimToady everything!
lizmat messing up a release
moritz (git checkout -b hotfix-2015.01 2015.01)
cherry-pick the commit(s) you want to include
lizmat only one, fortunately 21:35
moritz do the usual ',ale release'
with 2015.01.1 as the version
*'make release'
and when that one works, upload the tarball, add a tag, push the tag
TimToady you're not supposed to release ale, you're supposed to consume it
moritz isn't supposed to be typing, he should be sleeping 21:36
TimToady moritz++ for typing anyway
21:36 rurban joined
Kristien What if it was a bot? 21:36
lizmat will need someone else to do the upload
Kristien Or speach-to-text?
Would we be doomed?
agentzh how should i check the type of an expression in NQP? 21:37
there's no WHAT() nor .WHAT in NQP it seems.
tony-o_ that's what you get with your other 170ms :-)
agentzh hehe
so it's "no"? 21:38
jnthn There's a .WHAT, but no .gist/.Str distinction
say($foo.HOW.name($foo)) # gets you the typename 21:39
agentzh is trying.
jnthn But if you're try trying to do a type check then smartmatch should work, or nqp::istype($obj, TheType)
agentzh got it. 21:40
thanks
jnthn np
21:40 xfix left
lizmat moritz: building on parrot and running spectest now 21:40
Kristien I want to become a professional Perl programmer. 21:42
but I'm stuck with Python and PHP projects 3:
dalek p/6pe: 2c3db63 | jnthn++ | src/vm/parrot/ops/nqp.ops:
Start porting 6pe ops to Parrot; passes 6 tests.
jnthn agentzh: ooc, what Rakudo startup hit would be low enough for it to not be annoying for you? 21:43
21:43 PZt left, Rounin left
PerlJam Kristien: you have 2 options: 1) find some you can convince to pay to you code in Perl 2) start a business of your own where you use perl to solve real-world problems and people pay you for that. 21:44
:-)
agentzh jnthn: under 100ms :)
jnthn agentzh: OK :)
Kristien :v
I couldn't install Parrot because I have no ICU. Does Moar depend on ICU? 21:46
TimToady agentzh: get a machine that's twice as fast? :P
Kristien I'll try to install Moar so I can implement my monitoring tool. 21:48
jnthn Kristien: No, Moar doesn't depend on that.
Kristien: You should also find Moar to be faster :) 21:49
Kristien Parrot doesn't have Proc::Async so I cannot use it anyway :P
lizmat so tagging should be something like: git tag -a -m"tag release #84" 2015.01.01 21:50
moritz: ^^ ?
moritz lizmat: I wouldn't include #84 there 21:51
lizmat just nn?
moritz lizmat: rather 'tag bugfix release 1 for 2015.01' or so
rurban parrot —without-icu
this is even autodetected
Kristien it failed installing 21:52
because ICU was not on my machine
psch rakudo fails, not parrot
Kristien the obvious fix is installing ICU, ol
rurban I thought I had all the necessary workarounds already
21:52 anaeem1_ joined
rurban but yes, moar has aio already, parrot not 21:53
dalek p/6pe: 5feb90a | jnthn++ | src/vm/parrot/ops/nqp.ops:
Port 3 more 6pe ops; 11/14 tests passing.
rurban jnthn++ parrot work 21:55
21:55 denis_boyun_ joined, vendethiel left
dalek ast: 3a2be84 | raydiak++ | S02-lists/indexing.t:
Add test for out of range index failures

Tests pre-hotfix 2015.01 bug fixed by lizmat++ @ f120204aca0276564163c3bc157c0639870d5227		 21:56
lizmat moritz: *sigh* tar building fails because of missing files in t/spec 21:58
21:59 skids left
lizmat moritz: is it supposed to include t/spec in the tar file ? 21:59
moritz lizmat: yes 22:00
lizmat: rm -rf t/spec; make testable
lizmat: iirc the release process replaces t/spec/ with a non-git directory, or something weird 22:01
lizmat $ make testable
make: *** No rule to make target `testable'. Stop.
:-(
moritz make t/spec 22:02
lizmat: sorry, my bulid-fu is rusty
lizmat I just reconfigured and am rebuilding
ab5tract tony-o_: rakudo compiled fine for me on i686 archlinux 22:03
dalek p/6pe: 44e9eb2 | jnthn++ | src/vm/parrot/ops/nqp.ops:
Implement interning, to pass 6pe tests.
 22:05
lizmat moritz: after make t/spec, make release still fails because of missing files 22:09
like S32-list/uniq.t
22:12 denis_boyun_ left
jnthn lizmat: You already nuked your t/spec and got a fresh checkout of it? 22:14
22:14 vendethiel joined
lizmat yes 22:14
but is also complaining about missing files in src/core
jnthn o.O 22:15
22:16 denis_boyun_ joined, jmark left
lizmat a stub file for the newio branch 22:16
that I removed about 2 weeks ago?
src/core/IO/Locally
jnthn And it's not on disk? 22:18
lizmat nope
feels like it is somehow still at 2014.12 22:19
jnthn wonders where on earth it gets it from
lizmat well, the file used to exist at some point in the past
ls 22:24
I touched the missing files and got a tar now 22:25
22:25 telex left
lizmat testing that tar 22:25
22:26 telex joined 22:29 mvuets joined
Kristien m: Duration.new(1).Real 22:29
camelia ( no output )
Kristien m: say Duration.new(1).Real.WHAT 22:30
camelia rakudo-moar f12020: OUTPUT«(Duration)␤»
Kristien Shouldn't that be Real?
lizmat yes
Kristien Oh. Duration <: Real
jnthn m: say Duration ~~ Real
camelia rakudo-moar f12020: OUTPUT«True␤»
Kristien oops :P 22:31
lizmat jnthn: but the fact that it does the role Real, should make it of that type, no?
method Real(Real:D:) { self } # in Real.pm 22:32
masak ooh, we have literal syntax for object hashes now?
squee!
lizmat notices that masak does not read the p6w :-)
.oO( or much later :-)
masak I really should do that. 22:33
this was from just catching up on p6l.
lizmat jnthn: I think Duration needs it's own "method Real(Duration:D:) { self.Num } ? 22:34
*its 22:35
jnthn lizmat: No, it already *is* a Real
lizmat: And .Real on a Real should be identity
lizmat eh. but it *does* a Real 22:36
Kristien Cool, Promise#then doesn't autocollapse.
m: my ($s, $t) = Supply.new, Supply.new; my ($p, $q) = $s.Promise, $t.Promise; my $x = $p.then({ $q; }); $t.emit(3.14); $s.emit(42); say ($x.result.WHAT, $x.result.result.WHAT)
jnthn m: say 42.Real.WHAT
camelia rakudo-moar f12020: OUTPUT«(Promise) (Rat)␤»
rakudo-moar f12020: OUTPUT«(Int)␤»
22:37 denis_boyun_ left
lizmat jnthn: ah, ok 22:37
Kristien JS promises typically do, making it impossible to write generic code.
lizmat I guess there's more cases like that :-)
jnthn Kristien: That'd be a bit...magical...for my liking :)
lizmat Kristien: not much of a promise then :-)
jnthn: ok, so Duration.Real -> Duration
jnthn lizmat: Yes 22:38
lizmat gotcha
jnthn Dammit. r-p does a SEGV in the 6pe-mop branch :(
dalek p/6pe: 91f6f52 | jnthn++ | src/vm/parrot/6model/serialization.c:
Port basic 6pe serialization to Parrot.
 22:39
lizmat ok, tar file built ok, spectested ok (complaned a bit about missing test files, but there you go) 22:40
22:42 rurban left
jnthn gets a busted stack in the debugger also. wtf. 22:48
22:49 ajr joined 22:50 ajr is now known as Guest18017, Guest18017 is now known as ajr_
Kristien RxP6 would be fun to see. 22:51
jnthn Kristien: That's basically what Supply is :)
Note the *whole* Rx, but the observer bit and a bunch of the combinators. 22:52
FROGGS_ Not*
jnthn uh, yes
FROGGS_ :o)
Kristien Yeah.
22:52 adu left
dalek kudo/6pe-mop: 432c65a | jnthn++ | tools/build/Makefile-Parrot.in:
Add CoercionHOW into the r-p Makefile.
 22:53
Kristien ok let's use Supply instead of Channel for the ping monitor
Whooo it works shiny 22:54
ab5tract Kristien: can i see? :D 22:57
dalek kudo/hotfix-2015.01: f0b3a47 | lizmat++ | src/core/ (6 files):
< 0 exceptions should fail rather than throw

Spotted by FROGGS++ . Not sure of the ecosystem fallout of this: could warrant a hot fix update of the compiler release before R* ?
Kristien ab5tract: gimme a sec, fixing one more thing
FROGGS_ also spotted by moritz++ 22:58
lizmat Hotfix release is now on rakudo.org: rakudo.org/downloads/rakudo/
FROGGS_++ for helping out 22:59
23:01 doobi-sham-25434 joined
dalek Heuristic branch merge: pushed 16 commits to rakudo/newio by lizmat 23:01
lizmat and on that note, I'm going to call it a night
raydiak good night lizmat++ 23:02
FROGGS_ gnight lizmat 23:03
I really love precomp... the libxml tests take 5s precomp'd instead of 28s :o) 23:05
23:07 FROGGS_ left, ptc_p6 left
jnthn 'night, lizmat 23:08
Hm, think I'll consider myself done for the day also :) 23:09
japhb o/ 23:10
23:11 anaeem1_ left
TimToady but the night is yet young! 23:11
japhb The night is not yet night!
jnthn Not in this timezone :P 23:13
geekosaur it's pretty night here... 23:15
23:17 japhb left
dalek Heuristic branch merge: pushed 17 commits to nqp by jnthn 23:18
jnthn merged 6pe in nqp
All backends pass the new test file there now, so no longer a need to keep it out of master. :)
TimToady \o/
jnthn 6pe-mop in Rakudo could be a while longer; we'll see. 23:19
TimToady If it's turtles all the way down, we'll hope it's hares all the way up.
jnthn :) 23:20
23:21 ajr_ left 23:23 spollei joined 23:24 japhb joined, beastd left
Kristien Is there a pastebin with Perl 6 syntax highlighting? 23:26
23:26 doobi-sham-25434 left
Kristien Ah, pastebin.com 23:26
ab5tract: pastebin.com/udu7qCiK 23:27
23:27 rurban_ left
japhb What are they using for Perl 6 highlighting? 23:29
psch m: say "1241.12512" ~~ / \d ** 2 % '.'/
camelia rakudo-moar f12020: OUTPUT«｢1.1｣␤»
psch m: say "1241.12512" ~~ / [\d+] ** 2 % '.'/ 23:30
camelia rakudo-moar f12020: OUTPUT«｢1241.12512｣␤»
psch m: say "1241.12512" ~~ / \d+ ** 2 % '.'/
camelia rakudo-moar f12020: OUTPUT«===SORRY!===␤Quantifier quantifies nothing␤at /tmp/nojLF2Ee6_:1␤------> say "1241.12512" ~~ / \d+ ** 2⏏ % '.'/␤Missing quantifier on the left argument of %␤at /tmp/nojLF2Ee6_:1␤------> say "1241.12512…»
japhb m: say "1241.12512" ~~ / \d+**2%'.'/
camelia rakudo-moar f12020: OUTPUT«===SORRY!===␤Quantifier quantifies nothing␤at /tmp/0a7ESU76WQ:1␤------> say "1241.12512" ~~ / \d+**2⏏%'.'/␤Missing quantifier on the left argument of %␤at /tmp/0a7ESU76WQ:1␤------> say "1241.12512" ~~…»
TimToady gotta use brackets there 23:31
japhb OOC, why?
TimToady the **2 doesn't really know the % is coming :) 23:32
23:32 panchiniak left
TimToady maybe it could be taught 23:32
but it's not clear that it would be clearer 23:33
for the rest of regex, one atom and one quantifier is a pretty strong self-clocking feature 23:34
japhb Fair enough
Kristien m: class A { has Proc::Async $.proc; }; say A.new.proc.WHAT; 23:38
camelia rakudo-moar f12020: OUTPUT«Proc::Async is disallowed in restricted setting␤ in sub restricted at src/RESTRICTED.setting:1␤ in method gist at src/RESTRICTED.setting:33␤ in sub say at src/gen/m-CORE.setting:16915␤ in block <unit> at /tmp/iMGvfWdSua:1␤␤»
Kristien m: class A { has Supply $.proc; }; say A.new.proc.WHAT;
camelia rakudo-moar f12020: OUTPUT«(Supply)␤» 23:39
Kristien Hmm, interesting.
jnthn TimToady: Not sure the **2 not knowing the % is coming is the issue, it's more than the + was the quantifier, and the **2 then has nothing to quantify. 23:41
('cus we're onto the next atom by then)
masak er.
std: /a+ +/
camelia std f9b7f55: OUTPUT«===SORRY!===␤Quantifier quantifies nothing at /tmp/kwR60Sa920 line 1:␤------> /a+ +⏏/␤Check failed␤FAILED 00:00 135m␤»
masak ok, that seems to be the way it actually works. 23:42
fancy that.
jnthn has spent some time in the grammar that parses it :P
masak :)
I bet I used to know that. I once re-implemented PGE.
jnthn Well, but PGE didn't use a Perl 6 grammar to parse its regexes, iirc?
masak no, it used an expression parser with parameterized ops. 23:43
23:48 muraiki_ joined, adu joined 23:49 spider-mario left
jnthn 'night, #perl6 23:52
masak 'night, jnthn 23:53

Powered by Cro and the Raku® Programming Language.

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

Thank you!