»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg p6eval perl6: ... | irclog: irc.perl6.org/ | UTF-8 is our friend!
Set by sorear on 4 February 2011.
00:11 donri left
colomon sorear: is there any easy / good way to define constants in C#? My code would be a lot cleaner if I had constants for pi and i 00:31
sorear colomon: public static readonly Complex i = ...; 00:35
and Math.PI
colomon .PI, not .Pi?
sorear correct, PI
colomon gives C# designers a dirty look 00:36
sorear++
jnthn Well, it's not mandated by the language. But the C# cultural convention is that CONSTANTS SHOULD BE SHOUTED OUT LOUD.
:)
sorear yeah. I find int.MaxValue to be much wronger than Math.PI 00:37
jnthn heh, what, you want consistency too? :)
oh what, how's it this late...and I should actually get up to work in the morning... 00:38
sleep & 00:41
colomon jnthn: o/
oooo, autopromotes Int to Double. handy 00:45
sorear yeah C-family languages do that as a rule
colomon I still don't quite have a feel for when C# is slick like that and when it is clunky. and I tend to expect clunky. 00:47
00:51 gbacon joined, Chillance left 00:54 am0c joined, whiteknight left
colomon btw, the other thing that keeps killing me is I don't want to use () after method calls with no arguments. :) 01:11
sorear Heh
01:23 am0c left 01:24 am0c joined 01:31 am0c left, am0c joined
[Coke] tchrist-- # general (#&$ery on p5p. 01:34
dalek ast: ca8ada6 | Coke++ | S (4 files):
rakudo nom fudging.
 01:36
kudo/nom: 35ec9c7 | Coke++ | t/spectest.data:
track failures, run fudged tests.
 01:37
sorear [Coke]: wah? 01:41
01:49 wamba left
colomon oh, shoot. C#'s Math doesn't have all the trig functions. 01:49
that makes the super-simple version I had in mind not work. 01:50
01:58 orafu left, orafu joined 02:00 abercrombie joined
colomon does not enjoy hand correcting his autogenerated code. 02:03
dalek ecza/serialize: b62a6bb | sorear++ | / (5 files):
checkpoint 1 of executable code in new model
 02:06
02:10 bluescreen10 joined 02:12 miguet left 02:17 s_mosher joined
colomon sorear: got all the non-hyperbolic trig functions (except atan2) passing all spectests. but I need to reformulate the code generation a bit, and I want to check with the suggested implementation of the complex functions anyway. still, nice progress tonight. 02:17
s_mosher hey sorear, I want to express appreciation for the CLR integration work you've been doing lately with Niecza 02:19
sorear colomon++ # excellent 02:24
s_mosher: Thank you! I need to know that I'm doing useful things sometimes :)
02:40 gbacon left 02:41 daniel-s joined
s_mosher sorear, for sure. I think it's a great strength for Niecza to have. 02:55
I tend to work with Rakudo mostly, but I'm planning something that will be better suited to Niecza. (It wouldn't be remotely practical with Rakudo today.) 02:57
out for now 03:05
03:05 s_mosher left
colomon afk # bed 03:06
03:06 wolfman2000 joined 03:24 Reaganomicon left 03:49 envi_laptop joined 03:59 masonkramer left 04:01 packetknife joined 04:03 packetknife left, abercrombie left 04:10 [particle]1 left 04:14 birdwindupbird joined, satyavvd joined 04:18 REPLeffect joined 04:21 tokuhiro_ left 04:24 jamtech left 04:29 tokuhiro_ joined, woosley joined 04:31 araujo left
moritz good morning 04:32
woosley good *, #perl6 04:49
04:51 uasi joined 05:03 soh_cah_toa left 05:04 mberends left 05:05 kaleem joined 05:10 JimmyZ joined
sorear hello moritz 05:20
05:27 koban` joined, koban` left 05:30 mberends joined 05:31 uasi left
dalek ast: e2676ba | jamesrom++ | README:
Typo.
 05:33
ast: f3dfe0e | moritz++ | README:
Merge pull request #6 from jamesrom/patch-1

Typo.
05:45 am0c left, _jaldhar left 05:52 JimmyZ left 05:53 wtw joined 05:55 mberends left 06:00 packetknife joined 06:08 orafu left 06:10 orafu joined 06:11 REPLeffect left 06:12 wamba joined 06:13 mberends joined 06:25 REPLeffect joined 06:27 SHODAN joined, mberends left 06:45 static_perl left 06:47 agentzh joined 06:49 fridim_ left
agentzh hi, can nqp compile source files to pbc/pir files that can be run by parrot later directly? 06:50
sorear yes 06:52
this is in fact how nqp works
moritz agentzh: yes, that's how the bootstrap works
agentzh moritz: any doc on the command line syntax?
06:53 dayangkun joined, mishin joined
moritz agentzh: dunno about docs, but --target=pir -o out.pir is what you need 06:53
agentzh moritz: cool, thanks! i'll try it out :)
moritz and then parrot -o out.pbc out.pir
there are plenty of examples in the Makefile 06:54
agentzh moritz: cool, it works, though not making things faster ;) 06:55
06:56 uasi joined
agentzh moritz: i'll read Makefile in detail, thanks :) 06:57
Woodi do copy of rakudo/src/core/* to dir in PERL6LIB just work ?
06:58 mj41 joined
Woodi hmm, or it is consider builtin ? 06:58
*considered 06:59
06:59 drbean left
sorear bad idea 07:00
the stuff in src/core/* isn't Perl modules
they're setting fragments
subtly differen
Woodi problem is IO::Select::INET do not work for me and it is in core... 07:01
so how to use i-s-i ? 07:02
debian-rakudo here
sorear don't use it 07:04
just access it
b: say IO::Socket::INET.^methods 07:05
p6eval b 1b7dd1: OUTPUT«newexceptionperlmaybefailBooldefinedStrNumIntcanisadoeslistACCEPTSStrNumericSeqallanyonenonejoinelemsflatmapsortrotatefirstgrepreverseendminmaxminmaxpickrollclassifyreduceuniqkvkeysvaluespairspostcircumfix:<[ ]>at_pospostcircumfix:<{ }>at_keyclonedefineditemblessBUILDBU…
sorear and it's spelled Socket, not Select
Woodi ah :)
right
thanks sorear
sorear sleep 07:07
07:14 dayangkun_ joined 07:17 dayangkun left 07:19 lumi_ left 07:22 araujo joined, araujo left, araujo joined 07:32 woosley left, woosley joined, woosley left
ingy plobsing: a pegex parser applies a compiled pegex grammar to an input stream, and then calls a method in a "receiver" class for every matched subrule. the default receiver returns a hash that looks like {subrule_name => [terminals]} 07:42
plobsing: so... both :)
plobsing: by having your action methods in a receiver class (instead of inside the grammar rules themselves) you can use the same grammar against many differnt receivers 07:44
moritz it's the same concept as the 'actions' in Perl 6 grammars 07:45
and so far I haven't seen a single grammar with more than one actions class/receiver 07:46
ingy moritz: well I can easily imagine it :)
say you have a yaml grammar for instance
moritz ingy: it's not a problem of imagination :-) 07:47
ingy: I can think of many cases too, I just said that I've never encountered such a case in the wild
ingy then everyone who wants to work with yaml in any special way will simply define a receiving class
moritz since the structure of captures is a result of the implementation, it basically becomes part of the grammar's API 07:48
so it becomes hard to change a grammar later on, while retaining the structure of the captures
07:48 wamba left
ingy noted 07:49
moritz that's kinda obvious, but it's less obvious if there's a good solution to that 07:51
p6's capture aliasing makes it possible (but still hard) to maintain the structure
ingy i'm spending a lot of time trying to make my default structures as clean as possible, and adding a few simple syntax hints to pegex 07:54
i'm spending a lot of time trying to make my default structures as clean as possible, and adding a few simple syntax hints to pegex to make the structures cleaner 07:55
so that the receiving classes can be minimal
tadzik good morning
ingy o/
07:56 daxim joined
ingy moritz: gist.github.com/1226129 07:56
pester pester... :) 07:57
moritz ingy: I too find that I don't often use <a> ** <b> 07:59
ingy: it's occasionally useful though 08:01
08:01 packetknife left
ingy moritz: I'll write up a list of variations of that that I would like to handle nicely and maybe you can help me find a reasonable syntax 08:03
moritz ingy: ok
ingy moritz: I need a**b because that can return a list of [a+] whereas a [b a]* has to return a form of [a,[a*]] and that just cascades into complexity 08:05
moritz ingy: why can't multiple captures of the same name simply return a list? 08:06
mishin what is latest release racudo ro windows?
08:06 uasi left
ingy a: b? c b+ 08:07
moritz mishin: rakudo has source releases which you can compile on windows too
ingy I think combining the b's in that case is not what you want
moritz mishin: I think the last winidows binary was 2011.04 or so 08:08
ingy: hm, right. Perl 6 solves that by allowing you to write (b?) c b+ or (b)? c b+
ingy I want pegex rules to follow very obvious, er, rules
mishin thanks is it planned to compile new release?
moritz somebody does that occasinally. 08:09
mishin: you could do it too
and share the result with the community
ingy moritz: bbl
08:14 dayangkun__ joined
mishin @motitz: Good idea, it is necessary to understand only how it to make and that to me it is necessary for this purpose 08:14
08:18 dayangkun_ left 08:21 lumi_ joined
jnthn monring o/ 08:25
moritz \o jnthn 08:26
jnthn +# S32-temporal/DateTime-Instant-Duration.t # err: Method 'eval' not found for invocant of class 'Str'
Well, that looks like lhf. ;) 08:27
[Coke]++ # tracking failure modes
08:28 agentzh left 08:29 daniel-s left, daniel-s joined
ingy moritz: one thing to mention is that pegex grammars always assume '^ ...grammar... $', (since Pegex is intended to be a parser, not a regex replacement) 08:34
I only thought of that because I was just considering adding a parser option to not enforce the $ to do a partial parse :) 08:36
08:38 Trashlord joined
moritz ingy: makes sense 08:40
dalek Heuristic branch merge: pushed 484 commits to rakudo/nom-exceptions by moritz
moritz oops, accidnetally pushed to the wrong branch :-)
dalek kudo: d7f45ff | moritz++ | src/core/Cool.pm:
Cool.eval
 08:41
08:46 odoacre_ joined 08:55 dayangkun__ left 08:56 dayangkun__ joined, dayangkun__ left 08:58 replore left, REPLeffect left, REPLeffect joined 08:59 PZt left 09:04 woosley1 joined 09:06 am0c joined 09:07 kaleem left, kaleem joined 09:08 replore joined 09:13 drbean joined 09:17 am0c left 09:18 am0c joined, PZt joined 09:20 uasi joined
moritz nom: say now.perl.eval 09:32
p6eval nom 35ec9c: OUTPUT«Method 'eval' not found for invocant of class 'Str'␤ in <anon> at /tmp/seKsoO4hPv:1␤ in <anon> at /tmp/seKsoO4hPv:1␤␤»
moritz evalbot rebuild nom
p6eval OK (started asynchronously)
09:32 tokuhiro_ left 09:39 wamba joined 09:46 am0c left
Woodi perl6: class A { has $.b = 10; }; sub foo() { say self.b }; my $b = A.new; $b.foo() 09:46
p6eval niecza v9-32-g380d891: OUTPUT«===SORRY!===␤␤'self' used where no object is available at /tmp/FpMaFHP0Pk line 1:␤------> { has $.b = 10; }; sub foo() { say self⏏.b }; my $b = A.new; $b.foo()␤␤Potential difficulties:␤ &foo is declared but not used a…
..rakudo 35ec9c: OUTPUT«===SORRY!===␤Symbol 'self' not predeclared in foo (/tmp/F3btE1vtBR:1)␤»
..pugs: OUTPUT«*** No compatible multi variant found: "&foo"␤ at /tmp/cS5G0ZtE0I line 1, column 70 - line 2, column 1␤»
Woodi pls someone tell me how to call method in perl6 09:47
imho DWIM is heavy broken in perl6
moritz Woodi: well, it works just fine if you actually write a method, not a sub 09:48
arnsholt nom: class A { hsa $.b = 10; method foo() { say self.b; }; }; my $a = A.new; $a.foo
Woodi perl6: class A { has $.b = 10; }; method foo() { say self.b }; my $b = A.new; $b.foo()
p6eval nom 35ec9c: OUTPUT«Null PMC access in find_method('b')␤ in <anon> at /tmp/sCgQyi8yuW:1␤ in <anon> at /tmp/sCgQyi8yuW:1␤ in <anon> at /tmp/sCgQyi8yuW:1␤␤»
rakudo 35ec9c: OUTPUT«Useless declaration of a has-scoped method in mainline␤Method 'foo' not found for invocant of class 'A'␤ in <anon> at /tmp/FW3IOJlmdA:1␤ in <anon> at /tmp/FW3IOJlmdA:1␤␤»
..niecza v9-32-g380d891: OUTPUT«===SORRY!===␤␤Methods must be used in some kind of package at /tmp/hyx7Uu5g5b line 1:␤------> class A { has $.b = 10; }; method foo⏏() { say self.b }; my $b = A.new; $b.fo␤␤Potential difficulties:␤ 'method' declaration…
..pugs: OUTPUT«10␤»
moritz Woodi: you can't expect a programming language to do what you mean if it contradicts with what you tell it.
arnsholt Or something. My Perl 6 is a bit rusty 09:49
09:49 am0c joined
arnsholt At any rate, as moritz says, you need to place the method declaration inside the class 09:49
moritz arnsholt: probably works if you spell 'has' correctly :-)
arnsholt Ooooooh, right. That might help =D
Woodi moritz: DWIM usually allow to say what i mean and language gueses right :)
arnsholt nom: class A { has $.b = 10; method foo() { say self.b; }; }; my $a = A.new; $a.foo # Once more, with feeling
p6eval nom 35ec9c: OUTPUT«10␤» 09:50
arnsholt There we go.
moritz arnsholt++ # with feeling
09:50 envi_|2 joined
Woodi ah 09:50
arnsholt Woodi: But that'll explode messily when you have class A {}; class B {}
09:51 envi_laptop left
arnsholt It's the kind of dwimmery that'll only come back and bite you because figuring it out properly is an AI-complete problem (or close enough) 09:51
moritz DWIM must be applied con spirito, ma non troppo :-)
Woodi i just know how to write 'espresso' and 'cappucino' :)
perl6: class A { has $.b = 10; }; method foo() { say self.b; }; my $b = A.new; $b.foo() 09:53
p6eval rakudo 35ec9c: OUTPUT«Useless declaration of a has-scoped method in mainline␤Method 'foo' not found for invocant of class 'A'␤ in <anon> at /tmp/ar6dwzEOi9:1␤ in <anon> at /tmp/ar6dwzEOi9:1␤␤»
..pugs: OUTPUT«10␤»
..niecza v9-32-g380d891: OUTPUT«===SORRY!===␤␤Methods must be used in some kind of package at /tmp/rVGT1p0wCF line 1:␤------> class A { has $.b = 10; }; method foo⏏() { say self.b; }; my $b = A.new; $b.f␤␤Potential difficulties:␤ 'method' declaration…
09:53 JimmyZ joined
Woodi perl6: class A { has $.b = 10; }; sub foo($self) { say $self.b }; my $b = A.new; $b.&foo 09:54
p6eval pugs, rakudo 35ec9c, niecza v9-32-g380d891: OUTPUT«10␤»
moritz Woodi: if the method is meant to go into the clas, write it inside the class
Woodi but why it working
*is is *? 09:55
goot that works but strange
perl6: class A { has $.b = 10; }; method foo($self) { say $self.b }; }; my $b = A.new; $b.foo 09:56
moritz because methods and subs are really just the same thing, they are just being searched for in different ways
p6eval pugs: OUTPUT«*** ␤ Unexpected "};"␤ expecting ";", identifier, Doc block, block declaration, declaration, construct or expression␤ at /tmp/GjmwrntSVA line 1, column 64␤»
..rakudo 35ec9c: OUTPUT«Useless declaration of a has-scoped method in mainline␤===SORRY!===␤Confused at line 1, near "}; my $b "␤»
..niecza v9-32-g380d891: OUTPUT«===SORRY!===␤␤Methods must be used in some kind of package at /tmp/KpfJuDuNfV line 1:␤------> class A { has $.b = 10; }; method foo⏏($self) { say $self.b }; }; my $b = A.n␤␤Unexpected closing bracket at /tmp/KpfJuDuNfV…
arnsholt That's the monkey patching approach, essentially. Put the method declaration -inside- the braces of the class, and you can drop the .& syntax
Woodi perl6: class A { has $!b = 10; }; method foo() { say self.b }; }; my $b = A.new; $b.foo
jnthn The syntax you're looking for is "my method foo ..." then "$b.&foo"
p6eval rakudo 35ec9c: OUTPUT«Useless declaration of a has-scoped method in mainline␤===SORRY!===␤Confused at line 1, near "}; my $b "␤»
..niecza v9-32-g380d891: OUTPUT«===SORRY!===␤␤Methods must be used in some kind of package at /tmp/9eKhcynh6Z line 1:␤------> class A { has $!b = 10; }; method foo⏏() { say self.b }; }; my $b = A.new; $b␤␤Unexpected closing bracket at /tmp/9eKhcynh6Z…
..pugs: OUTPUT«*** ␤ Unexpected "};"␤ expecting ";", identifier, Doc block, block declaration, declaration, construct or expression␤ at /tmp/p8QqoC2JdS line 1, column 58␤»
moritz perl6: class A { has $.b = 10 }; my method foo() { self.b }; say A.new.&foo 09:57
p6eval niecza v9-32-g380d891: OUTPUT«===SORRY!===␤␤Methods must be used in some kind of package at /tmp/2pLDg_kLPp line 1:␤------> class A { has $.b = 10 }; my method foo⏏() { self.b }; say A.new.&foo␤␤Unhandled exception: Check failed␤␤ at /home/p6eval/ni…
..pugs, rakudo 35ec9c: OUTPUT«10␤»
JimmyZ moritz: there are not same if /$.b/$!b/
Woodi perl6: class A { has $!b = 10; method foo() { say self.b }; }; my $b = A.new; $b.foo
p6eval pugs: OUTPUT«*** No such method in class A: "&b"␤ at /tmp/WtRi0z4m_h line 1, column 45-52␤»
..niecza v9-32-g380d891: OUTPUT«Unhandled exception: Unable to resolve method b in class A␤ at /tmp/Nf0AdwQpJj line 1 (MAIN A.foo @ 2) ␤ at /tmp/Nf0AdwQpJj line 1 (MAIN mainline @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 2060 (CORE C968_ANON @ 2) ␤ at /home/p6eval/niecza/lib/…
..rakudo 35ec9c: OUTPUT«Method 'b' not found for invocant of class 'A'␤ in method foo at /tmp/XDwEmaEC3W:1␤ in <anon> at /tmp/XDwEmaEC3W:1␤ in <anon> at /tmp/XDwEmaEC3W:1␤␤»
moritz JimmyZ: right, but that's more of a difference in the compilation context than in the generated code object
09:58 am0c left
Woodi perl6: class A { has $!b = 10; method foo() { say self!b }; } my $b = A.new; $b.foo 09:58
p6eval niecza v9-32-g380d891: OUTPUT«===SORRY!===␤␤Strange text after block (missing comma, semicolon, comment marker?) at /tmp/7D5IP8Gk3q line 1:␤------> !b = 10; method foo() { say self!b }; }⏏ my $b = A.new; $b.foo␤␤Parse failed␤␤»
..rakudo 35ec9c: OUTPUT«===SORRY!===␤Confused at line 1, near "class A { "␤»
..pugs: OUTPUT«10␤»
moritz need a ; after the } that closes the class
or a newline
Woodi perl6: class A { has $!b = 10; method foo() { say self!b }; }; my $b = A.new; $b.foo 09:59
p6eval rakudo 35ec9c: OUTPUT«Method 'postcircumfix:<( )>' not found for invocant of class 'Mu'␤ in <anon> at src/gen/Metamodel.pm:3029␤ in method dispatch:<!> at src/gen/CORE.setting:644␤ in method dispatch:<!> at src/gen/CORE.setting:656␤ in method foo at /tmp/ak6_fPNeDy:1␤ in <anon> at …
..pugs, niecza v9-32-g380d891: OUTPUT«10␤»
Woodi perl6: class A { has $!b = 10; method foo() { say self.b }; }; my $b = A.new; $b.foo;
p6eval rakudo 35ec9c: OUTPUT«Method 'b' not found for invocant of class 'A'␤ in method foo at /tmp/Q7JQMfJEvA:1␤ in <anon> at /tmp/Q7JQMfJEvA:1␤ in <anon> at /tmp/Q7JQMfJEvA:1␤␤»
..pugs: OUTPUT«*** No such method in class A: "&b"␤ at /tmp/E2Bt12eIfM line 1, column 45-52␤»
..niecza v9-32-g380d891: OUTPUT«Unhandled exception: Unable to resolve method b in class A␤ at /tmp/UmTG8luWKq line 1 (MAIN A.foo @ 2) ␤ at /tmp/UmTG8luWKq line 1 (MAIN mainline @ 3) ␤ at /home/p6eval/niecza/lib/CORE.setting line 2060 (CORE C968_ANON @ 2) ␤ at /home/p6eval/niecza/lib/…
Woodi perl6: class A { has $!b = 10; method foo() { say $b }; }; my $b = A.new; $b.foo;
p6eval pugs: OUTPUT«*** ␤ Unexpected " };"␤ expecting "::"␤ Variable "$b" requires predeclaration or explicit package name␤ at /tmp/YEGH9P9njR line 1, column 47␤» 10:00
..niecza v9-32-g380d891: OUTPUT«===SORRY!===␤␤Variable $b is not predeclared at /tmp/TcEO80zcIS line 1:␤------> s A { has $!b = 10; method foo() { say ⏏$b }; }; my $b = A.new; $b.foo;␤␤Unhandled exception: Check failed␤␤ at /home/p6eval/niecza/boot/…
..rakudo 35ec9c: OUTPUT«A<70099768>␤»
Woodi i give up :)
JimmyZ rakudo: class A { has $!b = 10; method foo() { say self!b }; }; my $b = A.new; $b.foo # I think it's a bug
p6eval rakudo 35ec9c: OUTPUT«Method 'postcircumfix:<( )>' not found for invocant of class 'Mu'␤ in <anon> at src/gen/Metamodel.pm:3029␤ in method dispatch:<!> at src/gen/CORE.setting:644␤ in method dispatch:<!> at src/gen/CORE.setting:656␤ in method foo at /tmp/FherJzZ4W9:1␤ in <anon> at …
moritz Woodi: I recommend reading the chapter on objects and classes in github.com/downloads/perl6/book/2011.07.a4.pdf
Woodi: and perlcabal.org/syn/S12.html if you want the gory details :-)
JimmyZ LTA bug?
10:01 wamba left
Woodi must be book, S12 not contain any use of class... 10:01
moritz huh?
Woodi any small-complete-example
10:01 woosley1 left
moritz JimmyZ: should probably read 'Cannot find private method b in class A' or so 10:01
JimmyZ b: class A { has $.b = 10; method foo() { say self.b }; }; my $b = A.new; $b.foo 10:02
p6eval b 1b7dd1: OUTPUT«10␤»
JimmyZ b: class A { has $!b = 10; method foo() { say self!b }; }; my $b = A.new; $b.foo
p6eval b 1b7dd1: OUTPUT«Method '!b' not found for invocant of class 'A'␤ in 'A::foo' at line 22:/tmp/kH2gKWVHoK␤ in main program body at line 22:/tmp/kH2gKWVHoK␤»
JimmyZ b++ ;)
moritz beijing++ indeed :-)
JimmyZ submits LTA bug 10:03
moritz JimmyZ++
JimmyZ speak of bug, rt didn't proctect email address 10:06
10:06 JimmyZ left
moritz JimmyZ: are you logged in? 10:06
10:06 JimmyZ joined
JimmyZ moritz: no 10:07
moritz hm
well, in the end a copy goes to [email@hidden.address] and that doesn't protect the email either
or maybe not well
but that's how mailing lists work
JimmyZ would like to protect emails 10:08
szabgab hi, can anyone help me with the rakudo compilation on windows? 10:09
10:09 arthur-_ left
szabgab I sent the failure to perl6-users yesterday but no reaction so far 10:09
it starts by complaining that nqp.exe cannot be found and ends by a windows pop-up 10:10
moritz what does the windows popup say? 10:11
.\nqp.exe --target=pir --output=src\stage2\QRegex.pir src\stage2\QRegex.nqp
gmake: *** [QRegex.pbc] Error -1073741515
that doesn't look good :(
ah right, I see it in your email 10:12
szabgab: is there a libparrot.dll somehwere in the install/ dir?
szabgab nope 10:13
arnsholt szabgab: Win64 or win32?
szabgab 64
but I think strawberry is 32 bit
arnsholt Right. I think jnthn was wrangling some parrot-related problems on win64 the other day
Might be something similar you're running into
moritz szabgab: or in the parrot/ folder maybe? 10:14
szabgab yes I have parrot\libparrot.dll
10:14 arthur-_ joined
moritz szabgab: maybe try copying that to install\bin\ 10:15
szabgab www.nntp.perl.org/group/perl.perl6....g1540.html
moritz it seems that -shared 10:16
"C:\work\rakudo\install\bin\libparrot.dll"
appears in the makefile
so the build system expects it to be there
JimmyZ rm it
szabgab moritz: after copying libparrot.dll and running again perl Configure.pl --gen-parrot I get the same crash 10:17
10:17 wamba joined
szabgab oh in bin 10:17
jnthn Note that for some reason we don't always end up with libparrot.dll getting copied into the right place... 10:18
szabgab there is install\bin\libparrot.dll
jnthn I dunno why...
szabgab and still crashes
JimmyZ rm all libparrot.dll s
jnthn Yeah. If you run Rakudo from build directory it wants libparrot.dll in that directory too
Not a problem after make install, afaik. 10:19
I thougth the makefile had a line to copy it into the right place...will have to take a look and see what came of it.
szabgab I removed the whole install directory and started again 10:21
running perl Configure.pl --gen-parrot
adn I get the crash, should I run some make install now? 10:27
moritz won't do any good until nqp is properly build and installed 10:28
szabgab: what you could try: copy libparrot.dll into the nqp/ folder
jnthn When is it failing
?
szabgab www.nntp.perl.org/group/perl.perl6....g1540.html <- jnthn
moritz szabgab: and then in that folder, run 'gmake install' 10:29
jnthn OK, try what moritz++ suggested
ah, I think we never needed to do that in NQP before QRegex came along; it invokes nqp.exe. We didn't tend to invoke it during nqp build before.
szabgab so copy from install\bin\libparrot.dll to nqp\ ?
jnthn szabgab: yes 10:30
szabgab cd nqp; gmake install
moritz right
szabgab done
and now? 10:31
moritz did it work?
szabgab it seems so
moritz cd .. # back to rakudo
perl Configure.pl # without options
gmake
# and wait....
szabgab waiting :)
same popup crash 10:33
different output on console
gmake: *** [CORE.setting.pbc] Error -1073741515
moritz then also copy libparrot.dll into the rakudo dir
szabgab copy install\bin\libparrot.dll . 10:34
gmake
jnthn suspects we should just get Configure.pl to do this copying.
szabgab let's see if the copying helps at all :) 10:35
jnthn *nod* 10:36
I see you're building against Strawberry too, which I haven't tried. If it's got this far it's likely to work out though. 10:37
szabgab ok, gmake worked
what now?
gmake test 10:38
t\02-embed\01-load.t failed
ok, I have perl6.exe and it works when I am in the rakudo dir 10:40
after running gmake install it also works outside
thanks for your help
now trying gmake spectest 10:41
jnthn Great. I'll try and patch the Configure.pls this evening.
szabgab I see some of the tests in the spec are failing 10:43
is there a way I can easily report them?
10:44 odoacre_ left, odoacre left
szabgab or should I wait with that till thing get stable 10:44
?
10:45 odoacre joined
jnthn szabgab: I'd be interesting to know what fails. 10:45
I've been getting clean test runs here of late. 10:46
(Win64/MSVC)
szabgab t\spec\S02-literals\radix.rakudo .......................... Failed 3/131 subtests
t\spec\S02-magicals\args.t ................................ Failed 1/4 subtests
so far
jnthn Are you on 32 bit or 64? 10:47
szabgab 64 bit machine
Win7
but strawberry is 32 I think 10:48
10:50 wamba left
jnthn ah, then you've probably got a 32-bit build 10:50
We have some issues on 32 bit at the moment.
szabgab ok, I won't have much time to play with rakudo anyway, just wanted to make sure I can at least compile it :) 10:54
jnthn: it seems to be stuck on t\spec\S02-types\instants-and-durations.t ................. 11:02
11:02 breinbaa1 left
jnthn szabgab: Yeah, that's one of the 32-bit issues ;( 11:03
:(
szabgab ok
11:03 breinbaas joined 11:11 wamba joined 11:12 mishin left
snarkyboojum "stuck on an instant" - that's kinda deep 11:13
jnthn It gets stuck in an instant, for a long duration :P 11:15
11:27 donri joined 11:37 Patterner left 11:39 wamba left 11:40 JimmyZ left 11:41 Psyche^ joined, Psyche^ is now known as Patterner 11:44 geekosaur joined 11:47 jql joined
Woodi perl6: class A { has $!b = 10; method foo() { say $!b }; }; my $b = A.new; $b.foo; 11:48
p6eval pugs, rakudo 35ec9c, niecza v9-32-g380d891: OUTPUT«10␤»
Woodi tada :)
do IO::Socket::INET defaults to TCP ? 11:51
moritz I don't think anything else is implemented 11:53
Woodi looks like it depends on parrot implementation 11:55
ingy moritz: can action method return values change whether the match that invoked them is then considered a match? 11:56
moritz ingy: no 11:57
that's what assertions are for.
ingy moritz: good :) 11:58
I don't want to do it, but I thought maybe I was missing something important 11:59
and I already have assertions
=<foo> !<bar> <foo>
12:01 _jaldhar joined
ingy if an action return() the that branch is pruned, if return(undef) then the branch becomes the scalar (undef) 12:02
moritz letting actions influence the parse goes against the whole concept of separating actions and parsing
ingy moritz: right 12:03
very good point
jnthn Plus the action methods never return anything of interest.
They always "make".
moritz right, in p6 they make() stuff
not sure I like that part
ingy jnthn: that's not the case in pegex
tadzik yeah, I got stuck on that a few times 12:04
it feels weird, it's a normal method, but it doesn't return
ingy they are allowed to make stuff, but actually returning stuff looks like the better way out
moritz hey, I just got a nice idea 12:05
12:05 satyavvd left
moritz one can use signature unpacking on $/ in action methods 12:05
jnthn ooh!
tadzik woo
ingy each rule makes something by default, but an action method can change that 12:06
moritz method foo( $/ (:$bar, :$baz) ) { no need for $<bar> or $<baz> here ... }
tadzik looks magic
moritz does signature unpacking work in nom already? 12:07
ingy bbl :P
Woodi perl6: class A { has $!b = 10; method foo() { say $!b.perl }; }; my $b = A.new; $b.foo; 12:10
p6eval pugs: OUTPUT«\10␤»
..rakudo 35ec9c, niecza v9-32-g380d891: OUTPUT«10␤»
12:10 mtk joined
jnthn moritz: yes 12:14
moritz oh, but it doesn't do action methods yet :( 12:16
12:20 colomon left 12:31 packetknife joined 12:32 benabik left, JimmyZ joined
JimmyZ nom: sub foo($/) { say "$<hi>, $<world>"; }; foo({ 'hi'=>'hello', 'world'=>'#perl6'}); #like this? 12:34
p6eval nom 35ec9c: OUTPUT«hello, #perl6␤»
moritz no 12:35
12:35 bluescreen10 left
moritz nom: sub foo($/ (:$hi, :$world)) { say "$hi, $world" }; foo { hi => 'hello', world => '#perl6' } 12:36
p6eval nom 35ec9c: OUTPUT«hello, #perl6␤»
moritz excpt that I'd pass in a Match object, not a hash
JimmyZ b: ( -> { say $^hi, $^world; } )('hello', '#perl6'); 12:40
p6eval b 1b7dd1: OUTPUT«===SORRY!===␤Placeholder variable cannot override existing signature at line 22, near " )('hello'"␤»
JimmyZ couldn''t understand 12:42
12:43 woosley joined
moritz -> { is an empty signature 12:43
JimmyZ thanks 12:45
12:48 bluescreen10 joined 12:52 pernatiy joined 12:54 packetknife left 12:56 [particle] joined
felher rakudo: enum Thing ( large => 'l', small => 's') 12:58
p6eval rakudo 35ec9c: OUTPUT«===SORRY!===␤This type cannot unbox to a native integer␤»
felher Is that regression known?
moritz didn't know about it
felher++
jnthn did. 12:59
Just not doing non-numeric enums yet.
felher jnthn: k :) 13:00
Another thing: 13:02
rakudo: say (1 != 1 & 2); say ('a' !eq 'b' & 'c'); # where is the difference? 13:03
p6eval rakudo 35ec9c: OUTPUT«all(False, True)␤Bool::True␤»
moritz b: say (1 != 1 & 2); 13:05
p6eval b 1b7dd1: OUTPUT«Bool::True␤»
moritz I think that's a nom bug
rakudo: say (1 !== 1 & 2)
p6eval rakudo 35ec9c: OUTPUT«Bool::True␤»
felher moritz: They all should return junctions, shouldn't they? 13:07
moritz felher: no, since they are negated operators 13:08
(which are special-cased)
$a !op $b should be the same as not $a op $b
13:09 kaleem left
felher moritz: oh, okay, i see. So all(1,2) !== 3 is the same as '!all(1,2) == 3'? 13:09
moritz minus precedence issues, yes 13:10
!(all(1,2) == 3)
felher moritz: yeah, shold be 'not' instead of '!' :)
moritz: thnx :)
13:16 Holy_Cow joined, packetknife joined, Holy_Cow left
cognominal nom: eval qq"sub $_ (*\%_ ) \{ elt('$_', |\%_) \}" for < div >; 13:17
p6eval nom 35ec9c: OUTPUT«Lexical '$!' not found␤ in sub eval at src/gen/CORE.setting:421␤ in block <anon> at /tmp/K7bUV8kDof:1␤ in method reify at src/gen/CORE.setting:3660␤ in method reify at src/gen/CORE.setting:3565␤ in method reify at src/gen/CORE.setting:3565␤ in method gimme at sr…
moritz nom: say qq"sub $_ (*\%_ ) \{ elt('$_', |\%_) \}" for < div > 13:18
p6eval nom 35ec9c: OUTPUT«sub div (*%_ ) { elt('div', |%_) }␤»
13:18 lumi_ left
moritz nom: eval '1 2'; say $! 13:18
p6eval nom 35ec9c: OUTPUT«Confused at line 1, near "1 2"␤»
13:20 packetknife left 13:21 lumi_ joined 13:33 kaleem joined 13:37 jimmy1980 joined 13:48 jamtech joined 13:53 PacoLinux joined, uasi left, wamba joined, SHODAN left
[Coke] phenny, tell jnthn I have a 32-bit windows 7 box that I've built nom on before using strawberry & visual studio. ping me if you need testing. 14:00
phenny [Coke]: I'll pass that on when jnthn is around.
14:02 SHODAN joined 14:06 SHODAN left 14:11 thou joined 14:16 wolfman2000 left 14:17 jimmy1980 left
[Coke] yawns. 14:20
14:26 skangas__ joined, skangas left 14:27 wtw left 14:31 kaare_ joined 14:36 uasi joined 14:40 [particle] left 14:42 woosley left 14:45 [particle] joined 14:53 am0c joined 14:55 [particle]1 joined, [particle] left, sayu joined
moritz hates error messages from pdflatex 14:56
they are sometimes quite obscure, and point into completely wrong directions
TimToady phenny: tell ingy that if your grammar assumes ^ $ you can never use a grammar to parse a portion of a string, such as an embedded block in a different language 14:57
phenny TimToady: I'll pass that on when ingy is around.
felher Whats the easiest way to check if $value is in @array? $value == @array.any? 14:58
14:58 sivoais joined
moritz yes 14:59
felher moritz: thnx :)
14:59 Holy_Cow joined
moritz if $value is a number, and you want to compare numerically, that is :-) 14:59
14:59 Holy_Cow left
felher moritz: yes, i do :) 14:59
TimToady and if you don't want it fast :)
mux no more grep($value, @array) ? :)
TimToady you can do that too 15:00
moritz well, if you want it fast, and @array is big, you're doing something wrong :-)
TimToady thinks about content addressabe memories... 15:01
*able
moritz 's memory is also content addressable 15:03
15:06 thou left
cognominal as you get older, you don't loose much content but addressability becomes erratic. 15:08
sorear good * #perl6
cognominal: my addressability has always been erratic... 15:09
15:10 pjcj left, pjcj joined
sorear needs a way to have irssi ignore p6eval's niecza-mentions 15:11
15:11 daniel-s left
moritz sorear: should p6eval misspell niecza? :-) 15:11
(I'm not quite serious)
sorear moritz: hah. No, because then people I care about will start spelling it the same way. :) 15:12
moritz perl6: sub digitize($x) { $x.substr(0, 1) ~ ($x.chars - 2) ~ $x.substr(*-1) }; say digitize('niecza') 15:14
p6eval rakudo 35ec9c: OUTPUT«No applicable candidates found to dispatch to for 'Numeric'. Available candidates are:␤:(Mu, Mu %_)␤␤ in method Numeric at src/gen/CORE.setting:560␤ in sub infix:<<> at src/gen/CORE.setting:1929␤ in method substr at src/gen/CORE.setting:2636␤ in sub digitize at…
..pugs: OUTPUT«n4niecza␤»
..niecza v9-32-g380d891: OUTPUT«n4a␤»
TimToady /niecza.[^v]/ or so 15:15
highlighting goes wonky though 15:16
someone needs to rewrite irssi into P6 :)
doubtless there's some way to manage highlighting with a plugin, if you're devious enough 15:17
and have nothing better to do... 15:18
sorear anyone here played with smuxi? 15:23
15:27 y3llow left, y3llow_ joined, y3llow_ is now known as y3llow 15:29 am0c left 15:31 am0c joined 15:34 static_perl joined 15:36 mtk left 15:37 am0c left 15:40 am0c joined
TimToady jnthn: the fix for .Bool on // seems to need more help; I get "Cannot assign to non-container" in method Bool, which I presume is trying to set $/ 15:42
perhaps the // is closing over the wrong $/ somehow 15:43
also, some kind of similar fix is needed for //.ACCEPTS 15:44
maybe this is pmichaud++ territory
15:44 supernovus joined
TimToady more specifically, "when /(\d+)/ { $0 }" fails to set $0 15:45
15:45 mkramer joined
jnthn TimToady: Yeah, I realized that somehow I'd never quite followed through with implementing current spec on $/ and $! :S 15:48
phenny jnthn: 14:00Z <[Coke]> tell jnthn I have a 32-bit windows 7 box that I've built nom on before using strawberry & visual studio. ping me if you need testing.
15:48 mykhal joined
jnthn TimToady: I need to re-read it and then twiddle. I somehow thought I'd done that already... 15:48
TimToady: And yes, having a pmichaud++ around to look at it would be nice, but when that does happen then getting protoregexes back in place and NFA stuff integrated into nom is kinda higher priority. That's something I'd struggle with knowing how to do, whereas the $/ handling I can most likely figger out. 15:52
:)
TimToady just being a squeaky wheel...
15:52 mtk joined, uasi left
supernovus It seems the features chart is slightly out of date (last updated Aug 20th). I'm not sure how many of the items have been updated since, but I know Date/DateTime are supported in nom now. I can update that, but I'm sure there's other features that should be updated. I'm just not sure which ones. 15:54
15:54 Reaganomicon joined
TimToady feel free to fix the parts you know about, and ignore the rest, or bug other people to fix 15:55
supernovus TimToady: okay, works for me :-)
And since DateTime is working again, I'll get DateTime::Utils fixed up to work with nom. 15:59
jnthn I think we deserve a + on introspection in nom now. ng has one, and I consider nom at least as capable there, if not more so.
Anyone want to explain to me what "Inheritance manipulation" means in the MOP section, so I can know if we do it? :)
I mean, if you're constructing a class using the MOP, you can certainly .^add_parent 16:00
That's how the compiler implements it.
TimToady maybe they mean at run time
tadzik supernovus: having DateTime fixed would be nice too :)
TimToady or maybe they mean picking MRO 16:01
or something WALKish
jnthn class MyMROClassHOW is Metamodel::ClassHOW { method mro($obj) { ... } } # :-) 16:02
TimToady we could probably figger out who to blame for adding it
jnthn I'm quite sure you can implement your own MRO today, it's just diddling with the meta-object :)
Yeah. I don't mind it being there, I'd just like to know whether I can give us a + or only a +/- :)
supernovus tadzik: is it still broken? I haven't done much testing over here yet. I noticed I can at least create DateTime and Date objects again, and was going to throw DateTime up against strftime() and see what sticks ;-) 16:03
tadzik supernovus: there are like 2 failing tests in DateTime.t
TimToady looks like pmichaud++ added it
or last edited the line :) 16:04
probably just translated to json 16:05
supernovus tadzik: I'll look into it, I haven't run the spec tests in far too long. 16:06
tadzik: And since the spec tests aren't fully functional, I'll give nom the +- rating for DateTime for now.
16:06 colomon joined
TimToady yeah, would need to trace back through features.txt 16:07
tadzik sounds good
16:07 janosik joined
dalek atures: f38194f | (Timothy Totten)++ | features.json:
Changed DateTime/Date for nom to +-, changed Introspection for nom to +.
 16:08
16:12 mberends joined 16:13 mj41 left
sorear o/ colomon, mberends 16:17
colomon \o
sorear (janosik too if you want it)
16:18 replore left, pjcj left, pjcj joined 16:19 im2ee joined
im2ee Hi! :) 16:19
sorear Hello and welcome, im2ee.
mberends o/ sorear 16:20
16:23 replore_ joined
moritz nom: say (1 + *).WHAT 16:23
p6eval nom 35ec9c: OUTPUT«WhateverCode()␤»
16:24 am0c left
dalek atures: 1bc3ef4 | moritz++ | features.json:
nom does Whatever currying

also remove an entry where nobody else seems to know what it means		 16:25
supernovus well, there definitely is something weird with the current DateTime... DateTime.new(time) returns the proper UTC time. DateTime.now.utc is returning, err, okay, that's the problem, my current timezone is -0700 and DateTime.now is returning +0700... 16:27
pmichaud good morning, #perl6 16:29
mls_ hi pmichaud!
mberends o/ pmichaud
sorear hi pmichaud
16:31 kaleem left
[Coke] supernovus: checkout the recent todo'd spectests marked 'nom regression', that sounds familiar. 16:32
pmichaud mainly stopping in to say "I'm still around" -- things have been a little hectic around here the past few days 16:33
(all is relatively well, though)
sorear how are we for #phasers in 3040?
moritz nom: say 3040 div 24, ' ', 3040 % 24 16:34
p6eval nom 35ec9c: OUTPUT«126 16␤»
supernovus it seems the internal local-timezone class is doing something wrong... it's returning 25200 as my timezone offset instead of -25200 ... 16:35
moritz so, sign error?
flussence
.oO( didn't we have a divmod function? )
phenny flussence: 18 Sep 08:30Z <tadzik> tell flussence since nom is now default, please feel free to file every Pod-related issue you have to RT; external memory is better for me to remember about stuff
flussence tadzik: will do
tadzik flussence: thanks 16:36
16:36 static_perl left
jnthn pmichaud: o/ 16:37
[Coke] pmichaud: o/ 16:38
mls_ jnthn: here's an ugly patch for you to look at: gist.github.com/9be7df6c549eb2385b7e
16:38 kaleem joined, pjcj left
jnthn mls_: but...but...I like pretty things! :P 16:39
jnthn looks
mls_: My word, it's ugly :P
mls_ I don't see any better way to fix the :outer problem for CATCH blocks
jnthn mls_: What's the problem it's solving?
mls_: Well, I think (need to check the spec) we should be looking $/ up dynamically anyway
er, $!
16:40 pjcj joined
mls_ problem is, the CATCH block is already on the serialization queue 16:40
what about $_?
same thing?
jnthn Why doesn't CATCH have its outer being the block it's in?
That's what should happen really.
If we're fudging in some other frame at the moment, that really wants to go away, I suspect. 16:41
mls_ Ok, we can also do that.
(Currently there's a an extra block that creates the exception)
(maybe for historic reasons)
pmichaud ...there is? that seems odd. 16:42
jnthn is looking at the code and can't quite tell
But my brane is kind of a mess today, so :/
sorear nom is doing serialization now? I am so behind the times
jnthn sorear: Not in that sense...though hopefully, soon.
mls_ the first $handler := line creates it
pmichaud ah yes, I see the extra block. 16:43
jnthn *nod*
I'd rather see if we can not have that. Or at most, it only contributes to the dynamic scope, not the lexical one. 16:44
pmichaud perhaps the handler should add code to the catch block instead of wrapping it in a new one
jnthn That sounds preferable.
mls_ maybe it was added so that the handler doesn't change $_/$/...
pmichaud I suspect it was simply "get something working" 16:45
since many aspects of exceptions were still fuzzy at the time
mls_ (Btw, jnthn, if you want a 10% speedup for mandelbrot, just change "$z.abs > 2" to "$z.abs > 2e0") 16:47
mberends sorear: the Gtk text editor is beginning to work with plain text. Do you think we could get some syntax highlighting to work, and would you have time to help me when I get tuit (after next week's release)? Btw, I'll probably make that #phasers time on Wednesday via mobile connection during decommute.
PerlJam mls: that seems so backwards
mls_ yeah, that's something the optimizer should do ;) 16:48
16:48 birdwindupbird left 16:49 replore_ left
jnthn Hm...not sure if the optimizer can do that too easily. 16:49
It may be able to know that it's impossible for any of the same type candidates to match, mind.
mls_ Hmm, true, it can't assume that it's ok to promote a Int to Num in all cases 16:51
tadzik mls_: why does it give such a speedup?
jnthn It avoids a re-dispatch.
mls_ No more calls to .Bridge
tadzik ah 16:52
jnthn Num + Num ==> dispatch straight to the right candidate
Num + Int ==> coercion, then re-dispatch.
I suspect that kinda thing will only be made to look ever more ugly once we have an optimizer.
sorear mberends: ...why are you writing a text editor?
my C text editor is slow enough already.
pmichaud My gut still tells me that .Bridge is the wrong answer to this problem. 16:53
sorear is not thrilled with Bridge either
jnthn pmichaud: That may be so. But even if it isn't, there's still probably going to be a double-dispatch (the multi candidate that coerces, then the one that does the work). 16:54
pmichaud jnthn: I don't see a problem with special dispatch candidates for Int/Num variants.
jnthn pmichaud: Well, there's that option too :)
pmichaud: They do blow up exponentially
pmichaud I did a little playing around with some of these cases and managed to reduce things by one or two dispatches 16:55
jnthn But I guess a given program will probably only use to many variants.
s/to/so/
pmichaud part of the problem is that .Bridge only fixes one candidate at a time, when it should probably be done in pairs
sorear the blowup is only quadratic
jnthn sorear: ah, true.
sorear: That's bad enough.
16:56 supernovus left, thou joined
jnthn Int, Num, Rat, Complex... :) 16:56
Guess it ain't so bad.
pmichaud Str
unless we're willing to live with the redispatches there :)
jnthn pmichaud: I may be willing to live with those. :) 16:57
sorear pmichaud: I think that "multiple dispatch" is only a sensible approach to the numeric ladder if the dispatcher itself knows about the Int->Rat->Num->Complex subtyping relationship
jnthn I guess Rat + Int is going to be commonish.
sorear: Well, the problem is that it's not a subtyping relationship.
16:57 mberends left, pjcj left
jnthn nom: say Int ~~ Complex 16:57
p6eval nom 35ec9c: OUTPUT«Bool::False␤»
jnthn niecza: say Int ~~ Complex 16:58
p6eval niecza v9-32-g380d891: OUTPUT«Bool::False␤»
pmichaud I'm not certain about that relationship, fwiw.
sorear jnthn: exactly
jnthn: as the types are currently defined, multiple dispatch on + doesn't make sense
jnthn Well, in a sense, if people are trying to write fast code, they'll write in types. And probably use native ones anyway. 16:59
pmichaud well, we'd like it to be "fast" even if they don't. :-) 17:00
jnthn Well, multi-dispatch in terms of the decision making process is pretty fast in nom now. :-) It's just that our only current backend is kinda slow on the invoke part, so re-dispatch is a pretty painful cost. 17:01
mls_ hmm, let me run the profiler... 17:02
jnthn mls_: ooc, what does one feed the output of your profiler too? Or is it meant for direct human consumption?
mls_ no, it's meant for kcachegrind
jnthn oh!
OK :) 17:03
mls_ dunno if it exists for windows
jnthn I did manage to coerce it into running on Windows once before.
17:06 pjcj joined
mls_ invoke() cost seems to be 1% 17:06
17:07 JimmyZ left
moritz that's... surprisingly little bad 17:07
jnthn uh.
pmichaud ops or ticks?
moritz wait, is that PIR level invoke?
jnthn That's WAY off what you'll see if you look with a C profiler. 17:08
mls_ I mean the invoke for the Bridge calls ;)
jnthn oh.
mls_ ticks
jnthn Where is mandelbrot spending a lot of time, ooc?
mls_ on a hll or opcode level? 17:09
jnthn hll
17:10 plobsing left
mls_ well, the most time is spent in that inner block comparing z against 2 17:11
17:11 plobsing joined
mls_ but only 6.5% is spent with opcodes from that block, about 30% is infix<+>, infix<>>, infix<*>, abs 17:13
jnthn *nod*
PerlJam so ... all we need is a single op to do all that and it'll be *much faster ;) 17:14
mls_ ah, yes, a mandlebrot op!
mandel
jnthn PerlJam: Well, or to implementing inlining ;) 17:15
17:17 mberends joined
mls_ mandelbrot also spends 10% in @color_map[$_]. I wonder why that's so expensive. 17:17
pmichaud subscripting is generally expensive 17:18
it can likely be optimized, but there's a fair bit that happens there
afk, lunch
moritz mls_: too many redispatches, probably 17:19
17:19 MayDaniel joined, SHODAN joined
mls_ yes, many different functions involved 17:19
17:20 packetknife joined
jnthn food shop, bbiab 17:20
moritz postcircumfix, at_pos, exists, gimme, 17:21
17:22 sayu left
TimToady well, the thing Bridge should probably be doing is allomorphic caching 17:23
not to the extend P5 mushes its scalars, but enough to not do the same work over and over
*extent
and constants like 2 should certainly be as allomorphic as possible 17:24
well, within the bounds of Coolness
17:25 packetknife left
TimToady in the absense of type inference, the fastest run-time strategy will probably be do assume memoizing is useful on conversions, and disable it on values that are written as often as they are converted 17:26
(once that trend is noted)
P5 does some tricks like that
though mostly on regex optimizations 17:27
but since a constant can't be rewritten, memoization is a no-brainer 17:28
esp if it can be done at compile-time, but then it starts looking like type inference :) 17:29
sorear the complicating factors here are 1. multithreading 2. memory usage
TimToady both good reasons to figure it out at compile time :)
PerlJam no threaded compiles?
moritz if I change the $z.abs > 2 to $z.abs > 2e0, mandelbrot goes down from 48.6s to 42.1s 17:30
TimToady that's what I'm talking about
moritz nom: say (48.6-42.1) / 48.6
p6eval nom 35ec9c: OUTPUT«0.133744855967078␤»
im2ee What is exactly a "nom"?
PerlJam im2ee: it's the sound one makes while eating :)
17:30 mj41 joined
TimToady coincidentally, it stands for New Object Model 17:30
PerlJam im2ee: or it's just the New Object Model branch of Rakudo
im2ee TimToady, PerlJam , thanks. :) 17:31
TimToady sorear: the single-ownership model of mutables probably helps with memoization under threading 17:32
mls_ afk -> home...
TimToady if you're not the owner, you don't do it 17:33
with immutables, it doesn't much matter who does it, as long as it's idempotent
sorear who understands the single-ownership-model?
TimToady beats me, but that never stopped me before 17:34
I'm that fool that rushes in, y'know...
but I suppose you might think of it as a funny kind of monad 17:35
colomon as opposed to ingy, who is a funny kind of nomad? 17:37
TimToady
.oO(a 'mandelbrot op' is sorta what hotpath analyzers *do*, y'know...)
 17:41
jnthn knows :)
Just non-trivial (and fun ;-)) to build one :) 17:42
PerlJam As soon as M0 hits the streets, builing a mandelbrot op on-the-fly should be easier.
s/builing/building/
TimToady re subscript overhead, that's one of the reasons for the sigil policy, so that we can know that @a[$x] is a single subscript by definition, since $x can't flatten
so it can hotpath around the whole slice processor 17:46
P5 also had an optimization for literal subscripts, btw
ingy is nowmad 17:50
phenny ingy: 14:57Z <TimToady> tell ingy that if your grammar assumes ^ $ you can never use a grammar to parse a portion of a string, such as an embedded block in a different language
ingy TimToady: I'm not certain of that 17:51
I can invoke a parse with any start rule 17:52
which means that I can have actions that do launch entire new parses 17:53
17:53 mj41 left
ingy and then put the result of the sub parse into the ast that triggered them 17:53
.
17:56 mj41 joined 18:01 slavik1 left 18:04 slavik1 joined 18:07 molaf left 18:09 benabik joined 18:12 jamtech left 18:13 jamtech joined 18:14 packetknife joined 18:18 pernatiy left 18:22 PacoLinux left
sorear out 18:24
18:24 molaf joined 18:26 reokie joined 18:29 reokie left 18:30 envi_|2 left 18:38 packetknife left 18:40 PacoLinux joined 18:46 shinobicl_ joined 18:50 kaleem left 18:58 Caldrin joined 18:59 im2ee left, im2ee joined
moritz pmichaud: ping 18:59
19:14 masak joined
masak OH HAI #perl6 19:14
moritz lol it's masak! 19:15
jnthn o/ masak 19:16
mberends oh hai masak 19:18
TimToady o̐\ 19:19
(doffs hat)
jnthn :)
TimToady: So, there should be a $/ and $! per sub, and the mainline is a "sub" for this too. Is the lookup of them still lexical? 19:20
Something about the "gather and async must also define them" made me think there was a switch to a dynamic lookup, but I just realized that didn't really follow :) 19:21
TimToady last I heard :)
jnthn OK.
TimToady but it bears more thought... 19:23
jnthn If you say that I'm gonna put off doing anything until you've thunk more :P
19:26 daxim left
TimToady that's probably...wise... 19:28
ingy hi TimToady
PerlJam But then if you don't do something, you can't twiddle his thinking (which is probably why he needs to think more)
jnthn TimToady: Is it safe to assume the "one $! and $/ per sub" thing is here to stay? 19:29
Or, more likely than not? 19:30
Well, if I do it I guess I make some bugs go away...so provided it doesn't make more new ones on the way it's a win ;)
TimToady likely to stay, I think, but we do need to figure out how not to set $/ from things like grep: /pattern/ 19:31
jnthn Hm
yeah
That's why I was wondering about dynamic.
TimToady and grep: sub { /pattern/ } is not the answer
jnthn ew, no :)
It also fixes the subst case :) 19:32
TimToady well, even if we call it dynamically set, dynvars are still defined lexically
moritz grep could localize $CALLER::</>
jnthn ...wait, what? :)
TimToady: I menat dynamically looked up.
19:32 MayDaniel left
ingy TimToady: can you reexplain why I would not want to anchor grammars 19:32
19:32 REPLeffect left
TimToady jnthn: yes, but the setting of $/ from within a matcher doesn't have to use that mechanism 19:33
19:33 shinobicl_ left
jnthn $foo.subst(/blah/, { #`the $/ in here is the one set in subst` }); #`out here, $/ not touched` 19:33
TimToady what if you have a string that contains leading and trailing garbage you want to ignore; how do you parse that without substr()? 19:34
jnthn: yes, that argues for dynamic reads too 19:36
19:37 birdwindupbird joined, birdwindupbird left
jnthn TimToady: Yeah. I'm curious what the counter-examples for dynamic reads look like :) 19:37
TimToady I suppose something that wants to be transparent can do $/ := CALLER::<$/> or some such
ingy TimToady: well I feel like I'm creating an acmeist parsing language. it doesn't seem the job of a parsing language to write a json grammar that can pull json out of a garbage can. 19:38
but it I needed to do that I certainly could add garbage rules to the json-in-garbage grammar 19:39
TimToady the question is who's to be master, that's all...
that seems inside out to me 19:40
the P6 view is that there is no priviledged level of grammar, and every lexical scope can have its own language, as it were 19:41
ingy well I can certainly add a add-garbage-rules-to-my-grammar option in the parser
but I can't imagine ever needing it
unless I was wanted my grammars to also be general purpose regexen :D 19:42
cognominal nom: my ($a, $b) = "ab" ~~ /(a)(b)/; say "'$a' '$b'"
p6eval nom 35ec9c: OUTPUT«'ab' ''␤»
TimToady a parser at any level just starts at a point in the string, and ends at some other point; it's the outside of that that sets it going in the right starting spot, and verifies that the end spot is adequate
cognominal what am I doing wrong here?
jnthn nom: my ($a, $b) = ("ab" ~~ /(a)(b)/).list; say "'$a' '$b'" 19:43
p6eval nom 35ec9c: OUTPUT«'a' 'b'␤»
PerlJam cognominal: you're not flattening or it's a bug
jnthn I think you gotta .list it yourself. 19:44
ingy TimToady: sure but for json or any other dsl you always start at the start and end at the end, so if I publish 1000 grammars and they all have ^ ... $ obviously I mad a mistake
TimToady I suspect list assignment should do that for you
ingy: it's the thing that applies the grammar that checks that, not the grammar itself 19:45
or you just have alternate top rules, like the P6 grammar does 19:46
TOP checks, and comp_unit doesn't
TOP also also allows an alternate end to be specified from outside 19:47
ingy so if I define a rul called TOP I don't need ^ or $ in my grammar?
pegex has a TOP special rule too...
TimToady well, just don't make it impossible to work as an embedded parser, I don't much care how you don't do it :) 19:48
19:48 MayDaniel joined
cognominal thx jnthn++, TimToady++ 19:49
TimToady use inline python; python { ... }
ingy TimToady: did you see earlier: gist.github.com/1227402 19:50
19:50 im2ee left
ingy TimToady: and you didn't answer: 15:47 < ingy> so if I define a rul called TOP I don't need ^ or $ in my grammar? 19:51
does TOP imply ^ $? 19:52
the Pegex TOP is special in that if you start a parse with rule aaa the default ast is {aaa: ...} whereas if you start with TOP it is just ... 19:54
jnthn Well, switching to just one $! and $/ per routine, not per block, didn't break the build...let's see how spectest fares... 19:55
...oh dear. 19:56
TimToady ingy: no, TOP does not imply ^ $
ingy anyway I think that when you use a grammar/parser the normal case is that you want to parse an entire text. but parsing a substr should be a (non-default) option 19:57
ashleydev Couldn't you add something equivalent to Lisp macros to languages like Perl or Python?: www.paulgraham.com/raq.html
phenny ashleydev: 31 Aug 07:14Z <masak> tell ashleydev that yes, macros can be both AST-based and unhygienic. just defer all variable checks until macro-apply-time, and then have $x in the macro be the same as $x outside the macro. it's probably easier to implement than hygiene.
ingy but I'm not entirely sure of myself, so I show up on #perl6 to eat brains 19:58
ashleydev "Not without turning them into dialects of Lisp."
TimToady parsers should not assume how they will be used
arnsholt ashleydev: I read that too yesterday. I was amused ^_^ 19:59
ingy parser should assume how they will be used
arnsholt masak: Apparently you're going to turn Perl 6 into a dialect of Lisp =D
ashleydev haha
TimToady too late
ingy and provide alternatives when are needed :)
donri you can simulate macros with python's standard libs
arnsholt Simulate macros? 20:00
donri well, in that they're not a language feature, but it does provide stuff like ast parser/compiler and import hooks
ingy TimToady: anyway, thanks for your input. I don't think I'm missing anything at this point.
20:03 bluescreen10 left 20:04 packetknife joined, localhost left
cognominal nom: my ($a, $b) = ( "a \n<-> b" ~~ / ^ (.*?) ^^ '<->' (.*) / ).list ; say "'$a' '$b'" 20:05
p6eval nom 35ec9c: OUTPUT«'' ''␤»
cognominal it seems that there is no backtrack over ^^
20:06 localhost joined 20:07 mkramer left
jnthn OK, I fixed eval up and now the switch to $! and $/ being per sub doesn't look quite so disasterous for spectest. 20:07
cognominal cool, jnthn++, it seems I run into it today. 20:08
masak arnsholt: what TimToady said. you seem to be under the impression that Perl 6 isn't a Lisp. 20:09
20:10 Chillance joined, BooK joined
jnthn gist.github.com/1227465 # fallout from $! and $/ per sub 20:11
arnsholt masak: Well, I do agree that Perl is definitely very Lispy
TimToady 'course, it's also a Smalltalk, Java, Erlang, awk, sed, shell, C, C++, C#...
arnsholt But phrasing it as only languages that are "dialects of Lisp" can have macros sort of implies the whole parentheticalness to me 20:12
cognominal nice potpourri
TimToady arnsholt: it's assuming that homiconicity can only be done with parens 20:13
masak well, only *real* homoiconicity :P
TimToady or however you spell that
well, the Perl philosophy is to fake everything so well you can't normally tell the difference 20:14
masak :)
arnsholt Quite, quite
TimToady we don't, however, define our lists using cons cells by default, though you can do cons using => 20:15
cognominal that's a lapsus, when TimToady thinks Lisp, he thinks homicide :)
jnthn TimToady: Should eval bind its result to $!, or assign?
TimToady part of the reason for simplifying the scoping of $/ and $! is to make it not matter, as long as we're consistent 20:16
jnthn TimToady: ah, well, it's consistency we lack it turns out, now I see what's happening 20:17
TimToady: try was binging, eval assigning.
er, binding :)
so try { ... }; eval '...' # epic fail :)
jnthn switches eval to bind to see what happens
TimToady it might be better to switch everything to assignment, so we can do $/ := CALLER::<$/> tricks 20:18
jnthn heh
turns out Parrot defines a find_caller_lex but not store_caller_lex
20:18 _jaldhar is now known as jaldhar
jnthn So if I switch to assignment I don't have to go and hack C to add an op 20:18
So, assign wins ;)
TimToady \o/
20:20 bluescreen10 joined, janosik left
BooK TimToady: did you get my email? about a T-shirt 20:21
20:22 kshannon joined
TimToady ah, yes, I had overlooked it 20:23
jnthn TimToady: Looks dramatically better after switching both to assign. 20:25
Well, making try assign too.
Yay. Down to just two test files with regressions now, both in S05. 20:27
20:32 kaare_ left
TimToady if $! and $/ become dynamic, we'd need some way of declaring control subs that don't declare $! and $/ 20:33
or force them to do $/ := CALLER::<$/> by convention 20:34
jnthn fwiw, Rakudo's setting doesn't declare control-y things as subs.
TimToady innersting
that's anudda way
jnthn Well, depends what you mean by "controlly things" 20:35
my &redo := -> |$ { ... # for example
Not quite sure what the rationale was for it; you'd hafta ask pmichaud++
TimToady I'm thinking of things that take closures that used to close over OUTER::<$!> and such
your ifs and whiles and such
if we're going to look up CALLER::<$!> via their stack frames, they need to be transparent 20:36
jnthn ah, I see
Yeah, that's the big downside of them becoming dynamic. 20:37
How waterbedesque.
TimToady but control routines seem relatively rare, so can be de-huffmanized
jnthn True. Whereas .subst is quite common :)
OK, I have only 1 test file with regressions left. 20:38
TimToady or such routines can copy out their final $! or $/ if they want
.match would, .subst wouldn't, to the first approximation
or it's in the // s/// macro logic 20:39
jnthn .subst not doing so would make masak happy :)
ah, I think the failing test file is a victim of the changed semantics. 20:40
TimToady but then how does s/// set $/, if it's hidden down in .subst?
20:40 packetknife left
jnthn Maybe either s/// doesn't desugar to .subst (though it probably should) or it passes along an adverb. 20:41
{
my $str = "abcabcabc";
ok($str ~~ m:p/abc/, 'Continued match');
TimToady an optional rw param to .match and .subst seems messy
masak well, I hope you solve this ;) I'm going to bed.
jnthn Here, it expected a fresh $/.
masak g'nite, #perl6.
jnthn night, masak
20:41 masak left
tadzik night 20:41
dang!
jnthn :) 20:42
tadzik and he's talking about leaving too early!
TimToady tadzik: you need an autoresponder :) 20:43
jnthn: fix with a 'my $/' maybe
tadzik well, it's Sweden, the road to bed is probably snowy, freezing and full of Polarizzly bears. Better leave to soon than too early 20:44
dalek ast: b68a710 | jnthn++ | S05-modifier/pos.t:
Update a test for current $/-per-routine spec.
 20:45
jnthn TimToady: coulda done that too..
OK, final run through the spectests justin case, then I push. 20:46
20:46 pmurias joined, wallberg joined
pmurias ingy: if TOP implies ^ and $ how am i supposed to turn that off? 20:47
PerlJam pmurias: matach a subrule? 20:48
dalek kudo/nom: 7563146 | jnthn++ | src/ (3 files):
Switch to having a $! and $/ per routine, not per block. Also consistently use assignment, not binding, for $! and $/.
 20:49
BooK TimToady: thanks 20:50
pmurias PerlJam: and if TOP contains a lot of stuff? copy and paste it?
jnthn There we go...that wasn't so bad as I feared. Time for a beer.
tadzik what does it fix? 20:51
oh, I think I know
jnthn tadzik: The patch I just did? 20:53
tadzik: Our non-compliance with the spec ;)
Though maybe some other bits too :)
pmurias: No, TOP would just ^ <that_rule> $ I guess.
20:55 SHODAN left
TimToady cognominal: yes, niecza and ng seem to agree it's a bug in nom 20:56
perl6: my ($a, $b) = ( "a \n<-> b" ~~ / ^ (.*?) ^^ '<->' (.*) / ).list ; say "'$a' '$b'"
p6eval niecza v9-32-g380d891: OUTPUT«'a ␤' ' b'␤»
..pugs: OUTPUT«Error eval perl5: "if (!$INC{'Pugs/Runtime/Match/HsBridge.pm'}) {␤ unshift @INC, '/home/p6eval/.cabal/share/Pugs-6.2.13.16/blib6/pugs/perl5/lib';␤ eval q[require 'Pugs/Runtime/Match/HsBridge.pm'] or die $@;␤}␤'Pugs::Runtime::Match::HsBridge'␤"␤*** '<HANDLE>' trapped b…
..rakudo 35ec9c: OUTPUT«'' ''␤»
jnthn TimToady: I think nom still has general issues backtracking, fwiw. 20:57
TimToady well, a pm thing, likely
jnthn very much so
TimToady whoa, Method 'postcircumfix:<[ ]>' not found for invocant of class 'CallContext' 20:59
jnthn o.O
That should really not happen.
TimToady the line in question: $s ~= @row[^20]».fmt("%3s ").join.subst(/^' '/,''); 21:00
which doesn't tell you much :)
tadzik hehe 21:01
that reminds me of something :>
jnthn tadzik: yes. :( 21:02
tadzik: Wasn't it something like that we were seeing in wherefore-gc?
tadzik jnthn: certainly
jnthn :'(
TimToady and if I take out the » it segfaults
jnthn That's bad news.
tadzik also when I tried to fix multi-line docs
jnthn TimToady: bt may help. 21:03
TimToady: But I doubt it. It's almost certainly very action-at-a-distance.
(GC issue)
tadzik TimToady: try with parrot -G perl6.pbc 'file' maybe?
TimToady hope I have enough memory :) 21:04
tadzik if you can build nom you surely have lots of it :P
21:05 benabik left
TimToady well, it got a different error, after a great deal of thrashing on a 3gig system :) 21:11
tadzik :P
TimToady so yeah, probably GC corruption
21:11 MayDaniel left 21:16 wallberg left 21:27 sbp left, nsh left
sorear good * #perl6 21:54
jnthn: what's the last word on $/ and $!?
diakopter sorear: his last commit made them per-routine, afaik 21:56
jnthn sorear: Latest is that they're per-routine (and mainline gets them)
sorear: And whether lookup of them is dynamic or lexical is still up in the air
sorear: nom has it as lexical
(mostly because that was the current state)
I think niecza has them per-routine anyway? If so, in that regard I think we just converged a bit :)
Also, $/ and $! are now consistently assigned to, not bound. 21:57
(We did a bit of each before...) 21:58
(As in, nom did a bit of each)
sorear jnthn: I am tentatively -1 to assigning to $/ and $!. I have an optimization which saves two allocations per routine entry that requires $/ and $! to be bound, not assigned 22:00
jnthn sorear: Well, it should be all assignment or all binding, not the accidental mixture we had. TimToady++ preferred assignment as it makes $/ := CALLER::<$/> style things possible. 22:01
sorear jnthn: requiring allocating two containers in the prologue makes me :(
jnthn sorear: It'd in theory make me :( too but (1) it's per routine and (2) I allocate them lazily in nom. 22:02
sorear jnthn: I wasn't aware that there was a mixture - I thought it was all binding
jnthn: please elaborate on (2)
jnthn sorear: The mixture in nom was accidental, not intentional.
sorear: I have a "trap" for if the lexpad lookup comes back as null (which should never happen). If it hits that trap, it checks if it's one of the magicals being looked up and creates the container then. 22:03
So things that never need a $!, $_ and $/ never end up allocating a container.
I figure a smart inliner would be able to further cheapen that by tossing the check in a lot of cases. But for now a null check per lexical access is vastly cheaper than all the allocations. 22:04
22:05 bluescreen10 left
dalek kudo/nom: e062a4c | jnthn++ | src/Perl6/Metamodel/ (2 files):
Make role Foo[::T] does Bar[T] { ... } work.
 22:05
kudo/nom: 178ddf6 | jnthn++ | src/ (2 files):
Refactor so that we create a BUILDPLAN at a per-class level as well as for the whole hierarchy.
22:15 mj41 left 22:20 bluescreen10 joined
tadzik gah, I don't get it 22:26
token comment:sym<#=> { [ '#=' {say('almost there')} \h+ $<attachment>=[\N*] ] } prints 'almost there' 6 times 22:27
when I change that to [ ... ]+, it doesn't print it at all
22:30 jamtech left
jnthn tadzik: Confuses me too, fwiw. 22:34
Not sure why it'd be doing that.
tadzik maybe it causes the regex to be pre-compiled in a wrong way, for some reason
22:38 jamtech joined
jnthn mebb 22:38
*mebbe
dalek kudo/nom: d9b19d6 | jnthn++ | src/Perl6/Metamodel/MROBasedTypeChecking.pm:
Add missing .^does.
 22:39
kudo/nom: e97afdb | jnthn++ | src/core/ (2 files):
Ensure that when we mix in a role, its attributes are initialized appropriately (for example, any defaults are applied).
kudo/nom: 9256ee2 | jnthn++ | t/spectest.data:
We now pass S14-roles/parameterized-mixin.t.
kudo/nom: b10ef51 | jnthn++ | NOMMAP.markdown:
Remove completed tasks from nommap.
jnthn We can probably have a + on the parametric roles entry on features now too. :) 22:41
22:41 domm left 22:43 arnsholt left
diakopter jnthn: I'll update it 22:45
22:45 packetknife joined
jnthn diakopter: thanks! 22:46
Didn't get a clone of that on my new machine yet...
dalek atures: 8fae163 | diakopter++ | features.json:
17:41 < jnthn> We can probably have a + on the parametric roles entry on features now too. :)
 22:47
22:47 packetknife left
tadzik jnthn: why does BUILD_LEAST_DERIVED take arguments, some future plans? 22:53
jnthn yeah
$foo does Answer(42)
22:54 wamba left 22:55 pmurias left
tadzik mhm 22:56
sleepytime, g'night!
jnthn night, tadzik 22:58
23:01 jamtech left 23:02 jamtech joined 23:06 envi_|2 joined 23:08 buubot_backup left 23:11 molaf_ joined 23:16 molaf left 23:17 packetknife joined 23:18 bluescreen10 left 23:28 buubot_backup joined 23:30 bluescreen10 joined
jnthn sleep for me & 23:32
23:33 packetknife left 23:34 envi_|2 left 23:59 uasi joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!